{"fetchedAt":"2026-10-16T11:39:10.814581+00:00","source":"https://export.arxiv.org/api/query?search_query=cat:*&start=0&max_results=10&sortBy=submittedDate&sortOrder=descending","papers":[{"id":"2610.22372v1","title":"LIBERO-PeRM: Benchmarking Personalized Robotic Manipulation","abstract":"Personalization is essential for general-purpose robots to transition into household environments and achieve true practicality. While general robotic manipulation has been extensively studied, personalized robotic manipulation (PeRM) lacks a controlled benchmark for separating task completion from preference satisfaction. We introduce LIBERO-PeRM, a benchmark for preference learning in embodied agents. LIBERO-PeRM studies five questions: 1) how policies follow preferences in zero-shot settings; 2) how they learn single preferences from demonstrations; 3) how efficiently they acquire preferences as demonstrations increase; 4) how learned preferences generalize across layouts and tasks; and 5) how policies compose preferences and behave under conflicts. We propose a controlled formulation and taxonomy of observable preferences in simulation, covering motion, skill, and user-profile choices. This formulation is restricted to language-expressible preferences that can be instantiated in LIBERO without changing the task goal, evaluated automatically, and reused across tasks. Building on LIBERO, our procedural pipeline generates paired demonstrations with and without preferences. We create five task suites (312 tasks) with high-quality demonstrations, rich metadata, and preference-satisfaction metrics. Experiments show that current VLA policies struggle with zero-shot personalization, while SFT makes preferences learnable. Mixed training achieves better no-preference retention than sequential adaptation and improves performance with contextual prompts compared with preference-only training, but can weaken motion-preference learning; preference generalization, decomposition, and conflict handling remain difficult. The code, datasets, and demos can be found on the project website: https://libero-perm.github.io/.","published":"2026-10-15T17:59:59Z","updated":"2026-10-15T17:59:59Z","authors":["Zhixu Li","Keqian Tang","Litian Gong","Jingyu Yao","Wenqian Zhang","Tianze Xu","Zehao Wang","Yuping Wang","Junge Zhang","Jiachen Li"],"categories":["cs.RO","cs.AI","cs.LG"],"headline":"New benchmark tests how robots learn and follow personal preferences","summary":"Researchers introduced LIBERO-PeRM, a simulation-based benchmark designed to study how robots can learn and apply personal preferences while doing tasks. They created various tasks and demonstrations to evaluate robot policies on preference learning, generalization, and handling conflicting preferences. Their experiments show current methods struggle with some aspects of personalization, highlighting challenges in preference learning and adaptation. This work focuses on simulated preferences expressed in language without changing the task goals and provides data and tools for further study."},{"id":"2610.22371v1","title":"Gen4P: Video Generators are 4D Perceivers","abstract":"We introduce Gen4P, a unified framework for recovering 4D geometry and motion from monocular video. We show that a pretrained video generator can be repurposed into a feed-forward 4D perceiver without iterative denoising. Building on on-demand querying, Gen4P processes clean video latents in a single encoder pass and predicts the requested 4D attributes through a lightweight cross-attention decoder. The on-demand query formulation enables flexible spatial sampling, and our proposed continuous-time queries further provide temporal flexibility. To strengthen geometric knowledge, we propose a straightforward plug-in solution by incorporating a pretrained video-depth LoRA as a frozen geometry prior, improving overall performance without adding extra trainable parameters. Gen4P achieves state-of-the-art or competitive performance on tasks of 3D point tracking, video depth and camera pose estimation. Qualitative results on in-the-wild videos illustrate its capability of tracking complex motion while maintaining the delicate structure. Together, these results demonstrate the potential of video generative pretraining for unified 4D perception. See https://gen4p.github.io for more visual results.","published":"2026-10-15T17:59:48Z","updated":"2026-10-15T17:59:48Z","authors":["Yitong Xia","Zeren Jiang","Chuanxia Zheng"],"categories":["cs.CV"],"headline":"New AI method extracts 3D motion from single videos","summary":"Researchers developed Gen4P, a system that recovers 3D shape and movement over time from ordinary videos taken with one camera. They repurposed an existing video-generating AI to quickly estimate detailed 4D information without slow calculations. Their method improves accuracy by using a pre-learned depth understanding and works well on tasks like tracking points and estimating camera position. The findings are based on tests with videos, showing promise for better video analysis through this AI approach."},{"id":"2610.22370v1","title":"Which Tokens to Merge? Diffusion Dynamics for Efficient Image Generation","abstract":"Token merging has emerged as an effective training-free strategy for accelerating diffusion models by reducing redundant token computation while minimizing generation degradation. A key challenge is determining which tokens can be safely merged and which should be preserved for subsequent computation. Existing method addresses this by preserving prompt-aligned foreground regions, thereby protecting semantically relevant structures. However, this text-relevance criterion overlooks background objects, local details that are not explicitly mentioned in the prompt, leading to degraded structural coherence. In this work, we propose Dynamics-aware Token Merging (DaTMe), a training-free framework that grounds token merging in the underlying diffusion dynamics. Our key insight is that token importance should reflect whether a token remains unresolved during denoising and whether it carries structural information. To this end, DaTMe constructs a dynamics- aware importance map from the Tweedie denoised estimate. The map combines a temporal signal, measuring inter-step changes in the predicted clean image, with a structural signal, capturing local high-frequency content such as boundaries and fine details. Using this map, DaTMe adaptively assigns tokens into merging roles under a given compression ratio: important tokens are preserved as anchors, while tokens that are both temporally stabilized and structurally homogeneous are selected as safe merge candidates. Experiments on PixArt-a and FLUX demonstrate that DaTMe maintains visual fidelity across the entire image, highlighting diffusion dynamics as an effective token-level merging decision in efficient generation. Our code is available at https://github.com/SeungJuCha/DaTMe.","published":"2026-10-15T17:59:35Z","updated":"2026-10-15T17:59:35Z","authors":["SeungJu Cha","Ye-Chan Kim","HyunGee Kim","Sungho Koh","Dong-Jin Kim"],"categories":["cs.CV"],"headline":"New Method Improves Speed and Quality in AI Image Generation","summary":"Researchers studied how to speed up AI image generation by merging parts of an image that do not need detailed processing, without harming quality. They created a method called DaTMe that uses information about changes and details in the image during the generation process to decide which parts to merge and which to keep. Their experiments show that DaTMe helps maintain image quality while reducing computation, but this finding is based on simulated tests with specific datasets."},{"id":"2610.22369v1","title":"SoftCache: Adaptive Step Size Caching for Diffusion Transformers","abstract":"Diffusion and flow models generate a sample over iterative steps, each a full pass through a large network, and that pass dominates generation cost. Training-free caching accelerates them by reusing activations from recent evaluations instead of recomputing them. Dynamic caching methods decide when to refresh by accumulating a cheap probe of the network input into a staleness signal and spending a fresh evaluation once it crosses a threshold. They perform hard caching: the signal is a binary switch between refresh and reuse, and every step, fresh or cached, advances by the prescribed step size. We observe that the signal carries much more than this switch. How far it sits below the threshold, its headroom, predicts the error induced by reuse. We propose SoftCache, which performs soft caching: instead of a binary switch, the headroom sets how far each cached step goes. Where the headroom is large, the sampler jumps past several scheduled nodes, and as a refresh nears, its steps shrink back toward the original step. These adaptive jumps may land between the scheduled nodes, taking the sampler off its fixed grid. SoftCache requires no training and adds no per-backbone settings: on FLUX for images and on HunyuanVideo and Wan2.1 for video, at speedups from $3.5\\times$ to $6.9\\times$, it is ahead of every baseline on all four fidelity metrics.","published":"2026-10-15T17:59:34Z","updated":"2026-10-15T17:59:34Z","authors":["Shimon Malnick","Nadav Orenstein","Ohad Fried","Shai Avidan"],"categories":["cs.CV"],"headline":"New Method Speeds Up AI Image and Video Generation","summary":"Researchers studied how some AI models create images and videos step-by-step, which is slow because each step involves heavy calculations. They developed SoftCache, a technique that reuses parts of previous calculations more smoothly by adjusting how much each reused step contributes, instead of simply reusing or recalculating entirely. Tested on image and video tasks, SoftCache sped up the process significantly without extra training and improved quality measures compared to existing methods. These findings come from simulations and tests on specific datasets."},{"id":"2610.22368v1","title":"Value as Memory: Online Value-Guided Post-training for General Robot Policies","abstract":"Real-world post-training of generalist robot policies faces two challenges: partial observability makes action selection ambiguous, while offline post-training methods delay feedback from deployment experience. We present Value as Memory (VaM), a framework that addresses these challenges through value caching and online real-world reinforcement learning (RL). Value caching conditions actions on recent value predictions as task-progress memory, while online updates enable timely policy improvement by combining offline demonstrations, off-policy interventions, and on-policy rollouts for sample-efficient adaptation. As a result, VaM transitions from offline RL warmup to online autonomous learning with sparse rewards. Across long-horizon tasks involving deformable objects and contact-rich manipulation, VaM yields a mean absolute success-rate gain of $29.4\\%$ with value caching, while achieving a $32.8\\%$ gain with online autonomous learning over matched offline training. The evaluated online schedule performs $8\\times$ as many policy iterations under the same gradient-step budget, with more than $2\\times$ lower recorded time per gradient step with concurrent data collection and training.","published":"2026-10-15T17:59:31Z","updated":"2026-10-15T17:59:31Z","authors":["Wenye Yu","Yijin Chen","Kaiyi Xiong","Wenbo Han","Wendi Chen","Yang Jin","Han Xue","Jun Lv","Cewu Lu","Chuan Wen"],"categories":["cs.RO"],"headline":"New Method Helps Robots Learn Tasks Better and Faster","summary":"Researchers developed a method called Value as Memory (VaM) to improve how robots learn in real-world settings. VaM helps robots remember recent progress and update their skills efficiently by combining past examples with new experiences. Tests on tasks involving flexible objects and complex interactions showed that VaM improved success rates by about 30%, enabling quicker and more effective learning compared to previous approaches."},{"id":"2610.22367v1","title":"Discovering state-of-the-art quantum circuits with large language models","abstract":"Significant efforts are currently being directed toward the implementation of large-scale quantum information processors, but realizing useful quantum advantage also requires the development of efficient quantum algorithms. While for many practical problems the dominant resource costs are associated with a small set of computational primitives, optimizing them in the quantum setting presents unique challenges, including a reversibility constraint absent from ordinary classical computation. Here we introduce an evolutionary framework in which large language models (LLMs) propose circuit designs and an independent evaluator verifies and scores them. We apply this framework to tasks spanning numerical computation, chemistry, optimization, quantum simulation, and number theory. In seven of the eight tasks, the discovered circuits significantly reduce both Toffoli and qubit counts relative to the state of the art. In the eighth, the search discovers a new reversible architecture for prime-field modular inversion. By handling reversibility differently, the architecture makes divsteps -- a classical algorithm previously impractical for reversible integer arithmetic -- viable and substantially improves the Toffoli-qubit trade-off for modular inversion. These results open the door for evaluator-guided LLM co-design of new algorithmic structures for quantum computation.","published":"2026-10-15T17:59:20Z","updated":"2026-10-15T17:59:20Z","authors":["Andi Gu","Alek H. Kemeny","Susanne F. Yelin","Mikhail D. Lukin"],"categories":["quant-ph","cs.AI"],"headline":"Using AI to improve quantum computing circuit designs","summary":"Researchers developed a system where large language models suggest quantum circuit designs, and an evaluator checks and scores them. They tested this approach on various problems like chemistry and optimization, finding improved designs that reduce key resource needs in most cases. The study's results suggest AI can help create more efficient quantum algorithms, but findings are based on specific tasks and require further validation."},{"id":"2610.22366v1","title":"Uniform estimates for elliptic systems with critical Sobolev coefficients","abstract":"The $W^{1,p}$ estimate for uniformly elliptic operators of divergence form with VMO coefficients is classical. In particular, this provides $W^{1,p}$ estimates for elliptic operators with coefficients in $W^{1,n}$. However, this theory does not yield constants in the estimates controlled by the $W^{1,n}$ norm of the coefficients, since the total $W^{1,n}$ norm does not control the rate of decay for the mean oscillation. In this work, we establish uniform estimates for divergence-form elliptic systems whose coefficients lie in the critical Sobolev--Lorentz spaces $W^{1,(n,s)}$ with $1\\le s<\\infty$, have an $L^\\infty$-bounded spatially symmetric part, and satisfy an integral ellipticity condition. The constant in the estimate is controlled by the corresponding norms of coefficients and the ellipticity constant.","published":"2026-10-15T17:59:13Z","updated":"2026-10-15T17:59:13Z","authors":["Tian Lan"],"categories":["math.AP"],"headline":"New estimates for mathematical equations with challenging coefficients","summary":"The researchers studied a type of mathematical equation called elliptic systems, which are important in physics and engineering. They focused on cases where the equation's coefficients have a critical level of smoothness, measured in advanced function spaces. The team developed new uniform estimates that control solution behavior based on these coefficient norms, improving previous results that lacked such control. Their findings are theoretical and provide better understanding of these complex equations under specified technical conditions."},{"id":"2610.22365v1","title":"$^6$He structure from semiclassical 3D models based on $\\textit{ab initio}$ spatial correlations","abstract":"We construct semiclassical 3D models based on $\\textit{ab initio}$ no-core shell model calculated two-body spatial correlations, which represent the preferred spatial configuration of nucleons for the ground state of $^6$He. We demonstrate that the presence of the two valence neutrons distorts and flattens the $α$ core in $^6$He, marking a clear departure from the tetrahedral configuration of free $^4$He, and that the two valence neutrons and the center of mass of the $α$ cluster ($α$-CM) are roughly located at the three distinct vertices of an equilateral triangle. We compute the geometric parameters that commonly characterize $^6$He structure, 1) $θ_{nn}$, the opening angle of the two valence neutrons with respect to $α$-CM, and 2) $r_{nn}$, the root-mean-square separation between the two valence neutrons. Refining the model with uncertainty estimation, we obtain $θ_{nn}= 58.5^{+5.4}_{-6.2}$ degrees and $r_{nn}= 4.12^{+0.37}_{-0.43}$ fm, agreeing with those calculated via a phenomenologically modified continuum-discretized coupled-channels method, while an interferometry experiment obtained a marginally higher $r_{nn}$.","published":"2026-10-15T17:59:12Z","updated":"2026-10-15T17:59:12Z","authors":["Mengyao Huang","Tobias Frederico","Peng Yin","James P. Vary"],"categories":["nucl-th"],"headline":"Researchers model helium-6 nucleus shape and neutron positions","summary":"Scientists created 3D models based on fundamental calculations to understand the arrangement of particles in the helium-6 nucleus. They found that two extra neutrons change the helium-4 core shape and together with it form an equilateral triangle pattern. The researchers measured angles and distances between neutrons, with results supporting some previous theoretical predictions but differing slightly from one experiment. Their work helps describe helium-6's structure through improved spatial correlations, though findings come from models and comparisons rather than new experimental data."},{"id":"2610.22364v1","title":"T2-z: Test-Time Optimization of Environment Latents for Robotic World Models","abstract":"Predicting how the world responds to an action requires more than what can be seen: robots must account for environmental factors that are neither directly observable nor explicitly specified, from physical properties such as friction and mass to human-defined environmental rules. We introduce T2-Z for test-time optimization of environment latents, which augments a robotic world model with a learnable environment latent to condition world predictions. This compact latent captures underlying environmental variation without supervision of what environmental factors exist or what values they take. During training, T2-Z alternates between two stages: a world-model optimization stage that fixes the latents and learns general action-conditioned dynamics across environments; and an environment-latent optimization stage that freezes the world model and optimizes the conditioning latents, allowing latent structure to emerge purely through world prediction. At test time, the world model remains frozen and a newly initialized environment latent is optimized from one or more test-time interactions, adapting the model to previously unseen environmental configurations. Across 7 simulated settings and 4 real-world robotic tasks, T2-Z reduces object prediction error metric from 15.20 to 7.00 and raises action success from 33.73% to 72.96%. Its emergent latents capture structured representations of individual and compounded factors across diverse visual scenes. Ablations further show that T2-Z can efficiently infer environment latents from as little as a single test-time interaction trajectory.","published":"2026-10-15T17:59:11Z","updated":"2026-10-15T17:59:11Z","authors":["Chenyuan Zhou","Yining Hong","Huang Huang","Yunfan Jiang","Wenlong Huang","David Held","Vitor Campagnolo Guizilini","Yue Wang","Li Fei-Fei","Jiajun Wu","Leonidas Guibas","Yejin Choi"],"categories":["cs.RO","cs.AI","cs.CV"],"headline":"New Method Helps Robots Adapt to Unseen Environments Faster","summary":"Robots often struggle to predict outcomes because environmental factors like friction or rules aren’t directly visible. Researchers developed T2-Z, a method that lets a robot's internal model learn a simplified representation of these hidden factors without prior knowledge. By adjusting this representation during testing, the robot can better predict and act in new settings. Tests in simulations and real tasks showed T2-Z improved prediction accuracy and success rates, though these results are based on specific experiments described by the researchers."},{"id":"2610.22363v1","title":"Perceive, Predict, and Plan: a Unified World Action Model for Egocentric Dexterous Manipulation","abstract":"We introduce P$^{3}$-WAM, a unified world action model for egocentric dexterous manipulation that learns 3D perception, action-conditioned prediction, and goal-directed planning through three complementary modes: Perceive, which reconstructs articulated hand motion and camera trajectories from video; Predict, which generates visual futures under prescribed actions; and Plan, which jointly generates action trajectories and their anticipated visual consequences conditioned on an initial image, a language instruction, and an optional visual goal. Our model integrates a compact, structured 3D action representation into a pretrained video diffusion transformer through a mixture-of-transformers architecture with shared base weights, modality-specific adaptation, and bidirectional visual-action attention. Randomized modality visibility lets video and 3D actions serve as either conditioning context or prediction targets across the three modes. Experiments show that unified training improves all three capabilities over single-mode training, including gains in planning quality under a vision-language model evaluation. Through unified training, our model outperforms specialized baselines on 3D hand reconstruction and action-conditioned video prediction across four datasets, while producing goal-directed manipulation videos with stronger instruction adherence and action progression than video-generation baselines. Project Page: https://p3-wam.pages.dev/","published":"2026-10-15T17:59:07Z","updated":"2026-10-15T17:59:07Z","authors":["Zirui Wu","Chen Guo","Tianjian Jiang","Zeren Jiang","Zeming Li","Lingchen Yang","Longteng Duan","Ying Xue","Handi Yin","Shenyuan Gao","Martin R. Oswald","Jun Zhang","Ping Tan","Jie Song"],"categories":["cs.CV","cs.RO"],"headline":"New AI model improves hand motion prediction and planning from video","summary":"Researchers developed P³-WAM, a model that learns to perceive hand motions, predict future visual scenes based on actions, and plan actions to achieve goals using video and language instructions. It combines 3D hand data with video analysis in a unified system that trains these tasks together, improving performance compared to training each separately. Their tests show better hand reconstruction and video prediction, as well as goal-directed planning that follows instructions more closely, though the work is based on dataset experiments and simulations."},{"id":"2610.20611v1","title":"A global analysis of ${\\bar B}\\to DP$ and ${\\bar B} \\to {\\bar D} P$ decays","abstract":"Recently, a fit to the 15 measured branching ratios of ${\\bar B} \\to D P$ decays ($B = \\{B^+, B^0, B_s^0\\}$, $D = \\{D^0, D^+, D_s^+\\}$, $P \\in \\{π, K\\}$) was performed under the assumption of flavor SU(3) symmetry [SU(3)$_F$]. A terrible fit was found: there was a $> 5σ$ disagreement with the SU(3)$_F$ limit of the Standard Model [SM$_{\\rm{SU(3)}_F}$]. It was also found that this discrepancy can be removed by adding $\\sim 20\\%$ SU(3)$_F$-breaking effects. Now, the ${\\bar B} \\to D^0 P$ amplitudes can contribute to other observables (direct and indirect CP asymmetries) if they interfere with the amplitudes for ${\\bar B} \\to {\\bar D}^0 P$ decays. This happens if the $D^0$ and ${\\bar D}^0$ decay to the same final state (e.g., $K^+ K^-$, $π^+ π^-$, $K^\\pm π^\\mp$, etc.). In this paper, we perform a combined fit of ${\\bar B} \\to D P$ and ${\\bar B} \\to {\\bar D} P$ decays, including all observables. We again find a $> 5σ$ disagreement with the SM$_{\\rm{SU(3)}_F}$, and we find that the data can be explained if $\\sim 60\\%$ SU(3)$_F$-breaking effects are included.","published":"2026-10-14T17:59:58Z","updated":"2026-10-14T17:59:58Z","authors":["Bhubanjyoti Bhattacharya","Rida Boumris","David London"],"categories":["hep-ph"],"headline":"Researchers find large symmetry breaking in B meson decays","summary":"Scientists studied decays of B mesons into D mesons and lighter particles, assuming a symmetry called flavor SU(3). Their analysis, combining multiple decay types and measured data, revealed a strong disagreement with the predictions based on this symmetry alone. The researchers found that to match observations, the symmetry must be broken by about 60%, suggesting significant deviations from the simplest theoretical expectations within the Standard Model. This finding is based on fits to existing decay measurements and does not confirm new physics."},{"id":"2610.20610v1","title":"Can Learning to Answer Spatial Questions Improve Novel-View Generation?","abstract":"Generating novel views from limited observations requires preserving both the appearance and spatial structure of a scene. Although recent approaches exploit fine-grained geometric cues, they do not explicitly model high-level semantic and spatial properties such as object identity, attributes, and inter-object relationships. In this work, we investigate whether 3D language understanding can provide complementary supervision for novel-view generation. We introduce Ask4NV, built on a pretrained unified multimodal model with a Mixture-of-Transformers (MoT) architecture. Our model incorporates camera geometry into both VAE and ViT representations and leverages understanding features during generation through joint self-attention with a modality-aware attention masking strategy. Experiments demonstrate strong image quality and geometric consistency in novel-view generation, alongside competitive 3D language understanding. Further analysis shows that joint training improves cross-view correspondence in VAE representations and similarity in a 3D-aware feature space.","published":"2026-10-14T17:59:58Z","updated":"2026-10-14T17:59:58Z","authors":["Jahyeok Koo","Hyeonseo Yu","Seungho Jang","Jaewoo Jung","Jiwon Kang","Seokju Cho","Seungryong Kim"],"categories":["cs.CV"],"headline":"Studying if Language Helps Create New 3D Scene Views","summary":"Researchers explored whether understanding 3D spatial language can improve generating new views of a scene from limited images. They developed Ask4NV, a model combining camera geometry and language features to better capture object identity and relationships. Their experiments show this approach produces high-quality images with consistent 3D structure and improves how different views correspond to each other. These findings are based on computer simulations and model testing, highlighting potential benefits of combining spatial language with visual generation."},{"id":"2610.20609v1","title":"SafeDriveVLA: Navigation-Conditioned World Model Dreaming for Conflict-Aware End-to-End Autonomous Driving","abstract":"Vision-Language-Action (VLA) models are gaining attention in autonomous driving, inheriting commonsense priors from large language models (LLMs) and widening the navigation interface from a fixed command set to free-form natural-language instructions. However, this open-ended interface exposes a new vulnerability surface: an inattentive driver may issue instructions that are unsafe given the current scene, and a malicious user may issue dangerous ones on purpose. We present the first benchmark suite that evaluates navigation following and conflict awareness together, along with adversarial robustness. Existing VLAs fall short on both: SimLingo completes no language-instructed lane change in closed loop, and all models degrade sharply under conflicting signals. We propose SafeDriveVLA, the first driving VLA that (1) explicitly predicts a three-level navigation-compliance mode to decouple conflict reasoning from action generation, and (2) uses navigation-conditioned world model dreaming to inform this prediction. Unlike prior world models that roll out expert or planned actions, ours dreams of strictly following the navigation signal, showing when following is safe. SafeDriveVLA reaches 88.16 Driving Score (DS) and 68.64% Success Rate (SR) on Bench2Drive, lifts closed-loop language following by 49.3% and reduces collisions under unsafe instructions from 60 to 38 compared to the best SimLingo variant. Video demonstrations are available on our project page.","published":"2026-10-14T17:59:56Z","updated":"2026-10-14T17:59:56Z","authors":["Shaoyuan Xie","Zihan Zhang","Jingxuan Wang","Jiashu Qu","Xiaoqing Liang","Lingdong Kong","Junchi Lu","Henrik I. Christensen","Qi Alfred Chen"],"categories":["cs.RO"],"headline":"New AI Helps Self-Driving Cars Follow Safe Navigation Commands","summary":"Researchers studied how self-driving cars understand and follow spoken navigation instructions. They found current models struggle with unsafe or conflicting commands. To improve safety, the team created SafeDriveVLA, which predicts if following an instruction is safe before acting and simulates driving while strictly following commands. Their tests show SafeDriveVLA better follows language instructions and reduces accidents compared to earlier models. These results come from computer simulations and testing benchmarks designed by the researchers."},{"id":"2610.20608v1","title":"Representation Conditioning for Generation","abstract":"Representation-alignment methods effectively accelerate the diffusion transformer training. REPA, for example, uses a pretrained vision encoder to supervise an intermediate layer through an MLP projection head which is discarded after training. It aims to import semantics from the vision encoder to the diffusion backbone. While this objective is achieved at early layers by back-propagating the alignment loss, there is no guarantee that later layers, supervised by the velocity prediction loss alone, can well access such semantics. We propose a simple solution, representation conditioning (RECON), to improve semantics in later layers. During training, we feed the pretrained vision encoder tokens into later layers; during inference, because we no longer have the clean image, we reuse the MLP head (as a proxy to the vision encoder) and project its output tokens to the later layers. Our method adds no parameters and leads to much faster convergence: RECON is $63\\times$ faster than SiT and $25\\times$ faster than REPA. Code is available at https://github.com/saeidrazavi/RECON","published":"2026-10-14T17:59:52Z","updated":"2026-10-14T17:59:52Z","authors":["Saeed Razavi","Xingjian Leng","Liang Zheng"],"categories":["cs.CV"],"headline":"New method speeds up AI image generation training significantly","summary":"Researchers studied how to improve training speed for AI models that create images. They found that using a technique called representation conditioning (RECON), which feeds certain encoded information into later stages of the model, helps it learn faster and access important image features better. Their approach speeds up training by dozens of times compared to previous methods, though their findings are based on training efficiency improvements rather than final image quality alone."},{"id":"2610.20607v1","title":"Hierarchical Residual Policy Learning for Real-World Mobile Manipulation with Sparse Human Guidance","abstract":"Mobile Manipulation (MoMa) is a cornerstone of autonomous robotic systems, yet real-world fine-tuning remains a formidable challenge because of high-dimensional action spaces and intricate coupling between heterogeneous modules. In this paper, we propose HiRe-MoMa (Hierarchical Residual Mobile Manipulation), a systematic framework for stable and sample-efficient real-world fine-tuning of complex MoMa systems. At its core, HiRe-MoMa adopts a Hierarchical Residual Architecture that fine-tunes a frozen base policy. A Policy-Agnostic Residual learns additive corrections atop the base policy, while Hierarchical Decoupling splits the residual into mobile-base and manipulator-arm heads with unidirectional conditioning, mitigating gradient interference and aligning corrections with the platform's physical hierarchy. To improve sample efficiency in data-expensive real-world settings, we propose Expert-Guided Value Alignment, which leverages sparse human interventions with under 20% training coverage to reshape the critic's value landscape, effectively guiding autonomous exploration through expert priors. Extensive real-world experiments on the TIAGo++ platform demonstrate that HiRe-MoMa consistently enhances task performance with only around 30 minutes of online interaction, regardless of the base policy architecture. To our knowledge, this is among the first frameworks to enable efficient on-robot fine-tuning for mobile manipulation in the real world, providing a practical paradigm for deploying complex robotic systems in dynamic, human-centric environments. Further details can be found on our project website: https://hire-moma.github.io/","published":"2026-10-14T17:59:52Z","updated":"2026-10-14T17:59:52Z","authors":["Zhefei Gong","Hefeifei Jiang","Zehao Wang","Jiachen Li"],"categories":["cs.RO","cs.AI","cs.LG"],"headline":"New Method Improves Real-World Robot Learning with Less Human Help","summary":"Researchers developed HiRe-MoMa, a way to fine-tune complex robot control systems more efficiently in real-world settings. Their method adjusts an existing robot policy by adding corrections in a hierarchical structure and uses limited human guidance to steer learning. Tests on a mobile robot showed improved task performance with just 30 minutes of real use, suggesting a practical approach for real-world robot training, though the results are specific to their tested setup."},{"id":"2610.20606v1","title":"Modeling and Understanding Neural Weights of Humanoid Policies with Diffusion","abstract":"Recent advances in robot learning have driven rapid progress in humanoid control, enabling increasingly agile behaviors in the real world. However, humanoid control networks are typically obtained through conventional training procedures, while comparatively little attention has been given to directly modeling the distribution of their learned weights and generating functional networks in weight space. In this work, we investigate whether high-performing humanoid control networks can be generatively modeled directly through their weights. We adopt a latent diffusion framework that couples a linear autoencoder with a Diffusion Transformer. The autoencoder compresses actor network parameters into a compact latent space, and the Diffusion Transformer models the distribution of these latents. Trained on networks collected from reinforcement learning for the Unitree G1 robot, our approach generates deployable networks that perform various motions in the real world, including walking, dancing, and spin kick. Beyond generation, we further study the learned weight space to better understand the structure captured by these humanoid control networks. We identify a structural principle in the first-layer weights that consistently reflects the robot's kinematic prior. Notably, this kinematic prior emerges under both Reinforcement Learning (RL) and Imitation Learning (IL) training, suggesting that it is not specific to a particular optimization process but may instead reflect common knowledge associated with humanoid embodiment. Together, these results demonstrate the feasibility of generative modeling in humanoid network weight space, while providing initial insights into the information encoded by their learned weights. Additional results are available on our project website: https://humanoid-weight-diffusion.github.io/","published":"2026-10-14T17:59:52Z","updated":"2026-10-14T17:59:52Z","authors":["Huixuan Cheng","Bike Zhang","Shankar Sastry","Koushil Sreenath"],"categories":["cs.RO","cs.AI","cs.LG"],"headline":"Researchers model and generate humanoid robot control networks","summary":"The researchers studied how to directly model the internal parameters (weights) of robot control programs, specifically for humanoid robots, using a technique called latent diffusion. They compressed these parameters into a simpler form and trained a model to generate new functional control programs that can perform tasks like walking and dancing. Their work offers new ways to understand how these robot controllers work, revealing common structural patterns related to the robot's physical design, and shows it is possible to create new control networks directly from learned weight distributions."},{"id":"2610.20605v1","title":"Nodal nematic superconductivity in moiré graphene","abstract":"Unconventional superconductivity can emerge in various forms in strongly correlated systems. Historically, the most extensively studied examples include nodal and triplet pairing, which were experimentally detected in cuprates and superfluid helium-3, respectively. Beyond these canonical cases, there are also nematic superconductors, in which the rotational symmetry of the lattice is broken in the superconducting state. While there are several examples of nematic superconductors whose pairing emerges inside a preexisting nematic normal state, experimental evidence for superconducting states that intrinsically break rotational symmetry on their own has been exceedingly scarce. Here, we report the experimental realization and identification of an intrinsic multi-component nodal nematic superconducting state in magic-angle twisted trilayer graphene (MATTG). By integrating tunneling spectroscopy with quantum transport measurements, we directly probe the symmetry of low-energy superconducting quasiparticles. Deep inside the superconducting state, we observe a pronounced two-fold anisotropy in the zero-bias tunneling conductance under an in-plane magnetic field, which is strongly suppressed together with superconductivity as the temperature or perpendicular magnetic field is increased. Remarkably, the major axis of anisotropy rotates by up to $90^{\\circ}$ across the superconducting dome, indicating an intrinsic, field- and density-tunable symmetry of the low-energy excitations. These behaviors imply a nodal superconducting state in MATTG that spontaneously breaks the six-fold rotational symmetry of the moiré superlattice. Our results establish moiré graphene as a versatile platform for realizing and interrogating nodal nematic superconductivity, opening a new avenue for exploring unconventional symmetry-breaking superconducting states in two-dimensional materials.","published":"2026-10-14T17:59:50Z","updated":"2026-10-14T17:59:50Z","authors":["Shuwen Sun","Grgur Palle","Kenji Watanabe","Takashi Taniguchi","Rafael M. Fernandes","Pablo Jarillo-Herrero","Jeong Min Park"],"categories":["cond-mat.supr-con","cond-mat.mes-hall"],"headline":"Scientists Find Unusual Nematic Superconductivity in Twisted Graphene","summary":"Researchers studied a special form of superconductivity in twisted trilayer graphene, where the material conducts electricity without resistance while breaking the usual rotational symmetry of its lattice. Using tunneling and transport measurements, they detected directional differences in the superconducting state that change with magnetic field and electron density. Their experiments suggest this graphene system hosts a rare nodal nematic superconducting state, providing a new way to explore complex superconducting behaviors in two-dimensional materials."},{"id":"2610.20604v1","title":"VITAL: From Visual Motion to Precise Action","abstract":"World-action models aim to turn visual anticipation into driving decisions, yet typically train action experts on trajectory representations defined independently of visual dynamics. This motivates deriving action targets directly from visual motion. We first examine a simpler setting in which trajectories are reconstructed from recorded future images. Even without the need for future prediction, reconstruction captures the dominant maneuver but leaves centimeter-scale errors. We therefore introduce VITAL (Vision-Induced Tokens for Action Learning), an action tokenizer that couples a visually derived motion scaffold with a compact residual encoding the trajectory error left by that scaffold. This representation makes visual motion the principal prediction target and gives small metric corrections a separate, normalized learning scale. Using flow matching, the action expert jointly predicts both components from the current image, ego state, and navigation command, without generating future images at inference. The same residual coordinates support bounded reinforcement post-training that refines execution while keeping the predicted visual scaffold fixed. With only one action denoising step, VITAL achieves 90.8 PDMS on NAVSIM v1 navtest, within 0.13 points of its ten-step score at 3.29x the end-to-end inference speed. Without retraining, it achieves 91.1 EPDMS on NAVSIM v2 navtest and 38.6 on navhard. Project page: https://worldbench.github.io/vital","published":"2026-10-14T17:59:49Z","updated":"2026-10-14T17:59:49Z","authors":["Ao Liang","Youquan Liu","Mingxiao Li","Wei Chow","Lingdong Kong","Yue Liao","Dongyue Lu","Yu Yang","Wei Yin","Runnan Chen"],"categories":["cs.CV"],"headline":"New Method Uses Visual Motion to Improve Driving Decisions","summary":"Researchers developed VITAL, a system that turns visual motion directly into precise driving actions. Instead of relying on predefined trajectories, VITAL predicts motion from current images and corrects small errors separately, improving accuracy. Tests in simulated navigation environments show VITAL performs well and runs faster, though these results come from controlled simulations rather than real-world driving."},{"id":"2610.20603v1","title":"CommerceGen-Studio: An End-to-End Framework for Commercial-Grade Main-Image Generation","abstract":"We present CommerceGen-Studio, an end-to-end framework for generating commercial-grade main images for e-commerce Stock Keeping Units (SKUs). Our framework is organized along a three-stage pipeline of information extraction, semantic rewriting, and multi-condition generation: Prompt-Eval-Refine-Loop Agent (PERL-Agent) distills heterogeneous SKU metadata into concise selling-point text lists through an iterative closed loop of VLM and LLM roles scored by a reference-free rubric. Layout Thinker Optimization (LTO) aligns the rewriter with both caption and rendered-image quality via alternating domain-specific policy optimization. SKU-DiT, trained under Reinforce-Distill-Reinforce (RDR) progressive training, conditions jointly on the white-background product image, the trademark crop, and the structured layout to hold subject fidelity, trademark integrity, and character-level text accuracy together. The framework runs in production at a scale of millions of main images per day with a latency of a few seconds per image, which we believe makes it among the first systems of this kind to operate at commercial scale. We also release CommerceGen-Bench, a benchmark of hard cases collected from live traffic across all product categories. On it, our framework achieves the highest overall scores among the evaluated open-source baselines on both editing tracks, with and without prompt enhancement. It also matches much larger closed-source models, and beats several of them at Instruction Following and Product Integration - the two properties that determine whether an image can be published.","published":"2026-10-14T17:59:48Z","updated":"2026-10-14T17:59:48Z","authors":["Xiaoli Xu","Zhengtao Wu","Zhengze Xu","Tao Liu","Yongchao Du","Zuan Gao","Xingjian Wang","Hao Yan","Mengting Chen","Wei Li","Bo Zheng"],"categories":["cs.CV"],"headline":"New system creates high-quality product images for online stores","summary":"Researchers developed CommerceGen-Studio, a framework that generates main product images for online shopping. Their system uses a three-step process to extract info, rewrite descriptions, and create images that keep product details accurate. Tested on millions of images daily, the researchers report it matches or beats other models on image quality and accuracy, but these findings come from their own evaluations and benchmark tests."},{"id":"2610.20602v1","title":"Heavy-Heavy-Light-Light correlators from holography on the worldsheet","abstract":"We study correlation functions of light local operators and defect changing operators on the conformal line set by the 1/2 BPS Wilson line of $\\mathcal N=4$ super Yang-Mills. We consider the defect changing operators associated with geometrical and internal cusp angles, which at strong coupling have a large scaling dimension and back-react on the dual $AdS_2$ world-sheet. We set up a holographic computation of correlation functions, in which the light operators are described by fluctuations around the cusped classical string solution, so that correlators of two defect changing operators and an arbitrary number of light protected operators are obtained from the on-shell action of these fluctuations. We implement this prescription in detail for the case of certain 4-point functions, with two light and two defect changing operators. Finally, since for small cusp angles the defect changing operators become light, we manage to derive relations satisfied by integrated correlators of light operators, valid at any coupling. We verify these relations at strong coupling for both internal and geometrical cusps.","published":"2026-10-14T17:59:46Z","updated":"2026-10-14T17:59:46Z","authors":["Matías Colaneri","Diego H. Correa","Víctor I. Giraldo-Rivera","Nicolás A. Ivanovich"],"categories":["hep-th"],"headline":"Researchers analyze complex particle interactions using holography","summary":"The researchers investigate how certain operators called defect changing operators and light operators interact along a special line in a theoretical physics model known as super Yang-Mills. Using holography—a method linking gravity and quantum theories—they study these interactions by examining fluctuations around a classical string shape. Their work includes detailed calculations of specific four-point functions and reveals relationships between operators valid at any interaction strength, confirmed at strong coupling. The findings are limited to theoretical models and rely on advanced mathematics and simulations."},{"id":"2610.18874v1","title":"ChunkAct: Action Chunking for Agentic Robotics","abstract":"Agentic robotics places a large language model (LLM) agent in the robot control loop, where it composes perception, motion, and vision-language-action (VLA) tools under the guidance of a memory and thereby generalizes to new tasks without retraining. This generality comes at a steep price in efficiency. Every primitive requires a round of LLM reasoning over the full interaction history, so the cost of an episode grows with the number of primitives. We find this reasoning largely redundant: once a sub-goal is chosen, its primitives follow a routine already prescribed by memory. We present ChunkAct, which introduces action chunking to agentic robotics. Online, a single call launches a chunk of chained primitives, which executes autonomously and returns control only upon failure. Offline, ChunkAct compiles the memory into these chunks, so execution is memory-free at test time and the planner is reserved for decisions that require reasoning. LLM computation therefore scales with the number of decisions rather than the number of primitives. On LIBERO-Pro, ChunkAct improves task success by 5.0 points over the state of the art in agentic robotics while requiring 2.9x fewer LLM calls, 4.4x fewer input tokens, and 2.3x fewer output tokens.","published":"2026-10-13T17:59:57Z","updated":"2026-10-13T17:59:57Z","authors":["Bingxin Xu","Wenrui Bao","Xinxin Liu","Jiawei Zhou","Yuzhang Shang","Emilio Ferrara"],"categories":["cs.RO","cs.AI","cs.CL","cs.CV"],"headline":"New method speeds up robot decision-making with fewer actions","summary":"Researchers studied how robots that use large language models (LLMs) plan and act, noting that each small action requires costly reasoning, which slows down tasks. They created ChunkAct, a way to group sequences of actions into chunks that run automatically, reducing the need for repeated reasoning. Their tests show ChunkAct improves task success and greatly reduces the amount of LLM computation. These results are based on simulations and show potential efficiency gains in robot control, according to the researchers."},{"id":"2610.18873v1","title":"LoopWAM: Closed-Loop Generalist Policy Improvement from Deployment Rollouts","abstract":"Scaling expert demonstrations alone does not enable generalist robot policies to improve from deployment. Real-world rollouts contain successes, suboptimal behavior, and failures: imitating every recorded action can reinforce poor behavior, while filtering out failures discards evidence of physical interactions and their consequences. We introduce \\textbf{LoopWAM}, a closed-loop world-action model that learns directly from this full quality spectrum, using rollouts from both itself and other deployed policies. Source-dependent noise sampling unifies behavior learning and action-conditioned world modeling, while quality steering and quality-weighted supervision retain full video supervision across qualities and regulate action imitation. Video and action tokens interact throughout parallel denoising, allowing learned consequences to inform action generation directly. On RoboChallenge Table30-V2, three deployment-and-training rounds raise success from 23.67\\% to 43.33\\%, a 19.7-percentage-point gain, with a final score of 56.80. LoopWAM also achieves 94.5\\% average success on RoboTwin~2.0 and 49.5\\% overall success on RoboCasa365 without additional embodied pretraining, obtaining the best overall results among the compared methods across all three benchmarks. On challenging shirt folding task, two rounds of deployment-driven improvement after single-task post-training raise success from 13.20\\% to 55.56\\%. These results demonstrate closed-loop improvement from mixed-quality experience under both multi-task learning and single-task post-training. \\href{https://loop-wam.github.io/}{\\textcolor{orange}{Project Page}}","published":"2026-10-13T17:59:56Z","updated":"2026-10-13T17:59:56Z","authors":["Wenxuan Guo","Yingping Liang","Zhaoguo Wang","Ziyuan Li","Angyuan Ma","Zhenyu Wu","Chengkun Wang","Bingyao Yu","Ze Chen","Borui Zhang","Xiuwei Xu","Meng Zhang","Erjin Zhou","Jiwen Lu"],"categories":["cs.RO","cs.CV"],"headline":"New Method Helps Robots Learn From Mixed Success Experiences","summary":"The researchers studied how robots can improve their behavior by learning from real-world attempts that include both successes and mistakes. They developed LoopWAM, a method that uses all experiences—good and bad—to better understand actions and their outcomes. Tests showed that LoopWAM improved robot success rates on multiple tasks and benchmarks, suggesting robots can improve by learning directly from their mixed-quality real-world experiences. These findings are based on simulation and experimental robot tasks described by the researchers."},{"id":"2610.18872v1","title":"EgoExplorer: A Vision-Motion-Touch Observatory for Human Loco-Manipulation In Situ","abstract":"Learning humanoid loco-manipulation from human demonstrations rests on observing how locomotion, posture, and object interaction evolve together as tasks unfold. Scaling collection across diverse environments, in turn, favors lightweight, portable capture systems. Yet, existing datasets provide only partial views of this coordination. Thus, we introduce EgoExplorer, a 100-hour vision-motion-touch observatory of human loco-manipulation built with a wearable system. With an emphasis on professional activities, its continuous recordings align egocentric video with reconstructed whole-body motion, fine-grained finger articulation, and tactile observations, accompanied by step-level descriptions. To achieve this coverage with sparse body sensing, we develop the Calibrate-Complete-Constrain pipeline. Concretely, reusable calibration fixes participant geometry and sensor attachments before motion inference to disentangle mounting offsets from articulation. A temporal prior then completes unobserved joint motion, while layered constraints promote measurement agreement, temporal continuity, and geometric consistency, properties that motion plausibility alone does not guarantee. Against tracker-fitted references, the method reduces body-point error by 71.9% relative to the strongest baseline; matched evaluations also show gains in hand stability and contact estimation. Preliminary studies explore multimodal generation and conditioning alongside humanoid retargeting with whole-body control in simulation.","published":"2026-10-13T17:59:55Z","updated":"2026-10-13T17:59:55Z","authors":["Yunsong Zhou","Xuekun Jiang","Haiqi Shen","Hangxu Liu","Chaojian Shi","Yujie Liu","Jingyuan Li","Haoran Yang","Jiafei Cao","Dongzhan Zhou","Dong Wang","Yao Mu","Chunhua Shen","Weinan Zhang"],"categories":["cs.RO","cs.CV"],"headline":"New Wearable System Captures Detailed Human Movement and Touch","summary":"Researchers developed EgoExplorer, a wearable device that records how humans move, use their fingers, and interact with objects, focusing on professional tasks. Their system combines video, body motion, and touch data, using a special method to accurately track body positions even with limited sensors. Tests show this approach greatly improves motion capture accuracy compared to existing methods. The team also conducted early studies using this data to simulate and control humanoid robots, though these findings are preliminary and based on simulations."},{"id":"2610.18871v1","title":"Cepheid: Decoupling Scale from Structure for Cross-Family Motion Completion","abstract":"Sparse-to-dense motion completion recovers dense mesh motion from sparse vertex displacements. Residual priors trained on one geometry family frequently fail to beat parameter-free harmonic extension off-family. Moreover, cross-family training meets a scaling barrier: residual amplitude varies by more than an order of magnitude across geometry families and collapses as observations become denser. Since this scale depends on unknown target motion, it cannot be measured directly at test time. We decouple residual structure from physical scale by learning a dimensionless prior, estimating amplitude from observables and a lightweight correction fitted only on training sources. Our unified model surpasses harmonic extension across all evaluated families, including unseen cloth and legged bodies, without skeletons, registration, or cross-mesh correspondence. The same prior applies without retraining to point-tracked animal video and a robot grasping a soft body. Project page: https://su-terry.github.io/cepheid-website/","published":"2026-10-13T17:59:53Z","updated":"2026-10-13T17:59:53Z","authors":["Chih-Hai Su","Tsung-Wei Ke","Yu-Lun Liu"],"categories":["cs.CV","cs.GR","cs.LG"],"headline":"New Method Improves Motion Prediction Across Different Shapes","summary":"Researchers addressed the problem of predicting detailed motion from limited movement data, which is challenging when shapes vary widely. They developed a new approach that separates the shape’s motion pattern from its size, allowing better predictions across different object types. Tests showed their model outperforms existing methods on various shapes, including cloth and animal movements, without needing prior shape alignment or retraining, based on results reported in their paper."},{"id":"2610.18870v1","title":"Dog-Breed Text-to-image Flow Matching from Pretraining to Fine-Tuning on 4 Consumer GPUs","abstract":"We generally perceive that training text-to-image diffusion models requires heavy compute: it is true if we aim at a comprehensive frontier model. This understanding prevents people from training their own models when getting started in the field. We develop MiniDog, a framework that enables pretraining diffusion transformers in 55 minutes and fine-tuning in another 15 minutes on four 3090 GPUs. For data, we use 26k dog images from ImageNet and caption them for pretraining and use 2k images generated by FLUX.1-schnell for fine-tuning. We precompute the image and text features and cache them for efficiency. For architecture, we use E2E-INVAE for its semantic latent space, a small transformer with 22M parameters, a small Qwen model for text encoding, and in-context conditioning. For training, we use iREPA and an EUPE perception encoder for acceleration. For evaluation, our repository supports FID, HPSv2 score, PickScore, and various VAE reconstruction metrics.","published":"2026-10-13T17:59:53Z","updated":"2026-10-13T17:59:53Z","authors":["Reyhaneh Esmailizadeh","Xingjian Leng","Zhanhao Liang","Liang Zheng"],"categories":["cs.CV"],"headline":"Fast Dog-Breed Image Generation Using Affordable GPUs","summary":"Researchers created MiniDog, a method to train models that generate dog images from text much faster and on more affordable hardware than usual. They pretrained their model on 26,000 dog images and fine-tuned it on 2,000 generated images, completing the process in under two hours using four consumer-grade GPUs. Their work shows it is possible to develop text-to-image models quickly with limited resources, but the findings are based on their own tests and specific setup."},{"id":"2610.18869v1","title":"What Matters in Quality-Aware VLA Pretraining: A Benchmark and Empirical Study","abstract":"How should demonstration quality be used during vision-language-action (VLA) pretraining to improve performance after target-task fine-tuning? We compare weighting, filtering, and conditioning across reward definitions and label granularities using 5,000 naturally suboptimal human demonstrations and 24 target tasks. Within each VLA, we match demonstrations and training budgets. Additional comparisons examine pretraining data size and the roles of quality tags during fine-tuning and inference. Across three VLAs, we find that approach-aware episode conditioning improves mean success over plain behavioral cloning; conditioning and filtering favor episode and chunk labels, respectively, on average, with model-dependent reversals; and constant-tag fine-tuning improves mean success after conditioned pretraining with either tagged or untagged inference. We introduce AXIS-Bench to support these controlled comparisons. Built on crowdsourced simulation demonstrations, it provides recorded object and contact states for recomputing quality labels on shared trajectories. Replayable trajectories support workspace-aligned rendering and user-defined real-world target tasks. In our real-world experiment, simulation pretraining raises success from 9.6\\% to 31.7\\%. Project page: https://axis-bench.github.io/.","published":"2026-10-13T17:59:52Z","updated":"2026-10-13T17:59:52Z","authors":["Mingxuan Yan","Litian Gong","Dihong Huang","Zhixu Li","Tianyu Zhang","Wenqian Zhang","Hefeifei Jiang","Peihao Li","Jian Zhang","Mengfei Zhao","Yikai Tang","Hai Zhai","Zehao Wang","Ruijian Liang","Yanjia Huang","Lin Shao","Changliu Liu","Kaiyu Hang","Zhiwen Fan","Masayoshi Tomizuka","Jianfei Yang","Jiachen Li"],"categories":["cs.RO","cs.AI","cs.LG"],"headline":"Studying How to Use Demo Quality in Vision-Language-Action Training","summary":"Researchers investigated different ways to use the quality of human demonstration videos to improve training of vision-language-action models, which aim to perform tasks after fine-tuning. They tested methods like weighting, filtering, and conditioning on data from 5,000 suboptimal human demos across 24 tasks, introducing a new benchmark called AXIS-Bench for controlled comparisons. Their findings suggest certain conditioning and filtering techniques can improve success rates, and simulation pretraining notably increased success in real-world tests. These results support the value of quality-aware pretraining but are based on specific models and simulations."},{"id":"2610.18868v1","title":"Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards","abstract":"Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a capability essential to automated discovery and recursive self-improvement (RSI). Yet open-source models still struggle on long-horizon tasks, largely for lack of both training data and a working recipe, and scaling both is hard: long, complex tasks are difficult to curate and verify, and their rewards are sparse. We address both. First, we introduce Tev-Gym, the first long-horizon agent training dataset with verifiable, dense, progressive rewards -- 1,170 verified task environments across 7 domains. It is built on a latent-world-first approach that first constructs a latent world: a complete and correct hidden specification of the task domain, workflow, and expected state after each step. Each task is then progressively rendered from this specification into an agent-facing environment, enabling controllable complexity and a natural progress metric. Second, we turn this metric into a training recipe: via progress-potential reward shaping (PPRS), we convert progress into a dense per-step reward that improves training efficiency and makes otherwise-intractable tasks trainable. Training with this recipe yields Tev-27B, which achieves substantial gains on the Agents' Last Exam computing and math sciences domains (ALE-CM), matching GPT-6 Astra and Claude Opus 5 at Pass@5. Despite training only on ALE-related data, it generalizes to the Terminal-Bench (TB) 3.0 and 4.0 subsets, surpassing GLM 5.2 and Gemini 3.7 Flash, and reaches Claude Opus 4.8-level performance on middle-horizon tasks such as TB 2.1.","published":"2026-10-13T17:59:51Z","updated":"2026-10-13T17:59:51Z","authors":["Wenbo Hu","Shraman Pramanick","Krishna Kumar Singh","Nanyun Peng","Kai-Wei Chang","Yong Jae Lee"],"categories":["cs.CL","cs.AI","cs.CV","cs.LG"],"headline":"New Method Improves Training for Long-Term Autonomous Tasks","summary":"Researchers created Tev-Gym, a large collection of 1,170 task environments designed to train agents on complex tasks lasting hours or days, with detailed progress tracking to guide learning. They developed a training technique that turns progress into frequent rewards, helping agents learn tasks that were previously too difficult. Their trained model, Tev-27B, showed strong results on specific math and computing challenges, matching or exceeding some advanced AI models, and also performed well on other related tasks despite limited training data. The results are based on simulations and comparisons within test environments."},{"id":"2610.18867v1","title":"EmbodiedEvo: Self-improving Coding Agents for Physical Intelligence","abstract":"Bringing general-purpose agents into the physical world calls for an executable representation of physical intelligence. Compared with vision-language-action (VLA) models, which map observations and language instructions directly to actions, code makes control logic explicit and reusable. Code can serve not only as a policy representation but also as the workspace in which the agent solves a task. Rather than following a prescribed workflow of planning, execution, and verification, the agent can autonomously inspect the state, revise and execute code, and use feedback to choose the next step. The success of this SWE-style loop in digital environments motivates its extension to physical tasks. Yet physical execution remains prone to recurring failures. To retain and refine experience for later scenes, we introduce EmbodiedEvo, where a coding agent evolves reusable Skills and Lessons through feedback on a single Playground scene, then freezes the resulting harness for diverse new scenes. On RoboDojo-Sim, EmbodiedEvo achieves an average score of 44.48 and a success rate of 37.82\\%, surpassing all officially verified results. Paired tests show the largest gains on memory-dependent tasks and successful reuse of coordinated procedures, while fine contact control remains difficult. Experience from stronger models also improves weaker executors: Astra's experience raises Sol and Luna's success rates by 8.86 and 5.50 percentage points over no experience. Compared with raw interaction history, Skills and Lessons achieve comparable task performance at 53.0\\% lower execution cost.","published":"2026-10-13T17:59:51Z","updated":"2026-10-13T17:59:51Z","authors":["Wenyi Li","Yizhe Chi","Youjie Zheng","Chunyu Liu","Xiaoqiu Wang","Mingju Gao","Zheng Jiang","Deyao Hong","Calvin Xiao","Kaisen Yang","Qinhuai Na"],"categories":["cs.RO"],"headline":"New Coding Agents Learn Physical Tasks by Improving Themselves","summary":"Researchers developed EmbodiedEvo, a coding agent that improves its own control logic for physical tasks by writing and revising code based on feedback. Unlike traditional models that directly map observations to actions, this agent evolves reusable skills and lessons in a simulated environment to better handle new tasks. Experiments show EmbodiedEvo outperforms previous methods, especially in tasks needing memory and coordination, though some fine control remains challenging. These findings come from simulations and show potential but do not cover real-world testing."},{"id":"2610.18866v1","title":"AutoAgent0: An Agentic Runtime for Safe Closed-Loop Driving","abstract":"Safe autonomous driving requires more than accurate planning at a single frame: it demands reliable closed-loop execution in dynamic environments. Although end-to-end driving policies have advanced rapidly, the accumulated closed-loop errors, especially in unseen scenarios, can push vehicles into states from which the same policy struggles to recover. We address the challenge by treating closed-loop driving safety as a test-time execution problem. We introduce AutoAgent0, an agentic runtime and harness framework that augments existing driving experts with complementary detection skills, rule-based action verification, and feedback-driven recovery. To address inaccurate scene perception, an independent detection module constructs scene context using complementary perception skills, and a rule-based verifier checks proposed trajectories against this context before execution. When the nominal trajectories are unreliable, the orchestrator of the runtime activates Recovery Planning, where an agentic loop constructs alternative action proposals from action primitives and iteratively refines them using verification feedback. This design combines independent scene assessment with feasibility checks and online recovery, addressing both perception errors and the limited recovery capabilities of individual experts. Evaluations on Fail2Drive and NavSafe-$\\infty$ demonstrate improved closed-loop robustness over single-expert execution in long-tail and safety-critical scenarios. These results demonstrate that runtime supervision can improve closed-loop robustness across multiple driving experts without retraining the underlying policies.","published":"2026-10-13T17:59:50Z","updated":"2026-10-13T17:59:50Z","authors":["Jiageng Liu","Seth Z. Zhao","Zewei Zhou","Alejandro Huerta","Aidan Bayer-Calvert","Yuxin Bao","Haotian Lin","Robert Spataru","Zijun Cui","Yining Hong","Bolei Zhou","Jiaqi Ma"],"categories":["cs.RO"],"headline":"New system helps self-driving cars recover from unexpected errors","summary":"Researchers developed AutoAgent0, a system designed to improve the safety of autonomous vehicles during real-time driving. It adds extra detection, rule-based checks, and a recovery process to existing driving models to better handle errors and unexpected situations. Their tests show that this approach can help cars respond more reliably without needing to retrain the original driving software."},{"id":"2610.18865v1","title":"PLAYGROUND: Progressive Layout Generation with Render-Grounded Planning","abstract":"Generating editable graphic designs requires arranging visual and textual elements into a coherent composition while preserving their content and editability. Although recent MLLM-based approaches employ iterative generation and visual refinement to reflect the real-world design workflows, their high-level plans are predetermined prior to execution, limiting their adaptability as the design evolves. To address this limitation, we propose PlayGround, a progressive layout generation framework that dynamically adapts to the current rendered canvas. PlayGround establishes a progressive design loop comprising a Progressive Planner, a Layout Generator, and a Visual Verifier, guided by a Structure-Aware Retriever. Given an initial set of unordered elements, the Structure-Aware Retriever initially infers a coarse layout structure to retrieve structurally similar reference designs. Subsequently, at each turn, the Progressive Planner examines the current rendered canvas alongside the remaining elements, forms the next functional group, and plans its composition. The Layout Generator then predicts editable layout attributes, while the Visual Verifier evaluates rendered candidates and provides feedback when refinement is necessary. To train the Progressive Planner with step-wise supervision, we construct a two-stage data generation pipeline that extracts functional groups and turn-level planning trajectories from completed designs. Experiments on Crello and LICA demonstrate the effectiveness of PlayGround over existing methods. Our code and weights will be publicly released.","published":"2026-10-13T17:59:50Z","updated":"2026-10-13T17:59:50Z","authors":["Jaeho Lee","Junhwan Heo","Junghyun Park","WonJun Moon","Eunju Yang","Seungho Jang","Seongchan Kim","Seungryong Kim"],"categories":["cs.CV"],"headline":"New AI method improves editable graphic design layouts step-by-step","summary":"Researchers developed PlayGround, a system that arranges visual and text elements into editable graphic designs by adapting its plans as the design evolves. Unlike previous methods with fixed plans, PlayGround progressively plans and updates the layout based on the current state, using a cycle of planning, generating, and visually verifying designs. Tests on design datasets show PlayGround performs better than existing approaches. The findings are based on simulated experiments described in the abstract."},{"id":"2610.13948v1","title":"Can Language Models Play the Wikipedia Game?","abstract":"Language models increasingly support or automate information retrieval from document collections too large to index. This task is difficult when it requires specialist linguistic or scientific knowledge. We used the Wikipedia Game to test language models on an analogous information retrieval task. In the game, players must reach a target article using only article links. We compared sentence embedding models and LLMs across five Wikipedia languages and five academic-literature datasets. We construct very large local graphs of Wikipedia pages and of paper citations using OpenAlex and provide code to build and navigate them. This allows the study of language models' navigation capabilities without relying on third-party APIs. We compared the models with human and TF-IDF benchmarks. The Wikipedia Game format enables robust comparisons across techniques, languages, and domains. No model significantly outperformed humans, although the best models (GPT-o3 mini, GPT-4o, QwQ:32B, and DeepSeek-R1:70B) achieved comparable success. Performance and rankings varied across languages and topics, with implications for practitioners utilising these techniques.","published":"2026-10-09T17:59:56Z","updated":"2026-10-09T17:59:56Z","authors":["Alexander Hickey","Joanna Knight","James Bishop"],"categories":["cs.AI","cs.IR"],"headline":"Testing AI on Navigating Wikipedia Like Humans Do","summary":"Researchers explored whether language models can find specific Wikipedia articles by following linked pages, similar to a popular navigation game. They compared different AI models, including large language models, with humans across multiple languages and academic topics using large Wikipedia and citation networks. The study found that while no AI outperformed humans, some models performed similarly, though results varied by language and subject. This work helps understand AI's ability to navigate complex information, but findings are based on simulations without real-world testing."},{"id":"2610.13947v1","title":"Towards Physically Grounded Articulated 3D Assets: A Multi-Agent Pipeline and Benchmark","abstract":"In this paper, we study physically grounded articulated 3D asset generation---moving the goal from reproducing how objects look to capturing how they move and function. Inspired by the literature on procedural and agentic generation, we present Artisan, a system that extends the latest method Articraft by introducing a multi-agent pipeline with physical simulation feedback, enabling comprehensive attributes beyond geometry, appearance, and physically plausible articulation, such as collision-free motion, stability under gravity, and accurate mass and material properties. Furthermore, we construct a unified benchmark spanning four key aspects: spatial fidelity, physical plausibility, downstream task performance, and overall quality. The benchmark encompasses 100 reference assets from ArtVIP and LightWheel, 50 stability prompts, and 50 ABO objects with reference masses, with protocols evaluating various metrics. Through extensive evaluation of more than 10 representative models, Artisan achieves competitive spatial fidelity while attaining the lowest collision rates and highest stability rates among the evaluated methods. Together, our system and benchmark establish insights for advancing 3D generation towards physically plausible, functionally usable assets.","published":"2026-10-09T17:59:41Z","updated":"2026-10-09T17:59:41Z","authors":["Xudong Cai","Ruijie Zhu","Haoliang Shang","Ruining Li","Marc Pollefeys","Shangzhe Wu","Andrea Vedaldi","Chuanxia Zheng"],"categories":["cs.CV"],"headline":"New Method Improves 3D Models' Physical Realism and Movement","summary":"Researchers developed Artisan, a system that creates 3D objects not just by how they look, but by how they move and behave physically. Artisan uses multiple agents and physical simulations to ensure objects avoid collisions, remain stable, and have realistic mass and materials. They also designed a benchmark to evaluate these physical and functional qualities. Tests show Artisan performs well in shape accuracy and excels in reducing collisions and increasing stability compared to other methods. The findings are based on simulations and evaluations using existing 3D datasets."},{"id":"2610.13946v1","title":"An entangling Hubbard quasiparticle collider on a digital quantum processor","abstract":"Scattering of quasiparticles provides direct access to the interactions that govern strongly correlated systems. In this work, we realize the quasiparticle scattering experiment of the Fermi-Hubbard model on a digital quantum processor by colliding a doublon and a holon in the half-filled chain of 21 sites encoded on 42 qubits of the ibm_boston processor. For the first time, we measure the transmission probability of quasiparticles of the Hubbard model on quantum hardware, which agrees with the Bethe-ansatz scattering matrix without free parameters. Beyond the scattering probabilities, the collision generates entanglement by superposing the transmitted and reflected components, which we infer from the particle-number fluctuation. In sharp contrast, a control collision between two doublons shows neither the interaction dependence nor the number fluctuation, as expected for indistinguishable particles. Our work demonstrates that the ability to create quasiparticles of a chosen flavor on demand and program their motion and interactions makes a digital quantum processor a new experimental toolset for probing the entanglement structure generated by quasiparticle scattering.","published":"2026-10-09T17:59:06Z","updated":"2026-10-09T17:59:06Z","authors":["Chan Bin Bark","Hanbyul Kim","Yoonil Koo","SeungBeom Hong","Seik Pak","Kwang Jun Ahn","Sangmo Cheon","Youngseok Kim","Moon Jip Park"],"categories":["quant-ph","cond-mat.str-el"],"headline":"Simulating particle collisions on a quantum computer reveals interactions","summary":"Researchers used a quantum computer to simulate collisions between special particles called quasiparticles in a model that helps understand strongly interacting systems. They measured how these particles scatter and found results matching theoretical predictions, showing that their quantum processor can mimic these complex interactions. The study also shows that particle collisions create quantum entanglement, highlighting the potential of quantum processors as tools for studying fundamental particle behavior."},{"id":"2610.13945v1","title":"Splat2Act: From Captured Scenes to Personalized VLA Manipulation","abstract":"A user who brings a generalist manipulation policy into their home or office wants it to pick up their objects from their shelves, counters, and tables. Yet even a strong generalist like $π_{0.5}$ achieves 0% pickup success in our captured deployment scenes without adaptation, and collecting real demonstrations in every user's environment is prohibitively expensive. We present Splat2Act, an automated real-to-sim-to-real pipeline that personalizes a pretrained vision-language-action (VLA) policy to a specific scene and object set using only a handheld phone scan and images of the user's objects. Unlike prior real-to-sim work that reconstructs individual tabletops, Splat2Act operates on building-scale indoor environments: it builds a photorealistic digital twin with 3D Gaussian Splatting and collision geometry, automatically discovers tabletop and non-tabletop receptacles, constructs feasible pickup tasks, and generates successful demonstrations with a privileged simulation oracle, with only scene and object capture required from a human. Applied to 101 scanned campus environments with a median navigable area of 104.87 m$^2$ and 7 receptacles, the pipeline yields 852 receptacles and over 68,000 pickup demonstrations. We fine-tune $π_{0.5}$ in two stages: Stage 1 adapts to hundreds of public object assets, drawn either from the deployment scene alone or from all 101 scenes; Stage 2 personalizes to the user's own objects on the deployment scene's receptacles. Our strongest personalized policy reaches 62.5% aggregate pickup success in the target scene. Deployed on a real Stretch 3, the same policy achieves 63.9% success on a 12-task subset, closely tracking its 66.7% simulated success on those tasks. These results suggest that automated real-to-sim-to-real training can personalize VLA manipulation policies using only simulation-oracle data.","published":"2026-10-09T17:59:06Z","updated":"2026-10-09T17:59:06Z","authors":["Xiaomeng Ye","Krishna Jaganathan","Gyubeum Lim","Gunjan Chhablani","Muhammad Zubair Irshad","Zsolt Kira"],"categories":["cs.RO","cs.CV"],"headline":"Teaching Robots to Pick Up Your Stuff Using Phone Scans","summary":"The researchers tackled the problem that general robot pickup policies fail to grab objects in new home or office scenes without costly demonstrations. They developed Splat2Act, a method that uses phone scans and photos to create detailed digital copies of rooms and objects, then simulates many pickup attempts to retrain the robot's policies. Their approach improved robot pickup success from 0% to over 60% in simulated and real tests, showing that adapting robot skills using simulated data from personalized scans is promising. The results are based on simulated and limited real robot trials."},{"id":"2610.13944v1","title":"InstantFusion: A Shared Latent Interface for Heterogeneous Image Generators","abstract":"Diffusion/flow models have emerged as the de facto paradigms for image generation, giving rise to a diverse family of modern generators with increasingly heterogeneous architectures and representations. Despite their remarkable generation capabilities, the cross-model alignment of latent representations and the consistency of their underlying denoising dynamics remain underexplored. In this work, we uncover an intriguing cross-model compatibility: independently trained generators exhibit similar latent spatial organization, while directly handing off intermediate states between unaligned models can still preserve coarse visual content. Motivated by these observations, we propose InstantFusion, a lightweight shared latent interface that connects the denoising trajectories of heterogeneous image generators. Specifically, to enable flexible cross-model state translation, model-specific Latent AutoEncoders (LAEs) are introduced to project intermediate states from different generators into a shared latent space and decode them into the latent space of a target generator. This shared space is constrained by two complementary objectives: aligning intermediate states derived from the same image at matched noise levels and synchronizing their corresponding local denoising velocities. Once trained, InstantFusion allows denoising to seamlessly transition across frozen generators without requiring a dedicated mapping for every model pair. The versatility of InstantFusion is demonstrated across acceleration, multi-preference composition, cross-architecture on-policy distillation, and cross-model composition, achieving competitive performance across these diverse settings.","published":"2026-10-09T17:59:04Z","updated":"2026-10-09T17:59:04Z","authors":["Xuanlang Dai","Jiazi Bu","Yujie Zhou","Yuhong Liu","Beichen Zhang","Penghui Yang","Ziyu Liu","Yuhang Zang"],"categories":["cs.CV"],"headline":"New method links different AI image generators through shared space","summary":"Researchers studied how various advanced AI models create images differently but found their internal organization shares similarities. They developed InstantFusion, a tool that connects the internal processes of different image generators by translating their intermediate data into a common format. This allows image creation to switch smoothly between models without special adjustments for each pair. Their work, based on theoretical design and testing, shows promise for combining and speeding up diverse image generation techniques, but real-world effectiveness may vary."},{"id":"2610.13943v1","title":"Noise Space Learning for VLA Adaptation","abstract":"Pretrained vision-language-action (VLA) models encode broad manipulation priors, yet adapting them to specific tasks requires balancing limited robot interaction, computational resources, and safety constraints. Noise space learning (NSL) adapts flow-matching-based VLAs by learning a context-conditioned distribution over noise while keeping the pretrained VLA fixed. Building on prior noise-steering methods, we formulate NSL as an interface for VLA adaptation that accommodates different learning algorithms and instantiate it with PPO, SAC, and AWR. Theoretically, we provide an information-theoretic analysis of how NSL enables efficient adaptation. Empirically, we study these NSL algorithms on LIBERO, RoboTwin, and four real-robot manipulation tasks under interaction, computation, and execution constraints. Across these settings, we develop a suite of effective NSL variants and derive practical recommendations for choosing among them under different resource and deployment regimes.","published":"2026-10-09T17:58:50Z","updated":"2026-10-09T17:58:50Z","authors":["Junjie Lu","Chuheng Zhang","Yucen Wang","Yuhua Jiang","Xinyao Qin","Min Xu","Li Zhao"],"categories":["cs.RO"],"headline":"Improving Robot Task Learning with Noise Space Adaptation","summary":"The researchers address the challenge of adapting broad vision-language-action (VLA) models to specific robot tasks while managing limited interactions, computing power, and safety. They propose noise space learning (NSL), a method that keeps the original model fixed but learns to adjust noise patterns based on the task context. The study provides theoretical analysis and tests NSL variants on simulated and real robot tasks, offering guidance on choosing methods under various practical constraints. Their findings are based on both theoretical reasoning and experimental evaluation."},{"id":"2610.13942v1","title":"OGLE-2016-BLG-0523: A Low-mass Brown Dwarf or super-Jupiter in a $>100$ au Orbit","abstract":"To search for very wide-orbit binary-lens events, we cross-match microlensing events detected by OGLE-IV and KMTNet between 2011 and 2024. After removing contaminants from variable stars, we identify OGLE-2016-BLG-0523 as a microlensing event showing two short-lived bumps in its light curve, with the two separated by 8.36 yr. We investigate three scenarios to explain the two bumps, including a binary-lens single-source (2L1S), a single-lens binary-source (1L2S), and two unrelated single-lens single-source events (2$\\times$1L1S). The 1L2S interpretation is excluded due to its poorer fit to the data, while the 2$\\times$1L1S scenario is disfavored by the low event rate, unphysical total source flux, and unphysical color for one of the two putative sources. For the surviving 2L1S interpretation, the Einstein timescales of the two bumps are $8.343 \\pm 0.076$ days and $3.231 \\pm 0.047$ days, respectively, implying a binary mass ratio of $q = 0.15$. A Bayesian analysis indicates that the primary lens is most likely a low-mass M dwarf, while the companion is probably a low-mass brown dwarf or super-Jupiter. The inferred projected separation is about 450 au, making the companion the widest low-mass object discovered via microlensing to date. Our discovery highlights both the promise and challenges of identifying very wide, low-mass companions through microlensing. Although upcoming facilities such as Roman and Earth 2.0 space telescopes will improve sensitivity to such events, overcoming degeneracies and distinguishing them from variable phenomena will remain essential for building a robust statistical sample.","published":"2026-10-09T17:58:27Z","updated":"2026-10-09T17:58:27Z","authors":["Hongjing Yang","Weicheng Zang","Yoon-Hyun Ryu","Andrzej Udalski","Yuchen Tang","Michael D. Albrow","Sun-Ju Chung","Andrew Gould","Cheongho Han","Kyu-Ha Hwang","Youn Kil Jung","In-Gu Shin","Yossi Shvartzvald","Jennifer C. Yee","Sang-Mok Cha","Dong-Jin Kim","Hyoun-Woo Kim","Seung-Lee Kim","Chung-Uk Lee","Dong-Joo Lee","Yongseok Lee","Byeong-Gon Park","Richard W. Pogge","Przemek Mróz","Michał K. Szymański","Jan Skowron","Radoslaw Poleski","Igor Soszyński","Paweł Pietrukowicz","Szymon Kozłowski","Krzysztof A. Rybicki","Patryk Iwanek","Krzysztof Ulaczyk","Marcin Wrona","Mariusz Gromadzki","Mateusz J. Mróz","Jiyuan Zhang","Shude Mao","Hongyu Li","Qiyue Qian"],"categories":["astro-ph.EP","astro-ph.GA","astro-ph.SR"],"headline":"Discovery of a distant low-mass companion using microlensing","summary":"Researchers studied a rare microlensing event with two bumps separated by over 8 years, linked to a binary system. They ruled out some explanations and found the best fit suggests a low-mass star with a companion that could be a brown dwarf or super-Jupiter orbiting very far away. This is the widest low-mass object found by microlensing so far. The study shows both the potential and difficulty of detecting such distant companions, with future telescopes expected to improve observations but challenges remaining in analysis."},{"id":"2610.13941v1","title":"Images Talk Back: Guiding MMDiTs by Suppressing Image-to-Text Feedback","abstract":"Multimodal diffusion transformers (MMDiTs) allow the evolving image to update internal text representations through joint attention. These representations then guide subsequent image updates, raising a question: does this feedback help correct deviations from the prompt, or preserve them? We find that image-to-text (I2T) feedback can preserve the objects and attributes already represented in the evolving image, including errors. Transplanting I2T messages can rescue failures or disrupt successes. At matched intervention strengths, suppressing the influence of I2T feedback improves generation, whereas opposite controls degrade it. These findings motivate Semantic Feedback Guidance (SFG), which suppresses the influence of I2T feedback by guiding away from an auxiliary branch that strengthens this feedback and weakens text-to-image conditioning. The in-place variant, iSFG, directly suppresses I2T feedback and strengthens text-to-image conditioning, requiring no extra model evaluations. Temporal interventions motivate running the SFG branch during the first 20% of image sampling steps, requiring about $1.2\\times$ the base NFE without classifier-free guidance (CFG). Experiments across three image-generation backbones show gains in alignment and preference-model scores. On FLUX, adding SFG to CFG raises GenEval from $0.6661$ to $0.7479$; on SD3.5M, iSFG adds $+1.33$ HPSv2.1 over CFG at zero extra NFE. The control also transfers to text-to-video and image-to-video generation, improving VBench scores. Code is available at https://github.com/yyc1215lll-creator/SFG.","published":"2026-10-09T17:58:15Z","updated":"2026-10-09T17:58:15Z","authors":["Yucheng Yang","Xiaoxiao Sun","Shusheng Yang","Jiakang Chen","Serena Yeung-Levy"],"categories":["cs.CV"],"headline":"Improving AI Image Generation by Reducing Feedback Loops","summary":"Researchers studied how an AI image-generation tool uses feedback from its evolving images to update the guiding text. They found that this feedback can preserve errors instead of fixing them. By suppressing this feedback through a new method called Semantic Feedback Guidance (SFG), they improved the accuracy and quality of the generated images. Their experiments with different AI models showed better alignment to prompts and higher preference scores, with some improvements extending to video generation. Results are based on tested interventions during image creation steps."},{"id":"2610.13940v1","title":"Sandpile avalanches in two dimensions are finite","abstract":"In the abelian sandpile model on $\\mathbb{Z}^d$, a site with at least $2d$ grains topples by sending one grain to each neighbor. In a box $[-L,L]^d$ whose boundary absorbs grains, adding grains at random sites and toppling defines a Markov chain, and Athreya and Járai (2004) constructed the stationary state on $\\mathbb{Z}^d$ as the limit of its stationary distributions as $L\\to\\infty$. By a formula of Dhar (1990), when a grain is added at the origin of a sandpile in the stationary state, the expected number of topplings at a site in the resulting avalanche is proportional to the Green function of simple random walk from the origin. This Green function is finite for $d\\geq3$, where Járai and Redig (2008) proved that avalanches are almost surely finite, and infinite for $d=2$. We prove that for $d=2$ avalanches are nevertheless almost surely finite: the probability that an avalanche reaches distance $R$ is at most a negative power of $\\log R$. The proof follows the avalanche backward and reduces the question to an estimate for the uniform spanning tree, which we prove by a multiscale electrical argument.","published":"2026-10-09T17:57:56Z","updated":"2026-10-09T17:57:56Z","authors":["Ahmed Bou-Rabee"],"categories":["math.PR"],"headline":"Researchers Show 2D Sandpile Avalanches Always End","summary":"The study looks at a sandpile model where grains topple and spread to neighbors, sometimes causing avalanches. While previous work suggested avalanches might be infinite in two dimensions, the researchers proved that avalanches in two dimensions always stop eventually. They used mathematical tools involving random walks and spanning trees to show the chance of an avalanche reaching very far decreases rapidly. Their results are theoretical and rely on complex probability and electrical network arguments."},{"id":"2610.13939v1","title":"The Tate Conjecture for Ordinary Calabi-Yau Threefolds","abstract":"We use the theory of canonical lifts to prove the integral geometric Tate conjecture for ordinary Calabi-Yau threefolds over finite fields, and also for divisor classes on certain ordinary Calabi-Yau varieties.","published":"2026-10-09T17:57:50Z","updated":"2026-10-09T17:57:50Z","authors":["Lukas Brantner"],"categories":["math.AG","math.NT"],"headline":"Researchers study a math problem on special geometric shapes","summary":"The researchers addressed a mathematical problem called the Tate conjecture, focusing on special geometric objects known as ordinary Calabi-Yau threefolds over finite fields. They used a method involving canonical lifts to prove this conjecture in this context, and also for certain divisor classes on related shapes. Their findings are theoretical and specific to these complex geometric structures within abstract mathematics."},{"id":"2610.12470v1","title":"Dex-One2Many: Learning Dexterous Manipulation from a Single Human Demonstration","abstract":"While learning dexterous manipulation from a single human video offers a promising alternative to costly robot demonstrations, many recent methods predominantly imitate demonstrated motions. Such strict motion matching often limits generalization to initial object poses, goal poses, and grasps not shown in the video. Alternatively, discovering a policy via reinforcement learning (RL) allows for broad generalization, but without prior guidance, it struggles with high-dimensional exploration in complex, multi-stage tasks. To address these coupled generalization and exploration challenges, we present Dex-One2Many, a real-to-sim-to-real framework that learns a generalizable dexterous manipulation policy from a single human video. Our key insight is to abstract the video into sequential scene graphs that guide RL, enabling efficient exploration while preserving broad generalizability. The graphs serve as generative constraints for sampling diverse reset states and provide dense rewards for each stage. Because the graphs constrain relations rather than exact poses, these reset states cover object poses and grasps beyond the video, while initializing each stage from them with dense rewards keeps exploration short and guided. Trained entirely in simulation, Dex-One2Many transfers zero-shot to a real multi-fingered hand. Across five tool-use and manipulation tasks, Dex-One2Many exceeds baselines by 6.5% in seen configurations, while its robust generalization widens this gap to 71% in unseen scenarios.","published":"2026-10-08T17:59:58Z","updated":"2026-10-08T17:59:58Z","authors":["Jusuk Lee","Sungha Kim","Yeonsoo Park","Jonguk Cheon","Yoonkyo Jung","Yongjun You","H. Jin Kim","Jia-Bin Huang","Furong Huang","Youngseok Jang","Seungjae Lee"],"categories":["cs.RO","cs.CV"],"headline":"Teaching Robots Dexterous Tasks from a Single Human Video","summary":"The researchers addressed the challenge of training robot hands to perform complex tasks by learning from just one human video, rather than needing many demonstrations. They developed Dex-One2Many, a method that uses simplified scene descriptions from the video to guide learning in simulation, helping robots explore different object positions and grips beyond what the video shows. Their system, trained only in simulation, was tested on real robotic hands and showed better performance and generalization in manipulating tools compared to previous methods. These findings are based on simulated training and testing on several tasks."},{"id":"2610.12469v1","title":"Rubric-CEPR: Self-Evolving Image Editing via Reward-Verified Self-Distillation","abstract":"Instruction-guided image editors have become highly capable, yet improving them further still depends on human-edited training pairs or external reward models. Such supervision is costly to obtain and can reward plausible failures: a realistic output may leave the requested change undone or alter content that should be preserved. In this work, we strive to improve a pretrained image editor using only its own generations, without human-edited targets or an external training-time reward model. To this end, we propose a self-evolving framework, named Rubric-CEPR, that verifies the editor's own samples with its internal representations through a rubric-augmented Contrastive Edit-Preservation Reward (CEPR). A Planner proposes structured edit instructions from unlabeled images, the Editor samples multiple candidate edits, and a frozen Critic scores each candidate with decomposed rubric checks for edit realization, removal of the old state, and content preservation, using features already exposed by the editor. Non-compensatory gates reject infeasible candidates, and the best verified candidate is distilled into the editor through lightweight adapter training. On Qwen-Image-Edit, Rubric-CEPR improves ImgEdit from 4.36 to 4.60 (+5.5%), with a +24.9% gain on object isolation, and transfers to GEdit-Bench and Complex-Edit. The same procedure also improves Step1X-Edit by +7.8% on ImgEdit. We hope our approach will serve as a solid baseline for image editors that improve themselves from their own verified samples. Our code is publicly available at $\\href{https://riteshthawkar.github.io/Rubric-CEPR/}{\\text{this URL}}$","published":"2026-10-08T17:59:58Z","updated":"2026-10-08T17:59:58Z","authors":["Ritesh Thawkar","Shubham Patle","Shravan Venkatraman","Rao Muhammad Anwer"],"categories":["cs.CV"],"headline":"New method helps image editors improve themselves without extra data","summary":"Researchers developed Rubric-CEPR, a system that enables an image editing model to get better using only its own generated images, without needing human corrections or external feedback. It does this by having the model check its own edits against specific goals and keep the best ones to learn from. Their experiments show moderate improvements in editing quality on several test sets. However, these findings are based on specific tests and the method relies on the model's internal checks rather than external validation."},{"id":"2610.12468v1","title":"DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training","abstract":"We present DreamTrue, a multi-view, cross-embodiment robot world model for action-faithful and physically plausible video prediction. Training such a model on existing robot datasets faces two obstacles: imprecise calibration can impair action following, while limited coverage of unsuccessful interactions can bias predictions toward successful outcomes. To improve action following across embodiments, we render action trajectories into image-space conditions and introduce offline geometric calibration to align these conditions with the target videos. To broaden interaction coverage, we introduce counterfactual post-training, modifying recorded action trajectories and generating future videos under a wider range of actions and contact configurations. To provide feedback on these predictions without paired ground-truth futures, we construct a human-annotated video dataset covering robot, object, and interaction defects and use it to train an embodied video reward model. Its scores guide reinforcement-learning post-training toward more physically plausible interaction outcomes. On AgiBot, DreamTrue attains state-of-the-art action following, while reducing the human-assessed interaction defect rate from from 48.12% to 6.25%. Notably, our model ranks first in the world model track of the AgiBot World Challenge 2026. The project page can be found at https://brave-eai.github.io/DreamTrue.","published":"2026-10-08T17:59:51Z","updated":"2026-10-08T17:59:51Z","authors":["Junyan Li","Ruizhi Li","Yu Liu","Xiangshuo Liu","Mingchao Sun","Hongyu Pan","Mu Xu","Lue Fan","Zhaoxiang Zhang"],"categories":["cs.RO","cs.CV"],"headline":"New Robot Model Predicts Actions More Accurately and Realistically","summary":"Researchers developed DreamTrue, a robot world model that better predicts robot actions and interactions in videos. They addressed issues like camera calibration errors and lack of data on failed actions by adjusting action paths visually and simulating more varied scenarios. Using a specially created dataset with human feedback, they trained the model to improve the realism of its predictions. Tests showed DreamTrue follows actions more faithfully and reduces errors in predicted interactions, achieving top results in a recent robotics challenge."},{"id":"2610.12467v1","title":"CSF: Contextual Safety Filtering for Motion Generators","abstract":"Text-conditioned motion generators produce trackable whole-body motion, but they have no notion of scene-dependent safety: the same action may target an object or a person. Existing safeguards either inspect the prompt, require labeled motion data, or enforce geometric constraints; therefore, they do not directly account for how scene context changes a motion's meaning. We introduce contextual safety filtering (CSF), a training-free filter that grounds natural-language safety rules in safe and unsafe reference trajectories produced by the generator. For each active rule, safe and unsafe reference trajectories define an affine safety value that a safe reference tracking CBF-QP enforces. Across four pretrained generators with different architectures, CSF activates the intended rules in all explicit and scene-triggered unsafe cases and reduces the danger-event rate by up to 90%, while preserving 88-100% of benign motions. We demonstrate the complete system on a real-world Unitree G1, where it successfully prevents unsafe motions in a variety of scenarios, including interactions with humans and objects.","published":"2026-10-08T17:59:50Z","updated":"2026-10-08T17:59:50Z","authors":["Lizhi Yang","Yiling Hou","Yao Tang","Junheng Li","Daniel Weng","Blake Werner","Aaron D. Ames"],"categories":["cs.RO","cs.LG"],"headline":"New Method Helps Robots Avoid Unsafe Movements in Different Scenes","summary":"Researchers developed a safety filter called Contextual Safety Filtering (CSF) to help motion-generating systems understand when actions might be unsafe depending on the scene, such as targeting a person or object. Unlike past methods, CSF uses examples of safe and unsafe motions to guide the system without extra training data. Tests on several motion generators and a real robot showed CSF reduces unsafe actions by up to 90% while keeping most normal motions unchanged. These results are from controlled experiments and real-world demonstrations described by the researchers."},{"id":"2610.12466v1","title":"On the estimation and validity of AI time horizons---a statistical look at the METR plot","abstract":"METR's 50\\% time horizon measures the human completion time of software tasks that an AI solves with 50\\% probability, allowing AI capabilities to be expressed in interpretable units. On 228 tasks and 26 AIs, we recompute the time horizons using splines and item-response theory to relax the assumption that the AI difficulty of a task depends linearly on the log of human time. Our fitted spline can be interpreted as a function that \\emph{converts} human time to AI difficulty; it is nearly flat in a region from 2--30 min but close to linear elsewhere. Hence, a time-horizon jump from 3 min to 30 min is much easier than one from 30 min to 5 hours despite the same multiplier of $10 \\times$. Overall, we contribute time-horizon point estimates that perform better under a cross-validated suite of proper scoring rules, as well as diagnostic plots for assessing time horizons' construct validity. We suggest that time horizons be interpreted together with the diagnostic plots, especially as new time-horizon-based benchmarks are proposed or existing ones grow to include longer tasks.","published":"2026-10-08T17:59:50Z","updated":"2026-10-08T17:59:50Z","authors":["Drew T. Nguyen","William Fithian"],"categories":["cs.AI"],"headline":"Researchers improve measuring AI task difficulty using human time","summary":"The researchers studied a method called the METR plot, which estimates how long humans take to complete tasks that AI can solve about half the time. They analyzed data from 228 tasks and 26 different AIs and found that the relationship between human completion time and AI difficulty is not always a straight line as previously assumed. They improved the measurement by using more flexible mathematical tools and provided better estimates and diagnostic tools to interpret AI performance benchmarks more accurately. Their findings come from statistical analysis and suggest cautious interpretation of AI time horizons."},{"id":"2610.12465v1","title":"A Balanced Data Diet: Addressing the Exploration Bottleneck in Mega-Scale RL for Robot Control","abstract":"General-purpose robots must perform a wide range of tasks from agile locomotion to dexterous manipulation. While sim-to-real reinforcement learning (RL) has proven to be a useful tool for this goal, current RL pipelines depend on engineering-heavy, per-task structural priors such as shaped rewards and demonstrations. Recent work has shown that diverse simulator resets, combined with massively parallel simulation, can alleviate much of this engineering burden on several manipulation problems. However, we find that naively scaling this paradigm to more precise or dynamic problems remains non-trivial. While simulator resets can help with exploration, uniformly sampling over this distribution wastes a growing fraction of learning experience on task configurations the policy has already mastered or cannot yet attempt. This makes it challenging to see the expected benefits of scaling parallel environments for RL, since much of the learning signal in a batch is wasted during learning. To mitigate this, we introduce Success Guided Sampling (SGS), a simple adaptive sampler that concentrates RL training on task configurations around the frontier of the policy's capabilities. Doing so allows large-scale simulated RL to make the most out of the experience in a batch, enabling much more effective scaling to large-scale parallel simulation. Across experiments using up to $2^{20}$ (over one million) parallel environments, SGS enables RL to solve challenging multi-terrain quadruped locomotion and contact-rich assembly tasks that prior methods fail to solve. Finally, we distill the learned manipulation policies into RGB-based policies and demonstrate zero-shot transfer to several challenging assembly tasks on real hardware. Project website: https://sgs-rl.github.io/.","published":"2026-10-08T17:59:50Z","updated":"2026-10-08T17:59:50Z","authors":["Octi Zhang","Mateo Guaman Castro","Patrick Yin","Ignacio Dagnino","Abhishek Gupta","Rosario Scalise","Byron Boots"],"categories":["cs.RO","cs.LG"],"headline":"New Method Improves Robot Learning in Simulations","summary":"Researchers studied how general-purpose robots learn tasks like walking and object handling using simulated training environments. They found that randomly sampling tasks during training wastes effort on tasks robots already mastered or can't do yet. To fix this, they developed Success Guided Sampling (SGS), which focuses training on tasks just beyond the robot's current abilities. Using SGS, their simulated learning scaled to over a million parallel environments and solved tasks earlier methods could not. They also tested transferring these learned skills to real robots with promising results."},{"id":"2610.12464v1","title":"What 30,000 Hours of Ego-centric Video Does Not Teach","abstract":"World models offer a promising alternative to physics-based simulators, yet remain far from practical deployment. We ask how far scaling ego-centric human video takes them, using a dataset of 30,000 hours spanning over 1,000 scene types and 14,000 contributors. Rather than relying on opaque downstream metrics, we directly evaluate agent and object-interaction fidelity on a challenging out-of-distribution benchmark. Increasing training data by 100x improves both, but unevenly: the agent is modeled well, while object fidelity remains far lower and improves slowly. We show that the agent gains need not come from data, and a careful visual conditioning design saturates fidelity with a fraction of it, which lets us measure object fidelity on its own and discover its saturation point. We then introduce a supervision scheme that shifts capacity from scene appearance toward object dynamics, improving object fidelity though a substantial gap remains. Finally, our conclusions transfer to downstream humanoid modeling. Overall, our results suggest that scaling ego-centric data brings agent modeling close to its limit while leaving its effects on the world far behind, and that closing this gap will depend on how models are trained, not only on how much data they see.","published":"2026-10-08T17:59:49Z","updated":"2026-10-08T17:59:49Z","authors":["Jiahua Dong","Anurag Bagchi","Yash Jangir","Muhammad Zubair Irshad","Sergey Zakharov","Martial Hebert","Homanga Bharadhwaj","Yu-Xiong Wang","Vitor Campagnolo Guizilini","Pavel Tokmakov"],"categories":["cs.CV"],"headline":"Scaling Video Data Helps Agent Models, But Objects Remain Hard","summary":"Researchers studied how increasing large amounts of first-person video data affects computer models that predict human and object behavior. Using 30,000 hours of video, they found that while the models improved significantly at representing human agents, they struggled much more at accurately modeling how objects move and interact. They show that simply adding more data is not enough; improving object modeling requires better training methods. Their findings highlight limits in current approaches and suggest future improvements depend more on training design than data volume."},{"id":"2610.12463v1","title":"From Reactive Containment to Proactive Assurance: Lessons from OpenAI, Anthropic, and Google Agent Security Incidents","abstract":"In 2026, cybersecurity evaluations involving OpenAI, Anthropic, and Google agents reached real systems outside their authorized test scope. The paths were different. OpenAI agents exploited research infrastructure, coordinated across runs, and compromised parts of Hugging Face's production environment. Anthropic reported cases in which a misconfigured third-party environment exposed real systems to agents pursuing simulated cyber tasks. In a separately reported evaluation, Google's Gemini accessed three real organizations through an unintended internet route; Google stated that the model stopped in all three instances. Taken together, the cases show why an evaluation cannot rely on an assumed boundary. That boundary must be verified while the agent is operating. This comparative instrumental case study develops a Proactive Agent Security Assurance Cycle (PASAC) and a five-layer Boundary Assurance Stack. The framework combines risk-tiered task design, executable scope contracts, pre-run validation, least-capability access, independent egress enforcement, credential restrictions, cross-run monitoring, automatic stop conditions, and evidence-based reauthorization. A leading-indicator model, nine design propositions, and seven falsifiable hypotheses turn these lessons into a testable research program. Because the public Gemini record is limited to attributed statements and journalism, its detailed causal mechanism remains provisional. The central conclusion is straightforward: proactive agent security requires continuous assurance across the full execution system, not confidence in any single sandbox or safeguard.","published":"2026-10-08T17:59:49Z","updated":"2026-10-08T17:59:49Z","authors":["Abbas Raftari"],"categories":["cs.CR","cs.AI"],"headline":"New Framework for Safer AI Agent Security Developed","summary":"Researchers studied cybersecurity incidents involving AI agents from OpenAI, Anthropic, and Google that unintentionally reached real computer systems beyond their testing limits. They found that assuming safety boundaries without constant checks is risky. To address this, the researchers proposed a new security framework called the Proactive Agent Security Assurance Cycle (PASAC) with multiple layers of protections and continuous monitoring. Their findings emphasize the need for ongoing safeguards rather than relying on one-time security measures. Some details about Google’s case remain uncertain due to limited public information."},{"id":"2610.12462v1","title":"Disk Structure May Determine AGN Variability and Explain the Accretion Disk Size Problem: No Broad Line Region Required","abstract":"Variability is a crucial probe of active galactic nuclei (AGN) disks, but the impact of the structure and internal physics of the disk on this variability is poorly understood. We perform 3D multi-frequency radiation magnetohydrodynamic simulations of the $1300-5940~\\rmÅ$ emitting region of an AGN disk. This turbulent disk simulation happens to stochastically have different disk thickness above and below the midplane. Light curves emitted from the thin top of the disk are broadly consistent with traditional variability models, where the X-ray irradiation emitted from the corona is the main driver of variability in UV-optical light curves. Conversely, the UV-optical light curves emitted from the thicker bottom of the disk are not strongly correlated with the X-ray light curve and have disk continuum reverberation mapping lags that are $3-5\\times$ longer than the light travel time. While these features are inconsistent with traditional variability models, they are in good agreement with numerous recent observations. Our simulations allow us to interpret these recent observations with a new model for AGN disks, where the thicker disk leads to greater X-ray absorption, less X-ray reprocessing at larger disk radii, and a stronger signal from intrinsic variability driven by fluctuations in the AGN disk. In our simulations, we can measure the inflow timescale of these fluctuations as a long lag between light curves emitted from different disk radii. Understanding how disk structure impacts AGN variability and reverberation mapping can allow us to predict the thickness of AGN disks using UV-optical AGN light curves.","published":"2026-10-08T17:59:49Z","updated":"2026-10-08T17:59:49Z","authors":["Amy Secunda","Yan-Fei Jiang","Jenny E. Greene"],"categories":["astro-ph.HE","astro-ph.GA"],"headline":"New Disk Shape Model May Explain Black Hole Light Changes","summary":"Researchers studied the swirling disks around active black holes using detailed computer simulations. They found that differences in disk thickness can change how light varies over time, challenging older models that rely mainly on X-rays causing brightness changes. Their findings match recent observations and suggest that thicker parts of the disk absorb more X-rays and show delayed light signals. This work helps explain black hole brightness changes and may let scientists estimate disk thickness by studying light patterns."},{"id":"2610.12461v1","title":"OuroWorld: Bringing Any 3D World Alive as Diverse, Endlessly Looping 3D Cinemagraphs","abstract":"Recent 3D world models generate photorealistic, explorable scenes that remain frozen in time. OuroWorld is a mask-free framework that turns any static 3D Gaussian Splatting scene into a 3D cinemagraph: a dynamic scene with vivid, diverse motion looping seamlessly from any viewpoint. A vision-language model infers plausible dynamics and guides a video model to synthesize a reference video, which we lift and complete into multi-view videos. To learn from this imperfect supervision, we propose Inconsistency-Robust Periodic 4DGS: a Fourier-series deformation field guarantees looping by construction, while a Grounded Drift Field anchored at the reference view absorbs cross-view inconsistency. Unlike prior Eulerian methods limited to fluid-like motion, we capture general deformation, object motion, and illumination change. We introduce a ground-truth-free evaluation covering vividness, naturalness, loop seam coherence, and scene quality. On 39 reconstructed and generated scenes, OuroWorld outperforms all baselines and wins 70.8%-99.0% of user-study comparisons. Project page: https://ouroworld.userwei.com","published":"2026-10-08T17:59:46Z","updated":"2026-10-08T17:59:46Z","authors":["You-Zhe Xie","Ting-Wei Chou","Yu-Hsuan Li","Kaipeng Zhang","Zhixiang Wang","Yu-Lun Liu"],"categories":["cs.CV","cs.GR"],"headline":"Turning Static 3D Scenes into Lifelike, Looping Videos","summary":"Researchers developed OuroWorld, a method that transforms frozen 3D scenes into dynamic, endlessly looping videos viewable from any angle. They use language understanding to imagine natural motions, then create videos showing these movements, ensuring smooth loops through mathematical modeling. Their approach captures varied motions and lighting changes beyond simple fluid effects. Tests on many scenes showed their method creates more vivid and natural animations than other techniques, although the results are based on computational models rather than physical experiments."},{"id":"2610.10540v1","title":"Null string holography in AdS from local BRST descent","abstract":"We show that the radial sector of a Carroll-Weyl gauged null string in Poincaré $\\mathrm{AdS}_D$ is locally trivial in BRST cohomology. In a quotient-adapted BFV basis, the radial coordinate, its conjugate momentum, and the associated ghosts form a contractible quartet. An explicit contracting homotopy extends this reduction to the full algebra of local spatial jets, leaving precisely the $(D-1)$-dimensional ILST BRST complex. The same quartet carries zero net Virasoro central charge, so that the anomaly condition of the parent theory, $D=27$, descends to the ILST condition $d=26$.","published":"2026-10-07T17:59:59Z","updated":"2026-10-07T17:59:59Z","authors":["Sarthak Duary","Sourav Maji"],"categories":["hep-th"],"headline":"Researchers analyze null string behavior in curved space holography","summary":"The researchers studied a special type of string called a null string in a curved space known as Anti-de Sitter (AdS) space. They found that certain mathematical parts related to the string's radial direction become simple and do not contribute to complexity in the theory, using techniques from BRST cohomology. Their work clarifies how a 27-dimensional condition simplifies down to a well-known 26-dimensional condition in a lower-dimensional theory. This study is theoretical and based on advanced mathematical physics concepts."},{"id":"2610.10539v1","title":"Tetris3D: 3D Scene Generation With Objects That Fit Together","abstract":"We propose Tetris3D, a generative framework for single-image 3D scene reconstruction that recovers objects which are physically and geometrically coherent as a scene. Existing methods often generate objects independently or couple them implicitly, providing limited guidance for ensuring fine-grained spatial compatibility between neighboring objects that interact with one another. To address this, we explicitly condition the generation of each object on the geometry of surrounding objects and their physical relationships, guiding its shape and pose to remain geometrically and physically plausible within the scene. Moreover, we introduce ComOb, a physics simulation-based dataset of 1.2M scenes featuring physical interactions across diverse object categories, with per-object meshes and pairwise physical relation annotations. Comprehensive experiments on synthetic and realworld scenes show that Tetris3D recovers coherent object shapes and poses even when interacting regions are occluded, and achieves state-of-the-art performance in both generation quality and physical stability.","published":"2026-10-07T17:59:48Z","updated":"2026-10-07T17:59:48Z","authors":["Jaeyeong Kim","Jinhyuk Jang","Jongmin Lee","Kyehong Park","Seungryong Kim"],"categories":["cs.CV"],"headline":"New Method Creates Realistic 3D Scenes with Fitting Objects","summary":"Researchers developed Tetris3D, a system that generates 3D scenes from single images by creating objects that physically and geometrically fit together. Unlike earlier methods that treat objects independently, Tetris3D considers surrounding objects and their physical relationships to produce realistic shapes and positions. They also created a large dataset, ComOb, with simulated scenes and detailed object interactions. Tests show Tetris3D improves the accuracy and stability of generated scenes, though this is based on simulation and image data."},{"id":"2610.10538v1","title":"Never Look Back: Understanding Persistence in 3D Object Memory from Egocentric Videos","abstract":"As we move through the world and carry out everyday tasks, we encounter objects that may become relevant only later. We are capable of recalling where we left something or what was inside a container, even without knowing we would need it later. Here, we study how an embodied assistant can build a similar memory from egocentric videos, by observing a person's day-to-day activities. We present Ledger, a persistent 3D object memory that combines object locations, their histories, and contextual descriptions. It associates observations across the recording and retains objects after they leave the view, including those the person never touches. It clusters each object's observations by resting locations and records a move only after repeated evidence, reducing the effect of localization noise. Short descriptions preserve details such as an object's contents or supporting surface. It saves these records to later answer spatial questions without having to access the original images or video. Our memory raises HD-EPIC accuracy from 29.7% to 42.6%, UCS-Bench accuracy from 33.8% to 38.5% and localizes Ego4D objects with a 0.99 m median error on returned predictions. Our analyses identify complementary roles for temporal persistence, contextual descriptions, and retrieval. Our study on 100 stitched streams of multiple scenes each further exposes failures in both retrieval and construction. Per-scene construction partially recovers the performance lost across scene changes compared to that of single scene streams.","published":"2026-10-07T17:59:41Z","updated":"2026-10-07T17:59:41Z","authors":["Shravan Chaudhari","William Paul","Suchi Saria","Rama Chellappa","Homanga Bharadhwaj"],"categories":["cs.CV","cs.AI","cs.RO"],"headline":"Building Persistent 3D Object Memories from First-Person Videos","summary":"Researchers developed Ledger, a system that watches daily activities through wearable cameras to remember where objects are and their histories, even after they leave view. Ledger groups observations by object resting spots and records moves only with repeated proof, reducing errors. It uses short descriptions to keep details and answers spatial questions without re-watching videos. Tests showed Ledger improved accuracy in locating objects but also revealed challenges when scenes change, highlighting areas for further improvement."},{"id":"2610.10537v1","title":"Sharp Polynomial Upper Bounds for Anticanonical Volumes","abstract":"For every positive integer $n$, we prove that there is a constant $C_n$, depending only on $n$, such that $Vol(-K_X) \\leq C_n ε^{-(2^n-n-1)}$ whenever $X$ admits an $ε$-lc log Fano boundary. The exponent is optimal in every dimension at least two, even for toric Fano varieties of Picard number one. In particular, the optimal exponent for fourfolds is eleven. We also prove an anticanonical interpolation theorem with optimal exponent $2^n-1$. The proof combines signed discrepancy estimates under projection, finite morphisms to projective space, and the canonical bundle formula. A reduction to a base bounded independently of $ε$, followed by a volume estimate along a flag, yields the sharper volume exponent.","published":"2026-10-07T17:59:30Z","updated":"2026-10-07T17:59:30Z","authors":["Pinxian Bie","Peien Du","Zhengjie Yu"],"categories":["math.AG"],"headline":"New Sharp Limits Found for Volumes in Complex Geometry","summary":"Researchers found a precise upper limit on a certain geometric measurement called anticanonical volume for objects known as log Fano varieties, depending on their dimension and a parameter epsilon. They proved the best possible exponent in these limits for dimensions two and higher, including specific cases like four-dimensional shapes. Their approach involved advanced mathematical techniques, but the findings remain theoretical and describe relationships in abstract geometric spaces rather than practical applications."},{"id":"2610.10536v1","title":"Decoupling Exploration from Optimization in RLVR","abstract":"Modern language models undergo reinforcement learning with verifiable rewards (RLVR) on top of already-trained checkpoints. A key promise of RLVR is the discovery of new reasoning strategies. In principle, a model can sample novel ideas absent from its prior training data. In practice, however, augmenting RLVR with strong novelty incentives has seen limited success and can degrade model quality. Because verifiable rewards supervise only a narrow slice of the model's knowledge and behavior, such degradations are difficult to recover from. Instead, we decouple exploration from optimization in a framework we call Exploration-Distillation (ExpDis). We train one or more explorer policies with a novelty bonus in the reward, filter their trajectories for correctness and quality, and distill them into a separate student policy. The student policy is then trained without a novelty bonus. We repeat the above procedure for several rounds, alternating between exploration and optimization. This decoupling allows us to aggressively scale exploration without degrading the student policy. Across seven mathematical reasoning benchmarks and two model families, ExpDis outperforms DAPO at the same wall-clock budget. Moreover, we observe improved pass@$k$ scaling, indicating that ExpDis produces models that generate more diverse correct solutions.","published":"2026-10-07T17:59:26Z","updated":"2026-10-07T17:59:26Z","authors":["Saif Punjwani","Micah Goldblum"],"categories":["cs.LG","cs.AI","cs.CL"],"headline":"New Method Boosts Language Model Reasoning Without Quality Loss","summary":"Researchers studied how to improve language models' reasoning by trying new ideas during reinforcement learning. They found that mixing creativity incentives directly with model training can harm quality. To fix this, they separate exploring new ideas from optimizing performance through a process called Exploration-Distillation, where explorations are filtered and taught to a separate model. Their experiments on math reasoning tests show this approach performs better than previous methods and creates models that produce a more diverse set of correct answers."},{"id":"2610.10535v1","title":"A weakly modelled view of the joint compact-binary mass plane: population structure and spectral-siren cosmology","abstract":"We introduce a flexible mixture model for the joint source-frame component-mass distribution of compact binaries, inferred simultaneously with cosmology, together with a framework that identifies structures in the reconstructed density, quantifying their posterior stability, local prominence, and cosmological relevance. Applied to GWTC-5.0, the workflow clarifies several familiar and emergent structures. The $\\sim35~M_\\odot$ feature is classified as a clear near-equal-mass peak despite an ambiguous interpretation in the primary-mass projection. A weaker unequal-mass concentration near $(21,12)~M_\\odot$ is recovered frequently but with lower prominence. A tentative joint transition near $60~M_\\odot$ appears at mildly unequal masses, with a more robust manifestation in the primary-mass projection. For cosmology, the two-dimensional mass model constraint on $H_0$ is $33\\%$ narrower than a model that flexibly reconstructs only the primary mass, and only $\\sim~10\\%$ wider than analyses based on strongly parameterized one-dimensional spectral structure. At the feature level, we show that the $\\sim10~M_\\odot$ and $\\sim35~M_\\odot$ concentrations depend on $H_0$ mainly through a common scale which captures the entire $\\sim~(1+z)$ cosmological rescaling of both masses; the intermediate unequal-mass feature instead retains residual dependence in the orthogonal mass-ratio direction. More generally, we argue that population prominence, formal inclusion in an astrophysical feature catalogue, and cosmological relevance need not coincide. These results suggest that a genuinely two-dimensional view may be required to fully exploit spectral-siren information. Our workflow provides a reproducible way to detect population structure and identify which aspects of its geometry are associated with spectral-siren constraints. The feature finder is directly applicable to other population reconstructions.","published":"2026-10-07T17:59:04Z","updated":"2026-10-07T17:59:04Z","authors":["Viola De Renzis","Alessandro Agapito","Michele Mancarella"],"categories":["gr-qc"],"headline":"New model reveals detailed patterns in black hole masses and cosmology","summary":"Researchers developed a flexible model to study how pairs of dense objects like black holes vary in mass and how these patterns relate to the universe's expansion. Applying this model to recent gravitational wave data, they identified clear mass groups and measured how these relate to cosmological factors like the Hubble constant (H0). Their work shows that considering two masses together offers tighter constraints on cosmology than looking at one mass alone. These findings come from their data analysis approach and suggest more complex mass relations matter for understanding the cosmos."},{"id":"2610.10534v1","title":"RoboPrompt: Intuitive Robot Policy Steering with Sparse Human Input","abstract":"End-to-end robot policies trained through imitation learning remain constrained by limited data diversity, making reliable zero-shot deployment in real-world settings challenging. Shared-autonomy methods enable human correction through teleoperation, but specialized hardware and operator training hinder deployment at scale. Other approaches incorporate human guidance as additional policy inputs, often requiring architectural changes and dedicated training for steerability, which limits their applicability across policies. We present RoboPrompt, a general-purpose, lightweight robot policy steering system that enables users to guide policy behavior through intuitive, sparse inputs, including drawn traces, target points, and coarse directional instructions. RoboPrompt decouples human-intention translation from the underlying policy: a reusable module converts human guidance into action drafts, which are refined through the diffusion or flow-matching dynamics of the base policy. By controlling action generation in noise space, RoboPrompt balances human intent with the policy prior without modifying the base policy architecture or fine-tuning it for steerability. Experiments demonstrate effective steering across Diffusion Policy, $π_{0.5}$, and FastWAM. We further use steered rollouts for online policy improvement through DAgger. After 2-3 rounds of iteration, average success rates increase by 15.5\\% for $π_{0.5}$ across three tasks and by 21.3\\% across three policies(Diffusion Policy, $π_{0.5}$, FastWAM) on the Insert Bread task, while average human intervention counts decrease by 44.0\\% (2.86 to 1.60) and 81.9\\% (2.60 to 0.47), respectively.","published":"2026-10-07T17:58:53Z","updated":"2026-10-07T17:58:53Z","authors":["Yanwen Zou","Chenyang Shi","Guoxuan Xu","Wenye Yu","Wendi Chen","Ye Pan","Cewu Lu","Chuan Wen"],"categories":["cs.RO"],"headline":"New method helps humans guide robots with simple inputs","summary":"The researchers address the problem that robot policies trained by imitation learning struggle to perform well in new real-world situations. They developed RoboPrompt, a system that lets users steer robot behavior using easy, minimal inputs like drawings or simple directions without changing the robot's original control program. Their experiments show RoboPrompt improves task success and reduces the amount of human corrections needed, although results are based on specific tested policies and tasks."},{"id":"2610.10533v1","title":"EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory","abstract":"Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs) with limited additional computation. Beyond model scaling, this architecture has demonstrated the potential to decouple factual knowledge storage from general-purpose computation, offering a promising route to updating factual knowledge while keeping the Transformer backbone fixed. Realizing this potential is challenging because different expressions of a fact may activate different n-gram embeddings, while updating shared embeddings can unintentionally change the model's predictions about other facts. We propose EngramEdit for decoupled knowledge updates through conditional memory. EngramEdit first computes target memory representations that make the model predict the updated fact across multiple expressions. It then jointly updates the shared n-gram embeddings to match these targets across expressions and edits, penalizing updates to frequently reused embeddings more strongly to preserve unrelated knowledge. Experiments show that EngramEdit enables independent factual knowledge updates through conditional memory, achieving near-perfect editing success. Revised knowledge is usable across unseen expressions and in multi-hop reasoning, with nearly three times the strongest baseline's accuracy under chain-of-thought (CoT) prompting. Unrelated knowledge and general capabilities are largely preserved even as factual updates accumulate. These findings show that EngramEdit turns conditional memory into an editable knowledge interface, extending its role beyond model scaling to support decoupled knowledge updates.","published":"2026-10-07T17:58:52Z","updated":"2026-10-07T17:58:52Z","authors":["Hongru Cai","Ran Wei","Wenjie Wang","Chengfa Wu","Ning Song","Yongqi Li","Wenjie Li"],"categories":["cs.CL"],"headline":"New method updates facts in language models without retraining","summary":"Researchers studied a way to update factual knowledge in large language models without changing the entire model. They developed EngramEdit, a technique that adjusts specific memory components to revise facts while preserving other knowledge. Tests showed EngramEdit effectively updates facts even in varied expressions and complex reasoning, keeping unrelated information intact. These results come from experiments with a special memory-based model architecture and suggest a new approach to managing knowledge in language models."},{"id":"2610.10532v1","title":"Predictive Rerouting of Connected and Automated Vehicles Using Traffic and Charging Demand Forecasts","abstract":"In this paper, we consider the problem of predictive rerouting of connected and automated vehicles (CAVs) in mixed traffic with electric vehicle charging demand. We provide a framework that combines a diffusion convolutional recurrent neural network (DCRNN) with a routing policy that accounts for congestion, CO2 emissions, route length, and charging demand. The DCRNN uses historical network observations to forecast traffic conditions and charging demand. These forecasts are then used to evaluate feasible alternative routes for eligible CAVs. A route change is accepted when the alternative preserves connectivity to the original destination and improves the prescribed route cost. We evaluate the proposed framework in SUMO under controlled traffic disruptions at five CAV penetration levels, ranging from 5% to 45%. We compare its performance with K-shortest-path routing, predictive-density routing, V2X proactive routing, and a reference scenario without rerouting. In the considered scenarios, the proposed framework reduces the average travel-time index by approximately 1.8% relative to the reference scenario and achieves the lowest average travel-time index, highest average speed, and lowest aggregate CO2 emissions among the active routing methods. Across all five penetration levels, it accepts 41 route changes, compared with 146 for K-shortest-path routing and 153 for V2X proactive routing. The reference scenario retains lower aggregate emissions and distance traveled, illustrating the tradeoff between congestion reduction and the additional travel associated with rerouting.","published":"2026-10-07T17:58:52Z","updated":"2026-10-07T17:58:52Z","authors":["Tony Kinchen","Ting Bai","Andreas A. Malikopoulos"],"categories":["cs.NI","eess.SY"],"headline":"Smart Rerouting for Automated Cars Cuts Travel Time and Emissions","summary":"Researchers developed a new system to predict traffic and electric vehicle charging needs, helping automated cars choose better routes. They used a special neural network to forecast conditions and tested their method in simulated traffic with different levels of automated car use. The findings show small improvements in travel time, speed, and emissions compared to other routing methods, but also highlight a tradeoff as rerouting sometimes increases total distance traveled. The results are based on simulations, not real-world trials."},{"id":"2610.10531v1","title":"Efficient Estimation of Logical Sensitivities Through Fault-Counting","abstract":"Quantum error-correcting circuits are affected by multiple physical noise mechanisms, whose contributions to logical failure must be understood to evaluate code performance and guide improvements in hardware. The individual error budget contributions for an error type $i$ can be characterized by its logical sensitivity $ν_i = \\frac{\\partial p_L}{\\partial p_i}$, which measures the response of the logical error rate $p_L$ to each physical noise parameter $p_i$. Normally, $ν_i$ is measured using linear fits such as finite differences, where $p_L$ is measured at two or more values of $p_i$ to calculate partial derivatives. In this paper, we develop a differentiable estimator to obtain all components of $\\nabla_{\\mathbf p}p_L$ simultaneously from a single Monte Carlo data set at one noise configuration, using information about the underlying fault configurations. In surface code simulations with circuit-level noise, the estimator agrees with conventional finite differences while requiring one to two orders of magnitude fewer shots to achieve the same variance. We apply this technique to quantify error budgets, resolve sensitivities at the individual qubit level, and infer effective code distance. Finally, we incorporate the sensitivities into Newton root finding to locate and trace threshold contours in multidimensional noise models. These results provide an efficient method for extracting and applying logical sensitivity information from standard quantum error correction simulations.","published":"2026-10-07T17:58:42Z","updated":"2026-10-07T17:58:42Z","authors":["Winston Fu","J. Wilson Staples","Jeff D. Thompson"],"categories":["quant-ph"],"headline":"New Method Speeds Up Error Analysis in Quantum Computing","summary":"Researchers developed a method to more efficiently measure how different physical noise sources affect errors in quantum computers. Instead of the usual approach that requires many tests, their method uses one set of simulation data to estimate all error sensitivities at once. This helps analyze and improve quantum error correction, although the findings so far come from computer simulations, not experiments."},{"id":"2610.08791v1","title":"World Models' Last Exam in Physics","abstract":"Video world models can produce visually convincing yet physically inconsistent sequences, raising concerns about their reliability for prediction and planning in embodied AI systems. Existing evaluations often rely on model-based judgments or reference videos, while direct physical tests largely focus on mechanics. We introduce World Models' Last Exam in Physics, a measurement-based benchmark for evaluating physical consistency in video world models. The benchmark comprises 40 controlled tasks spanning mechanics, optics, fluids, thermal and phase-change phenomena, electromagnetism, and surface tension. Each task pairs an initial image and a generation prompt with predefined physical criteria, enabling interpretable tests of observable physical relationships without requiring reference videos. Its evaluator combines task-observability screening with task-specific quantitative physical measurements. Experiments on eight video generation models across 1,280 videos reveal persistent physical inconsistencies and substantial variation across tasks, with the best model achieving an overall score of 57.76 out of 100. Evaluation on synthetic videos with known physical relationships provides evidence for the validity of the measurement module under controlled conditions. The evaluator also achieves higher agreement with human judgments than a direct vision-language model baseline in both within-task rankings and pairwise comparisons. By combining coverage across physical domains with scores grounded in measurable evidence and explicit measurement limitations, the benchmark provides an interpretable basis for diagnosing physical inconsistencies and tracking progress toward physically consistent video world models.","published":"2026-10-06T17:59:56Z","updated":"2026-10-06T17:59:56Z","authors":["Mingju Gao","Qingle Liu","Yuzhao Peng","Xinjie Lin","Ziming Qin","Zheng Jiang","Wenyi Li","Calvin Xiao","Youjie Zheng","Kaisen Yang","Qinhuai Na"],"categories":["cs.CV"],"headline":"New Test Checks Physical Accuracy of AI Video Predictions","summary":"Researchers created a new benchmark called World Models' Last Exam in Physics to measure how well AI systems generate videos that follow real physical laws. The test involves 40 tasks covering areas like mechanics, optics, and fluids, each with clear physical criteria and no need for comparison videos. They evaluated eight video models on over a thousand videos, finding many physical errors and differences across tasks. Their approach agrees better with human judgments than previous methods and helps identify where AI video models struggle with physical consistency."},{"id":"2610.08790v1","title":"Building Rome from a Single Image","abstract":"Single-image scene generation aims to produce a complete 3D scene mesh from a single image, including surfaces the camera did not observe. While pretrained 3D object generators encode a strong shape prior, they are mainly designed for isolated objects in a fixed canonical volume and focus mostly on indoor scenes, since diverse 3D data for outdoor scenes are quite limited. In this work, we present a method that redesigns such an object-centric generator, e.g., Trellis 2, to work on both indoor and outdoor scenes while retaining its prior. We accomplish this by (a) partitioning the scene into adaptive chunks that scale relative to the distance to the camera; nearby chunks have a smaller size to keep the finer detail, while distant structures, e.g., buildings, are covered by large chunks; (b) making the generator capture explicit 2D-3D correspondence by lifting image features and making the model aware of the free space, observed surface, and unobserved region; (c) synthesizing around 4,000 outdoor scenes to broaden the training data, as existing scene datasets are largely indoor. Experiments on Tanks and Temples, ScanNet++, and in-the-wild images show that our method outperforms all baselines in geometric accuracy and perceptual quality across both indoor and outdoor scenes.","published":"2026-10-06T17:59:50Z","updated":"2026-10-06T17:59:50Z","authors":["Jiraphon Yenphraphai","Fang Li","Tianshuo Xu","Depu Meng","Quentin Herau","Yihan Hu","Raymond A. Yeh","Wei Zhan"],"categories":["cs.CV"],"headline":"Creating 3D Scene Models From a Single Photo","summary":"Researchers worked on a way to generate complete 3D models of scenes from just one image, including parts not visible in the photo. They improved existing methods by dividing scenes into different sized chunks based on distance, teaching the system to better link 2D image details to 3D shapes, and training it on thousands of outdoor scenes. Their tests showed their approach better captures the shape and appearance of both indoor and outdoor environments compared to earlier methods."},{"id":"2610.08789v1","title":"QF3: Fast Flow RL with Filtered Q-Gradients","abstract":"Flow policies have become a standard policy class for learning robot behaviors from demonstrations, but reinforcement learning is still critical for improving pre-trained flow policies or learning them from scratch through interaction. We introduce QF3 (Fast Flow RL with Filtered Q-Gradients), an online off-policy RL algorithm that trains a flow policy with flow matching plus the critic's action gradient, backpropagated through a one-step prediction of the flow's output. To keep updates where the critic and this prediction are reliable, QF3 applies the critic gradient only to action dimensions that stay near the replay action. To our knowledge, QF3 is the first off-policy flow RL method to train humanoid locomotion policies from scratch and transfer them zero-shot to hardware. Paired with a high-throughput off-policy training recipe, it trains humanoid locomotion and motion-tracking policies with a 10x wall-clock speedup over FPO++, a recent on-policy flow RL method. We further apply QF3 to fine-tune pretrained flow-based manipulation policies on both ABC-Sim and Robomimic tasks. These results suggest that QF3 can both learn robot policies from scratch and refine those acquired from demonstrations. Website: https://qf3-rl.github.io/","published":"2026-10-06T17:59:34Z","updated":"2026-10-06T17:59:34Z","authors":["Chung Min Kim","Brent Yi","David McAllister","Hongsuk Choi","Himanshu Gaurav Singh","Jinkun Cao","Ken Goldberg","Pieter Abbeel","Carmelo Sferrazza","Angjoo Kanazawa"],"categories":["cs.RO","cs.LG"],"headline":"New method speeds up robot learning from demonstrations and interaction","summary":"Researchers developed QF3, a reinforcement learning algorithm that improves robot behavior policies by combining flow matching with selective use of value gradients. QF3 focuses updates on reliable action areas and achieves much faster training speeds, reportedly about 10 times faster than a recent similar method. The researchers demonstrated QF3 can train humanoid walking from scratch and apply learned policies directly to real robots, as well as fine-tune pre-trained manipulation skills, based on simulation and benchmark tasks."},{"id":"2610.08788v1","title":"Chiral Central Charge from Real-Space Twist Operator Correlator","abstract":"Recent years have witnessed substantial progress in the entanglement-based characterization of quantum phases of matter, with growing interest in real-space formulas due to their theoretical and experimental relevance. For two-dimensional gapped quantum systems, we propose a universal real-space formula for the chiral central charge based on the twist-operator correlator $Z_{g,h}=\\langleψ^{(N)}\\rvert\\mathcal{T}_A(g)\\mathcal{T}_B(h)\\mathcal{T}_C(g)\\mathcal{T}_C(h)\\lvertψ^{(N)}\\rangle$, where $\\lvertψ^{(N)}\\rangle=\\lvertψ\\rangle^{\\otimes N}$ is the $N$-copy state and $g,h\\in S_N$ are permutations. The twist operator $\\mathcal{T}_R(g)$ permutes the $N$ replicas within region $R$ according to $g$. For pairs $(g,h)$ satisfying the spherical condition, we show that the phase of the correlator is determined by the chiral central charge $\\mathfrak{c}_{-}$ through $Z_{g,h}/\\left\\lvert Z_{g,h}\\right\\rvert=(θ_gθ_h/θ_{gh})^{\\mathfrak{c}_{-}}$, where $θ_g^{\\mathfrak{c}_{-}}$ is the topological spin of the twist defect labeled by $g\\in S_N$. The construction applies to both bosonic and fermionic systems. We derive the formula within a field-theoretic framework under suitable assumptions and further test it numerically in microscopic lattice models. Our results establish a universal connection between real-space entanglement structures and the chiral central charge, providing a systematic framework for characterizing chiral topological phases directly on the lattice.","published":"2026-10-06T17:59:31Z","updated":"2026-10-06T17:59:31Z","authors":["Yunlong Zang"],"categories":["cond-mat.str-el"],"headline":"New way to identify a key quantum property from real-space patterns","summary":"Researchers propose a universal formula to measure a property called the chiral central charge in two-dimensional quantum systems using patterns of entanglement in real space. They use a special mathematical tool called a twist operator correlator to relate this quantum property to measurable quantities. The team derived the formula theoretically and tested it with computer models, suggesting a new method to study certain quantum phases directly on lattice structures, though the approach depends on some assumptions in their theory."},{"id":"2610.08787v1","title":"Algebraic Tensor Network Renormalization and Holographic duality","abstract":"Emergent generalized symmetries and the holographic principle play very important roles in understanding quantum phase transitions. In recent years, tensor-network renormalization with generalized symmetry and the corresponding fixed-point tensor formulation have been proposed as a discrete spacetime framework for reformulating conformal field theory. Nevertheless, a complete holographic formulation within this framework is still lacking, and the current formulation does not directly accommodate sector partition functions that are modular covariant in general. In this work, we formulate $\\mathcal{R}$-TNR, which builds the categorical constraints of a fusion category $\\mathcal{R}$ into local tensor spaces and preserves these constraints under Loop-TNR. For suitable choices of $\\mathcal{R}$, the initial tensors constructed here flow under $\\mathcal{R}$-Loop-TNR to stable gapless fixed points. In the Ising, tricritical Ising, and three-state Potts conformal field theory examples, a broad range of initial parameters leads to the same gapless fixed point in each example. The method resolves conformal towers by anyon sector and yields scaling dimensions, conformal spins, and magnitudes of sector-resolved structure constants with extremely high accuracy compared to previous methods. These results realize a topological bootstrap based on stable symmetry-topological-order in which finite categorical and geometric input, together with an initial tensor that need not be fine-tuned, suffices to recover the infrared conformal field theory and its sector-resolved conformal data.","published":"2026-10-06T17:59:20Z","updated":"2026-10-06T17:59:20Z","authors":["Chenqi Meng","Tian Lan","Zhengcheng Gu"],"categories":["cond-mat.stat-mech","math-ph"],"headline":"New method links complex math to quantum phase transitions","summary":"Researchers developed a mathematical approach called $\\mathcal{R}$-TNR to better understand quantum phase transitions and related theories. This method uses advanced algebraic structures to improve precision in identifying key properties of physical systems, demonstrated in examples like the Ising model. Their work remains theoretical and uses simulations, showing promise in connecting symmetry and geometry with quantum phenomena without requiring exact starting conditions."},{"id":"2610.08786v1","title":"Energy-constrained two-way capacity bounds for noisy Gaussian channels","abstract":"A gap between achievable rates and converse bounds persists for energy-constrained bosonic communication in the presence of excess noise. We derive a unified adaptive weak-converse bound for thermal attenuation, noisy amplification, and additive Gaussian noise under an unconditional mean transmitted-photon-number constraint. The bound applies to arbitrary adaptive protocols with quantum memories and constrains the two-way quantum, entanglement-distribution, private, and secret-key capacities. It vanishes throughout the entanglement-breaking region and approaches the corresponding PLOB bound at infinite energy. At finite energy, it improves the evaluated Gaussian squashed-entanglement and PLOB bounds in relevant parameter regimes, while comparison with hashing rates quantifies the remaining gap to achievability. The result provides a common energy-dependent benchmark for noisy bosonic Gaussian channels and extends finite-energy converse methods beyond the quantum-limited setting.","published":"2026-10-06T17:59:12Z","updated":"2026-10-06T17:59:12Z","authors":["Stefano Pirandola"],"categories":["quant-ph","math-ph","physics.optics"],"headline":"New Energy Limits Found for Noisy Quantum Communication Channels","summary":"The researchers studied how much information can be reliably sent through noisy quantum systems that use light particles (photons) under energy limits. They developed a new mathematical bound that applies to various types of noise and complex communication methods, improving on previous limits at realistic energy levels. Their findings offer a better benchmark for understanding communication limits in noisy quantum optical channels, but the results are theoretical and based on mathematical modeling rather than experiments."},{"id":"2610.08785v1","title":"Conformal Prediction Sets Quantify Information Gain: A Theoretical Perspective","abstract":"Conformal prediction is a popular tool for uncertainty quantification that outputs prediction sets with finite-sample coverage guarantees. While prediction set size is commonly used as a heuristic measure of uncertainty, the information-theoretic basis for this interpretation remains poorly understood. In this work, we provide such a foundation using a decision-theoretic generalization of entropy tailored to set-valued prediction. In particular, we introduce a family of generalized information measures based on the size and coverage of conformal prediction sets. Notably, Shannon mutual information admits an exact integral representation in terms of these measures. We then show that, in standard classification settings, the reduction in conformal set size from additional information (i) is sandwiched between calibration-dependent members of this family and (ii) obeys a data processing inequality, both up to finite-sample calibration and model error terms. Together, our results formally relate conformal prediction to classical information-theoretic quantities and justify using set-size reduction as an information gain metric. Empirically, we validate our theory across 11 classification settings and show that set-size reduction and Shannon mutual information can rank features differently in a greedy feature selection experiment.","published":"2026-10-06T17:59:11Z","updated":"2026-10-06T17:59:11Z","authors":["Kevin Zhang","Stephen Bates"],"categories":["cs.LG"],"headline":"Researchers link prediction set size to information gained in uncertainty estimates","summary":"The researchers studied how measuring the size of prediction sets—a way to express uncertainty in forecasts—relates to how much information is gained when making predictions. They developed a new theoretical framework connecting prediction set size to ideas from information theory, showing this size reduction reflects information gain under certain conditions. Their findings, supported by tests on various classification tasks, explain why set-size reduction can be a meaningful measure of information, though results depend on model accuracy and calibration."},{"id":"2610.08784v1","title":"PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation","abstract":"Pretrained robotic policies can suffer substantial performance degradation under out-of-distribution (OOD) conditions encountered during deployment, motivating post-training through real-world interaction. However, reinforcement-learning (RL)-based post-training typically requires substantial environment interactions, a burden that is especially significant in manipulation, where each trial can be slow, costly, or destructive. Therefore, we present PEARS, a physics-prior-guided hybrid RL framework for sample-efficient online adaptation of pretrained policies with tactile feedback. After each episode, its physics-guided force reasoning (PFR) module uses physical priors encoded in a vision-language model (VLM) to diagnose failures from the visual outcome and tactile interaction history and update task-appropriate contact-force bounds. A high-frequency hybrid force-position controller then enforces these bounds during contact. Complementarily, tactile-conditioned diffusion steering reinforcement learning adjusts the latent noise of the frozen flow-matching policy to correct errors in free-space motion and contact timing without updating the base model. In simulation, PEARS improves success rates by 12.4-37.4 percentage points over the strongest per-task baselines. PEARS also reduces the number of interaction episodes required for a certain success threshold by up to 53.2% relative to the fastest baseline. In real-world experiments, PEARS achieves success rates of 95% on Whiteboard Erasing and 90% on Pipette Liquid Aspiration. These results show that combining the PFR module with policy steering can accelerate adaptation while reducing costly interactions. The project website is available at https://song-kun.github.io/pears.","published":"2026-10-06T17:59:03Z","updated":"2026-10-06T17:59:03Z","authors":["Kun Song","Yiming Wang","Yilin Chen","Tianyi Ding","Jiaxin Tian","Tianqi Gong","Daolin Ma","Jia Pan"],"categories":["cs.RO"],"headline":"New method helps robots adapt quickly to tough tasks using touch","summary":"Researchers developed PEARS, a system that helps robots improve their performance during tasks that involve touch, especially when facing unexpected challenges. It uses physical knowledge and visual-tactile feedback to understand mistakes and adjust force during interactions, plus a special learning approach to correct motion errors without retraining the whole model. In tests, PEARS increased success rates and reduced trial numbers in both simulations and real-world tasks. These findings suggest PEARS can help robots adapt more efficiently while minimizing costly trial-and-error, though results are based on specific experiments."},{"id":"2610.08783v1","title":"Quantum twisting microscopy as a momentum-resolved probe of quantum geometry and sublattice symmetries","abstract":"In two-dimensional (2D) flat-band systems, as engineered, for instance, by stacking and twisting van der Waals materials, the momentum dependence of the Bloch states is a key factor in determining the interaction-induced emergence of different phases. Hence, methods of probing this form of ''quantum geometry'' are crucial to understanding these systems. Yet, extracting the momentum dependence of the underlying quantum geometric tensor from measurements is challenging in 2D. Here, we propose employing a quantum twisting microscope (QTM) with a sublattice-polarized tip to fill this gap. We show how expectation values of sublattice Pauli matrices can be probed across the scanned momentum path. This allows for the direct measurement of crucial symmetries characterizing the different candidate instabilities in graphene-based moiré systems. We furthermore show that the quantum metric component along the scan direction can also be extracted; the full quantum geometric tensor is also accessible, given that more than one sublattice polarization can be used. Our work illustrates the potential and extends the versatility of the QTM approach. If implemented in future experiments, this would provide unprecedented access to the symmetry signatures of the correlated phases and to the underlying quantum geometry in twisted bilayer graphene and related systems.","published":"2026-10-06T17:59:02Z","updated":"2026-10-06T17:59:02Z","authors":["Lucas V. Pupim","Mathias S. Scheurer"],"categories":["cond-mat.mes-hall","cond-mat.str-el"],"headline":"New microscope idea probes quantum geometry in 2D materials","summary":"The researchers address the challenge of measuring how electron states change with momentum in layered two-dimensional materials, which is important for understanding their unusual properties. They propose using a quantum twisting microscope with a specialized tip to directly measure key symmetry features and parts of the quantum geometry in these systems. Their work is theoretical and shows how such measurements, if realized experimentally, could reveal important details about electronic phases in twisted graphene layers."},{"id":"2610.08782v1","title":"4D-HOF: Hand-Object Flow Matching for Feed-Forward 4D Interaction Reconstruction","abstract":"Existing methods for 4D hand-object reconstruction often rely on costly per-sequence optimization, while generative approaches typically synthesize interactions from random noise, which can lead to unstable interaction prediction. We introduce 4D-HOF, a feed-forward framework that reconstructs 4D hand-object interactions from coarse but informative estimates produced by vision foundation models. Concretely, we learn a conditional flow matching model that transports foundation-model-derived hand-object states toward an interaction manifold, allowing the model to correct errors in translation, rotation, and alignment in a feed-forward manner. A key advantage of our generative formulation is that it naturally enables test-time guidance within the transport process. Rather than applying a separate post-hoc optimization after reconstruction, we directly steer the evolving generative states using physical interaction constraints and observed 2D evidence, allowing the reconstruction to be refined as part of the generative process itself. By training the generative model on diverse datasets, 4D-HOF generalizes robustly to challenging in-the-wild scenarios. Experiments on out-of-domain benchmarks show that 4D-HOF achieves state-of-the-art performance, producing more stable and accurate 4D hand-object reconstructions.","published":"2026-10-06T17:59:02Z","updated":"2026-10-06T17:59:02Z","authors":["Shiqi Li","Sean Cho","Yijie Li","Fengzhi Guo","Bowen Wen","Cheng Zhang"],"categories":["cs.CV","cs.AI","cs.GR"],"headline":"New method improves 3D hand-object interaction reconstruction","summary":"Researchers developed 4D-HOF, a faster approach to recreate how hands interact with objects in 3D over time, using initial rough estimates from existing vision models. Their method adjusts errors directly during generation, guided by physical and visual cues, avoiding slower step-by-step corrections after reconstruction. Tested on varied datasets, the team found 4D-HOF produces more accurate and stable results than previous methods, though the findings rely on the datasets and setups they used."},{"id":"2610.06853v1","title":"On the Dirichlet problem for the degenerate $2$-Hessian equation","abstract":"In this paper, we establish global $C^{1,1}$ solvability of the Dirichlet problem for the degenerate $2$-Hessian equation on bounded strictly mean convex $C^{3,1}$ domains $Ω$, with general $C^{3,1}$ boundary values and nonnegative right-hand sides $f\\in C^{1,1}(\\overlineΩ)$, resolving the $k=2$ case of a longstanding open problem. We uncover a global semiconvexity structure for general $2$-admissible solutions by proving a lower bound for $σ_{3}[D^{2}u]$ independent of $\\inf_Ωf$. For $3\\leq k\\leq n-1$, we also establish global semiconvexity for $k$-admissible solutions on the unit ball with $C^{3}$ boundary values and $f^{1/(k-1)}\\in C^{1,1}(\\overline{B}_{1})$, with a bound independent of $\\inf_{B_{1}}f$. A counterexample shows that the $C^{3,1}$ boundary value assumption for global $C^{1,1}$ solvability is sharp in the Hölder scale.","published":"2026-10-05T17:59:55Z","updated":"2026-10-05T17:59:55Z","authors":["Yasheng Lyu"],"categories":["math.AP"],"headline":"Researchers solve a special mathematical boundary problem","summary":"The researchers addressed a difficult math problem involving the 2-Hessian equation, which describes certain shapes or surfaces, on well-behaved regions with specific boundary conditions. They proved that solutions exist with good smoothness properties and uncovered structural features of these solutions. Their findings also extend to related equations with different parameters. A counterexample shows their assumptions about smooth boundaries are necessary. These results come from theoretical analysis and resolve a long-standing open question in this area."},{"id":"2610.06852v1","title":"One Figure, Every Canvas: Editable Flowchart Relayout via Agentic Pipeline","abstract":"Pipeline figures in ML papers must be repurposed across many canvases, including paper columns, 16:9 slides, portrait posters, 1:1 social teasers, 9:16 phone previews. Each format imposes a different aspect ratio on the same computational graph, where any silently broken connection misrepresents the method. We formulate aspect-ratio-adaptive flowchart relayout as a distinct task: given a raster flowchart and a target ratio, produce a structurally faithful, hallucination-free, editable layout. Existing methods fail characteristically: image-to-image models stretch blocks and reject extreme ratios, text-to-image agentic systems hallucinate content, and parse-then-render systems mis-route edges. We propose an agentic pipeline factored into Parse, Style, and Layout stages, each pairing a main agent with a critic that combines deterministic constraint checks with VLM visual feedback so connectivity is explicitly checked and prevented from being silently broken. Outputs are draw.io-editable mxGraph XML. On a curated benchmark of 100 flowcharts at five aspect ratios, evaluated by Gemini 3.1 Pro and validated against human judgments, our method reaches 68.6% Content Fidelity versus 11.2-41.4% for prior work. Project page: https://onefigureeverycanvas.vercel.app/","published":"2026-10-05T17:59:54Z","updated":"2026-10-05T17:59:54Z","authors":["Shih-Chen Tseng","Chih-Hsuan Chen","Ryan Yang","Hsi-An Chen","Chun-Wei Tuan Mu","Yu-Lun Liu"],"categories":["cs.CV","cs.AI"],"headline":"New method improves resizing of flowchart images for different formats","summary":"Researchers addressed the challenge of adapting flowchart images from machine learning papers to various display formats without losing or misrepresenting information. They developed a three-stage process combining parsing, styling, and layout adjustments with careful checks to preserve connections and structure. Tested on 100 flowcharts across multiple aspect ratios, their approach better maintained content accuracy compared to previous methods, according to automated and human evaluations."},{"id":"2610.06851v1","title":"Base Models Can Reason By Taking a Cue From Training Data","abstract":"In this paper, we study how training data creates associations between the tokens at the start of a base model's response and the reasoning behavior that follows. First, we demonstrate that fixing particular starting token cues makes a base model's performance competitive with that of its reinforcement learning (RL)-trained counterparts on math and coding. For instance, the cue \".\\n\\nOkay\" raises Olmo-3-7B's MATH-500 pass@1 accuracy from 42% to 78%, while \"Alright,\" raises Qwen3-14B's from 72% to 87%. Second, RL makes these cues more likely, while fixing them recovers much of its performance gain over the base model. Third, we trace the reasoning effects of token cues to the training data. We perform causal data interventions to turn an arbitrary word, such as \"chicken\", into an effective reasoning cue, or remove an existing cue's effect. A similar edit makes the prompt instruction \"Think duck duck goose\" as effective as \"Think step by step\" at eliciting reasoning. We also find that the hidden state representations induced by different cues correlate with different document types from the training set. Finally, we extend our study of token cues with a case study in language model safety, finding that different cues elicit distinct refusal and compliance behaviors that correspond to different types of training data.","published":"2026-10-05T17:59:52Z","updated":"2026-10-05T17:59:52Z","authors":["Sophie L. Wang","Amil Dravid","Rulin Shao","Kevin Farhat","Sewon Min","Alexei A. Efros"],"categories":["cs.LG","cs.AI","cs.CL"],"headline":"Training Data Tokens Influence AI Reasoning Abilities","summary":"Researchers studied how specific starting words in AI model responses affect the model's reasoning on tasks like math and coding. They showed that fixing certain words can greatly improve performance, matching results from more complex training methods. The team traced these effects back to patterns in the training data and even altered random words to become effective reasoning triggers. They also found that different starting words influence how AI responds to safety prompts, linking behavior to training data types. These findings are based on experiments and data analyses by the researchers."},{"id":"2610.06850v1","title":"InterMimicGen: Scaling Humanoid Loco-Manipulation through Self-Evolving Motion Imitation","abstract":"Captured human-object interactions provide rich supervision for humanoid loco-manipulation, but they are sparse, heterogeneous, and not directly executable by robots. We introduce InterMimicGen, a self-evolving motion-imitation framework in which robot motion data and a tracking policy improve each other. First, we consolidate motion-captured human-object interaction datasets and retarget them into humanoid robot references while preserving whole-body coordination and dexterous hand-object relationships. This produces a large and diverse humanoid robot reference collection for dexterous whole-body loco-manipulation. Second, we train a physics-based generalist tracker that executes these references in simulation on a humanoid with dexterous hands, covering a scale and diversity beyond prior humanoid tracking systems for loco-manipulation. Third, we close a data flywheel: each round makes small, task-preserving changes to where an interaction takes place and how the body performs it, fine-tunes the tracker on them, and keeps only the variants whose simulated execution completes the task, which seed the next round. With more iterations, these small edits compound into broader coverage around the sparse original demonstrations while preserving task semantics and motion quality. Experiments show contact-preserving retargeting across robot configurations, broad tracking with a single generalist policy, executable motions that keep growing over augmentation rounds, and transfer to real robots. InterMimicGen provides a unified path from heterogeneous human demonstrations to a continually expanding motion resource for humanoid robot learning.","published":"2026-10-05T17:59:50Z","updated":"2026-10-05T17:59:50Z","authors":["Yucheng Zhang","Sirui Xu","Jinhong Li","Liuyu Bian","Anatulya Nandi","Derek Zhang","Xiangchen Liu","Xueting Li","Umar Iqbal","Yu-Xiong Wang","Liang-Yan Gui"],"categories":["cs.RO","cs.CV","cs.GR"],"headline":"New method helps robots copy complex human movements better","summary":"Researchers developed InterMimicGen, a system that improves how humanoid robots imitate human motions involving objects. They combined diverse human movement data, adapted it for robots, and used a feedback process where the robot's performance refines itself through repeated simulations. This approach creates more varied and executable robot motions over time, helping robots learn complex movements more effectively. The findings are based on computer simulations and tests transferring motions to real robots, showing progress toward better robot motion imitation."},{"id":"2610.06849v1","title":"Boundary-accessible work in a collision-charged topological quantum battery","abstract":"Quantum batteries store energy in nonequilibrium quantum states. The ergotropy sets the work extractable by an arbitrary global unitary, but a device usually exchanges energy through a single local port. It remains unclear how much stored work such a port can deliver, and whether topological edge modes help. We study a dimerized spin chain of Su-Schrieffer-Heeger (SSH) type. Excited two-level ancillas charge it at one end by repeated collisions, and an initially empty two-level load discharges it at the same site. An exact free-fermion mapping yields a bound on the load work for all number-conserving quadratic extraction protocols. The load gains extractable work only if a single-particle mode of the battery holds more than half an excitation. Resonant charging fills the topological edge mode above this threshold. A single boundary pulse then delivers 98.6% of the maximal load work allowed by the bound for representative parameters. Under the same six-collision controls, the trivial-phase load remains passive. Within the protocol that selects the load pulse for maximum delivered work, cycle power peaks at a finite collision number while delivered work continues to increase over the scanned range. Over the tested range, the advantage survives hopping disorder, which preserves chiral symmetry. On-site disorder breaks this symmetry and leaves a finite fraction of loads passive. Topology thus supplies a localized access channel, while spectral matching and symmetry decide whether stored energy becomes usable local work.","published":"2026-10-05T17:59:49Z","updated":"2026-10-05T17:59:49Z","authors":["Osman Orçun Keskin","Özgür E. Müstecaplıoğlu"],"categories":["quant-ph"],"headline":"Topological quantum battery improves local energy delivery","summary":"The researchers studied how much useful work a single local connection can extract from a quantum battery, focusing on a special kind called a topological quantum battery. They modeled energy charging and discharging using quantum spins arranged in a chain and found that when a topological edge mode is properly charged, a local pulse can deliver nearly all the possible work. Their theoretical results show this advantage depends on certain quantum symmetries and remains robust against some types of disorder, offering insight into how topology affects usable energy extraction in quantum systems."},{"id":"2610.06848v1","title":"TranScope: What the Software Hides About LLM Training Data, the Hardware Reveals at Scale, and Accelerators Magnify","abstract":"Membership is the root privacy primitive in machine learning: to date, no hardware-based out-of-distribution detection on black-box models has been demonstrated against constant-time, static neural networks with masked confidence. This paper performs the first cycle-level examination of how large language models and vision transformers interact with various modern microarchitecture components, including integrated accelerators, as LLMs scale in size and answers the question of whether the data that a model was trained on affects its execution footprint even without any input-dependent branch, dynamic optimization, or early exit and in constant-time models. The results confirm that the answer is yes and identify which modern hardware components, such as TLBs or on-core accelerators, reveal or amplify that effect. The results also answer whether the signal is informative enough to reliably classify the in-/vs/out-of-distribution property of membership. To understand why, we perform a systematic root cause analysis and find that the transformer's tokenization steps, which happen during training, alter the locality of the accesses the model makes to fetch the vocabulary token later during inference and, as a result, change the page table access patterns and TLB in a previously unknown data-dependent way, causing microarchitectural state to vary significantly based on whether or not the input was in the distribution of the transformer training data. Building on the above observation, we introduce TranScope: the first microarchitecture tool for detecting membership information with low cost, no need for a surrogate model, and significantly higher robustness, e.g., 0.6 AUC for PETAL (best previously reported) vs 0.9 AUC (ours). This reintroduces hardware as both an opportunity, e.g., a tool for checking copyright violation for the first time, and a new channel for inferring membership (MIA).","published":"2026-10-05T17:59:42Z","updated":"2026-10-05T17:59:42Z","authors":["Joshua Kalyanapu","Darsh Asher","Kaushal Mhapsekar","Bita Aslrousta","Rushiraj Chaitanyakumar Sheth","Maharshi Mukeshkumar Oza","Achyuta Kannan","Samira Mirbagher Ajorpaz"],"categories":["cs.CR"],"headline":"Hardware reveals if data was used to train AI models","summary":"Researchers studied how large language models interact with computer hardware during use and found that a model's training data affects its hardware activity patterns, even in constant-time models designed to hide information. They discovered that specific hardware parts, like memory components, show changes depending on whether input matches training data. Based on this, they developed TranScope, a tool that can detect if data was in a model's training set more reliably than before, using hardware signals without needing extra models. Results come from detailed hardware analysis and simulations."},{"id":"2610.06847v1","title":"S2PD: Serial-to-Parallel Diffusion for Physically and Logically Consistent Video Generation","abstract":"Bidirectional video diffusion models denoise entire videos in parallel, yet when trained on effectively unlimited in-distribution data from procedural generators, continue to violate physical laws and simple symbolic rules. We introduce Serial-to-Parallel Diffusion (S2PD), which performs autoregressive diffusion at high noise before switching to parallel diffusion at low noise. The autoregressive phase provides the serial computation needed to coordinate interdependent events and produce valid state transitions while the parallel phase jointly refines the entire video and reduces sampling time relative to fully serial generation. We implement S2PD with two architectures: a pixel-space diffusion transformer trained from scratch and a pretrained video model adapted through LoRA fine-tuning with causal attention. Across games, physical simulations, and real video, S2PD follows rules more reliably than matched bidirectional baselines and generates videos with greater temporal stability and sampling efficiency than other serial methods.","published":"2026-10-05T17:59:35Z","updated":"2026-10-05T17:59:35Z","authors":["Jeffrey Hu","Daniel Olmeda Reino","Ayush Tewari"],"categories":["cs.CV"],"headline":"New Method Improves Video Generation for Realistic and Stable Results","summary":"Researchers developed a video generation approach called Serial-to-Parallel Diffusion (S2PD) that combines step-by-step and all-at-once techniques. This method helps the model better follow physical and logical rules and produce smoother videos. Their tests on games, simulations, and real videos show S2PD creates more accurate and stable videos than existing methods, though the findings rely on comparisons within their experiments and do not guarantee all real-world scenarios."},{"id":"2610.06846v1","title":"BiasFlow: Geometric Monitoring and Backbone Regularization for Spurious Feature Reliance","abstract":"Worst-group accuracy (WGA) evaluates a trained predictor but does not characterize how its frozen backbone behaves when a new head is learned. We introduce BiasFlow, a hook-based toolkit for monitoring class-attribute centroid alignment (IBMI), within-class centroid separation (W-IBMI), and feature-projection sensitivity. IBMI is confounded by class-attribute correlation and is not a measure of causal feature reliance. We pair these diagnostics with BiasFlow Regularization (BFR), a supervised, composable class-conditional centroid-alignment penalty. W-IBMI verifies the quantity BFR optimizes; it is scale dependent and does not independently establish attribute removal. Across the reported small-scale benchmarks, adding BFR improves or preserves mean WGA, with gains up to +26.0 pp on UrbanCars. The principal independent stress test freezes CelebA-Std backbones and trains fresh heads on biased data: BFR+GroupDRO improves WGA from 40.7% to 64.1%, while Male probe accuracy decreases from 92.5% to 72.2%. Attribute information remains recoverable, and cross-task results are mixed. A controlled synthetic-watermark ImageNet experiment additionally improves watermark-shift accuracy by +23.0 pp under matched training. These results support evaluating centroid geometry and resistance to biased head retraining alongside WGA, within the tested protocols.","published":"2026-10-05T17:59:28Z","updated":"2026-10-05T17:59:28Z","authors":["Haojin Deng","Zhiping Lin","Yimin Yang"],"categories":["cs.AI"],"headline":"New Toolkit Monitors and Reduces Bias in AI Feature Use","summary":"Researchers developed BiasFlow, a tool that tracks how AI models use features connected to class labels and biases. They introduced a regularization method (BFR) to reduce reliance on spurious features, improving worst-group accuracy in tests. Their findings suggest evaluating feature geometry and retraining resistance offers insights beyond traditional accuracy, but attribute removal and full bias elimination remain challenging. The results are based on specific benchmarks and controlled experiments, indicating potential but limited certainty."},{"id":"2610.06845v1","title":"Rewire and reset are all you need: Minimal resources for manipulating non-Abelions","abstract":"Braiding and fusing anyons are fundamental to topological quantum computation, yet existing circuit protocols often employ ancillas and measurements. We determine the minimal circuit resources for performing such anyon operations in quantum double models of generic finite groups. Our approach braids anyons through a sequential unitary circuit without ancillas. Intuitively, this is obtained by deforming the underlying graph on which the anyons live. More generally, we introduce graph rewiring, a diagrammatic calculus that assigns a geometric interpretation to individual two-qudit gates and tracks their action on stabilizers. For fusion, we prove a general no-go theorem excluding purely unitary protocols for fusing unknown anyon types without introducing ancillas. Remarkably, we show such \"input-agnostic\" fusion is enabled by simply using a qudit reset. Hence, contrary to folklore, ancilla resources are not needed to braid or fuse non-Abelian anyons (non-Abelions). This work has implications in experimental efforts in error correction: our fusion protocol highlights reset as an underutilized primitive for error correction without the spacetime-overhead of ancillas and measurements.","published":"2026-10-05T17:59:17Z","updated":"2026-10-05T17:59:17Z","authors":["Chiu Fan Bowen Lo","Rohith Sajith","Anasuya Lyons","Ashvin Vishwanath","Ruben Verresen"],"categories":["quant-ph","cond-mat.str-el"],"headline":"Simpler methods proposed to manipulate exotic quantum particles","summary":"The researchers studied how to control special particles called non-Abelian anyons, important for quantum computing. They developed a method to move (braid) these particles using only basic operations without extra helper systems (ancillas), by changing the connections in their underlying structure. They also found that to combine (fuse) unknown anyons, you generally need more than just unitary operations, but can do it by simply resetting a quantum bit. Their work suggests simpler approaches could improve future quantum error correction, though this is a theoretical result based on models."},{"id":"2610.06844v1","title":"Learning to Read the Contextual Tokens in Diffusion Transformers","abstract":"Multimodal Diffusion Transformers (MM-DiTs) jointly process visual and textual representations throughout generation. These models repeatedly update the text tokens through multimodal attention, forming dynamic contextual tokens whose function is not well understood. In this work, we introduce a framework for reading this contextual space through natural-language interrogation. We train a lightweight bottleneck network that maps intermediate contextual tokens into the input space of a frozen Large Language Model (LLM), allowing the LLM to answer questions about the emerging image directly from these hidden representations. Our reader reveals that contextual tokens encode a rich, global representation of the emerging scene: generation-specific semantics, including attributes left underspecified by the prompt, are accessible surprisingly early in denoising, while increasingly fine-grained details become readable over time. Remarkably, this information remains decodable even when the MM-DiT receives an empty prompt, showing that contextual tokens accumulate substantial image-specific information from the evolving visual representation itself. We further find that generations with more readable contextual representations tend to receive higher human-preference scores. Building on these observations, we introduce Contextual Alignment, a training technique that explicitly reinforces the visual-semantic information encoded in the contextual tokens, improving generation quality and distributional coverage. Together, our results establish contextual tokens as both an interpretable view into the internal dynamics of MM-DiTs and an effective target for improving generative models.","published":"2026-10-05T17:59:17Z","updated":"2026-10-05T17:59:17Z","authors":["Omer Dahary","Etai Sella","Hadar Averbuch-Elor","Daniel Cohen-Or","Or Patashnik"],"categories":["cs.CV","cs.AI","cs.GR","cs.LG"],"headline":"Understanding Visual and Textual Tokens in AI Image Generation","summary":"Researchers studied a type of AI model that creates images based on both text and visuals, focusing on special tokens that change as the image emerges. They developed a method to translate these internal tokens into questions a language model can answer, revealing that these tokens contain rich information about the scene early on and grow more detailed over time. Their findings, from analyzing model behavior, suggest these tokens can help improve image generation quality, though the results are based on interpreting the model’s internal states rather than direct real-world tests."},{"id":"2610.03719v1","title":"Mirror symmetry for gCICY threefolds (I): a $[3,-1]$ block and a non-Gorenstein toric phase","abstract":"We construct and analyze a mirror family for the generalized complete intersection Calabi-Yau threefold $X_+=[\\mathbb{P}^1\\,|\\,3\\ {-1};\\ \\mathbb{P}^4\\,|\\,2\\ 3]$, with $(h^{1,1},h^{2,1})=(2,46)$. An auxiliary-variable presentation turns the negative-degree section into a variation of toric GIT. Crossing one wall flops 16 disjoint $(-1,-1)$-curves and yields a Calabi-Yau threefold $X_-$, a nef complete intersection in a toric $\\mathbb{Q}$-Fano eightfold, which carries a genus-one fibration of index 4. Because this ambient space is not Gorenstein, the Batyrev-Borisov construction does not apply. We use the Hori-Vafa equations only to select a 2-parameter Laurent family, and construct a smooth projective crepant compactification $Y$ of the Laurent model with $(h^{1,1},h^{2,1})=(46,2)$; for very general parameters the Mordell-Weil group is $\\mathbb{Z}/4\\mathbb{Z}$. The periods of $Y$ satisfy a rank-6 Picard-Fuchs system with two maximally unipotent boundary points, whose indicial algebras are isomorphic to the rational even cohomology rings of $X_-$ and $X_+$. The resulting genus-zero predictions agree with independent counts of 176 vertical lines and 100 vertical conics on the genus-one fibration $X_-\\to\\mathbb{P}^2$.","published":"2026-10-02T17:59:29Z","updated":"2026-10-02T17:59:29Z","authors":["Atsushi Kanazawa"],"categories":["math.AG","hep-th","math.SG"],"headline":"Researchers explore complex shapes called Calabi-Yau threefolds and their mirrors","summary":"The researchers studied a specific type of geometric object known as a generalized complete intersection Calabi-Yau threefold, focusing on one called X_+. They developed and analyzed its mirror shape, called Y, using advanced mathematical tools including equations and geometric transformations. Their work deals with spaces that are not covered by standard methods, and they found certain mathematical properties that match predictions from counting particular curves on these shapes. Their results come from theoretical constructions and mathematical analysis, not physical experiments."},{"id":"2610.03718v1","title":"Dark matter bound-states made easy","abstract":"A light force carrier can mediate long-range interactions between dark matter particles and facilitate the formation of bound states. Models incorporating these features have attracted considerable attention in recent years, with significant advances in both the theoretical understanding of such bound states and their impact on the parameter space consistent with the observed dark matter relic abundance. However, the relevant results and their consistent implementation in Boltzmann equations remain technically intricate, requiring careful treatment or the use of specialized tools. This complexity has hindered broader adoption of these findings and restricted their use to a small number of expert groups. To address this challenge, we propose a simple prescription based on the analytic limits of the relevant rates, which allows standard freeze-out computations to be refined through a straightforward semi-analytical prescription that captures the main features of the full calculation. We illustrate this approach for a $U(1)$ dark matter model with a light force carrier, including Sommerfeld enhancement and bound-state effects.","published":"2026-10-02T17:59:19Z","updated":"2026-10-02T17:59:19Z","authors":["Simone Biondini","Stefan Vogl"],"categories":["hep-ph"],"headline":"Simplifying Dark Matter Bound-State Calculations","summary":"The researchers focus on how light force carriers can cause dark matter particles to form bound states, which affects theories about dark matter. They note that existing methods to include these effects in calculations are complicated and limited to experts. To make this easier, they propose a simpler, semi-analytical method that approximates the complex details well. They demonstrate their method using a specific dark matter model, showing how it includes key effects like enhanced interactions and bound states."},{"id":"2610.03717v1","title":"Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis","abstract":"This paper examines the role of Novel View Synthesis (NVS) in geometric representation learning. In principle, NVS should reason about 3D scene structure, thereby enabling transferable multi-view geometric representations. Yet, existing encoder-based NVS methods yield poor representations. This is not because of a lack of supervisory signal, but rather due to inconspicuous architectural choices: \\textit{spatially expressive decoders} that dilute representational capabilities of the scene encoder, and \\textit{low-level pixel-space targets} that hinder feature learning. We present SNAP, a self-supervised encoder-decoder transformer that addresses both through a pose-conditioned local decoder and a latent-space reconstruction objective. SNAP is task agnostic, and we show that it is competitive with special-purpose geometry-supervised methods. SNAP also performs competitively against self-supervised representations across five tasks: visual localization, pose estimation, point correspondence, depth estimation, and robot manipulation. Remarkably, SNAP's patch features exhibit emergent viewpoint invariance that approaches heavily supervised models despite lower compute and data budgets. Under camera shifts where standard 2D representations collapse, SNAP degrades more gracefully, revealing that restricting decoder expressivity actively prevents the suppression of transferable geometric structure. https://snap-nvs.github.io","published":"2026-10-02T17:59:14Z","updated":"2026-10-02T17:59:14Z","authors":["Keerthi Kaashyap","Dennis Anthony","Akshay Krishnan","Nhi Ngoc Nguyen","Jeremy Collins","James Hays","Shreyas Kousik","Animesh Garg"],"categories":["cs.CV","cs.AI","cs.RO"],"headline":"New method improves 3D scene understanding using less complex decoders","summary":"Researchers studied how computers can learn 3D scene structures from images, a process called Novel View Synthesis (NVS). They found that using simpler decoders and focusing on compressed data rather than raw pixels helps preserve important 3D information in the learned representations. Their new method, SNAP, showed competitive results on multiple visual tasks and maintained useful features even with limited data and computing power. These findings come from their experiments and highlight architectural choices affecting learning, not real-world testing yet."},{"id":"2610.03716v1","title":"MoSE3: Learning World-Space SE(3) at Every Pixel","abstract":"Dense 3D point tracking has been a prominent paradigm for modeling motion in dynamic scenes, but a point track is just a 3-DoF translation curve per pixel: it captures where pixels go, not the rotation of the underlying part, nor which pixels move together as one body. We propose MoSE3, the first feed-forward model that predicts dense SE(3) motion from monocular RGB video, producing full 6-DoF rigid transforms at every pixel in world space. Per-pixel SE(3) motion offers a richer view of how a scene moves: rotation, translation, and grouping all at once. Directly predicting SE(3) is challenging: rotations lie on a curved manifold that is ill-suited to Euclidean regression, and annotations for SE(3) are particularly difficult to acquire. To address these challenges, MoSE3 predicts per-pixel SE(3) through two jointly learned intermediates, 3D point tracks and rigidity embeddings, and recovers SE(3) by differentiably fitting transforms within each soft rigid cluster, enabling end-to-end prediction and supervision. To close the data gap, we introduce Art-Kubric, a large-scale synthetic dataset with dense SE(3) and rigidity labels for articulated objects with rich physical interactions. MoSE3 achieves state-of-the-art SE(3) estimation at pixel, part, and object levels on both rigid and articulated benchmarks, and state-of-the-art average 3D point tracking accuracy across three datasets, while showing strong generalization to real-world videos despite being trained solely on synthetic motion data.","published":"2026-10-02T17:58:52Z","updated":"2026-10-02T17:58:52Z","authors":["Jiahuan Cheng","Zhiyi Li","Tian Xia","Ruojin Cai","Yilun Du","Qianqian Wang"],"categories":["cs.CV"],"headline":"New Model Tracks Detailed 3D Movements from Video Pixels","summary":"Researchers developed MoSE3, a model that predicts full 3D movements—including rotation and translation—at every pixel in a video, offering a richer understanding of how scenes move. They overcame challenges by using intermediate steps to estimate motion and introduced a synthetic dataset to train their model. Their results show improved accuracy in tracking motion on various tests, with good performance on real videos despite training only on simulated data."},{"id":"2610.03715v1","title":"4DCodeBench: Benchmarking Agents on Inverse Graphics of Dynamic Scenes","abstract":"We introduce 4DCodeBench, a benchmark for 4D inverse graphics through code generation, in which agents reconstruct dynamic scenes from video as executable graphics programs. To accomplish this, agents must translate visual observations into compact representations of scene structure and dynamics, by implementing abstractions such as physical simulations to reproduce complex behavior. To evaluate this capability, we curate a set of real-world videos and construct synthetic scenes spanning diverse physical phenomena, including deformation, fluid flow, and fracture. We perform extensive benchmarking of frontier models, finding that strong static reconstruction capabilities do not yet translate into reliable reconstruction of complex dynamics. 4DCodeBench provides a testbed for tracking progress toward agents that can interpret the dynamics of the world through code. Our benchmark is available at https://github.com/4DCodeBench/4DCodeBench","published":"2026-10-02T17:58:49Z","updated":"2026-10-02T17:58:49Z","authors":["Ruihong Shen","Žiga Kovačič","Peter Kulits","Xingrui Wang","Zizhang Li","Joshua B. Tenenbaum","Alan Yuille","Jieneng Chen","Jiajun Wu"],"categories":["cs.CV","cs.AI","cs.GR"],"headline":"Testing AI's Ability to Understand and Recreate Moving Scenes","summary":"Researchers created 4DCodeBench, a test that measures how well AI agents can turn videos of moving scenes into computer programs that recreate those scenes. The challenge involves capturing both the structure and movement in videos, including complex behaviors like fluid flow and fractures. Their tests show that while AI handles static images well, it struggles with accurately representing dynamic, changing scenes. This benchmark helps track progress in teaching AI to understand the changing world through code."},{"id":"2610.03714v1","title":"SpiderCSS: Scalable Fault-Tolerant CSS State Preparation","abstract":"Fault-tolerant preparation of logical states is a critical primitive for the realization of large-scale fault-tolerant quantum computing. This paper introduces SpiderCSS, a scalable compilation pipeline that generates highly-optimized, fault-tolerant state preparation circuits for arbitrary Calderbank-Shor-Steane (CSS) codes. Starting from an idealized specification of a CSS state, we use the ZX-calculus and fault-equivalent rewrites to construct a diagram that prepares the given state while preserving the fault tolerance of the initial ideal ZX-diagram. From this diagram, SpiderCSS constructs a fault-tolerant CSS-state preparation circuit using a modular construction that is CNOT-minimal both within each component and also in the way these components are connected together. By only using fault-equivalent rewrites, the final circuit is provably fault-tolerant without needing expensive verification. Monte Carlo simulations across a variety of CSS codes up to distance 15 show that SpiderCSS scales efficiently and constructs circuits with substantially reduced depth and qubit footprint compared to existing scalable and heuristic methods, while simultaneously achieving an average decrease of 33.7% in logical error rate and an increase of 9.5% in acceptance rate.","published":"2026-10-02T17:58:44Z","updated":"2026-10-02T17:58:44Z","authors":["Boldizsár Poór","Andrey Boris Khesin","Sarah Meng Li","Benjamin Rodatz","John van de Wetering","Richie Yeung"],"categories":["quant-ph"],"headline":"New Method Improves Quantum Computing Error Correction Circuits","summary":"Researchers developed SpiderCSS, a new way to create circuits that prepare error-correcting states for quantum computers. Using advanced mathematical tools, SpiderCSS builds circuits that are both fault-tolerant and more efficient, minimizing errors without costly checks. Simulations showed it reduces errors and resources compared to previous methods, though these findings are based on theoretical models and computer simulations rather than experiments."},{"id":"2610.03713v1","title":"What Should World Models Forget? Stratified Retention for Continual Adaptation","abstract":"Continual learning treats degradation on previously seen data as evidence of failure, a convention inherited from settings with a stationary prediction target, where a correct label remains correct indefinitely. World models do not satisfy this condition. Their prediction target is the environment, which changes, so knowledge that was accurate when acquired may later become false, and discarding it is required behavior rather than a defect. Non-stationary ground truth is well studied in the concept drift literature and in the temporal factuality of language models, but has not been formulated for world models, which are distinctive in that they also encode knowledge that must never be revised. We argue that continual world models require retention stratified by invariance timescale, separating invariants such as physics and object permanence, which must never be revised, from instance-level facts that should be revised as soon as the environment changes. Standard forgetting metrics cannot distinguish a world model that has correctly revised outdated knowledge from one that has suffered catastrophic forgetting, and consequently rank a frozen model highest, while existing physical-reasoning benchmarks evaluate only frozen checkpoints. We propose differential retention, which reports invariant regression testing across the adaptation stream jointly with revision latency, without aggregation.","published":"2026-10-02T17:58:14Z","updated":"2026-10-02T17:58:14Z","authors":["Nishit Anand","Ramani Duraiswami","Dinesh Manocha"],"categories":["cs.LG","cs.AI","cs.CV","eess.IV","eess.SP"],"headline":"Improving World Models by Knowing What to Forget and Keep","summary":"The researchers studied how world models—AI systems that predict changing environments—should handle old knowledge. Unlike models with fixed facts, world models face changing truths, so forgetting some outdated info is necessary. They suggest separating knowledge into permanent facts, like physics, and changeable facts that need updates. They also propose new ways to measure if models properly revise information instead of simply forgetting. Their work highlights challenges in evaluating world models and offers methods for continual learning in changing settings."},{"id":"2610.03712v1","title":"RNADyn: A Benchmark for Generating and Understanding RNA Dynamics","abstract":"Ribonucleic acid (RNA) functions through conformational changes that are not fully captured by static structures. However, large-scale standardized RNA dynamics data remain limited, and existing approaches typically treat trajectory generation and dynamics understanding as separate objectives. Here, we introduce RNADynBench, a standardized RNA molecular dynamics (MD) benchmark with 2585 quality-controlled 100-ns all-atom trajectories and leakage-controlled splits. Building on RNADynBench, we develop RNADynNet, a unified model for RNA dynamics learning that uses a shared backbone for both trajectory generation and dynamics fingerprint extraction from a single conformer. It combines coordinate denoising, single-frame-to-trajectory alignment, and physical grounding to connect all-atom trajectory generation with dynamics representation learning. Physical grounding improves both generated dynamics and the physical information recoverable from these fingerprints. Across both test sets, including the high-flexibility challenge set, the generated trajectories achieve RMSF correlations of 0.875 and 0.766, while single-conformer predictions show comparable agreement with MD-derived dynamics. RNADynBench and RNADynNet together establish a benchmark and unified modeling framework for generating and understanding RNA dynamics.","published":"2026-10-02T17:58:08Z","updated":"2026-10-02T17:58:08Z","authors":["Yiming Huang","Lennart Bastian","Hanqun Cao","Luis Vollmers","Tolga Birdal"],"categories":["cs.LG","physics.bio-ph","q-bio.BM"],"headline":"New Benchmark and Model Help Study RNA Movements","summary":"Researchers highlight that RNA works by changing shape, but current data and methods don't fully capture these movements. They created RNADynBench, a large, standardized collection of RNA movement data, and RNADynNet, a model that both generates RNA movement patterns and extracts dynamic features from a single RNA shape. Their approach links movement generation with understanding dynamics, showing promising agreement with detailed simulations. The work provides tools for studying RNA dynamics, based on simulations and statistical results, without testing on real-world experiments yet."},{"id":"2610.03711v1","title":"How to Build Pseudorandom Unitaries in Microcrypt","abstract":"We provide the first evidence relative to a classical oracle that pseudorandom unitaries (PRUs) can exist without quantum-computable one-way functions. To obtain this result, we first prove that five independent random diagonal phase layers interleaved with Hadamard transforms \\[ U=F_5HF_4HF_3HF_2HF_1 \\] form a strong PRU with security under adaptive, controlled access to the unitary, its inverse, transpose, and complex conjugate. Our main result is that, when the phase layers are implemented using a classical random oracle O, the construction remains secure against uniform BQP adversaries with coherent access to suitable Boolean completions of classical-input $QMA^{PH}^{O}}$ decision problems. Such an adversary can invert (even quantum-computable) one-way functions and solve efficiently verifiable one-way puzzles defined relative to O. Our result therefore provides classical-oracle evidence that PRUs do not imply quantum-computable one-way functions, in the form of a black-box separation. We substantially strengthen the classical-oracle result of Kretschmer, Qian, and Tal (STOC 2025) by obtaining PRUs secure against these QMA-aided adversaries. Our construction also consists of a simple, efficient circuit that queries a random function, suggesting a route to concrete implementation via the random-oracle heuristic.","published":"2026-10-02T17:58:06Z","updated":"2026-10-02T17:58:06Z","authors":["Aditya Gulati","Dakshita Khurana","Kabir Tomer"],"categories":["quant-ph"],"headline":"Researchers Show New Way to Build Complex Quantum Operations","summary":"The researchers explored a type of quantum process called pseudorandom unitaries (PRUs), which are important for quantum cryptography. They demonstrated, using a theoretical model called a classical oracle, that PRUs can exist even without certain hard-to-compute functions previously thought necessary. Their work involves a specific pattern of quantum transformations and shows security against advanced quantum attacks in this theoretical setting. These findings are based on mathematical results and simulations, not practical experiments, and suggest new directions for future quantum security research."},{"id":"2610.03710v1","title":"EyeRobot 2.0: Active Gaze for Precise Manipulation without Wrist Cameras","abstract":"Inspired by human vision, we introduce a framework using active gaze to enable fine-grained bimanual manipulation with only a single stereo camera. EyeRobot 2.0 physically attends to a 3D fixation point in the scene by swiveling two eye viewpoints to center their gaze on it. The resulting images are processed foveally by allocating more visual tokens to the image centers, focusing computation on task-relevant features. Such Active Visual Fixation (AVF) requires carefully coordinated gaze during task execution, which we accomplish hierarchically by first training a low-level gaze servoing policy conditioned on a goal object, then training a target selector which emits fixation goals based on task progress. Both modules are trained with RL on real-world data: the first is trained with a dense geometric reward and the second co-trains with the BC gripper policy which allows it to discover fixation sequences that can resemble a human's fixation sequence while performing the task. EyeRobot 2.0 further takes advantage of fixation by canonicalizing gripper information into a fixation-relative SE(3) frame, which compacts the size of the action distribution to learn. We collect teleoperation data for 7 real-world and 6 simulated tasks, and conduct over 1000 physical and 1800 simulated robot trials comparing EyeRobot 2.0 against passive stereo and ego + wrist camera policies trained on the same data. Removing wrist cameras is costly for standard policies: with only passive stereo, real-world success drops from 52% to 27%. EyeRobot 2.0 closes this gap with only stereo, outperforming passive stereo by 40% in real and 20% in sim. It matches ego + wrist policies when their wrist views are clear (69% vs. 64%), and more than doubles their success when grasped objects occlude the wrist cameras (48% vs. 22%)","published":"2026-10-02T17:57:51Z","updated":"2026-10-02T17:57:51Z","authors":["Kush Hari","Justin Kerr","Nidhya Shivakumar","Samarth Mahapatra","Carmelo Sferrazza","Jiahui Lei","Jitendra Malik","C. Karen Liu","Ken Goldberg","Angjoo Kanazawa"],"categories":["cs.RO","cs.AI"],"headline":"Robot Uses Active Eye Movements for Precise Tasks Without Wrist Cameras","summary":"Researchers developed EyeRobot 2.0, a robot system that mimics human eye movements to focus its visual attention on task-relevant points using a single stereo camera. By actively moving its 'gaze' and concentrating processing on important image areas, it improves fine manipulation without needing wrist-mounted cameras. Experiments with real and simulated tasks show that EyeRobot 2.0 performs better than traditional passive stereo camera methods and matches or exceeds the success of robots using wrist cameras, especially when the wrist view is blocked."},{"id":"2610.02210v1","title":"Moore, Escher, Penrose: A Conformal Golden Braid","abstract":"I don't think I have ever done anything as peculiar in my life. Among other things, it shows a young man looking with interest at a print on the wall of an exhibition that features himself. How can this be? Perhaps I am not far removed from Einstein's curved universe.'' So wrote M.C. Escher about his 1956 lithograph Print Gallery. Nearly half a century later, a mathematical analysis related its geometry to an untwisted source image through a conformal power map $z \\mapsto z^α$, $α\\in \\mathbb{C}$. Building on this construction, we use a frozen text-to-image diffusion model to generate new self-referential scenes. Prompting alone does not enforce the recursion, while a post-hoc transformation can leave structures poorly connected. Applying the transformation during sampling is also insufficient: the denoiser may \"repair\" the intended distortion or drift out of the prescribed geometry. We construct a generalized inverse $T^\\dagger$ of the non-invertible image transformation $T$, adapted to its recursive constraint. In the idealized formulation, the Penrose identity $TT^\\dagger T = T$ makes $TT^\\dagger$ an idempotent projection onto geometrically admissible images. Yet denoising only the transformed image remains an out-of-distribution task, even with projection. We therefore braid denoising steps with $T$ and $T^\\dagger$: source-space steps develop the untwisted scene, while transformed-space steps refine its appearance and connections in the final geometry. We generate Print Gallery-like compositions and explore further transformations. Rather than distorting a finished image, we let the scene and its distortion develop together.","published":"2026-10-01T17:59:59Z","updated":"2026-10-01T17:59:59Z","authors":["Sophia Feldman","Assaf Shocher"],"categories":["cs.CV"],"headline":"Creating self-referential images inspired by Escher's Print Gallery","summary":"Researchers studied the unusual geometry in M.C. Escher's 1956 artwork Print Gallery, where a picture contains a smaller version of itself in a twisted loop. They used mathematical transformations and AI-based image generation to create new scenes with this self-referential, twisted structure. The team developed a method combining transformations and denoising steps to maintain the complex geometry during image creation. Their work is theoretical and based on simulations, exploring how to generate images that evolve together with their geometric distortions."},{"id":"2610.02209v1","title":"Leading gravitational dressing of operators and states in de Sitter space","abstract":"We describe construction of gravitationally dressed observables in de Sitter (dS) space, to leading non-trivial order in Newton's constant. We show that for an underlying field theory observable to be \"gravitationally dressable,\" it must be dS invariant. If it satisfies this condition, its gravitational dressing can be constructed in terms of certain Green functions, by solving the condition that the dressed observable commute with the constraints. We provide a construction of the Green functions relevant for dressing observables lying on a time-symmetric slice. A simple example of a dS invariant observable is an operator creating a two-particle state of scalars. In the large mass limit, the particles are shown to be antipodally located, and the dressing of this operator (or corresponding state) can be explicitly constructed. Its behavior is diagnosed by certain correlators involving this state, and we check explicitly that these correlators then reproduce a linearized version of the Schwarzschild-dS solution. The underlying dS-invariant observables may be thought of as relational, and we argue can be generalized to include systems with properties of observers, which then could be likewise gravitationally dressed. We also briefly describe how such dressed observables can be related to alpha vacua, and discuss other questions regarding a more complete construction and role of such operators.","published":"2026-10-01T17:59:58Z","updated":"2026-10-01T17:59:58Z","authors":["Steven B. Giddings","Zi-Yue Wang"],"categories":["hep-th"],"headline":"How to Add Gravity Effects to Observables in Expanding Space","summary":"Researchers studied how to modify basic measurements or operators in a model of expanding universe called de Sitter space to include gravity's influence, working to the first significant level in gravity's strength. They found these measurements must have a special symmetry, and they developed a method using mathematical functions to create gravitationally adjusted versions that fit the model's rules. Their example with two particles showed how the adjusted measurement matches known gravity solutions, but the work is theoretical and based on simplified assumptions."},{"id":"2610.02208v1","title":"Sphere Encoder 2","abstract":"Sphere Encoder is an autoencoder that generates images by decoding random points from a high-dimensional latent sphere. We identify two limitations of the original formulation that reduce its generation quality. First, random points concentrate near the equator relative to the pole on an encoded latent, but the training rotation never reaches this region, leaving a gap that limits one-step generation. Second, training for generation with pixel-wise reconstruction loss encourages the decoder to average over plausible images, producing blurry images that lack high-frequency details. We present Sphere Encoder 2 to address both limitations, substantially improving image generation quality while maintaining the speed and simplicity of a autoencoder. Models are released at \\href{https://github.com/kaiyuyue/sphere2}{github.com/kaiyuyue/sphere2}.","published":"2026-10-01T17:59:58Z","updated":"2026-10-01T17:59:58Z","authors":["Kaiyu Yue","Sean McLeish","Ruchit Rawal","Brian Bartoldson","Menglin Jia","Tom Goldstein"],"categories":["cs.CV"],"headline":"Improving Image Generation with Sphere Encoder 2","summary":"Researchers studied Sphere Encoder, a tool that creates images from points on a high-dimensional sphere but found two main problems limiting image quality. The original method missed certain sphere regions during training and used a technique that produced blurry images lacking sharp details. They developed Sphere Encoder 2 to fix these issues, improving image quality while keeping the process fast and simple. Their findings are based on testing this improved version and the models they created are publicly available."},{"id":"2610.02207v1","title":"One Basis to Animate Them All: Gaussian Blendshape Distillation for Real-Time Avatars","abstract":"3D Gaussian avatars support fast rendering, however, their real-time animation is often challenged by the costly neural inference. We address this bottleneck and show that the animation of pretrained avatar models can be closely approximated by a linear combination of identity-independent blendshapes. Building on this finding, we introduce GALA (Gaussian Animation via Linear Approximation), a distillation method that replaces per-frame heavy neural decoding with a shallow coefficient predictor and a linear blend. To improve fidelity and reduce memory requirements, we propose to construct the basis using block-local PCA under a rendering-aware metric and a memory budget. Our method learns a shallow MLP network to predict blendshape coefficients and applies to various animation architectures without retraining original models. We validate GALA by accelerating the inference of three distinct avatar models for 3D animation of facial expressions and full-bodies with clothing dynamics. Across these models, our distillation generalizes to held-out identities and reduces CPU animation cost by up to three orders of magnitude while preserving most of the rendering quality. Excellent results of our method confirm the shared linear structure of learned avatar representations and enable highly efficient and accurate animation at frame rates reaching up to 60fps on mobile devices. Project page: https://ramazan793.github.io/gala/","published":"2026-10-01T17:59:58Z","updated":"2026-10-01T17:59:58Z","authors":["Ramazan Fazylov","Stamatis Lefkimmiatis","Ivan Laptev"],"categories":["cs.CV","cs.AI","cs.HC","cs.LG"],"headline":"New Method Speeds Up Real-Time 3D Avatar Animation","summary":"Researchers developed GALA, a technique to make 3D avatar animation much faster by simplifying complex neural processing into a mix of basic shapes. Their approach predicts how these shapes blend to create expressions and movements without needing to retrain existing models. Tests show GALA significantly cuts computing time—by up to 1000 times—while keeping most visual quality, allowing smooth animations even on mobile devices. This work supports the idea that many avatars share common underlying structures, but results are based on specific models and simulations."},{"id":"2610.02206v1","title":"KaliBench: A Fine-Grained Benchmark for Cybersecurity Tool Use on Kali Linux with Runtime-Free Verifiable Rewards","abstract":"LLMs are increasingly applied to cybersecurity workflows, where they are expected to translate analysts' intent into tool invocations. However, existing evaluations focus on knowledge-based assessments or end-to-end agentic tasks, and do not directly measure LLMs' ability to generate executable commands for real-world cybersecurity tools. This gap is critical because cybersecurity operations rely on strict command-line interfaces (CLIs), where minor syntax errors, incorrect flag--value bindings, or argument misordering can invalidate execution. We introduce KaliBench, a fine-grained benchmark and dataset for natural-language--to--CLI translation on Kali Linux, comprising 8,504 query--command pairs spanning 1,642 tools across 23 capability dimensions and 5 security phases. KaliBench is constructed via a manuscript-grounded pipeline with deterministic canonicalization and alias-aware evaluation, enabling precise and reproducible assessment of tool selection and argument construction. To ensure both semantic correctness and practical executability, we develop a multi-stage verification pipeline that combines LLM-based validation, sandboxed terminal execution, and human-in-the-loop refinement. Building on these fine-grained, deterministic signals, KaliBench further enables runtime-free verifiable rewards for training. Across three evaluation modes and 24 configurations of general-purpose and security-focused open-weight models, no open-weight model exceeds 42% exact-command accuracy in the unrestricted setting, highlighting the difficulty of accurate CLI-based cybersecurity tool use without explicit tool hints. We further show that supervised fine-tuning and reinforcement learning with verifiable rewards derived from KaliBench significantly improve an 8B model and achieve performance comparable to a 685B MoE model.","published":"2026-10-01T17:59:55Z","updated":"2026-10-01T17:59:55Z","authors":["Pengfei Li","Naufal Suryanto","Sicheng Zhang","Muzammal Naseer"],"categories":["cs.CL","cs.AI","cs.CR"],"headline":"New Test Measures AI Command Skills for Cybersecurity Tools","summary":"Researchers created KaliBench, a detailed test to see how well AI models can turn natural language into exact commands for cybersecurity tools on Kali Linux. They built a method to check that commands are both correct in meaning and actually work when run. Testing various AI models, they found none were very accurate without hints, but training with KaliBench data improved performance. The findings highlight challenges in generating precise tool commands and show the benchmark helps guide better AI training."},{"id":"2610.02205v1","title":"ROWBench: Do Video Models Render What the Program Specifies?","abstract":"Programmable world models separate executable dynamics from visual generation, offering a promising foundation for next-generation game engines. However, their visual adherence to explicit rules and interactions remains insufficiently evaluated. Existing benchmarks assess visual quality, controllability, and instruction or physical adherence, but rarely test fidelity to fine-grained, program-specified world events. We introduce PROWBench, comprising 170 programmatically constructed episodes and 600 proxy videos covering diverse scenes and interactions. PROWBench logs entity states and timestamped events, including those outside the camera's field of view, as replayable world records, from which it renders synchronized views and proxy representations. This enables generated videos to be checked against the observable consequences of program execution. An extensible framework constructs scenes, controls behaviors, and can render each camera view in different representations, such as coarse 3D, and bounding boxes. The benchmark covers first- and third-person perspectives, with synchronized multi-view observations available for a subset of episodes. Grounded in these records, PROWBench evaluates entity control, long-horizon memory, and, with two VLM-based metrics, Logic-Render Alignment and Interaction Success Rate, adherence to the prescribed timeline and the visual realization of timestamped engine-recorded events.","published":"2026-10-01T17:59:53Z","updated":"2026-10-01T17:59:53Z","authors":["Zheng-Hui Huang","Guixu Lin","Yu-Ju Tsai","Jian-Kai Zhu","Fengbo Lan","Yu-Lun Liu","Yung-Yu Chuang","Kaipeng Zhang","Zhixiang Wang"],"categories":["cs.CV"],"headline":"New Test Measures Accuracy of Video Models in Following Game Rules","summary":"Researchers developed PROWBench, a new benchmark with 170 program-created video scenes to check if video models correctly follow detailed rules and events specified by a program. By comparing generated videos with recorded world states and timed events, PROWBench tests whether models visually represent what the program dictates, including actions outside the camera view. The researchers use this to evaluate models' control, memory, and alignment between logical events and visual output, but the results focus on simulation and model evaluation rather than real-world applications."},{"id":"2610.02204v1","title":"Reconstruct, Practice, Go Real: Guided Self-Improvement for Embodied Agents","abstract":"Building reliable robot capabilities across diverse tasks requires substantial human effort to develop and maintain skills, design rewards, and integrate perception with control. We present Reconstruct, Practice, Go Real (RPG), a framework for autonomous improvement of robot execution systems without updating model weights. RPG identifies manipulation capabilities in an offline dataset and constructs related practice tasks in simulation. During practice, RPG uses execution feedback, privileged simulator state, and available dataset videos to diagnose failures. It develops new reusable symbolic skills, refines existing skills, and revises the system prompt based on these diagnoses. Cross-task evaluation tests individual candidate changes and merged revisions before they are retained for reuse. At test time, a multimodal LLM uses the resulting system prompt and skill library to coordinate perception and robot control. On held-out initializations of 22 manipulation tasks, RPG improves task success from 28.6% after the first practice round to 95.0% after 15 rounds, outperforming all evaluated baselines, including ASPIRE (75.5%) and CaP-Agent0 powered by GPT-6 Astra Pro (60.0%). After a common calibration and hardware-adaptation procedure, the frozen system succeeds in all 30 physical trials, with ten trials on each of three tasks. Project Website: https://rpg-robot.github.io/","published":"2026-10-01T17:59:50Z","updated":"2026-10-01T17:59:50Z","authors":["Yen-Jen Wang","Haozhe Jiang","Shuying Deng","Haoru Xue","Weirui Ye","Rocky Duan","Nika Haghtalab","S. Shankar Sastry","Pieter Abbeel","Haozhi Qi"],"categories":["cs.RO","cs.AI","eess.SY"],"headline":"New Method Helps Robots Learn Tasks More Reliably on Their Own","summary":"Researchers developed a system called Reconstruct, Practice, Go Real (RPG) that allows robots to improve their ability to perform tasks without changing their core software. RPG helps robots practice tasks in a simulated environment, learn from mistakes, and build new skills automatically. Tests showed that robots using RPG improved their success rate from about 29% to 95% over time and performed well in real-world trials. These findings are based on simulation and controlled experiments reported by the researchers."},{"id":"2610.02203v1","title":"Embedding Prediction Helps Image Generation","abstract":"In diffusion transformers, a class label or a text prompt is embedded once, and the same condition is reused at every denoising step. We ask whether predicted embeddings can serve as this condition instead. Next-Embedding Predictive Autoregression (NEPA) trains a Transformer to predict the next continuous embedding in a sequence. In generation, the clean image follows the noisy image, so its embeddings are the next embeddings after the condition and the noisy image. We train a NEPA model to predict them all at once with Multi-Embedding Prediction, and in Embedding Conditioned Generation, a DiT generator is conditioned on these predictions, recomputed at every denoising step, so the conditioning signal adapts to the current noisy state. Experiments on class-conditional ImageNet $256\\times256$ study the condition of the generator, the design of Multi-Embedding Prediction, and the scaling of both models. The NEPA model adds a second network to every sampling step; with it, and combined with REPA, our final model, NEPA-DiT-XL, reaches an FID of 1.32 using about a third of the training compute of REPA.","published":"2026-10-01T17:59:49Z","updated":"2026-10-01T17:59:49Z","authors":["Sihan Xu","Ji Xie","Zilin Wang","Hui Shen","Stella X. Yu"],"categories":["cs.CV","cs.LG"],"headline":"New Method Improves Image Generation by Predicting Embeddings","summary":"Researchers explored using predicted embeddings as changing guides during image generation, instead of fixed inputs. They developed a method called NEPA that trains a model to anticipate the next embedding in a sequence and uses these predictions to better condition the image generator at each step. Their experiments on ImageNet images show that this approach, combined with another method, achieves strong results while using less training computation. The findings are based on simulations and model performance tests."},{"id":"2610.02202v1","title":"ScholarCatalyst: A Benchmark for Retrieving Papers That Inspire New Research","abstract":"What makes great scientists great? Even as AI systems start to make progress on open problems, scientists remain far ahead of them at sensing which prior idea, buried in an ever-growing archive of research, a new problem needs. To study this skill, we draw on researchers who know firsthand which earlier work advanced their completed projects, with papers serving as pointers to the ideas within. Using our automated pipeline that makes author annotation scalable, we build ScholarCatalyst by having 184 lead authors of 207 recent computer science papers label which candidates did or could have advanced their project, each with a detailed rationale. We introduce a retrieval task with author-provided judgments: given an initial research question, retrieve these papers from only the literature available when the project began. Agentic search does no better than embedding retrieval (0.42 vs. 0.48 Recall@20) despite calling that same retriever as a tool. Even an agent built on Claude Fable 5.1, which may have seen the completed papers during training, reaches only 0.51 R@20. These results highlight the need for new training recipes that equip models with expert intuition for searching broad corpora. We envision ScholarCatalyst as a step toward scientific agents that can take a half-formed idea and point to the prior research it needs.","published":"2026-10-01T17:59:47Z","updated":"2026-10-01T17:59:47Z","authors":["Sohyeon Kim","Yoonho Lee","Bo Liu","Dayoon Ko","Rulin Shao","Seungone Kim","Graham Neubig","Pang Wei Koh","Aakanksha Chowdhery","Akari Asai","Omar Khattab","Yejin Choi","Gunhee Kim","Chelsea Finn"],"categories":["cs.AI","cs.CL","cs.IR"],"headline":"New Benchmark Helps Study How Scientists Find Inspiring Papers","summary":"Researchers created ScholarCatalyst, a dataset where 184 scientists identified past papers that helped their recent work. Using this, they tested computer programs to see how well they can find important prior research from the literature available when the projects started. The programs performed moderately but did not match expert judgment, showing current AI struggles to replicate scientists’ intuition for useful papers. The study suggests that better training methods are needed for AI to assist in scientific discovery."},{"id":"2610.02201v1","title":"SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation","abstract":"High-resolution 3D generation increasingly relies on voxel latents and multi-stage pipelines that first predict active structure and then synthesize local geometry. While effective, this design fragments continuous surfaces into many local tokens, inflates generation cost, and often weakens topological consistency for thin or highly connected shapes. We introduce SILSA, a topology-aware 3D generation framework that represents shapes with compact sliding-window slice latents. Instead of generating expensive voxel tokens, SILSA uses a fixed set of overlapping slices along the three canonical axes, where each token summarizes a local depth window to preserve cross-sectional continuity and support single-stage rectified-flow generation. A Slice VAE encodes oriented surface samples into multi-axis slice latents and reconstructs them with a sparse volumetric decoder, while a Volumetric Anchor Lattice coordinates directional slice streams through a shared 3D workspace. To preserve structural correctness, we introduce slice-level topology supervision that matches persistence diagrams and aligns Betti transitions across neighboring slices. Experiments show that SILSA improves structural fidelity while substantially reducing generation cost. SILSA improves PSNR by $8.7\\%$, coverage by $5.96$ absolute points, and Betti error by $9.2\\%$ over the strongest baseline, while using $70.0\\%$ fewer tokens than the next-most compact baseline and over $98\\%$ fewer tokens than sparse or hierarchical tokenizers, effectively reducing training memory by $40.4\\%$ and inference time by $58.5\\%$. Qualitative results further show improved preservation of thin structures, repeated components, and long-range connectivity.","published":"2026-10-01T17:59:46Z","updated":"2026-10-01T17:59:46Z","authors":["Tianjiao Yu","Xinzhuo Li","Yifan Shen","Ying Shen","Kiet A. Nguyen","Adheesh Sunil Juvekar","Ismini Lourentzou"],"categories":["cs.CV","cs.AI","cs.LG"],"headline":"New Method Improves 3D Shape Generation While Cutting Costs","summary":"Researchers developed SILSA, a new way to create detailed 3D shapes that keeps surfaces connected and reduces computing effort. Unlike older methods that break shapes into many pieces, SILSA uses overlapping slices to maintain continuous structure and supports a single-step generation process. Tests show SILSA creates more accurate shapes with fewer parts, improves structural consistency, and speeds up training and use, though these findings are based on the experiments described by the researchers in their paper."},{"id":"2609.40363v1","title":"Probing the Reionization History and Bubble Sizes with JWST Lyman-$α$ Fraction Measurements","abstract":"Recent James Webb Space Telescope (JWST) measurements quantify the fraction of ultraviolet-selected galaxies showing detectable Lyman-alpha (Ly-$α$) lines at $z \\sim 6-14$. These Ly-$α$ fractions are partly shaped by Ly-$α$ scattering from neutral hydrogen in the surrounding intergalactic medium (IGM), and so the observed Ly-$α$ fraction evolution can inform our understanding of the reionization history and the size distribution of ionized bubbles during reionization. We compare the JWST measurements with Ly-$α$ transmission calculations based on both the THESAN reionization simulations and a flexible semi-analytic model. In each case, we calibrate a sub-grid model for the intrinsic Ly-$α$ emission lines emerging from the interstellar medium (ISM) and circumgalactic medium (CGM) of the host galaxies using empirical estimates of the Ly-$α$ equivalent width (EW) distributions at $z \\sim 6$. We find that the reionization history and bubble sizes in THESAN yield good agreement with current JWST Ly-$α$ fraction observations, with a reduced $χ^2_ν= 0.4-1.2$. Our semi-analytic models and THESAN Ly-$α$ transmission calculations agree well, allowing us to explore a wide range of reionization models. We derive $1σ$ bounds on the average mass-weighted neutral fractions of: $\\langle x_{\\mathrm{HI}} \\rangle$ = $0.38^{+0.14}_{-0.38}$ ($0.52^{+0.09}_{-0.21}$), $0.57^{+0.23}_{-0.57}$ ($0.68^{+0.12}_{-0.39}$), and $0.93^{+0.05}_{-0.17}$ ($0.89^{+0.11}_{-0.15}$) at $z \\sim$ 7, 8, and 11, based on the Ly-$α$ fraction measurements with EW $>25\\,\\mathring{\\rm A}$ ($>10\\,\\mathring{\\rm A}$).","published":"2026-09-30T17:59:59Z","updated":"2026-09-30T17:59:59Z","authors":["Aritra Kundu","Adam Lidz","Guochao Sun"],"categories":["astro-ph.GA","astro-ph.CO"],"headline":"Studying Early Universe Hydrogen Using JWST Galaxy Observations","summary":"Researchers used new data from the James Webb Space Telescope to study how much neutral hydrogen filled the early universe between about 7 and 11 billion years after the Big Bang. They compared the fraction of distant galaxies showing a specific light signature, called Lyman-alpha emissions, with computer simulations of how hydrogen affected this light. Their findings provide estimates of the hydrogen's neutral fraction and suggest the size of ionized bubbles during that era, but results are based on models and observations with inherent uncertainties."},{"id":"2609.40362v1","title":"Multimodal Flow: Unified Flow Modeling of Language and Vision in Embedding Spaces","abstract":"We present Multimodal Flow, a fully continuous generative model of language and vision. Most unified multimodal models either model both language and quantized images as discrete tokens or combine discrete language prediction with continuous image generation. The former introduces a visual quantization bottleneck. The latter requires modality-dependent objectives and sampling procedures. Fully continuous modeling avoids these trade-offs and enables a shared generative process, but remains underexplored for multimodal pretraining. Multimodal Flow introduces a unified continuous architecture that integrates multimodal continuous representations with a shared chunk-causal flow backbone. It organizes text blocks and images as ordered continuous hyperchunks, preserving textual token order and visual spatial structure. The backbone learns a single vector field over these hyperchunks through Flow Matching. Joint attention enables cross-modal interaction, while modality-specific feed-forward networks process each modality. The model predicts multiple target chunks in parallel during training and generates hyperchunks sequentially at inference. We instantiate MF-1 and pretrain it on multimodal data. Across 0.6B, 1.2B, and 1.6B scales, continued pretraining consistently improves multimodal modeling. With only 150B pretraining tokens, MF-1 achieves an average score of 82.8 across GenEval and DPG-Bench and 75.3 across VQAv2, MMBench, and POPE, remaining competitive with unified models trained on substantially more data. Under matched data, optimization, and parameter budgets, Multimodal Flow further outperforms representative hybrid and discrete models. These results establish continuous chunk-based embedding flow modeling as a new fully continuous paradigm for unified multimodal modeling. The related code and model are publicly released at https://github.com/hustvl/Multimodal-Flow.","published":"2026-09-30T17:59:57Z","updated":"2026-09-30T17:59:57Z","authors":["Hongyuan Tao","Xinggang Wang","Lianghui Zhu","Yongkang Li","Yunchao Wei","Bin Feng","Shaoyu Chen","Qian Zhang","Chang Huang","Kai Yu"],"categories":["cs.CV"],"headline":"New Model Combines Language and Vision Without Data Loss","summary":"Researchers developed Multimodal Flow, a new type of AI model that processes both language and images as continuous data rather than discrete pieces. This approach avoids common problems like losing image detail or needing different methods for each type of data. Their model, MF-1, showed strong performance on various tests using less training data than other models, suggesting this continuous method could be a promising way to unify language and vision understanding in AI. Results are based on pretraining and testing with simulated data."},{"id":"2609.40361v1","title":"Ranking-Aware Prompt Optimization for Multimodal Clinical Diagnosis","abstract":"Multimodal large language models (MLLMs) are rapidly advancing clinical diagnosis, yet their adaptation pipelines remain anchored to accuracy-based objectives. Clinical data are heavily class-imbalanced: a constant-majority predictor can score above 90% accuracy while being clinically useless. We therefore evaluate and optimize for AUROC, a threshold-free score that ranks positives above negatives and is invariant to class balance. We focus on prompt optimization in MLLMs. Reflective methods such as GEPA use a binary scores matrix with one row per evaluation instance and one column per candidate prompt; cells record per-instance correctness, so the column average is accuracy and drives candidate selection. We introduce pair-level Pareto prompt evolution (Ranking-PE), which replaces each correctness row with a pairwise-ordering row over (positive, negative) instance pairs: the cell is 1 if the candidate scores the positive higher than the paired negative. The column average then equals empirical AUROC (by the Wilcoxon-Mann-Whitney identity). We apply this swap at all three layers the prompt evolution search reads from - the scores matrix that decides Pareto dominance, the per-example feedback to the reflection LM, and final candidate selection - at no extra model calls and with no surrogate loss. Across three diseases on MIMIC, accuracy-based prompt evolution can degrade ranking; Ranking-PE reverses this, beating the accuracy-based recipe by +5.8 AUROC pp on fine-tuned Qwen3-VL-8B and +16.2 pp on MedGemma-4B. Ablations examine each design component and show that a medical-grade visual backbone - via vision-encoder-tuned SFT or medical pretraining - is a prerequisite that prompt search cannot replace - our recipe extends reflective prompt evolution from text-only data to multimodal clinical decision-making.","published":"2026-09-30T17:59:56Z","updated":"2026-09-30T17:59:56Z","authors":["Tian Xia","Minghao Liu","Yiqing Liang","Laixi Shi","Jiayun Wang"],"categories":["cs.LG","cs.CL","cs.CV"],"headline":"Improving Clinical AI by Focusing on Better Ranking, Not Just Accuracy","summary":"The researchers identify that accuracy alone can be misleading in clinical diagnosis AI, especially when some conditions are much rarer than others. They developed a new way to optimize prompts for multimodal AI models that focuses on ranking positive cases higher than negatives, measured by AUROC, a score less affected by class imbalance. Testing on clinical data, this new method improved model ranking performance significantly, but the researchers note high-quality medical imaging components remain essential and cannot be replaced by prompt tuning alone."},{"id":"2609.40360v1","title":"Semifactual Credit-Augmented Policy Optimization","abstract":"Reinforcement learning with verifiable rewards (RLVR) has improved the reasoning capabilities of large language models (LLMs), yet their predictions remain sensitive to task-irrelevant prompt features. We investigate this sensitivity through semifactual prompt interventions that preserve the underlying problem and its answer. Our analysis reveals substantial variation in token-level sensitivity and shows that suppressing high-drift token candidates during decoding improves reasoning accuracy without updating model weights. These findings highlight a limitation of Group Relative Policy Optimization (GRPO), which assigns the same outcome-derived advantage to every response token and may reinforce potential spurious dependence alongside useful reasoning. Motivated by this observation, we introduce Semifactual Credit-Augmented Policy Optimization (SCAPO), a causally inspired variant of GRPO that incorporates semifactual stability into token-level credit assignment. SCAPO measures token probability drift for fixed responses under semifactual interventions and uses normalized stability scores to reduce advantages for relatively unstable tokens during early training, while granting no additional credit for stability alone. On Qwen3-4B-Base and Qwen3-1.7B-Base, SCAPO improves AIME 2024-2026 accuracy over GRPO by 5.63 and 4.17 percentage points, respectively. At both model scales, SCAPO achieves the best results on most evaluated mathematics benchmarks and all evaluated out-of-distribution benchmarks among the compared methods. These results suggest that semifactual stability provides an effective training signal for improving reasoning and generalization through finer-grained credit assignment in RLVR. The code is available at https://github.com/DtYXs/SCAPO.","published":"2026-09-30T17:59:56Z","updated":"2026-09-30T17:59:56Z","authors":["Junshu Pan","Zhizhang Fu","Shulin Huang","Yiran Ding","Zifan Cheng","Wenqi Shao","Qiaosheng Zhang","Yue Zhang"],"categories":["cs.LG","cs.AI","cs.CL"],"headline":"New Method Improves Language Model Reasoning by Focusing on Stable Words","summary":"Researchers studied how large language models sometimes rely on unimportant parts of a question when reasoning. They found that some words cause unstable predictions and that limiting these during response generation can improve accuracy. Based on this, they developed a new training technique called SCAPO that gives less credit to unstable words, improving reasoning and generalization in tests. Their findings come from testing with math benchmarks and suggest a way to better assign credit in model training, though the results are limited to the models and tasks studied."},{"id":"2609.40359v1","title":"Removing Timing Shortcuts Improves Non-Invasive Brain-to-Text","abstract":"We find that major reported improvements in decoding words from non-invasive brain recordings are largely reproducible without any brain data. In the influential work of d'Ascoli et al. (2025), time series of brain activity from subjects perceiving continuous speech are segmented into fixed-length windows starting at each word. A neural network then generates predictions for all of the words in a sentence together. Neighbouring windows partially overlap, implicitly revealing the interval between words. Since these intervals indicate the duration of the words spoken, and different words tend to have different durations - for example, \"the\" is much shorter than \"supercalifragilisticexpialidocious\" - the neural network can improve its predictions of words without relying on the underlying brain activity. Consistent with this, the method reaches 22.0% balanced accuracy on synthetic signals containing no brain information, compared with 22.3% on real brain recordings. To prevent the network from learning this shortcut, we make a single, simple change. Instead of jointly encoding all windows in a sentence, we process each independently. As a result, the neural network achieves better performance by learning underlying word-specific information from brain recordings. This makes two existing strategies become much more effective than before. Both aggregating predictions from distinct neural responses to the same word and using a pretrained LLM as a linguistic prior now substantially improve results. On our perceived speech benchmark, this simple recipe (SimpleB2T) achieves a word error rate of 36.6% with five observations per word, approaching past invasive speech decoding performance, albeit under different conditions. The results in this work expose an important shortcut in brain-to-text decoding and show that removing it leads to a simple and considerably more effective strategy.","published":"2026-09-30T17:59:52Z","updated":"2026-09-30T17:59:52Z","authors":["Dulhan Jayalath","Oiwi Parker Jones"],"categories":["cs.LG","q-bio.NC"],"headline":"Fixing timing tricks improves brain-based word decoding","summary":"Researchers found that earlier impressive results decoding words from brain activity used hidden timing clues, not the brain signals themselves. They changed the method to analyze each word's brain data separately, forcing the system to rely on real brain information. This improved performance, making word prediction more accurate and showing that removing timing shortcuts leads to better brain-to-text decoding methods. These findings are based on tests with non-invasive brain recordings and simulations."},{"id":"2609.40358v1","title":"Physis-Lang: Self-Evolving Language as a Physical Representation for Video World Model","abstract":"Video world models are expected to predict how the physical world evolves, yet they often produce visually plausible videos that violate basic physical principles. Existing approaches commonly assume that natural language is insufficient to represent the physical knowledge required for reliable generation, and therefore introduce additional visual, latent, numerical, or planning-based signals. We revisit this assumption and introduce Physis-Lang, a self-evolving framework that treats physical language as a shared and optimizable representation across data curation, model training, and video generation. Physis-Lang represents physical processes through language that describes their relevant entities, causes, interactions, governing principles, temporal evolution, and effects. To improve this representation, we construct PhysCapBench, which decomposes physical processes into atomic assertions and evaluates captions using recall and precision. An agentic loop iteratively analyzes assertion-level errors and refines the instruction used to produce physical captions. Physis-Lang further converts model deficiencies into textual descriptions and uses language-guided retrieval to identify visually diverse videos that cover missing physical processes. Experiments on four widely used physical video benchmarks with Wan and Cosmos backbones demonstrate consistent improvements in physical plausibility. Notably, starting from open-source Cosmos3-Nano backbones, our Physis-Lang-enhanced models surpass the leading proprietary Veo 3.1 model.","published":"2026-09-30T17:59:51Z","updated":"2026-09-30T17:59:51Z","authors":["Liming Lu","Xianzheng Ma","Wenkun He","Guanqi Zhan","Yilin Zhao","Junyu Chen","Mengyao Xu","Jiaojiao Fan","Wenhang Ge","Yuchao Gu","Yunze Liu","Boyi Li","Zhen Dong","Victor Prisacariu","Ming-Yu Liu","Song Han","Han Cai"],"categories":["cs.CV"],"headline":"New Language-Based Method Improves Physical Accuracy in Video Models","summary":"Researchers studied how video models often create videos that look real but break physical rules. They developed Physis-Lang, a system that uses evolving language descriptions to better represent physical processes in videos. By breaking down and refining these language descriptions and using them to guide data selection and training, the researchers showed improvements in how well videos follow physical principles across several tests. Their approach outperformed previous top models in making videos that respect physics, based on simulation experiments."},{"id":"2609.40357v1","title":"Singularity of harmonic measures for hyperbolic lattices","abstract":"Let $Γ<\\mathrm{Isom}(\\mathbb H^n)$, $n\\geq2$, be a cocompact lattice containing a convex cocompact codimension-one subgroup in the sense of Sageev. We prove that every finitely supported admissible random walk on $Γ$ has hitting measure singular with respect to Lebesgue measure on $\\partial\\mathbb H^n$. The proof extends the method of Kosenko--Tiozzo, replacing Fourier analysis for a cyclic subgroup by a von Neumann dimension argument that also applies to nonabelian subgroups. As corollaries, we prove the singularity conjecture of Kaimanovich and Le Prince for cocompact hyperbolic lattices admitting a proper cocompact cubulation, cocompact Kleinian groups, and cocompact hyperbolic lattices with totally geodesic sublattices of codimension one. In particular, our result covers all cocompact hyperbolic reflection groups, cocompact arithmetic lattices of simplest type and cocompact nonarithmetic lattices arising from the hybrid and inbreeding constructions.","published":"2026-09-30T17:59:35Z","updated":"2026-09-30T17:59:35Z","authors":["Nikolay Bogachev"],"categories":["math.DS","math.GR","math.GT","math.PR"],"headline":"Research on random walks in hyperbolic spaces shows new singularity results","summary":"The researchers studied certain random walks within special groups acting on hyperbolic spaces, focusing on how likely these walks are to hit points on the boundary of the space. They proved that the probability distributions (“hitting measures”) for these walks are singular, meaning they differ in a fundamental way from standard geometric measures. Their work extends previous methods and confirms a conjecture for various complex groups, but their results rely on advanced mathematical arguments and apply within theoretical frameworks rather than practical scenarios."},{"id":"2609.40356v1","title":"ViTeX-Bench: Benchmarking High-Fidelity Video Scene Text Editing","abstract":"Recent video generation is increasingly realistic and controllable, yet video editing remains less developed, particularly for precise local edits that must preserve the original scene dynamics. Video scene text editing replaces text on scene surfaces, such as storefront signs, whiteboards, and product labels, while preserving the surrounding content, motion, and camera dynamics. Although scene text editing is well studied for images, video scene text editing that achieves high visual quality, temporal consistency, and edit locality remains underexplored. Existing resources offer limited paired real-video data, and general video-editing metrics do not directly measure whether the requested text remains correct over time. We introduce ViTeX-Bench, a benchmark suite comprising ViTeX-Dataset and a three-axis evaluation protocol. The dataset contains 387 real-world 720p videos with text-region masks and editing instructions: 230 provide reviewed, pipeline-generated paired edits for training, and 157 form a frozen evaluation split. The protocol evaluates text correctness, visual and temporal quality, and edit locality through 13 metrics, with one primary metric per axis and a Pareto comparison of their trade-offs. OCR calibration, human evaluation, and annotation-sensitivity analyses support the interpretation of these scores. Across eight baselines from four editing families, accurate text, temporal stability, and scene preservation remain difficult to achieve together. We also release ViTeX-Edit-14B, an open-source reference editor fine-tuned on the paired training split with motion-aligned glyph-video conditioning. It achieves CharAcc 0.688, the highest mean among the evaluated video-native editors, and the lowest comparable text-crop Warp among raw editor outputs. ViTeX-Bench provides a reproducible foundation for studying these trade-offs in video scene text editing.","published":"2026-09-30T17:59:33Z","updated":"2026-09-30T17:59:33Z","authors":["Xinghao Chen","Xiangbo Gao","Jiongze Yu","Yuheng Wu","Zhengzhong Tu"],"categories":["cs.CV","cs.AI"],"headline":"New Benchmark Tests Video Scene Text Editing Quality and Consistency","summary":"Researchers studied the challenge of editing text in videos, like changing signs or labels, while keeping the scene’s motion and look unchanged. They created ViTeX-Bench, including a dataset of real videos with text edits and a detailed evaluation method measuring text accuracy, visual quality, and how well edits blend over time. Testing eight existing methods showed it’s still hard to balance these goals. They also introduced a new editing tool that performed best in their tests, aiming to help future research in this area."},{"id":"2609.40355v1","title":"Quantum hypothesis testing of non-mixed-unitarity: A multifaceted hierarchy of quantum channel discrimination","abstract":"Given multiple uses of an unknown quantum channel, we determine whether it is a specified non-mixed-unitary channel or belongs to the set of mixed-unitary channels. Formulating this as a composite channel hypothesis testing problem, we characterize the Stein exponents achievable by parallel strategies under progressively weaker restrictions on probe states and auxiliary memory. In particular, we consider block-i.i.d. probes, which allow arbitrary correlations within blocks of fixed size while remaining i.i.d. across blocks, and derive finite-letter expressions of the corresponding Stein exponents. Varying the block size, which interpolates between fully i.i.d. and arbitrary probes, together with further restrictions on intra-block correlations and access to auxiliary memory, results in a multifaceted hierarchy. We establish several strict separations within this hierarchy. Without auxiliary memory, we prove that fully i.i.d. probes yield a vanishing Stein exponent for every unital non-mixed-unitary channel, while blocks of three product probes suffice to obtain a strictly positive exponent for the qutrit Werner-Holevo channel. For block size two, although product probes seem to be insufficient for O(3)-covariant channels, an entangled probe achieves a strictly positive exponent for the qutrit Werner-Holevo channel, even though every maximally entangled probe fails. In contrast, auxiliary memory makes fully i.i.d. probes sufficient by yielding a strictly positive exponent for every non-mixed-unitary channel, even without requiring input-reference entanglement, and this exponent admits a lower bound determined by the diamond-norm distance from the mixed-unitary set. Moreover, for odd-dimensional Werner-Holevo channels, we find that this exponent is infinite, with every full-Schmidt-rank pure state being optimal.","published":"2026-09-30T17:59:20Z","updated":"2026-09-30T17:59:20Z","authors":["Pratik Ghosal","Pritam Halder","Ayan Patra","Aditi Sen"],"categories":["quant-ph","math-ph"],"headline":"Researchers study identifying special quantum channels using advanced tests","summary":"The researchers explored how to tell if an unknown quantum channel is a specific non-mixed-unitary type or a mixed-unitary one by testing it multiple times. They developed a framework comparing different testing strategies that vary in complexity and resource use, finding that certain strategies perform better under specific conditions. Their findings are theoretical and based on mathematical characterizations rather than experiments, highlighting nuanced differences in the efficiency of various quantum testing methods."},{"id":"2609.40354v1","title":"Design of Mechanically Compliant Membrane Reflectors for Giant Optical Phase Nonlinearity","abstract":"Optical nonlinearities arise when the response of a photonic system depends on light intensity. Radiation pressure can create such a nonlinearity by moving a mechanically compliant reflector, thereby shifting the phase of the reflected light. Based on measured properties, we predict a giant phase responsivity of 263 rad W$^{-1}$ and a device-equivalent $n_{2,\\mathrm{eff}} = 5.4 \\times 10^{-7}$ m$^2$ W$^{-1}$ for a silicon nitride membrane trampoline with serpentine springs. We introduce strength-range-aperture metrics for reflective phase elements. Among reported mechanical resonators compared under a common direct-reflection protocol, this trampoline is an outlier, combining high compliance, practical optical accessibility, and 99.85% retention of its small-signal phase responsivity through a full $2π$ reflected-phase shift. Practical implementation requires optical and thermal co-design. Our framework indicates design priorities for future mechano-optical phase elements.","published":"2026-09-30T17:59:16Z","updated":"2026-09-30T17:59:16Z","authors":["Lior Michaeli","Harry A. Atwater"],"categories":["physics.optics","cond-mat.mes-hall","physics.app-ph"],"headline":"Designing Flexible Membranes for Strong Light Phase Changes","summary":"The researchers studied how a flexible membrane reflector can change the phase of reflected light depending on light intensity, a type of optical nonlinearity caused by radiation pressure moving the membrane. They predict a very large phase response from a silicon nitride membrane with special springs, based on measured properties. Their work, using theoretical predictions and comparisons, highlights design factors and challenges for making such devices practical, but actual application requires careful optical and thermal design."},{"id":"2609.38180v1","title":"Point2Part: Unified 3D Partitioning from Point Prompts","abstract":"Existing 3D part decomposition methods do not necessarily partition the original shape into non-overlapping parts that collectively cover the entire shape, allowing overlaps or gaps that hinder downstream part-level applications. We instead formulate part decomposition as a joint partitioning of the entire shape, where the predicted parts are non-overlapping and jointly recover the entire shape. Our key insight is that part decomposition should consider all desired parts jointly, rather than modeling each part independently. To this end, we develop a promptable model for 3D part decomposition from images or meshes. Users can specify desired parts through 3D point prompts for controllable decomposition. Given one point prompt per desired part, our model produces the corresponding parts as a complete partition of the entire shape. We build on a pretrained 3D generation model and first obtain a shape latent from either an input image or mesh. We then introduce a prompt encoder that maps each 3D point prompt to a part token while attending to the shape latent. To decode the desired parts, we propose a novel part decoder jointly scoring the entire shape against all part tokens in a coarse-to-fine manner, assigning every position within the shape volume to exactly one part. We perform part decomposition in this shared shape latent space, enabling a unified model for image-to-part generation, mesh-to-part generation, and part segmentation. Our method outperforms existing works on all part-quality metrics across all three tasks, and improves compatibility among parts by an order of magnitude over previous SOTA methods. Code and models will be released.","published":"2026-09-29T17:59:58Z","updated":"2026-09-29T17:59:58Z","authors":["Hao-Tang Tsui","Yu-Rou Tuan","Xiaoxuan Ma","Nicolas Ugrinovic","Takaaki Shiratori","Kris Kitani"],"categories":["cs.CV"],"headline":"New method divides 3D shapes into exact, non-overlapping parts","summary":"Researchers developed a technique that breaks down 3D shapes into separate parts without overlaps or gaps, improving accuracy for shape analysis. Their model uses user-provided point prompts on the shape and processes the entire object jointly, rather than treating parts separately. Tested on images and meshes, this approach better covers the whole shape and outperforms previous methods, though results are based on computational models rather than physical tests."},{"id":"2609.38179v1","title":"LYRA: The formation and growth of nuclear star clusters in dwarf galaxies","abstract":"Nuclear star clusters (NSCs) are dense massive star clusters ubiquitous in galactic centres across various mass regimes, present in up to $\\sim30$ per cent of dwarf galaxies with $M_\\star<10^{7}\\, \\mathrm{M}_\\odot$. Although their main growth pathways are well studied, there is still ongoing debate regarding their formation. We use the LYRA cosmological hydrodynamical simulations to study the formation and growth of NSCs in a sample of six dwarf galaxies, where we identify NSCs with masses $M_\\mathrm{NSC} \\sim 10^4 - 10^6 \\, \\mathrm{M}_\\odot$ in all galaxies in our sample. The NSCs in our sample emerge at early times as a compact component, with the host galaxy growing around them. This is consistent with a scenario in which NSCs are among the earliest structures to form in a galaxy. We characterise the origin of the stars that constitute the NSCs and explore the properties of the stellar populations in each of these channels. We find that most of the NSC mass originates from in-situ star formation within the NSCs themselves. While this might appear at odds with current observations, the bulk of this star formation occurs at early times, resulting in a significant population of old and metal-poor stars, in agreement with an early NSC formation. We also identify galaxy-galaxy mergers as a significant contributor to the NSC mass, with usually one merger dominating this contribution. Our results highlight the complexity in the growth history of NSCs in dwarf galaxies and highlight the importance of considering additional growth channels for NSCs in dwarfs galaxies.","published":"2026-09-29T17:59:57Z","updated":"2026-09-29T17:59:57Z","authors":["Joaquin Sureda","Azadeh Fattahi","Sownak Bose","Mariya Lyubenova","Jessica E. Doppel","Thales Gutcke","Katja Fahrion","Rüdiger Pakmor"],"categories":["astro-ph.GA"],"headline":"How dense star clusters form in dwarf galaxies' centers","summary":"Researchers used computer simulations to study dense star clusters called nuclear star clusters (NSCs) found in the centers of small galaxies. They found that NSCs form very early, with most of their stars created right inside these clusters long ago, making them old and metal-poor. The study also showed that merging with other galaxies adds to the NSC's growth, indicating that NSC formation in dwarf galaxies is complex and involves multiple processes. These findings come from simulations and help explain how NSCs develop over time."},{"id":"2609.38178v1","title":"Skill-Space Shooting for Autonomous Robot Policy Improvement","abstract":"Robots deployed in the physical world must be able to improve beyond their initial training as they encounter new situations and failures. For this improvement to scale across tasks, it must make effective use of experience without requiring human demonstration of each correction. Recent agentic systems offer a way to reduce this reliance on human effort by using foundation models to autonomously compose learned behaviors to complete tasks. Yet completing tasks this way does not itself teach a task policy to overcome its own failures; that requires turning these behaviors into learnable corrections for the policy. Our insight is that many such corrections are familiar short behaviors, or skills: they recur across tasks and describe actions that foundation models can reason about from a scene. We introduce skill-space shooting, which uses foundation model guidance to explore corrections through these reusable skills and turn successful trials into policy improvement. Real-world experiments show repeated improvement in policies acting autonomously, while skills can also be shared to reduce the teaching needed to improve on new tasks. By making reusable skills a source of corrective supervision, skill-space shooting enables scalable and generalizable policy improvement within and across tasks. Additional results and videos at https://skill-space-shooting.github.io.","published":"2026-09-29T17:59:55Z","updated":"2026-09-29T17:59:55Z","authors":["Zihang Rui","Renhao Wang","Haoxu Huang","Yang Gao"],"categories":["cs.RO","cs.AI","cs.LG"],"headline":"Helping Robots Learn from Mistakes Using Reusable Skills","summary":"This research addresses how robots can get better at their jobs after initial training, especially when facing new challenges or failures. The researchers developed a method called skill-space shooting, where robots use known short actions, or skills, guided by advanced models to find and apply fixes on their own without human help. Their experiments show this approach helps robots improve repeatedly and share skills to learn new tasks more easily, though results come from controlled tests and further real-world verification may be needed."},{"id":"2609.38177v1","title":"Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering","abstract":"Reasoning about the 3D world from multi-view images remains a fundamental challenge for Multimodal Large Language Models (MLLMs). While modern MLLMs handle single-image inputs effectively, they struggle to integrate evidence across viewpoints into a coherent 3D understanding. A growing body of work attempts to close this gap by injecting 3D awareness into MLLMs, either by boosting fine-grained pixel-level cross-view correspondence or by fusing features from 3D geometry foundation models, yet a substantial gap to human reasoning persists. In this work, we revisit human spatial reasoning, which suggests that rather than relying on fine-grained geometry cues, humans roughly identify common objects across views, infer the relative geometry between viewpoints, and assemble a coarse 3D layout of the scene. Inspired by this process, we introduce Imagine3D-LLM, an MLLM that learns to assemble a similar compact 3D representation of the scene and conditions its answer on this representation. Concretely, we append a small set of learnable summary tokens after the image tokens, decode them into a compact 3D Gaussian Splatting representation supervised by a photometric reconstruction loss, and train jointly with the standard next-token prediction objective. Notably, although only the summary tokens receive direct reconstruction supervision, this objective also induces stronger cross-frame correspondence within the LLM's underlying image features, suggesting that learning to reconstruct propagates 3D-aware signals throughout the model. As a result, Imagine3D-LLM consistently outperforms prior approaches across multiple spatial reasoning and 3D understanding benchmarks, suggesting that imagining the scene can be more effective than being told its pixel-wise geometry.","published":"2026-09-29T17:59:52Z","updated":"2026-09-29T17:59:52Z","authors":["Jaewoo Jung","Hyeonseo Yu","Honggyu An","Jisang Han","Mungyeom Kim","Minkyeong Jeon","Heeseong Shin","Wonjun Moon","Federico Tombari","Daniel Barath","Marc Pollefeys","Seungryong Kim","Sunghwan Hong"],"categories":["cs.CV","cs.CL"],"headline":"New AI Learns to Visualize 3D Scenes Like Humans Do","summary":"Researchers addressed how AI models struggle to understand 3D scenes from multiple images. They created Imagine3D-LLM, an AI that mimics human reasoning by recognizing objects across views and building a simple 3D layout before answering questions. This approach, trained using a method that encourages the AI to reconstruct scenes, helped it perform better on tests of spatial and 3D understanding, though the findings are based on their specific training methods and benchmarks."},{"id":"2609.38176v1","title":"Breakdown of Local Denoising as Semantic Speciation","abstract":"The dynamics of generative models exhibit two apparently distinct temporal windows: a speciation window, in which a sample commits to a semantic class, and a nonlocality window, in which local context windows become insufficient for generation. Motivated by evidence of their near-concurrence in a variety of frontier models, we investigate their relationship through the spatial distribution of semantic information. Under a \"common cause\" hypothesis, we prove that the nonlocality window must lie in the speciation window. This hypothesis postulates that semantic labels explain a fraction of the correlations between distant tokens, a condition that is natural for many real datasets. We further give conditions under which both windows shrink to a single limiting time as system size grows, defining a \"phase transition\", and verify this behavior analytically in Gaussian mixtures. Together, these results identify conditions under which semantic information explains the concurrence of speciation and nonlocality, connecting two complementary perspectives on the emergence of semantic structure in generative modeling.","published":"2026-09-29T17:59:50Z","updated":"2026-09-29T17:59:50Z","authors":["Guangkuo Liu","Mert Okyay","Yifan F. Zhang","Fangjun Hu","Rahul Nandkishore","Xun Gao"],"categories":["cs.LG","cond-mat.dis-nn","cond-mat.stat-mech"],"headline":"Study Links Semantic Grouping and Context Limits in AI Models","summary":"Researchers explored how generative AI models decide on meaning categories (speciation) and when local information stops being enough (nonlocality). They proposed that both happen together because meaning explains distant relationships in data. Through mathematical analysis and testing on simplified models, they showed conditions where these two phases merge into one transition. The findings suggest a connection between how AI models form meaning and handle context, but the results are theoretical and based on specific assumptions."},{"id":"2609.38175v1","title":"Proximal causal inference through cross-proxy balancing","abstract":"Proximal causal inference identifies causal effects in the presence of unmeasured confounding by drawing on two sets of proxy variables. Identification typically utilizes bridge functions, defined as solutions to integral equations. However, the mechanism by which fitted bridge functions correct for confounding bias remains opaque, offering little to interpret, inspect or stress-test. We show that the defining equation of a treatment bridge function is already a balance condition, with a cross-proxy form: the weights are functions of the treatment confounding proxies and covariates, and they balance the outcome confounding proxies and covariates (i.e., reweighting the distribution in a particular treatment arm to match the distribution across treatment arms). Several existing identification paths via a treatment bridge function can then be interpreted as providing conditions under which balance on the outcome proxies implies balance on the unobserved confounders, which we call balance propagation. Leveraging this framing, we show that estimation of the treatment bridge function is a type of balancing weight estimation. Finally, we show that an outcome-weighted estimator form can also be obtained for a large class of proximal estimators, including those that utilize an outcome bridge function. Using this framing, we provide conditions under which common estimators are numerically equivalent.","published":"2026-09-29T17:59:48Z","updated":"2026-09-29T17:59:48Z","authors":["Grace V. Ringlein","Trang Q. Nguyen","Elizabeth A. Stuart","Harsh Parikh"],"categories":["stat.ME"],"headline":"New method clarifies hidden bias correction in causal studies","summary":"Researchers studied how to identify true causal effects when some influencing factors are unmeasured, by using proxy variables (stand-ins). They found that a key step—called a treatment bridge function—acts by balancing these proxy groups across treatment options, which helps correct bias. Their work explains how this balancing works, links several existing methods, and shows when different estimators behave similarly. Their findings come from mathematical analysis of these functions and help make the correction process more understandable."},{"id":"2609.38174v1","title":"Imaging phase winding in topological superconductors with a fork-tip Josephson STM","abstract":"The complex phase winding in a topological superconductor remains challenging to access by conventional real-space probes. We propose a scheme to probe this winding with a fork-tip Josephson STM near an impurity, where two superconducting tips form an interferometer sensitive to phase differences between two positions. The signal remains accessible even if the relative phase between the fork-tip and sample fluctuates, establishing a route towards a local phase-sensitive STM probe of topological superconductivity. We numerically compute the interference pattern to demonstrate that the magnitude and phase of the order parameter can be reconstructed as a function of angle, providing a route towards resolving the longstanding open experimental problem of directly detecting phase winding in topological superconductors.","published":"2026-09-29T17:59:46Z","updated":"2026-09-29T17:59:46Z","authors":["Vladislav Poliakov","Archisman Panigrahi"],"categories":["cond-mat.supr-con","cond-mat.str-el"],"headline":"New STM method may reveal phase patterns in topological superconductors","summary":"Researchers address the difficulty of measuring phase winding—a complex pattern—in topological superconductors using standard methods. They propose a new microscopic technique involving a fork-shaped superconducting tip in a scanning tunneling microscope to detect phase differences near impurities. Their computer simulations suggest this approach could reconstruct the superconducting phase pattern, offering a possible way to observe an unresolved experimental challenge. These findings are theoretical and based on numerical modeling, not yet experimental proof."},{"id":"2609.38173v1","title":"In-context Robot Learning Made Simple: A Democratized Recipe for Manipulation Tasks","abstract":"We study robotic in-context learning (ICL), an emerging paradigm that enables robots to infer and execute tasks from visual demonstrations. Despite its growing promise, the problem itself remains under-defined: a visual demonstration simultaneously conveys action trajectories, object semantics, manipulation affordances, spatial relations, and task goals, making it unclear what information the robot is actually expected to follow. In this work, we first provide a clear problem definition of robot ICL that explicitly defines its learning target and resolves this fundamental prompt ambiguity. Building on this definition, we develop a minimalist and reproducible ICL framework (SimpleICL) with a visual prompt encoder and a low-cost data collection protocol. Without massive pre-training or specialized data infrastructure, our framework achieves strong performance in both simulation and real-world environments. Extensive experiments further reveal several key properties of robot ICL, including action, semantic, composition, and affordance discrimination. We will fully open-source our data and training pipeline to facilitate systematic and reproducible research on robot ICL. The project page can be found at https://simpleicl.github.io/simpleicl.","published":"2026-09-29T17:59:45Z","updated":"2026-09-29T17:59:45Z","authors":["Minxing Li","Minghao Han","Weizhi Zhao","Hanwen Wang","Xiangshuo Liu","Shuyao Shang","Jingxiang Zhou","Mingchao Sun","Hongyu Pan","Mu Xu","Yu Liu","Lue Fan","Zhaoxiang Zhang"],"categories":["cs.RO"],"headline":"Making Robot Learning from Visual Examples Simpler","summary":"Researchers studied how robots learn tasks by watching visual demonstrations, a complicated problem because demonstrations show many types of information at once. They clarified what exactly robots should learn and created a simple, easy-to-replicate method called SimpleICL that works well without needing huge amounts of data or complex training. Their experiments, done both in simulations and real settings, highlight important ways robots understand actions and object relationships. They plan to share all their data and code for others to use."},{"id":"2609.38172v1","title":"Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation","abstract":"Teaching humanoids loco-manipulation skills, such as carrying diverse objects, via visual imitation is a promising path toward generalist robots. However, collecting diverse, high-quality interaction videos, such as clips that clearly show a person's full body and unoccluded interactions with objects, poses a practical barrier to scaling this approach. We propose PRISM, a real-to-sim-to-real framework that overcomes this limitation by amplifying a handful of real videos into a large, diverse training set. PRISM first generates hundreds of diverse \"counterfactual\" human-object interaction videos via video-to-video (V2V) generation from a few exemplar real videos. Our contact-anchored real-to-sim pipeline then reconstructs both human and object motions, retargeting this imperfect video data into physically plausible trajectories. The intra-class variability across these counterfactual videos lets us train a single policy that generalizes to unseen objects within each category. We demonstrate the full pipeline by deploying this policy on a real robot without any real-world fine-tuning. Using only onboard depth observations, our humanoid picks up, carries, and drops objects, including boxes, barrels, bins, and balls, across novel instances, sizes, and initial configurations.","published":"2026-09-29T17:59:45Z","updated":"2026-09-29T17:59:45Z","authors":["Zihan Wang","Zhen Wu","Pieter Abbeel","Rocky Duan","Jitendra Malik","Carmelo Sferrazza","C. Karen Liu","Guanya Shi","Angjoo Kanazawa"],"categories":["cs.RO","cs.CV","cs.GR"],"headline":"Creating Diverse Robot Training Videos from Few Real Clips","summary":"The researchers address the challenge of teaching robots how to handle various objects by imitation, which usually requires many high-quality videos of human-object interactions. They developed a method called PRISM that generates hundreds of varied synthetic videos from a small number of real ones, transforming these into realistic motion data for training robots. Their method helped train a robot to pick up and carry different unseen objects without additional real-world training, though the approach is demonstrated through simulated video generation and controlled tests."},{"id":"2609.38171v1","title":"Classification of topological phases of matter in stochastic systems","abstract":"Topological phases of matter support edge states protected from noise and perturbations, and the classification of which symmetry and dimension support such phases was developed for quantum systems and analogous platforms. While topological phases have also been discovered in stochastic systems and posited as mechanisms for biochemical processes, a classification consistent with their Markovian constraints remains lacking, impeding generalization to new scenarios. These constraints alter the matrix space preventing the application of previous classification methods, while hosting new properties such as the necessity of non-Hermiticity for non-trivial topological phases. We introduce new methods including a new homotopy approach and redefinition of the point gap and find that only two symmetry classes remain robust: no symmetry and pseudo-Hermiticity. In these symmetry classes, we identify the dimensions with topologically non-trivial phases and their group structure, creating a rigorous framework for predicting robust behavior in active and living matter.","published":"2026-09-29T17:59:43Z","updated":"2026-09-29T17:59:43Z","authors":["Dexin Li","Evelyn Tang"],"categories":["cond-mat.stat-mech"],"headline":"New framework classifies topological phases in random systems","summary":"Researchers studied special states in stochastic, or random, systems that remain stable despite noise, similar to quantum systems but with different constraints. They developed new mathematical methods to classify which symmetries and dimensions allow such stable phases, focusing on systems that follow specific rules called Markovian constraints. Their findings identify only two key symmetry classes with robust phases and detail the dimensions where these occur, providing a foundation for understanding stability in complex active or biological materials. The results are theoretical and based on mathematical analysis."},{"id":"2609.35771v1","title":"From Density to Mass: A New Parametrized Framework for Dark Matter Environment Around Black Holes","abstract":"Astrophysical black holes (BHs) are not isolated vacuum objects but are expected to reside within complex environments, such as dark matter (DM) halos and baryonic distributions. Such environments can modify the spacetime geometry surrounding the BH and leave observable imprints on gravitational-wave (GW) signals, including shifts in the quasinormal mode (QNM) spectrum, tidal Love numbers (TLNs), and the inspiral dynamics of extreme-mass-ratio inspirals (EMRIs). We develop a general framework for describing DM-dressed BH spacetimes by parameterizing the geometry directly in terms of the enclosed mass function, rather than the traditional density-based profile. This mass-based formulation is naturally connected to the total gravitational field, accommodates several common halo profiles as limiting cases, and incorporates basic physical consistency requirements such as regularity, causality, and appropriate asymptotic behavior. Within this framework, we investigate the environmental imprints on QNMs, TLNs, and GW fluxes from EMRIs. We find a complementary sensitivity to the halo structure: QNMs predominantly probe the inner regions of the environment, whereas static TLNs are more sensitive to its outer structure. EMRI fluxes, in contrast, probe the matter distribution through its influence on the orbital motion and GW propagation. Our results establish the enclosed-mass formulation as a flexible and physically controlled framework for connecting astrophysical environment models with precision GW observables.","published":"2026-09-28T17:59:59Z","updated":"2026-09-28T17:59:59Z","authors":["Rajes Ghosh","Nicholas Speeney","Avijit Chowdhury","Chiranjeeb Singha","Miguel A. S. Pinto","Emanuele Berti"],"categories":["gr-qc","astro-ph.GA"],"headline":"New Method Links Dark Matter Around Black Holes to Gravitational Waves","summary":"Researchers developed a new way to describe how dark matter environments influence black holes by focusing on the total mass enclosed around the black hole, rather than just the density. This method aligns better with gravity's effects and physical rules. Using this framework, they studied how dark matter affects signals from gravitational waves, finding different measurement types highlight different parts of the surrounding matter. These findings are theoretical and help connect dark matter models with future gravitational wave observations."},{"id":"2609.35770v1","title":"FurE: Efficient Instance-Specific 3D Fur Reconstruction without Animal-Fur Datasets","abstract":"Realistic and editable animal fur reconstruction from multi-view images is challenging due to fine-scale detail, self-occlusion and obfuscation, and, unlike human hair, the lack of animal-fur datasets. Fur usually covers most of an animal's body, with large inter-species and intra-species variability. We present FurE, an efficient strand-based animal fur reconstruction method that recovers a per-strand, editable groom by optimizing a root-conditioned latent field, decoded into strand geometry via a PCA-based decoder. We reconstruct a defurred animal body using local fur-thickness cues from a surface-constrained Gaussian Frosting representation together with part-based priors. We further show that a PCA-based decoder learned from human-hair strand data can alleviate animal-data scarcity while enabling substantially faster optimization. FurE achieves a 10x speedup in strand training over current SOTA dense per-strand optimization while retaining strand fidelity and generalizing across synthetic and real-world sequences, with quantitative and qualitative validation despite the reduction in training time.","published":"2026-09-28T17:59:58Z","updated":"2026-09-28T17:59:58Z","authors":["Srinjay Sarkar","Prakhar Kaushik","Soumava Paul","Alan Yuille"],"categories":["cs.CV","cs.AI","cs.GR"],"headline":"New Method Efficiently Reconstructs Animal Fur from Photos","summary":"The researchers addressed the challenge of creating realistic 3D models of animal fur, which is difficult due to detailed hair strands, overlapping fur, and a lack of animal-fur data. They developed FurE, a method that reconstructs individual fur strands from multiple images and allows editing. FurE uses techniques borrowed from human hair data to overcome limited animal data and runs about ten times faster than previous methods while maintaining quality, as shown in tests with synthetic and real footage."},{"id":"2609.35769v1","title":"Telescopic Language Models","abstract":"One deployed language model must often serve many compute budgets, yet serving each budget still means a separate training or compression run per point. We train a Telescopic Language Model (TLM) to be that continuum: a nested-capacity Transformer supervised by stochastic prefix supervision with a full anchor. At every step, one randomly truncated prefix of the capacity axis is trained against the full next-token target, alongside one full-capacity pass, so the trained artifact is a valid language model at every depth. Two forward-backward passes per step, no architectural change, nothing extra at inference. Fixed-exit suites such as Matryoshka Language Model Suites (MLMS) occupy one point in this design space, and the point has a cost: supervising only a few fixed exits leaves the nested model at chance level everywhere else (perplexity 10^2-10^5 in our baselines). On a 200M proxy suite (20B FineWeb-Edu tokens, identical data stream for all methods), a single TLM run is a valid language model at every one of its twenty layer prefixes, in perplexity and on perplexity-sensitive downstream tasks, reducing the area under the quality-budget curve by 43-44% relative to the fixed-exit suites while matching them at full capacity, at ~12% lower GPU cost per run. The prefix sampling density is a dial: concentrating it on a few depths recovers fixed-exit quality there at the price of the continuum, so the operating points become a training-time choice rather than an architectural one. These results indicate that the training objective, not the nesting itself, is what makes a model elastic.","published":"2026-09-28T17:59:53Z","updated":"2026-09-28T17:59:53Z","authors":["Zhilin Guo","Boqiao Zhang","Hakan Aktas","Kyle Fogarty","Nursena Koprucu Aslan","Wenzhao Li","Canberk Baykal","Albert Miao","Siyu Hong","Yixiao Liu","Adam Wu","Ashish Kumar Singh","Sakar Khattar","Chenliang Zhou","Weihao Xia","Cristina Nader Vasconcelos","Cengiz Oztireli"],"categories":["cs.CL","cs.AI"],"headline":"New Language Model Adjusts Its Size to Save Computing Power","summary":"The researchers studied a language model that can change its size during use to fit different computing limits without retraining each version. They trained their model, called Telescopic Language Model (TLM), to perform well at every smaller size by teaching it with partial versions during training. Their experiments show that TLM works well across many sizes, uses less computing power during training, and performs better than previous fixed-size models. These findings come from tests using layered model prefixes and focus on training methods rather than model structure changes."},{"id":"2609.35768v1","title":"PDMD: Projected Distribution Matching Distillation for Video Diffusion Models","abstract":"Modern video diffusion models require tens of denoising evaluations over long spatiotemporal token sequences. Distribution Matching Distillation (DMD) reduces the number of function evaluations (NFE) to just a few. However, DMD samples can degrade during training, exhibiting progressive oversaturation and artifacts. We trace this instability to critic errors, which enter successive student updates and accumulate over time. We introduce Projected Distribution Matching Distillation (PDMD) to filter critic errors. PDMD projects out the component of the DMD update parallel to the student-critic endpoint residual. At a fixed noisy query, we prove that this residual is an unbiased estimate of the critic's endpoint error. Under high-dimensional assumptions, this projection removes a constant fraction of critic error while discarding only a vanishing fraction of ideal DMD signal. Empirically, the projection stabilizes training and improves sample quality where DMD degrades and develops unnatural textures. PDMD requires only a one-line code change to DMD, with no extra loss, network, data, model pass, or multi-stage training. With Wan2.1, PDMD achieves a VBench total score of 83.73 at 4 NFE, surpassing matched DMD by 1.03 points. On MiniMax-H3 joint video-audio generation, PDMD achieves a VideoGen-Eval visual total score of 83.17, 0.41 points above the strongest distilled baseline. PDMD also achieves the best performance on all six audio metrics among the compared 4-NFE models. Qualitative comparisons and user studies favor PDMD over the distilled baselines in visual quality, motion, and audio quality. Code and models are available at https://pdmd2026.github.io/.","published":"2026-09-28T17:59:52Z","updated":"2026-09-28T17:59:52Z","authors":["Zimo Wang","Junkun Yuan","Angtian Wang","Haotian Yang","Canyu Zhang","Siyuan Yuan","Xingchang Huang","Bo Liu","Yizhi Wang","Yiding Yang","Chongyang Ma","Gordon Guocheng Qian"],"categories":["cs.CV","cs.LG"],"headline":"New method improves training of faster video diffusion models","summary":"Researchers studied a problem in video diffusion models where reducing the number of steps causes quality issues like oversaturation and artifacts. They identified errors in the training feedback (critic errors) as the cause and developed Projected Distribution Matching Distillation (PDMD) to filter out these errors during training. Their method, proven mathematically under certain assumptions and tested through computer experiments, stabilizes training and improves video and audio quality without added complexity. The findings are based on simulations and evaluations reported by the researchers."},{"id":"2609.35767v1","title":"Learning Native Reflection in Unified Models with Interleaved Reinforcement Learning","abstract":"Unified multimodal models can both look at and render images, so in principle they can repair their own generations: diagnose what an image gets wrong, revise it, observe the result, and diagnose again. Whether a revision helps is known only after it is rendered, so the reflection text and the image generation must be learned jointly, over the whole loop. Supervised fine-tuning (SFT) on reflection trajectories gives a cold start but does not find the high-success repair paths, and naive RL that optimizes only the renderer or only one head leaves most of the gain untapped. We introduce UMM-Reflection, which applies reinforcement learning (RL) to complete reflection trajectories inside one unified model: sibling trajectories share one initial image, so the group-relative advantage compares reflection strategies, and one trajectory-level advantage updates both the reflection tokens and the flow-based revisions, avoiding the combinatorial blow-up of per-round credit assignment. Unlike single-round editing or pipelines with an external critic, credit flows across rounds and to both roles of the same model, and no verifier is needed at inference. On BAGEL, UMM-Reflection improves GenEval by 12.05 points over SFT, and the gains transfer to WISE (+10.97), OneIG-Bench (+3.48), and T2I-CompBench++ (+4.63), none of which is used in training.","published":"2026-09-28T17:59:36Z","updated":"2026-09-28T17:59:36Z","authors":["Yijia Fan","Ziqi Huang","Zhongang Cai","Yan Li","Zimo Wen","Wanqi Yin","Haiwen Diao","Ziwei Liu"],"categories":["cs.CV","cs.AI"],"headline":"Teaching AI to Improve Its Own Image Creations","summary":"Researchers developed a method called UMM-Reflection that helps unified AI models diagnose and fix mistakes in images they create. Using a special type of learning, their approach trains the AI to make better revisions over multiple steps without needing an external reviewer. Tests show this method improved the AI's image quality on several benchmarks compared to earlier techniques, although these results are based on the researchers' experiments and reported metrics."},{"id":"2609.35766v1","title":"Hamiltonicity of mildly pseudorandom regular graphs","abstract":"We show that if an $(n,d,λ)$-graph satisfies $λ\\leq (1-δ)d$ and $d\\gg δ^{-6}(\\log n)^{3}$ for some $δ>0$, then it is Hamiltonian. A qualitatively similar result was recently proven by Bradač and Janzer. Our proof here is shorter and gives better quantitative bounds. \\par In our proof, as in earlier work of Ferber, Han, Mao, and Vershynin, we use a random matrix inequality to show that a mild spectral gap is typically preserved after randomly sampling an appropriate proportion of the vertices. This allows us to deduce that typical balanced bipartite subgraphs of pseudorandom graphs contain perfect matchings. To convert a collection of perfect matchings into a Hamilton cycle, we use a variant of the sorting network method.","published":"2026-09-28T17:59:30Z","updated":"2026-09-28T17:59:30Z","authors":["Alp Müyesser"],"categories":["math.CO"],"headline":"Researchers find new conditions for special looped paths in graphs","summary":"The researchers studied certain graphs called mildly pseudorandom regular graphs and showed these graphs contain a special loop visiting every point exactly once, known as a Hamiltonian cycle, under specific conditions. They improved on previous work by providing a shorter proof and better numerical limits. Their approach involves mathematical tools from random matrix theory and methods to combine smaller connections into a complete loop. The results rely on theoretical analysis using these advanced techniques."},{"id":"2609.35765v1","title":"Retrieving Biblical Intertextual References in Karen Blixen's Seven Gothic Tales","abstract":"Identifying intertextual references is central to literary scholarship, but computationally difficult when source material is transformed through paraphrase, allusion, historical language, and translation. We investigate this problem through biblical intertextuality in Karen Blixen's Seven Gothic Tales. Drawing on the commentary to a critical edition, we construct a benchmark of 189 annotated references and evaluate retrieval against all 31,170 verses of historically plausible Danish Old and New Testament translations. We compare TF-IDF and BM25 with multilingual and Danish sentence encoders, examine the effect of linguistic normalization, and fine-tune a Danish encoder using hard negatives and five-fold cross-validation. We analyze performance across automatically derived lexical-overlap strata representing quotations, paraphrases, and allusions. Linguistically normalized BM25 provides a strong zero-shot baseline, attaining an overall R@10 of 0.365 and retrieving every quotation within its ten highest-ranked verses. The best zero-shot dense model achieves a comparable overall score of 0.360 while performing better on allusions. Fine-tuning DFM-large raises its overall R@10 from 0.265 to 0.508 and more than doubles its performance on allusions, from 0.138 to 0.339. However, evaluation against editorial annotations alone understates the model's scholarly usefulness: a literary scholar judged seven of 30 selected rank-one predictions counted as false positives to be meaningful additional references. These findings show both the potential and the epistemic limits of computational intertextual retrieval. Rather than treating scholarly annotations as exhaustive or model outputs as discoveries, we propose retrieval models as heuristic co-readers that recover documented references and generate candidates for expert-led close reading.","published":"2026-09-28T17:59:27Z","updated":"2026-09-28T17:59:27Z","authors":["András Kovács","Alexander Conroy","Daniel Hershcovich","Jens Bjerring-Hansen"],"categories":["cs.CL"],"headline":"Computers help find Bible references in Karen Blixen's stories","summary":"Researchers tackled the challenge of identifying biblical references in Karen Blixen's Seven Gothic Tales, which can be hard to spot due to paraphrasing and old language. They tested several computer methods to find 189 known references within over 31,000 Bible verses translated into Danish. The study showed some models are good at finding direct quotes, while others better detect subtle allusions, especially after training. However, human experts still play a key role since computers may miss or misclassify some references, so these tools serve as helpful aids rather than definitive discoverers."},{"id":"2609.35764v1","title":"Reliability-Gated Fusion of Consumer Head and Foot IMUs for Lower-Body 3D Pose","abstract":"Sparse inertial pose estimation promises camera-free motion capture from consumer devices, but consumer sensors are unreliable: firmware-fused orientations are biased, mounting varies between sessions, and streams drift or drop out. On a new 35-take single-subject benchmark pairing an earbud head inertial measurement unit (IMU) with two smart-insole foot IMUs (SAM-3D-Body pseudo-ground-truth labels), we show the reliability problem is channel-level: a channel ablation isolates foot acceleration as the most informative input (66.6 mm vs. 79.0 mm head-only) and the firmware-fused foot orientation as the liability that destroys the gain. We therefore let the model learn how much to trust each channel of each stream: one temporal gate per stream per channel block, trained with an auxiliary reliability objective on synthetically corrupted pretraining data. The channel-gated model is the most accurate of our learned fusion arms on clean data (69.4 mm vs. 83.7 static, 86.6 ungated) and under every simulated fault (bias in training; drift, dropout eval-only); its gates suppress the natively biased foot-orientation channels on clean real data without test-time supervision and flag dropout bursts at 0.92-0.999 AUROC. Two contrasts: dropping a channel known a priori to fail is flat across foot faults but collapses when an unanticipated stream fails (head dropout: 92.9 vs. 79.3 mm); and a fine-tuned HMD-Poser is more accurate on clean data (64.4 mm) and nominally under drift, with no significant paired difference under bias or dropout, but a larger worst-case degradation from clean (+16.1 vs. +3.5 mm, single seed). Learning to gate reliability instead of sensor count is the lever for deployable sparse inertial capture. Code is available at https://github.com/ZhilinGuo/reliability-gated-imu-fusion.","published":"2026-09-28T17:59:22Z","updated":"2026-09-28T17:59:22Z","authors":["Zhilin Guo","Boqiao Zhang","Oszkár Urbán","Josef Bengtson","Hakan Aktas","Wenzhao Li","Siyu Hong","Kyle Fogarty","Chenliang Zhou","Ali Senguel","Cengiz Oztireli"],"categories":["cs.CV","cs.HC"],"headline":"Improving Body Pose Tracking Using Consumer Motion Sensors","summary":"Researchers studied how to better estimate lower-body 3D pose using common wearable sensors like earbuds and smart insoles, which can be unreliable due to errors and signal loss. They developed a method where a model learns to trust or ignore certain sensor data based on its reliability, improving accuracy even when sensors give faulty or drifting data. Their results, based on tests with simulated faults and real recorded data, suggest this reliability-gating approach helps overcome common problems in motion tracking with consumer devices, though the findings are limited to their specific data and tested scenarios."},{"id":"2609.35763v1","title":"Unifying Distributional Training for One-Step Visual Generation","abstract":"\\emph{Distributional training} provides collective supervision for one-step visual generation by matching real and generated features in frozen representation spaces. We introduce \\emph{a unified theoretical framework} that separates distribution modeling from matching discrepancy and connects global objectives to pointwise feature updates through Wasserstein gradient flow. Under this framework, FD-Loss and Gaussian-kernel Drifting are recovered through Gaussian optimal transport and kernel-density-based KL matching, respectively. The framework motivates \\textbf{MGFlow}, which models feature distributions with Gaussian mixtures at an adjustable granularity between global moments and sample-based representations. MGFlow supports both optimal transport and score-based matching, and couples mass-constrained sample assignment with paired component updates to address mode collapse that mixture expressivity alone does not resolve. On ImageNet $256\\times256$, MGFlow substantially surpasses the FD-Loss baseline, achieving state-of-the-art results with \\textbf{1.45} $\\mathrm{FDr}^6$ on pMF-H and \\textbf{1.64} on JiT-H. For text-to-image generation, MGFlow post-trains FLUX.2 [klein] 4B into a one-step generator that outperforms the original four-step model on both GenEval and PickScore. Project page: https://shihaoyang0423.github.io/MGFlow-website/","published":"2026-09-28T17:59:14Z","updated":"2026-09-28T17:59:14Z","authors":["Chi Zhang","Haoyang Shi","Yueyi Liu","Ruichuan An","Junkang Zhou","Chang Li","Xiuyuan Lu","Yichi Zhang","Bo Wang","Yuhang Wu","Sen Cui","Miao Liu"],"categories":["cs.LG"],"headline":"New framework improves one-step image generation quality","summary":"Researchers developed a theoretical framework that unifies how computer models learn to generate images in a single step by comparing features of real and generated images. They created a method called MGFlow that models image features more flexibly and addresses common problems like mode collapse. Testing on high-resolution images and text-to-image tasks, MGFlow showed improved performance over previous approaches, though the findings are based on mathematical modeling and simulations."},{"id":"2609.35762v1","title":"Peppy: An AI-Assisted Workflow for Tight Convergence Analysis of Optimization Algorithms","abstract":"This paper presents Peppy, an AI-assisted workflow for discovering tight, analytic convergence proofs for first-order optimization algorithms. Generic approaches to using LLMs to conduct mathematical research target an unspecified, broad spectrum of problems and sometimes use the Lean 4 proof assistant for formalization. On the other hand, Peppy leverages domain-specific knowledge more heavily and is thereby capable of constructing the proofs in a more structured manner, which allow a minimal and accessible verification through SymPy. We experimentally demonstrate through examples that Peppy provides a rigorous, practical, and reproducible paradigm for AI-assisted theorem synthesis in optimization. We further highlight its capability of closing several open problems on tight convergence analysis of first-order optimization algorithms, including conjectures for Nesterov's FGM. Overall, Peppy is designed to turn the art of optimization algorithm analysis into a science.","published":"2026-09-28T17:59:12Z","updated":"2026-09-28T17:59:12Z","authors":["Jaewook J. Suh","TaeHo Yoon","Edward Duc Hien Nguyen","Bicheng Ying","Shiqian Ma"],"categories":["math.OC"],"headline":"AI Tool Helps Prove Mathematical Results in Optimization Algorithms","summary":"Researchers developed Peppy, an AI-assisted method to create tighter, clear proofs about how certain optimization algorithms perform. Unlike broad AI approaches, Peppy uses specific knowledge to build structured proofs that can be verified easily with existing software. The team demonstrated that Peppy can rigorously and reliably produce results, even solving some open questions about a well-known optimization method called Nesterov's FGM. These findings come from their experiments and relate to theoretical analysis, not real-world testing."},{"id":"2609.31620v1","title":"FuseReg: Regularizing Layer Fusion Mitigates the Reconstruction-Generation Gap in Representation Autoencoders","abstract":"Representation autoencoders (RAEs) reuse features from a pretrained visual encoder as reconstruction and diffusion latents, integrating strong visual representations into image generation. However, RAEs still need to decide which encoder layers form the shared latent space for the generator and pixel decoder. This choice involves a trade-off. Shallower layers tend to preserve fine pixel details better, while deeper layers tend to yield better generation metrics. A fixed heuristic layer fusion therefore couples two stages that benefit from different information. We introduce FuseReg, which replaces heuristic feature selection with training over random subsets of encoder layers. We theoretically analyze the underlying mechanism: subset sampling explicitly penalizes sensitivity to cross-layer disagreement. On ImageNet-256 with DINOv3-L, a single FuseReg decoder reconstructs from full, sparse, and single-layer fusions without retraining, achieving higher PSNR than decoders specialized to fixed fusions. This flexibility also benefits generation: decoder replacement alone reduces unguided gFID by 27% with an unchanged RAEv2 DiT-XL generator. The same regularization principle extends to diffusion training, with joint regularization of both stages reducing unguided gFID by 29% on DiT-Base. These results show that training downstream models for layer-fusion robustness narrows the reconstruction-generation gap without modifying the pretrained encoder.","published":"2026-09-25T17:59:59Z","updated":"2026-09-25T17:59:59Z","authors":["Hongyang Du","Yunfei Xie","Junjie Ye","Jiawei Yang","Xiaoyan Cong","Haodong Zhang","Yongchao Huang","Haiyu Wu","Zongxia Li","Shihang Gui","Dawei Liu","Runhao Li","Jingcheng Ni","Chen Wei","Randall Balestriero","Yue Wang"],"categories":["cs.CV"],"headline":"Helping image-making AI use both detail and the bigger picture","summary":"Image-making AI needs to understand both tiny details and the overall scene. The researchers trained part of a system using changing combinations of these features, instead of one fixed mix. In their image tests, this improved both how faithfully it rebuilt pictures and how well it generated new ones."},{"id":"2609.31619v1","title":"Learning to Stop without Learning to Stop: Self-Supervised Confidence Training Improves Reasoning Efficiency","abstract":"Reasoning models often generate very long reasoning traces, making inference computationally expensive. Existing approaches typically improve efficiency either through inference-time early-stopping mechanisms or by explicitly encouraging shorter reasoning during training, for example through reinforcement learning with length penalties. We show that substantial efficiency gains can instead emerge from a different kind of supervision: \\textit{confidence}. Using a self-supervised procedure, we fine-tune reasoning models to predict their confidence in the answer at intermediate points along their own reasoning trajectories using only 600 training problems. Confidence is used only as a training target: the loss contains no objective for reasoning length, efficiency, or stopping. At inference, the fine-tuned models use the standard generation procedure, with no confidence elicitation or early-stopping mechanism. Despite this, self-supervised confidence fine-tuning makes reasoning more efficient, reducing generated tokens by up to 25\\% at matched accuracy across Gemma, Qwen, Nemotron, and GPT-OSS models on mathematical, scientific, and coding reasoning benchmarks, with efficiency gains comparable to methods that explicitly optimize for shorter reasoning. Analysis of reasoning episodes further shows that confidence supervision largely preserves the base models' high-level reasoning composition rather than selectively suppressing particular behaviors. Our results suggest that efficient reasoning may emerge as a downstream consequence of learning metacognitive signals, without being directly optimized.","published":"2026-09-25T17:59:52Z","updated":"2026-09-25T17:59:52Z","authors":["Parsa Hosseini","Akasha Tigalappanavara","Sumit Nawathe","Chenrui Fan","Sourya Basu","Genta Indra Winata","Anirban Das","Soheil Feizi","Nima Chitsazan"],"categories":["cs.AI","cs.CL","cs.LG"],"headline":"Teaching AI confidence could help it spend less time thinking","summary":"AI can spend a lot of computing time working through an answer. The researchers trained several models to judge how confident they were partway through a problem. In their tests, the models then used up to 25% less text to work things out while maintaining similar accuracy—even without being told to stop sooner."},{"id":"2609.31618v1","title":"Quantum Representation Selects Neutral Skyrmion Molecules","abstract":"Skyrmions experience a sideways gyroscopic force when they move. Bound molecules can cancel this force; we call them gyroscopically neutral. We show that the quantum states available on each site of a magnet select the composition needed for cancellation. In two models of $SU(3)$ magnetism, this composition changes from a pair to a trimer. Both models assign exactly the same energy to every configuration of their common classical order parameter, isolating the effect of the local quantum states. Neutral molecules can also unwind when additional local states are allowed. Mean field calculations nevertheless find both compositions bound and locally stable, with the neutral trimer surviving sampled interaction changes. Comparing their motion reveals the changed cancellation rule, and prepared neutral molecules translate over short distances with little deformation.","published":"2026-09-25T17:59:49Z","updated":"2026-09-25T17:59:49Z","authors":["Vishnu Jejjala"],"categories":["cond-mat.str-el","cond-mat.mes-hall","hep-th","nlin.PS"],"headline":"Why tiny magnetic swirls can travel better in groups","summary":"Tiny swirls in a magnet tend to veer sideways as they move. This theoretical study finds that groups can cancel that sideways push, but whether it takes two swirls or three depends on the quantum states available inside the magnet. Calculations suggest these groups can stay together and move short distances without changing shape much."},{"id":"2609.31617v1","title":"Singularity models for the Bernoulli free boundary problem from isoparametric hypersurfaces","abstract":"We develop a general construction of homogeneous solutions to the Bernoulli free boundary problem, as well as general extremal domains on the sphere, from isoparametric foliations of the sphere. Our construction produces rich families of infinitely many new examples with sophisticated topologies connected to minimal surfaces of the sphere by smooth families of interpolating capillary surfaces, which include novel singularity models in low dimensions and recover most known homogeneous one-phase solutions. We also introduce a geometric representation for every such homogeneous map and establish a geometric rigidity theorem: for every prescribed isoparametric foliation, and in particular for cohomogeneity-one subgroups of $O(n)$, its radial geometry uniquely determines the corresponding homogeneous solution. Finally, we study the density of the constructed solutions and conjecture that the lowest density among non-flat cones in a given dimension $n \\geq 5$ is attained by an $O(k) \\times O(n-k)$-invariant cone.","published":"2026-09-25T17:59:16Z","updated":"2026-09-25T17:59:16Z","authors":["Benjy Firester","Raphael Tsiamis","Zihui Zhao"],"categories":["math.AP","math.DG"],"headline":"Finding new shapes where smooth boundaries develop sharp points","summary":"Some mathematical problems ask where a boundary will form, rather than specifying it in advance. The researchers developed a way to build many new solutions with complicated shapes, including places where the boundary stops being smooth. Their work helps describe which shapes these irregular points can take; one proposed rule about the simplest shapes remains a conjecture."},{"id":"2609.31616v1","title":"A Unified Spin-Fermion Framework for Magnetic Diversity in Chromium Monopnictides","abstract":"Isoelectronic compounds are generally expected to exhibit related electronic, structural, and magnetic behavior. Chromium monopnictides CrX (X= Sb, As, P), however, display a striking evolution from high-temperature altermagnetism in NiAs-type CrSb, through double-helical antiferromagnetism in MnP-type CrAs, to the absence of resolved long-range magnetic order in MnP-type CrP. Here, combining first-principles calculations in the non-collinear Curie-paramagnetic state, structural analysis, magnetic phase-diagram calculations, and exchange-parameter extraction, we establish a unified microscopic framework for this evolution. All three compounds share a formal high-spin $d^5$ configuration, nominally Cr$^{1+}$, coupled to itinerant states of predominantly pnictogen character. From Sb to P, chemical pressure enhances ligand itinerancy and drives the structural evolution from the high-symmetry NiAs lattice to increasingly distorted MnP-type networks. The accompanying reconstruction of competing Cr-Cr exchanges favors A-type antiferromagnetism in CrSb and double-helical order in CrAs, while placing CrP near a frustrated regime with closely competing magnetic tendencies. Our results establish chemical-pressure-driven exchange reconstruction as a general mechanism through which a common spin-fermion electronic structure can generate contrasting magnetic phases in strongly correlated materials.","published":"2026-09-25T17:59:05Z","updated":"2026-09-25T17:59:05Z","authors":["Ruoshi Jiang","Bartomeu Monserrat"],"categories":["cond-mat.str-el","cond-mat.mtrl-sci"],"headline":"Why similar chromium compounds behave like very different magnets","summary":"Three chromium compounds have similar electron counts but very different magnetic behavior. Using calculations, the researchers traced these differences to how their atoms are arranged and how electrons move between them. Their model suggests that changing chromium’s partner element reshapes the competing interactions that determine the magnetic pattern."},{"id":"2609.31615v1","title":"On Simon's area-minimizing hypersurfaces with fractal singular sets","abstract":"We verify that the stable minimal hypersurfaces with prescribed singular sets constructed by Leon Simon are area-minimizing. Thus, Simon's work settles the existence of area-minimizing hypersurfaces with fractal singular sets.","published":"2026-09-25T17:58:42Z","updated":"2026-09-25T17:58:42Z","authors":["Zhenhua Liu"],"categories":["math.DG"],"headline":"The smallest possible surfaces can have surprisingly rough patches","summary":"Think of a soap film trying to cover as little area as possible, but in higher-dimensional mathematics. This paper verifies that certain previously constructed surfaces really do minimize area, even though their irregular points can form intricate fractal patterns. The abstract is brief and does not describe how the proof works."},{"id":"2609.31614v1","title":"Gap-free Differentially Private PCA for Gaussian Data","abstract":"We give a gap-free differentially private algorithm for the principal component analysis (PCA) problem with Gaussian data.","published":"2026-09-25T17:58:38Z","updated":"2026-09-25T17:58:38Z","authors":["Alina Ene","Huy L. Nguyen"],"categories":["cs.DS","cs.LG"],"headline":"Finding patterns in data while protecting individual privacy","summary":"This paper presents a method for finding the main patterns in data while limiting what the results reveal about any one person. It applies to data following a bell-shaped statistical model and removes a technical requirement that the leading patterns be clearly separated. The short abstract does not report practical performance or test results."},{"id":"2609.31613v1","title":"Galois groups of twisted reciprocal polynomials: a uniform asymptotic","abstract":"We study the Galois group $G_f$ of a random polynomial $f$ in the family of polynomials of degree $2n$ satisfying the twisted reciprocal relation $f(x) = x^{2n}/b^n \\cdot f(b/x)$. We use a Euclidean height adapted to this relation. Our main result is an asymptotic theorem of van der Waerden--Bhargava type: for fixed $b \\neq 0$ and $n \\geq 4$, the number of polynomials of height at most $H$ whose Galois group is not the full hyperoctahedral group $S_2 \\wr S_n$ is an explicit constant times $H^n\\log H$, with an error of order $H^n$. We determine the dependence of the leading constant on $b$; the leading-order group $G_1$ is of index $2$. This paper is a sequel to a recent paper by Anderson, Bertelli, and the author addressing reciprocal polynomials (i.e. the case $b = 1$).","published":"2026-09-25T17:58:22Z","updated":"2026-09-25T17:58:22Z","authors":["Evan M. O'Dorney"],"categories":["math.NT"],"headline":"Counting rare exceptions in the symmetry of equations","summary":"Polynomial equations can have hidden symmetries among their solutions. For a particular family of equations, this paper estimates how often those symmetries are smaller than the largest possible group. It gives a more precise count of these exceptions as the range of equations considered grows."},{"id":"2609.31612v1","title":"First-Order Stationarity of Reverse Diffusions","abstract":"Recent literature has shown a strong connection between optimization and sampling. We develop the corresponding first-order theory for diffusion models. First, the SDE-based reverse-time flows of overdamped and underdamped Langevin diffusions contract relative Fisher divergences at explicit exponential rates whenever the stationary potential of the forward process is strongly convex---a condition on the noising process one chooses, not on the data. This is a unique advantage of SDE-based reverse diffusion, absent in the reverse process based on ODEs. Second, we incorporate discretization and establish averaged first-order stationarity bounds---the sampling analog of averaged gradient-norm guarantees in nonconvex optimization---for samplers of both overdamped and underdamped diffusion models. As in nonconvex optimization, the convexity-free certificate is local: it guarantees score consistency, not global mode weights.","published":"2026-09-25T17:57:43Z","updated":"2026-09-25T17:57:43Z","authors":["Zhifeng Chen","Chenyang Jiang","Yazhen Wang"],"categories":["stat.ML","cs.LG"],"headline":"Checking how well AI learns to reverse random noise","summary":"Some AI systems generate data by gradually undoing a process that adds random noise. The researchers developed mathematical checks for certain versions of this process and for their computer approximations. These checks establish a local kind of consistency, but do not guarantee that the system produces every type of outcome in the right proportions."},{"id":"2609.31611v1","title":"A Multidimensional General-Relativistic Boltzmann Solver for Neutrino Transport: Implementation, Discretization and Optimization","abstract":"We present the implementation of a multidimensional general-relativistic Boltzmann solver for neutrino transport using the finite volume method. We extend the general-relativistic magnetohydrodynamics solver \\texttt{Gmunu} to discretize the full $6$D phase space. We discretize the momentum space in spherical coordinates in the comoving frame, and the position space in the lab frame in Cartesian, cylindrical or spherical coordinates. As in the M1 scheme, we expand the interaction kernels up to first order in a Legendre series. We present a discretization scheme that ensures consistency between the number-conservative and non-conservative formulations in empty flat spacetime, multiple dimensions and coordinate systems, conserves energy to $\\sim 1\\%$ with $20$ energy bins in 1D tests. We discuss optimizations of the implicit solver for stiff source terms to minimize its computational cost. In particular, we introduce a method that leverages the Legendre expansion of the kernels to reduce the dimensionality of the problem. In a 1D core-collapse supernova snapshot, our method yields a speed-up factor of $300$ compared to a full-matrix LU method for $14$ angular bins when including energy- and species-coupling interactions. Finally, we validate our implementation on standard test cases and report quadratic convergence in space, energy and propagation angles. We compare our solver to the M1 scheme in simplified 1D configurations and find differences of about $10\\%$ in the free-streaming luminosities in relaxation test cases, which we attribute primarily to the M1 closure relation. The average energies, on the other hand, agree between the two methods. We compare both methods on a core-collapse supernova test case and find an overall good agreement in the fluid variables' profiles at the time of core bounce for $20$ energy bins.","published":"2026-09-25T17:57:00Z","updated":"2026-09-25T17:57:00Z","authors":["Arthur Offermans","Harry Ho-Yin Ng","Patrick Chi-Kit Cheong","Anthony Mezzacappa","Tjonnie Guang Feng Li"],"categories":["astro-ph.HE","gr-qc","hep-ph"],"headline":"Simulating elusive particles inside exploding stars more efficiently","summary":"Neutrinos are particles that carry energy through exploding stars, and tracking them in a simulation is computationally demanding. The researchers built and tested a solver that accounts for their motion and interactions, including effects of gravity. One calculation was 300 times faster than a comparison method in a specific simplified test; that is not a speed claim for an entire star simulation."},{"id":"2609.31610v1","title":"Optimal regularity for the arrival time equation","abstract":"For a convex mean curvature flow in $\\mathbb{R}^{n+1}$, the corresponding arrival time function solves a degenerate elliptic equation that becomes singular at extinction. We prove that it is $C^{24,α}$ for any $0<α<1$ in the plane, with a logarithmic modulus for its twenty-fourth derivatives. We prove that this bound is sharp combining regularity obstruction in our companion paper. We obtain asymptotic expansions of the rescaled flow to any prescribed order, including derivative estimates for the remainder. Combined with the work of Sesum for $n\\ge 2$, this also gives the sharp $C^{2,2/n}$ regularity for convex arrival time functions in higher dimensions.","published":"2026-09-25T17:56:53Z","updated":"2026-09-25T17:56:53Z","authors":["Yiqi Huang","Jingze Zhu"],"categories":["math.DG","math.AP"],"headline":"Researchers find precise smoothness limits for a math equation on shapes","summary":"The researchers studied a special equation that describes how convex shapes change over time using mean curvature flow. They showed that in two dimensions, the solution is very smooth—up to 24 derivatives with some controlled irregularity—and that this level of smoothness cannot be improved. They also expanded their results to higher dimensions using related work. Their findings are based on mathematical proofs and detailed analysis, focusing on the behavior near the point where the shape disappears."},{"id":"2609.31609v1","title":"Exact series formulas for the capacities of the amplitude damping channel","abstract":"We derive explicit, absolutely convergent series formulas for the quantum, unassisted classical, and entanglement-assisted classical capacities of the qubit amplitude damping channel, eliminating residual optimizations and implicit roots. We also obtain analytical upper and lower bounds on two-way assisted quantum and private communication capacities by evaluating a balanced-squashing bound and optimizing the reverse coherent information. Our results characterize the optimal input populations, establish convergence and truncation properties, and provide analytical benchmarks for quantum and classical communication over dissipative channels.","published":"2026-09-25T17:56:34Z","updated":"2026-09-25T17:56:34Z","authors":["Stefano Pirandola"],"categories":["quant-ph","math-ph","physics.optics"],"headline":"Researchers find exact formulas for quantum communication limits","summary":"The researchers address the problem of determining how much information can be sent through a specific type of noisy quantum channel called the amplitude damping channel. They derived exact series formulas that precisely calculate different capacities of this channel without needing complex optimizations. Their work also provides bounds on other communication capacities and clarifies optimal input strategies. These findings provide theoretical benchmarks for quantum communication but are based on mathematical analysis rather than practical experiments."},{"id":"2609.31608v1","title":"Formal groups and $(\\varphi,Γ)$-modules","abstract":"Let $K/E$ be a finite unramified extension of $p$-adic local fields, and let $H$ be a one-dimensional formal $\\mathcal{O}_E$-module of finite height over $\\mathcal{O}_K$. We introduce the exponential period map of $H$ and use it to construct a complete regular local ring $R_{H,K}$ with imperfect residue field, an endomorphism $\\varphi_E$, and a commuting action of $Γ=\\mathrm{Gal}(K(H[p^\\infty](\\overline{K}))/K)$. We prove an equivalence of categories between finitely generated $\\mathcal{O}_E$-modules with a continuous $\\mathrm{Gal}_K$-action and étale $(\\varphi_E,Γ)$-modules over $R_{H,K}$. This recovers the classical cyclotomic and Lubin-Tate equivalences in the corresponding cases. In general, $R_{H,K}$ can have Krull dimension greater than one, and $\\varphi_E$ need not lift the $q_E$-power Frobenius modulo $π_E$. The proof uses $F$-dynamical systems over $\\mathcal{O}_E$, which combine contraction modulo $π_E$ with Frobenius on the residue field. For every flat $F$-dynamical system, we establish an equivalence of categories between étale $\\varphi$-modules and continuous representations of the absolute Galois group of the residue field on finitely generated $\\mathcal{O}_E$-modules.","published":"2026-09-25T17:55:56Z","updated":"2026-09-25T17:55:56Z","authors":["Daishi Kiyohara"],"categories":["math.NT","math.AG"],"headline":"New connections in number theory link formal groups and Galois actions","summary":"Researchers studied certain mathematical objects called formal modules over fields related to p-adic numbers. They introduced a new tool called the exponential period map to build a complex ring with special properties and showed how this setup connects algebraic structures with symmetries called Galois actions. Their work generalizes known equivalences in number theory and uses advanced systems combining two types of operations. The findings are theoretical, relying on abstract algebraic frameworks without direct computational examples."},{"id":"2609.31607v1","title":"Statistical attribute alignment for black-box generative AI via output post-processing","abstract":"Generative AI systems are increasingly used, but aligning their outputs with user requirements poses a continuing challenge. Here, we aim to ensure that the distribution of an attribute of an AI-generated output aligns with a user-specified target. This is motivated by examples such as fairness, where we want to ensure that a protected attribute (e.g., gender, race, or age categories) follows a desired distribution, and synthetic data generation, where we want the generated data to be representative of a target distribution. We study the practically important black-box access setting, where a user can repeatedly query a generative AI model. The goal is to return $m\\ge 1$ outputs whose joint attribute distribution is as close as possible to this target. For both exact and approximate alignment, we develop algorithms that minimize the expected number of queries to the generator, and we further demonstrate their optimality as the number of requested outputs $m \\rightarrow \\infty$. Experiments on text-to-image generation and geocoded persona generation tasks show that our post-processing algorithms improve statistical attribute alignment, complementing prompting-based interventions.","published":"2026-09-25T17:55:49Z","updated":"2026-09-25T17:55:49Z","authors":["Kevin Jiang","Morgane Austern","Edgar Dobriban","Jason M. Klusowski"],"categories":["stat.ME","cs.AI","cs.LG","math.ST"],"headline":"Improving AI Output to Match Desired Attribute Distributions","summary":"Researchers address the challenge of making AI-generated content meet specific attribute targets, like fairness in gender or race distribution. They developed methods that adjust outputs from generative AI models without needing to change the model itself, focusing on cases where users can only query the AI multiple times. Their algorithms aim to produce multiple outputs whose combined attributes closely match a user-defined distribution. Experiments show these post-processing techniques improve alignment alongside other methods, though results come from controlled testing and simulations."},{"id":"2609.31606v1","title":"Learning Robot Policies from Sparse Success Signals via STL-Guided Stein Variational Policy Gradient","abstract":"Learning robot policies for tasks with sparse success signals is challenging when completion depends on coordinated actions, precise contact outcomes, or satisfying several conditions together. Intricate physical interactions with the world further complicate these requirements. Prior work using conventional reward shaping mechanisms provides dense feedback but local progress might not translate into eventual task completion. We present Signal Temporal Logic-guided Stein Variational Policy Gradient (STL-SVPG), a population-based method that uses smooth STL robustness as a trajectory-level training objective. Differentiating this objective through the dynamics assigns credit to policy actions according to their effect on the complete task specification, rather than local progress alone. We evaluate the approach on six quadcopter and manipulator tasks that involves event-triggered responses, strictly ordered behavior, responses within specified deadlines, and physical interaction with the world. STL-SVPG achieves the highest mean success rate among the compared methods on five of six benchmarks. Simulation-trained policies trained in simulation transfer temporal and contact task behavior to the real world.","published":"2026-09-25T17:55:01Z","updated":"2026-09-25T17:55:01Z","authors":["Hongrui Zheng","Cristian Ioan Vasile","Antonio Loquercio","Rahul Mangharam"],"categories":["cs.RO"],"headline":"New method helps robots learn tasks with few success signals","summary":"Researchers developed a method called STL-SVPG to help robots learn tasks where success signals are rare and complex. Their approach uses a mathematical logic tool to guide learning based on the whole task rather than just small local steps. Tested on multiple robot tasks in simulations, STL-SVPG showed better success rates and even helped policies trained in simulations work in the real world. The findings are based on simulations and robot experiments described in the abstract."},{"id":"2609.31605v1","title":"Measurement-Based Uncomputation from an Error Correction Perspective","abstract":"Coherent uncomputation can be as costly as the computation itself, creating substantial spacetime overheads that can be prohibitive on early fault-tolerant quantum computers. We develop an algebraic framework for measurement-based uncomputation (MBU) through a quantum error-correction picture: measurements of the garbage register induce relative phases analogous to errors causing non-trivial syndromes. Accepting only zero-syndrome outcomes achieves post-selected MBU, removing the need for an uncomputation oracle at the cost of a probabilistic success rate determined by the affine rank of the populated garbage support. This rank can be much smaller than the number of garbage qubits, as illustrated through an example of the HHL algorithm. We then derive a structural condition under which the original uncomputation oracle can be reused to correct the phase for all measurement outcomes. No correction is required for the zero syndrome, while corrections for non-zero syndromes can be further adapted to individual measurement outcomes for simpler circuits, as illustrated by existing modular-arithmetic, QROM, and QRAM MBU constructions. Phase-correction circuits can be further simplified by co-optimising their implementation with the representation of the measurement-induced phases. Selecting which syndrome classes to correct interpolates between post-selected, partially corrected, and deterministic MBU, enabling trade-offs among correction cost, workspace, and repetitions.","published":"2026-09-25T17:54:52Z","updated":"2026-09-25T17:54:52Z","authors":["Minjun Jeon","Po-Wei Huang","Zhenyu Cai"],"categories":["quant-ph"],"headline":"New Method Reduces Overhead in Quantum Computation Cleanup Steps","summary":"The researchers address the high cost of a process called uncomputation in quantum computing, which cleans up leftover data but can be as expensive as the computation itself. They propose a measurement-based method framed through quantum error correction, using measurements to identify and handle leftover data with certain conditions affecting success rates. Their approach allows different trade-offs between correction effort, workspace, and repetition, illustrated with examples from quantum algorithms. These findings are theoretical and based on algebraic and error-correction frameworks, not yet tested in real quantum hardware."},{"id":"2609.31604v1","title":"Fourier duality on compactified Prym fibrations","abstract":"We study Fourier-Mukai duality for a class of compactified Prym fibrations including moduli spaces of $\\mathrm{SL}$ Higgs bundles over the elliptic locus. This leads to a shadow of the Hausel-Thaddeus conjecture, proof of the Corti-Hanamura motivic decomposition conjecture for these fibrations, and multiplicativity of the perverse filtration. Our approach can be described as a generalization of the Maulik-Shen-Yin package for compactified Jacobian fibrations to the case when the dual abelian fibration is a stack. This forces us to move beyond the case of full supports. Technical tools include a new pullback identity for the stacky Grothendieck-Riemann-Roch tau functor introduced by Toën, results on descent of (Arinkin-)Poincaré sheaves, and a comparison of Poincaré sheaves of the Prym varieties of families of smooth and nodal curves along the lines of Franco-Hanson-Horn-Oliveira.","published":"2026-09-25T17:54:33Z","updated":"2026-09-25T17:54:33Z","authors":["Anne Larsen"],"categories":["math.AG"],"headline":"Researchers Explore Advanced Duality in Mathematical Spaces","summary":"The researchers investigate a complex form of duality called Fourier-Mukai duality within specialized mathematical objects known as compactified Prym fibrations. Their work extends existing methods to more complicated cases involving stacks and leads to partial support for notable theoretical conjectures. They introduce new technical tools and comparisons to handle these complex structures, but the findings are theoretical and rely on abstract mathematical frameworks rather than practical experiments."},{"id":"2609.31603v1","title":"User Model Extraction via Belief Self-Distillation","abstract":"Large language models (LLMs) implicitly infer attributes of their users and adapt their behavior accordingly, yet these beliefs remain difficult to inspect and causally manipulate. We introduce Belief Self-Distillation (BSD), a unified read-write framework that bridges linear and causal probing by learning a compact user representation that can be both decoded and written back into the model. The frozen LLM acts as its own teacher, distilling beliefs from natural conversations without external annotations. Unlike conventional probing, BSD isolates not only information present in activations, but a state whose causal role can be directly tested. Across multiple model families, BSD faithfully recovers user beliefs and enables substantially stronger interventions than matched hidden-state steering. Crucially, we find that refusal depends not only on the request, but on the model's inferred user intent: changing this belief alters refusal while holding the request fixed. We further uncover a striking cross-model regularity: independently trained LLMs converge on a shared geometry for representing their users. Together, these results reveal implicit user models as readable and causally writable internal states with direct implications for AI safety, shaping how models condition safety decisions on whom they believe they are interacting with.","published":"2026-09-25T17:54:32Z","updated":"2026-09-25T17:54:32Z","authors":["Ali Holmov","Yiran Huang","Kirill Bykov","Zeynep Akata"],"categories":["cs.LG","cs.CL"],"headline":"New method reveals and changes what AI models believe about users","summary":"Researchers developed a technique called Belief Self-Distillation (BSD) to uncover and modify what large language models think about their users. BSD lets the AI model learn and update a simple summary of user traits from conversations without needing extra labels. The study shows that these user beliefs inside the AI can be read and changed in ways that affect how the model responds, especially in refusing requests. The findings suggest these internal beliefs are important for understanding and guiding AI behavior, but results are based on testing within models, not real-world application."},{"id":"2609.31602v1","title":"Prescribed projections for a general class of maps and efficient covering by variable plane curves","abstract":"We prove a prescribed projection theorem for any class of non-linear projections defined on open subsets of the plane which satisfies some mild geometric conditions. This generalizes Davies's classic result on prescribed projections for orthogonal projections, as well as a recent result of authors, Chang, McDonald, and Taylor, on a specific class of non-linear projection maps. We also give several geometric applications, including a proof that certain classes of plane curves can be used to cover arbitrary measurable sets with measure zero error.","published":"2026-09-25T17:54:27Z","updated":"2026-09-25T17:54:27Z","authors":["Alan Chang","Robert Fraser","Alex McDonald","Mark Meyer","Krystal Taylor"],"categories":["math.CA"],"headline":"New theorem on projecting plane shapes with broad applications","summary":"Researchers proved a new theorem that applies to many types of projections—ways of mapping parts of the plane to lines or curves—under mild geometric conditions. This work extends earlier results that were limited to specific kinds of projections. Their findings also show that certain families of curves in the plane can closely cover any measurable set with almost no leftover space, illustrating practical uses of their theoretical results."},{"id":"2609.31601v1","title":"Exact Characterization of the Holevo Bound by a Quantum Fisher Information Family","abstract":"The quantum Cramér-Rao bound constrains the precision of parameter estimation through the symmetric logarithmic derivative quantum Fisher information (SLD QFI). It is asymptotically achievable for regular single-parameter estimation models, but generally not in the multiparameter setting, where optimal measurements for different parameters may be incompatible. For multiparameter estimation, allowing collective measurements, the corresponding asymptotically achievable precision limit is the Holevo bound, whose standard formulation is an optimization over auxiliary Hermitian operators. We bridge this apparent difference in formulation by giving an exact characterization of the Holevo bound in terms of a family of QFIs interpolating between the SLD and right logarithmic derivative (RLD) QFIs. Specifically, for any locally identifiable finite-dimensional estimation problem, we prove that the operator-feasible region in the standard variational definition of the Holevo bound coincides with the intersection of the regions defined by the corresponding Cramér-Rao-type matrix constraints for the entire QFI family. This characterization also applies to rank-deficient states and does not require any prior choice of weight matrix. Consequently, the conventional weight-dependent Holevo bound is recovered by minimizing the corresponding weighted cost over the resulting common feasible region.","published":"2026-09-25T17:54:15Z","updated":"2026-09-25T17:54:15Z","authors":["Koji Yamaguchi","Hiroyasu Tajima"],"categories":["quant-ph","cond-mat.stat-mech","math-ph"],"headline":"Researchers Clarify a Complex Quantum Measurement Limit","summary":"The researchers address a challenge in measuring multiple quantum parameters simultaneously, where existing methods limit accuracy due to incompatible measurements. They provide an exact description of the Holevo bound, a known limit on measurement precision, by linking it to a family of quantum information measures. Their work applies broadly to quantum systems without needing certain prior assumptions. These findings help clarify theoretical limits but are based on mathematical characterizations rather than experimental results."},{"id":"2609.31600v1","title":"New LoRA Skills Should Read but Never Write","abstract":"Low-rank adapters (LoRA) make it cheap to fine-tune a large language model once per task, but combining several independently trained adapters into one model remains difficult: merging the updates in weight space causes interference, retraining on all task data is expensive, and routing between separate adapters gives up the goal of a single combined model. We trace the difficulty to two choices that every composition method makes implicitly. A LoRA update admits infinitely many equivalent factorizations; the choice among them is invisible while an adapter serves alone, but it determines what a learned interaction between adapters can see. A coupling between an old skill and a new one can likewise point in either direction, and the direction decides whether the old skills keep computing what they computed before. We introduce READ (Read-only Expansion of Adapter Deltas), which fixes both choices: each adapter is rewritten into a balanced canonical form that preserves its update exactly, and the coupling grows in one direction only, so a new skill can read the input subspaces of old skills but cannot write into their output subspaces. The only trainable object at each append is the new skill's row of the coupling matrix, and the composed update folds into the base weights with no inference cost, routing, or task-specific rules. We evaluate READ across four benchmark suites and two model families, adding skills one at a time. Across several families, READ improves every suite average over the strongest published baselines built from the same adapters---by more than twenty points on SuperGLUE and more than seven points on the domain suite---and nearly all complete addition sequences end above every direct baseline. Factor coordinates and coupling direction, which a lone adapter never exposes, are what decide whether composed skills survive.","published":"2026-09-25T17:54:01Z","updated":"2026-09-25T17:54:01Z","authors":["Zeyan Li","Panqi Yang","Qirong Guo","Shengda Zhuo","SIyuan Qiu","Hu Xu","Chun Li","Jianfeng Xu"],"categories":["cs.LG"],"headline":"A new method helps combine AI language skills without conflicts","summary":"Researchers studied how to combine multiple specialized updates (called adapters) to large language models without interference between them. They developed a method named READ that rewrites each adapter to avoid conflicts by allowing new skills to read but not alter old ones. Testing on several benchmarks and models, the researchers found READ improved performance notably compared to previous methods. Their findings highlight how preserving certain mathematical properties and controlling interaction direction between skills affect success, based on their experiments and not yet on real-world tasks."},{"id":"2609.31599v1","title":"General regularity obstructions for the arrival time equation","abstract":"We study the regularity of the arrival time function associated with mean curvature flow, formulated as a degenerate elliptic equation encoding the singular structure of the flow. We identify a systematic mechanism obstructing higher regularity, arising from higher-order asymptotic expansions near spherical singularities. We construct uncountably many low-regularity arrival time functions in all dimensions. In particular, we settle the question of smoothness in the planar case by showing that even for convex curve shortening flow the arrival time need not be $C^{25}$. Our companion paper proves that every such arrival time is $C^{24,α}$ for any $0<α<1$. Hence the obstruction at order $25$ gives the optimal regularity threshold. Our approach introduces new analytic ingredients, including a precise correspondence between elliptic asymptotic expansions and parabolic long-time asymptotics of the associated rescaled mean curvature flow, a complexification of the arrival time equation, and a method for prescribing higher-order asymptotic expansions.","published":"2026-09-25T17:53:50Z","updated":"2026-09-25T17:53:50Z","authors":["Yiqi Huang","Jingze Zhu"],"categories":["math.DG","math.AP"],"headline":"Researchers identify limits to smoothness in a geometric flow equation","summary":"The researchers examined a mathematical function called the arrival time function, which describes how shapes evolve under mean curvature flow—a process related to smoothing shapes over time. They discovered a fundamental reason why this function cannot be perfectly smooth at certain points, especially near spherical singularities, and demonstrated that in two dimensions the function's smoothness has a precise limit. Their findings rely on complex mathematical analysis and provide sharp bounds on how regular these functions can be, based on simulations and theoretical work."},{"id":"2609.31598v1","title":"Scalable Hybrid Device Architecture on Thin-Film Lithium Tantalate for Long Distance Quantum Network Nodes with Atomic Frequency Comb Quantum Memories","abstract":"Scalable photonic circuits are essential for realizing practical quantum networks beyond proof-of-principle demonstrations. Although diamond-based integrated photonics has been extensively investigated over the past two decades for quantum networking applications, its integration into scalable photonic platforms remains challenging. Quantum memories based on rare-earth-ion atomic frequency combs (AFCs) offer an attractive alternative; however, most demonstrations to date have relied on bulk free-space crystals, limiting their scalability for practical implementations. In this work, we systematically investigate the potential for hybrid integration of AFC-based quantum memories with a scalable integrated photonic platform. We propose a device architecture combining thin-film lithium tantalate (TFLT) nanophotonic circuits with ytterbium-doped yttrium orthosilicate (Yb:YSO) crystals for long-distance quantum communication. Our analysis shows that the hybrid TFLT-YSO platform offers a promising and scalable pathway toward integrated quantum memories and, ultimately, large-scale quantum networks.","published":"2026-09-25T17:53:11Z","updated":"2026-09-25T17:53:11Z","authors":["Ayed Sayem","Towsif Taher"],"categories":["quant-ph","physics.optics"],"headline":"New Hybrid Device Design Aims to Improve Quantum Network Nodes","summary":"Researchers explored combining tiny photonic circuits made from lithium tantalate films with special crystals doped with ytterbium to build scalable quantum memory devices. Current quantum memories often use large, hard-to-scale crystals, so this approach could help integrate memories into compact circuits. Their analysis suggests this hybrid design might support building larger quantum networks, though the findings are based on theoretical investigation and device design rather than experimental demonstration."},{"id":"2609.31597v1","title":"Galois groups of integer polynomials with prescribed coefficients","abstract":"Fix a degree $n$, a nonzero integer constant term, and some further coefficients of a monic integer polynomial, leaving $s$ coefficients to vary in $[-H,H]$. We first prove that, when only the leading and constant coefficients are fixed, the number of polynomials whose Galois group is not $S_n$ is of order $H^{n-2}$ for every $n\\ge4$. We then prove a half-dimensional extension: for $n\\ge6$ the corresponding exceptional count is of order $H^{s-1}$ whenever $s>n/2+1$, provided the coefficient slice meets the higher-multiplicity loci with a specified dimension bound. This geometric condition holds for every prescribed value if the free coefficients form an initial or final block, and for a nonempty Zariski-open set of prescribed values for arbitrary coefficient positions. It also holds for every value of one additional prescribed interior coefficient. Thus, in degrees seven and eight, any one interior coefficient can be fixed in addition to the leading and constant coefficients. The proof combines a fixed-norm version of Bhargava's primitive-group sieve with sparse Hermite interpolation. A separate imprimitive estimate, using a minimal intermediate field, gives $O(H^{β(n)})$ outside rational partial-product hypersurfaces, where $β(n)=\\ell+n/\\ell-2\\le n/2$ for composite $n$ with smallest prime divisor $\\ell$, and has no logarithmic loss. The fixed-constant quartic and quintic endpoints require separate local arguments: exceptional Fourier directions for $A_4$ quartics and a quadratic Gauss sum for quintics. We isolate further even-degree endpoint results, explain the distinct cubic obstruction, and conclude with conditional improvements under the upper-bound form of Malle's conjecture.","published":"2026-09-25T17:53:07Z","updated":"2026-09-25T17:53:07Z","authors":["Evan M. O'Dorney"],"categories":["math.NT"],"headline":"Studying special symmetry groups in certain integer polynomials","summary":"The researchers examined families of integer polynomials where some coefficients are fixed and others vary within a range. They focused on how often these polynomials have a particular type of symmetry group, called the full symmetric group, versus more exceptional groups. Their theoretical work provides estimates on how many polynomials fail to have the full symmetric group under various fixed-coefficient conditions, using advanced mathematical tools and geometric conditions. These results are mathematical and conditional on certain assumptions, without direct practical applications."},{"id":"2609.31596v1","title":"Oracle Distillation","abstract":"Learning and sensing are among the most promising applications of quantum technology, often with provable quantum advantages. In any realistic experiment, however, noise threatens to erase these advantages as the problem size grows. We introduce oracle distillation, a procedure that distills many queries to a noisy oracle into a single high-fidelity oracle. Here the oracle, an unknown unitary from a known family, models the interface through which a quantum device learns about an unknown system. We construct an explicit protocol that distills any Boolean oracle, both under adversarial noise that corrupts a constant fraction of the input qubits and under i.i.d. depolarizing noise at a constant rate on every qubit. At the heart of the protocol is the weak query, in which the device deliberately queries the oracle only weakly, trading response strength for the ability to correct errors. Our protocol is near-optimal among all protocols that distill the oracle while keeping its errors correctable, revealing a fundamental tension between these two requirements. An important implication of the protocol is a threshold theorem: every Boolean oracle problem whose quantum advantage is polynomial in the domain size, including Grover search, $k$-forrelation, and Simon's problem, retains a quantum advantage whenever the per-qubit depolarizing rate is below a constant threshold. In particular, the quadratic quantum advantage of Grover search, long believed fragile under noise, survives nearly intact at low error rates even when almost every query involves errors on one or more qubits. We further extend the protocol to fractional and continuous-time Boolean oracles. Quantum error correction can thus protect not only prescribed operations but also unknown dynamics. Our results make a large family of quantum advantages in learning robust against noise and open a path toward robust computational sensing.","published":"2026-09-25T17:52:52Z","updated":"2026-09-25T17:52:52Z","authors":["Ruohan Shen","Soonwon Choi"],"categories":["quant-ph"],"headline":"New Method Helps Quantum Learning Resist Noise","summary":"Researchers developed a method called oracle distillation to reduce errors when quantum devices query noisy systems. They created a protocol that combines many noisy queries into one accurate query, even under significant noise, allowing certain quantum algorithms to maintain their advantage. Their findings suggest some quantum advantages, like in Grover search, can survive noise better than previously thought, but results are theoretical and rely on specific noise models."},{"id":"2609.31595v1","title":"GraphWrit3R: End-to-End 3D Scene Graph Writing","abstract":"3D scene graphs provide a structured representation of complex environments by encoding objects, their semantic attributes, and the spatial and functional relationships between them. Current approaches for 3D scene graph generation suffer from several fundamental limitations. They rely on complex multi-stage pipelines with explicit intermediate representations, making systems fragile and prone to error propagation. They assume access to ground-truth object annotations during inference, which deviates from real-world scenarios. They depend on proprietary models, hindering open-source deployment, or incur prohibitively slow inference. We present GraphWrit3R, a simple end-to-end method that takes a 3D point cloud, Gaussian Splats, or a combination of both as input, and directly outputs a complete scene graph as a structured JSON script. The graph lists all objects, their semantic attributes, and the relationships between them, while avoiding all of the above mentioned limitations. The choice of multiple input modalities is purely for versatility, allowing a single set of weights to handle diverse scenarios. Point cloud inputs are encoded via Sonata and Gaussian Splat inputs via Chorus, with both modalities projected onto a shared voxel grid and fused through a novel per-voxel contrastive alignment loss before being decoded by a large language model. As a natural consequence of the LLM, GraphWrit3R also supports open-vocabulary querying. On the 3DSSG benchmark, our method achieves state-of-the-art performance on object class, predicate, and triplet recall, outperforming methods that rely on ground-truth object annotations during inference. We further provide qualitative results and analyze different input modality configurations, contrastive loss formulations, and token fusion strategies.","published":"2026-09-25T17:52:39Z","updated":"2026-09-25T17:52:39Z","authors":["Luka Milivojevic","Nikola Popovic","Sayan Deb Sarkar","Sebastian Koch","Iro Armeni","Luc Van Gool","Danda Pani Paudel"],"categories":["cs.CV"],"headline":"New Method Creates 3D Scene Graphs from Raw Data Easily","summary":"Researchers developed GraphWrit3R, a simpler way to create detailed 3D scene graphs from raw 3D data without needing complex steps or extra ground-truth information. Their approach uses a combination of input types, processes data with novel techniques, and outputs complete scene descriptions as structured scripts. Tests show this method performs better than existing ones on benchmark tasks, but the findings are based on specific datasets and simulations described in their research."},{"id":"2609.31594v1","title":"From Source Code to Network Profile: Automated and Traceable MUD Profile Generation for IoT Devices","abstract":"The Manufacturer Usage Description (MUD) standard allows IoT manufacturers to define expected network behaviors in a MUD file. This file can be translated into enforceable access-control policies, restricting compromised devices to operate solely through manufacturer-defined communication patterns. However, practical adoption of MUD depends on profiles that are accurate, complete, and maintainable. Existing approaches use traffic-based automation but require device deployment and prolonged monitoring, capturing only behavior exercised during observation. Rare, failure-triggered, or configuration-dependent communications may remain absent, producing incomplete policies that disrupt legitimate operation and offer limited insight into the software components responsible for each rule. We present AutoMUD, a source-code-driven tool that generates traceable MUD profiles for IoT devices from their firmware and software source code. AutoMUD combines static and syntactic extraction, retrieval-grounded language-model reasoning, and deterministic validation and compilation to recover the communication behavior characterizing an IoT device and translate eligible endpoints into policy rules. By analyzing code-level evidence, AutoMUD exposes rarely exercised and conditional communication paths, links every generated rule to its source-level provenance, and preserves excluded findings with explicit reasons for review. Our evaluation on a Linux-based repository demonstrates that AutoMUD recovers complete communication behavior, consolidates validated behavior into semantic endpoint groups, and generates structurally valid MUD profiles. Through a controlled semantic fault-injection campaign, we demonstrate that AutoMUD enables analysts to detect, localize, explain, and correct propagated errors, recovering policies semantically identical to their clean counterparts.","published":"2026-09-25T17:52:09Z","updated":"2026-09-25T17:52:09Z","authors":["Alessandro Lotto","Abdulla R. A. Almenhali","Savio Sciancalepore","Alessandro Brighente","Mauro Conti"],"categories":["cs.CR","cs.SE"],"headline":"Automated Network Profiles from IoT Device Code Improve Security","summary":"Researchers developed AutoMUD, a tool that creates detailed network behavior profiles for IoT devices by analyzing their source code instead of just monitoring traffic. This approach helps uncover rare or conditional network actions often missed by traditional methods, linking each network rule directly to its code origin for clarity. Their tests on Linux-based devices showed AutoMUD produces complete, accurate, and reviewable profiles, potentially helping analysts find and fix policy errors more effectively. The findings are based on code analysis and simulation, not real-world deployment."},{"id":"2609.31593v1","title":"LAT: a Latinized aperiodic tiling for any sample size","abstract":"Quasi-Monte Carlo methods allow computer experiments to be run with far fewer simulations than crude Monte Carlo. A digital net in base two, such as Sobol', is however only balanced when the number of samples is a power of two, and Latin Hypercube Sampling (LHS) accepts any sample size but only controls the one-dimensional margins. This work proposes a space- filling design defined for any sample size, referred to as LAT for Latinized aperiodic tiling. The unit hypercube is cut recursively across its longest edge following the golden section into N cells of equal volume. One point is then placed in each cell, and the margins are made Latin while every point stays inside its own cell. The construction only uses integer splits and costs O(dNlog N). The recursion is shown to follow the Fibonacci word and its aperiodicity is analysed. LAT is assessed with four L2-discrepancies and with the integration error on analytical functions and engineering emulators, and compared to Monte Carlo, LHS, Halton, Sobol' and a rank-1 lattice. LAT is better than Monte Carlo and LHS as soon as the integrand has interactions. Sobol' remains more accurate at the powers of two, but its error is one to two orders of magnitude larger at other sample sizes. The accuracy of LAT does not depend on the sample size. Finally, the cells form a partition of the hypercube for any N. This allows one to refine the design locally, to search for an optimum by splitting cells and to sample non-rectangular regions.","published":"2026-09-25T17:51:55Z","updated":"2026-09-25T17:51:55Z","authors":["Pamphile T. Roy"],"categories":["stat.ME"],"headline":"New sampling method improves computer experiment accuracy for any sample size","summary":"Researchers developed a new way to pick sample points for computer experiments called LAT, which works for any number of samples, unlike some existing methods that require specific sample sizes. LAT divides the space into equal parts using a special pattern and places points accordingly to better capture interactions in the data. Their tests show LAT performs better than some traditional methods when sample size isn't fixed, though another method is still best at certain fixed sizes. The method also allows flexible refinement of the sampling design."},{"id":"2609.31592v1","title":"Completely monotone negative powers of hyperbolic polynomials","abstract":"Scott and Sokal asked whether every homogeneous polynomial with the half-plane property has a completely monotone negative power. We show that it does, and prove the positivity conjectures of Michałek--Sturmfels--Uhler--Zwiernik and Kozhasov--Michałek--Sturmfels. For a homogeneous hyperbolic polynomial in $n\\ge2$ variables, positive on its hyperbolicity cone, every power $p^{-α}$ with $α\\ge4096n^2$ is completely monotone, independently of the degree and coefficients of $p$. When the cone has pointed closure, its Riesz kernel is strictly log-concave in the interior of the dual cone, with relative Gaussian comparison error at most $512n^2/α$ and explicit Hessian bounds. We also prove that positive exponents of complete monotonicity can accumulate at zero only if $p$ is a product of real linear forms. The specialized V'amos quartic provides an example with no definite determinantal representation of any positive integer power.","published":"2026-09-25T17:48:39Z","updated":"2026-09-25T17:48:39Z","authors":["Dongsheng Wei"],"categories":["math.CO","math.AG","math.CA"],"headline":"Researchers study special powers of certain multidimensional polynomials","summary":"The researchers investigated a type of polynomial known as hyperbolic polynomials, focusing on whether raising these polynomials to negative powers produces a property called complete monotonicity. They showed that for these polynomials in multiple variables, negative powers with sufficiently large exponents do have this property. Their results confirm previous mathematical conjectures about positivity and provide detailed bounds related to these polynomials’ behavior. The findings come from theoretical proofs and clarify conditions under which these properties hold, without direct practical applications discussed."},{"id":"2609.31591v1","title":"Stacked Fluid Metasurfaces: Mutual-Coupling-Aware Modeling and Optimization","abstract":"Stacked intelligent metasurfaces process the transmitted field layer by layer, and fluid antennas make the position of every radiator a design variable. Combined, they pack radiators at sub-wavelength spacings within and across layers, where mutual coupling governs the physics that current models omit. This paper develops a coupling-consistent model of a fluid transmit layer illuminating a stack of passive fluid beyond-diagonal layers. The multiport impedance matrix of any port constellation is shown to be the impedance kernel sampled at the three-dimensional separations, in one closed spherical-Hankel form whose in-plane restriction is intra-layer coupling and whose axial restriction replaces the scalar Rayleigh--Sommerfeld propagator. Only the transmitter is driven: the layer currents are induced, and the familiar cascade is the single-pass limit of one matrix inversion. In the wavenumber domain the light circle still separates radiation from reaction, now with a plane-wave propagator attached, every Fourier mode sees a transmission line loaded by the layers, and an efficiency identity shows that what the stack costs in efficiency is set by its total induced-current norm, not by its layer count. Because the load is never assumed layer-diagonal, interconnections may join atoms on different sheets, and vertical networks repeated at every site are spectrally local. Alternating optimization designs precoders, positions, and loads with closed-form gradients. Numerically, fluid ports on a $λ/8$ grid perform within a few per cent of continuously movable ones, while a design based on the conventional cascade model can do worse than open-circuiting the stack.","published":"2026-09-25T17:45:27Z","updated":"2026-09-25T17:45:27Z","authors":["Giovanni Iacovelli","Chandan Kumar Sheemar","Symeon Chatzinotas"],"categories":["eess.SP"],"headline":"New model improves design of stacked fluid metasurfaces with mutual coupling","summary":"Researchers studied stacked layers of fluid antennas, where antenna positions can be adjusted and mutual electromagnetic effects are important. They developed a mathematical model that accurately accounts for interactions between antennas across and within layers, improving over previous models. Their approach helps optimize antenna arrangements and signal processing to better predict system performance, based on simulations and theoretical analysis rather than physical experiments."},{"id":"2609.31590v1","title":"AgentWorld: Benchmarking Long-Horizon Collaboration of Multi-agent LLMs","abstract":"Existing multi-agent benchmarks primarily test in competitive settings, short-horizon interactions under 20 steps, or simply aggregate individual performance, failing to isolate and highlight genuine collaboration capabilities of LLM-based agents. We introduce AgentWorld, a benchmark of 100 human-annotated tasks (with 100 augmented variants) for evaluating long-horizon, multi-agent collaboration. Tasks span 50+ interaction rounds across a rich MMORPG sandbox and require 3-20 agents with asymmetric roles and abilities to coordinate through communication, joint planning, and resource sharing under a blackbox setting where each agent acts independently without access to others' internal states. To quantify collaboration effectiveness in addition to conventional binary task success, we propose Causal Collaboration Effectiveness (CCE), a graph-based metric that traces causal dependencies between agent actions and measures what fraction of a team's effort actually contributed to the outcome. Experiments with Gemini 3 Flash, Claude Haiku 4.5, GPT-5 Mini, and DeepSeek R1-70B show that even the best model achieves only 52.0% task success, with systematic failure modes including communication breakdowns, role confusion, and inability to maintain shared plans across rounds. AgentWorld is fully open-source.","published":"2026-09-25T17:44:03Z","updated":"2026-09-25T17:44:03Z","authors":["Raphael Shu","Yusen Zhang","Young Min Cho","Jin Mo Yang","Yuan Yuan","Wenliang Zheng","Sharath Chandra Guntuku","Lyle Ungar","Zhou Yu","Rui Zhang"],"categories":["cs.MA"],"headline":"New Test Measures Teamwork in AI Agents Over Long Tasks","summary":"Researchers created AgentWorld, a new test to evaluate how well multiple AI agents work together over many steps in complex tasks. Unlike earlier tests, it focuses on long-term collaboration among 3 to 20 agents who must share roles and communicate without knowing each other's internal thoughts. They also introduced a new metric, Causal Collaboration Effectiveness, to better understand each agent's contribution. Trials show even top AI models struggle with teamwork issues like poor communication and confusing roles. The test is open-source for further research."},{"id":"2609.31589v1","title":"Common-Mode Collapse and Recovery in Direct Feedback Alignment","abstract":"Direct feedback alignment (DFA) trains hidden layers through fixed random projections of output error. With tanh hidden units and independent sigmoid outputs, plain stochastic gradient descent can stall near the loss of a constant predictor of class frequencies. We trace this stall to the error's common mode, the component shared across inputs. An exact mean-covariance decomposition separates a rank-one update formed by the mean teaching signal and mean presynaptic activity. Its leading component drives tanh units toward saturation. At initialization, random feedback provides no systematic correction of the shared error on average; readout learning limits its duration. A reduced model initialized from the network, without fitted parameters, predicts the concentration of activation sensitivity across 48 settings. On MNIST, class decodability largely survives collapse, but readout learning remains slow at a fixed learning rate. Adam learns faster despite deeper collapse. Calibrating the baseline readout to the class prior suppresses collapse and speeds learning; weaker feedback trades less collapse for slower learning. Replacing errors by their signs sustains collapse; subtracting the signal's batch mean prevents sustained collapse and improves learning in the tested setting. Related effects occur in deeper and convolutional networks and on CIFAR-10, with severity and cost depending on the readout, optimizer and input statistics.","published":"2026-09-25T17:43:40Z","updated":"2026-09-25T17:43:40Z","authors":["Varun Reddy","Bernardo L. Sabatini","Houman Safaai"],"categories":["cs.LG","cs.NE"],"headline":"Researchers explore learning stalls in a neural network training method","summary":"The researchers studied a training method called direct feedback alignment (DFA) used in neural networks, which can experience stalls in learning due to a shared error component called the common mode. They analyzed how this causes neurons to saturate and slow training, and developed a simplified model to predict this effect. Their tests on datasets like MNIST showed that while some learning remains, training speed is reduced. They explored adjustments to reduce this stall, noting effects vary with model and training setup. The findings come from analysis and experiments, with some limitations to specific conditions."},{"id":"2609.31588v1","title":"RePlay: Retrieval-Based Voice Playback for Multi-Turn spoken dialogue","abstract":"Many voice interaction applications require exact control over both the content and delivery of responses, typically using pre-recorded lines. Recent full-duplex models respond with low latency but cannot guarantee exact content or reproduce a specific recorded performance, while cascaded systems can be constrained to predefined responses at the cost of additional latency. We propose RePlay, a spoken dialogue system adapted from PersonaPlex that handles multi-turn conversations by retrieving and playing pre-recorded lines. Using probing, we identify the layer and frame at which the upcoming response becomes recoverable, and use this hidden state as the retrieval query. RePlay retains only the layers up to that point and replaces text and speech generation with lightweight turn-taking and retrieval heads. In simulated multi-turn interviews, RePlay reaches a median latency of 383 ms, 3 to 7 times lower than ASR-LLM cascades of comparable dialogue quality, at the cost of lower exact-line accuracy. In a user study, participants preferred RePlay in 63% of ratings versus 12% for a fast cascade with a small LLM (p = 0.008), and showed a non-significant preference (46% vs. 21%) over a slower cascade with a stronger LLM.","published":"2026-09-25T17:43:21Z","updated":"2026-09-25T17:43:21Z","authors":["Sathvik Udupa","Naveen Kumar","Ryan Folmsbee"],"categories":["eess.AS","cs.SD"],"headline":"New system plays pre-recorded voice replies faster in conversations","summary":"Researchers developed RePlay, a voice dialogue system that retrieves and plays exact pre-recorded lines to manage multi-turn conversations. By identifying when a response can be predicted in the system's process, RePlay quickly selects the right line without generating new speech, reducing delay significantly compared to other methods. Tests showed RePlay is faster but less precise at matching exact lines, and users generally preferred it over some alternative systems, though the findings come from simulated chats and a limited user study."},{"id":"2609.31587v1","title":"Compact Documentation for Coding Agents: A Benchmark, an Optimizer, and Why It Does Not Transfer","abstract":"We investigate whether natural-language documentation helps coding agents resolve software issues, and we build the tools to construct and evaluate it. We introduce a roundtrip benchmark that scores code descriptions by whether code regenerated from them passes the original tests, and show that completeness, not length, drives a description's fidelity. Using the benchmark as an optimization signal, we discover a description-writing prompt that reaches full fidelity and generalizes to unseen files. We then test the hypothesis that motivated the work: that better documentation helps an agent resolve real repository issues. Across two model families and ten repositories, and against a positive control confirming that our evaluation can detect a genuine improvement, we find that it does not. When the source is present, neither static compact documentation nor retrieved context beats the issue alone. We report this negative result together with the benchmark and the optimizer, and we characterize the boundary at which documentation helps.","published":"2026-09-25T17:42:22Z","updated":"2026-09-25T17:42:22Z","authors":["Md Shohel Arman","Igor Molybog"],"categories":["cs.SE","cs.AI","cs.CL"],"headline":"Studying How Short Documentation Affects Coding Problem Solving","summary":"Researchers explored whether brief, natural-language documentation helps computer coding agents fix software problems. They created a test to measure how well code descriptions allow regenerating code that passes original tests, finding that thoroughness matters more than length. Using this test, they developed a prompt that produces accurate descriptions for new files. However, when applied to real software issues, improved documentation did not help coding agents perform better. The researchers share these findings, tools, and insights on when documentation might be useful."},{"id":"2609.31586v1","title":"Trust Guided Decision Transformer","abstract":"Decision Transformer performance degrades on long rollouts because the conditioning context drifts out of the training distribution. We show that this drift is visible through the model's own next state prediction error, which rises during rollout and stays elevated, giving a direct signal of when context has become unreliable. We introduce Trust Guided Decision Transformer (TGDT), which selects context before applying value guidance. At each step, TGDT evaluates several recent context suffixes using rolling next state prediction error, calibrated against held out offline data via split conformal prediction. It keeps only suffixes whose error stays within the calibrated threshold, then uses a frozen critic to choose the highest value action among the trusted suffixes. This reverses the order used by value only elastic selection, where the critic may choose an action generated from a context the model itself has flagged as unreliable. Experiments on D4RL navigation and locomotion tasks show that state prediction, critic guidance, and hard context reset each solve only part of the problem. TGDT reduces persistent high error runs and improves return over vanilla Decision Transformer, reset based context control, and value only context selection.","published":"2026-09-25T17:42:19Z","updated":"2026-09-25T17:42:19Z","authors":["Chainesh Gautam","Raghuram Bharadwaj Diddigi","Chandramouli Kamanchi","Pankaj Dayama","Sumanta Mukherjee","Kameshwaran Sampath"],"categories":["cs.LG"],"headline":"Improving AI Decision-Making by Trusting Reliable Context","summary":"The researchers studied why Decision Transformer models perform worse during long tasks, finding that their predictions become unreliable over time. They developed Trust Guided Decision Transformer (TGDT), which checks how reliable recent information is before making decisions, using its own prediction errors and a safety threshold. TGDT then chooses the best action from trusted contexts. Tests in navigation and movement tasks show TGDT reduces errors and improves performance compared to previous methods. The findings come from experiments and highlight the importance of context reliability in AI decision-making."},{"id":"2609.31585v1","title":"Transition Radiation Detector Upgrade for the GlueX-III Charmonium Program","abstract":"The GlueX-III experiment at Jefferson Lab, planned to run in 2027-2029, will enhance the laboratory's capability to explore charmonium production near threshold and related aspects of QCD in the non-perturbative regime. A central upgrade for GlueX-III is a large-scale triple-GEM Transition Radiation Detector (TRD) optimized for electron identification and pion suppression in the hadron-rich environment of fixed-target photoproduction. This document describes the motivation for and efforts related to such an upgrade, including the design, construction, and performance of a 720x528mm$^{2}$ triple-GEM-TRD prototype that was tested in the existing GlueX-II experimental acceptance during the 2025 run. Results validate the prototype's integration, timing performance, and electron identification capability. Its projected impact in the GlueX-III setting is discussed, with background reduction in a $J/ψ$ candidate sample demonstrated. The development and use of a modern TRD technology in a high-background environment at Jefferson Lab may inform detector design choices as a future upgrade path for the EIC.","published":"2026-09-25T17:41:03Z","updated":"2026-09-25T17:41:03Z","authors":["Lauren Kasper","Alexander Austregesilo","Cody Dickover","Sean Dobbs","Sergey Furletov","Yulia Furletova","Ilya Larin","Lubomir Pentchev","Saheli Rakshit","Nizar Septian","Simon Taylor","Julia Velkovska"],"categories":["physics.ins-det","nucl-ex"],"headline":"Upgrading a Detector to Find Particles More Clearly at Jefferson Lab","summary":"Researchers working on the GlueX-III experiment are improving a detector to better identify electrons and reduce interference from other particles during charmonium studies. They built and tested a new large detector prototype in 2025, confirming it works well for timing and distinguishing electrons. The team expects this upgrade will reduce background noise and improve data quality in upcoming experiments. Findings are based on prototype tests, and the work may guide future detector designs in similar research environments."},{"id":"2609.31584v1","title":"Numerical Simulation of Electrical Properties in Cortical and Trabecular Bone: A Simplified Model","abstract":"The electrical properties of biological tissues depend on their composition and microstructure and determine their response to applied electric fields. In bone tissue, these properties are closely related to its composition, porosity, and microstructural organization. However, simplified models that link bone microstructure (trabecular and cortical) to its effective electrical properties while accounting for the electrical anisotropy of cortical bone remain scarce. In this work, a multiscale approach was developed, combining finite element simulations with Bruggeman-based homogenization models. Trabecular bone was modeled using microCT derived geometries, and its effective conductivity was analyzed as a function of BV/TV and free water content. Cortical bone was represented by a simplified three-dimensional network of Haversian canals (axial = z) and Volkmann canals (transverse), incorporating their directional organization through an anisotropic Bruggeman formulation. The models captured the conductivity range reported experimentally for trabecular bone (approximately 25-200 mS/m) and reproduced the anisotropic behavior observed in the cortical FEM simulations, with higher conductivity in the axial direction. In the generated cortical geometries, 65.5% of the total porosity was associated with axial Haversian canals. At the macroscale, the tibia model with lower trabecular and cortical BV/TV and reduced cortical thickness showed an approximately 38-39% reduction in impedance relative to the reference model. These results indicate that microstructural variations in bone tissue are reflected in its effective electrical properties and in the macroscopic impedance of the model, providing a physical and computational basis for future investigations into the sensitivity of electrical measurements to changes in bone microstructure.","published":"2026-09-25T17:41:00Z","updated":"2026-09-25T17:41:00Z","authors":["María José Cervantes","Catalina A. Cely-Ortíz","C. Manuel Carlevaro","Ramiro M. Irastorza"],"categories":["physics.med-ph","math.NA"],"headline":"Researchers Model How Bone Structure Affects Its Electrical Properties","summary":"The researchers studied how the tiny structures inside bone tissue influence its electrical behavior. Using computer simulations and simplified 3D models, they examined the electrical conductivity of two bone types—trabecular (spongy) and cortical (dense)—and found their models matched known experimental ranges. Their work shows that changes in bone microstructure can affect overall electrical impedance, offering a computational tool to explore how electrical measurements might detect such changes in bone tissue."},{"id":"2609.31583v1","title":"A unified sensitivity kernel for coherent wave tomography in radar, seismic, ultrasound, and optical imaging","abstract":"Imaging an inhomogeneous medium with waves rests on the sensitivity of a boundary measurement to a change inside. Radar, seismology, medical ultrasound, and diffuse optics cast this sensitivity as a kernel of area-specific form. Here we show that all of them are built from two classes. The classes stem from the complex material parameter of the underlying wave equation, its real part determining the wave speed and its imaginary part the loss. Every kernel is the product of a forward and an adjoint field, and a real measurement retains one quadrature of that product. These quadratures are the two classes, and all kernels in ballistic and diffusive transport are combinations of them. We place the areas on one diagram of scattering strength and transport depth, exposing a shared computational limit set by wavefield storage, and propose a broadband on-shell compression that can relax that limit.","published":"2026-09-25T17:40:47Z","updated":"2026-09-25T17:40:47Z","authors":["Alexey Yamilov"],"categories":["physics.optics"],"headline":"Researchers unify wave-based imaging methods with a common sensitivity model","summary":"The researchers study how different imaging techniques like radar, ultrasound, and seismic imaging detect changes inside materials by measuring waves at their boundaries. They found that all these methods use sensitivity kernels made from two basic types related to wave speed and loss. Their work reveals a shared computational challenge in these imaging methods and suggests a new way to reduce this problem using broadband compression. The findings are theoretical and focus on improving understanding and computation in wave imaging."},{"id":"2609.31582v1","title":"Free semigroups of power series","abstract":"Given power series $f_1,\\dots,f_r\\in z^2K[[z]]$ over a field of characteristic zero, we show that the semigroup $\\langle f_1,\\dots,f_r\\rangle$ generated by the $f$'s under composition is free of rank $r$ whenever the multiplicative semigroup $\\langle\\mathrm{lc}(f_1),\\dots,\\mathrm{lc}(f_r)\\rangle$ in $K^\\times$ is free commutative of rank $r$; here $\\mathrm{lc}(f)$ denotes the first nonzero coefficient of $f$. To do this, we associate an affine linear transformation to every $F\\in\\langle f_1,\\dots,f_r\\rangle$ and then apply a ping-pong lemma. In particular, combining this with work of Pappalardi, Sha, Shparlinski, and Stewart, it follows that most semigroups of polynomials over a number field $K$ are free, and we make this statement precise through a counting argument.","published":"2026-09-25T17:40:23Z","updated":"2026-09-25T17:40:23Z","authors":["Wade Hindes"],"categories":["math.NT"],"headline":"Researchers study special groups formed by combining power series","summary":"The researchers explore a mathematical problem about combining certain infinite sums called power series in a particular way to form structures known as semigroups. They show that under specific conditions related to the first nonzero term of these series, the resulting structure is 'free,' meaning its elements combine without unexpected relations. They use a method involving transformations and a known mathematical tool called the ping-pong lemma. Their findings, supported by earlier work, suggest that most groups formed from polynomials over some number systems have this 'free' property."},{"id":"2609.31581v1","title":"Emergent frustrated magnetism in strain-patterned graphene","abstract":"Geometrically frustrated magnetism conventionally arises from pre-existing magnetic moments on lattices whose geometry prevents their interactions from being simultaneously satisfied, giving rise to highly degenerate states and rich collective behavior. Creating such frustration in an intrinsically non-magnetic material presents a fundamentally different challenge, requiring both the magnetism and the competing interactions to emerge from correlated electrons. Here we show that this can be realized in graphene through lithographically programmable strain engineering. Patterning strain and the associated pseudo-magnetic field (PMF) into superlattices creates a correlated electronic system with flat bands and strong interactions. Transport measurements reveal interaction-driven insulating behavior, anisotropic magnetic hysteresis, and slow relaxation dynamics reminiscent of spin freezing, qualitatively captured by Monte Carlo simulations of competing magnetic moments on the PMF-defined ruby superlattice. Cryogenic magnetic force microscopy further reveals magnetic textures associated with the PMF landscape. These observations demonstrate frustrated magnetism emerging from correlated electrons in otherwise non-magnetic graphene. Our results establish lithographic strain engineering as a general and scalable route to flat bands and correlated states in van der Waals materials, providing a versatile platform for programmable quantum matter.","published":"2026-09-25T17:40:09Z","updated":"2026-09-25T17:40:09Z","authors":["Yu-Chiang Hsieh","Wen-Han Kao","Christophe De Beule","Sheng-Zhu Ho","Ru-Long Gou","Bo-Nian Chen","Kuan-Yu Chou","Kuo-En Chang","Chin-Chia Chang","Ying-Mei Yang","Ching-Hua Kao","Hao-Chien Chiang","Jyun-Lin Chen","Sheng-Chin Ho","Kenji Watanabe","Takashi Taniguchi","Ming-Hao Liu","Ching-Hao Chang","Yi-Chun Chen","Ying-Jer Kao","Tse-Ming Chen"],"categories":["cond-mat.mes-hall","cond-mat.str-el"],"headline":"Strain in Graphene Creates Unusual Magnetic Behaviors","summary":"Researchers studied graphene, a material that is usually non-magnetic, and used finely controlled stretching patterns to create magnetic-like effects. By applying these strain patterns, they induced competing interactions among electrons that led to unusual insulating states and magnetic behaviors similar to those seen in naturally magnetic materials with frustrated magnetism. Their experiments and simulations support the idea that such magnetic properties can emerge from electron interactions in graphene, suggesting new ways to design quantum materials through mechanical strain."},{"id":"2609.31580v1","title":"Inclination of Miranda as a Constraint on the Tidal Dissipation of Uranus","abstract":"The five major moons of Uranus record a complicated geological history characterized by both tectonic resurfacing and heavy impact cratering, connecting their evolution to dynamical processes in the Solar System. Miranda, the smallest and closest major moon to Uranus, has an unusually large orbital inclination ($4.3^\\circ$) and an remarkable surface geology dominated by coronae and tectonic structures. In this paper we explore whether Miranda's inclination and geological history can be explained through past orbital resonances within the Uranian satellite system. Using direct $N$-body integrations with the code SIMPL, we explore the tidal and dynamical evolution of the five major moons while assuming a fast tidal dissipation of Uranus and varying the internal responses of the satellites. Our simulations confirm that Miranda's inclination can be excited through resonant interactions involving Ariel and Umbriel, even when Miranda mean motion is not involved in that resonance. The Ariel-Umbriel 5:3 mean-motion resonance appears to be the most recent major dynamical event capable of driving Miranda's inclination to its current value. We compute the tidal heating associated with these resonance encounters and show that, for a strongly dissipative Uranus ($Q_U\\sim600$), the resonance could cause heat fluxes in Miranda approaching those required to form its coronae. These results could provide new constraints on Uranus's tidal dissipation.","published":"2026-09-25T17:39:46Z","updated":"2026-09-25T17:39:46Z","authors":["Maryame El Moutamid","Matija Ćuk"],"categories":["astro-ph.EP"],"headline":"Miranda's Tilt Sheds Light on How Uranus Absorbs Orbital Energy","summary":"Researchers studied why Uranus's moon Miranda has an unusual orbit tilt and strange surface features. They used computer simulations to see how interactions with other moons might have changed Miranda's orbit and caused internal heating. Their findings suggest these moon interactions, especially between Ariel and Umbriel, could explain Miranda's tilt and surface changes, offering clues about how Uranus absorbs energy from its moons' movements. However, the results depend on assumptions about Uranus's internal energy loss and are based on models, not direct observations."},{"id":"2609.31579v1","title":"Optimal stability hierarchies of the Heisenberg Uncertainty Principle for solenoidal fields and of the second order Caffarelli--Kohn--Nirenberg inequalities","abstract":"In 2018, V. Maz'ya proposed 75 open problems in analysis and PDEs. One of them is about the best constant for the Heisenberg Uncertainty Principle for solenoidal vector fields motivated by questions in hydrodynamics. This was answered by Cazacu, Flynn and Lam in dimension two and subsequently by Hamamoto in all dimensions $N\\ge3$ by establishing sharp inequalities with explicit constants. Nevertheless, the much harder stability problem remains open in all dimensions. The first main result in this paper is to establish a sharp stability hierarchy by proving that its deficit controls the distance to the set of extremals, with the sharp stability constant $\\frac12(N-\\sqrt{N^2-4N+12})$ for $N\\ge4$ and $1$ for $N=3$, together with chains of remainder terms measuring the distance to explicit larger families of poloidal and toroidal fields. The second main result is the second order $L^2$-Caffarelli--Kohn--Nirenberg inequality with the weights $|x|^{-2a}$ and $|x|^{-2b}$ on the line $1+a+b=0$, for which we obtain the stability constant $\\frac12\\min\\{4(1+a),\\sqrt{(N+2a)^2+4N-4}-(N+2a)\\}$. This CKN inequality is sharp whenever the second number is the smaller one. We develop a novel method: the fourth order one-dimensional problems attached to the spherical modes are transformed, by a Fourier--Hankel transform of real order, into first order inequalities whose sharp stability is that of weighted Gaussian Poincaré inequalities. The inverse transform produces the extremals in terms of Kummer's function. As a byproduct of our new approach, we also obtain substantially simpler proofs, with sharp remainder terms and equality cases, of Hamamoto's one-dimensional inequality and of his sharp uncertainty principle for solenoidal fields.","published":"2026-09-25T17:39:05Z","updated":"2026-09-25T17:39:05Z","authors":["Anh Do","Tuan Duong","Nguyen Lam","Guozhen Lu","Van Hoang Nguyen"],"categories":["math.AP","math.CA"],"headline":"Researchers improve understanding of stability in uncertainty and inequality problems","summary":"The researchers studied open problems involving precise mathematical inequalities related to fluid flow and uncertainty principles, originally proposed by others. They focused on quantifying how close certain vector fields are to ideal cases by finding sharp stability constants and remainder terms in multiple dimensions. Using a new method involving special transforms and functions, they also simplified earlier proofs. Their results support stronger stability estimates for these complex inequalities, but the work is theoretical and involves advanced mathematical tools."},{"id":"2609.31578v1","title":"Algorithmic trading and stochastic integration","abstract":"We study simple predictable processes whose coefficients are represented by neural networks. On finite measure spaces, we establish density results for neural networks in Orlicz spaces. For filtrations generated by a stochastic process, measurable random variables, including at stopping times, can be approximated by neural networks depending on finitely many observations. Every stochastic integral with respect to a semimartingale can then be approximated, in the semimartingale topology, by integrals of such simple predictable processes. We show that restricting trading strategies to this class leaves the minimal mean-variance hedging error under partial information unchanged and obtain a no-free-lunch characterization in terms of equivalent martingale measures. Finally, the Bichteler-Dellacherie characterization of semimartingales remains valid even upon restricting the predictable integrands to those whose coefficients are represented by neural networks.","published":"2026-09-25T17:38:40Z","updated":"2026-09-25T17:38:40Z","authors":["Aleksandar Arandjelovic","Uwe Schmock"],"categories":["math.PR","q-fin.MF"],"headline":"Using Neural Networks to Model Trading Strategies in Finance","summary":"The researchers explored how neural networks can represent certain financial trading strategies within a mathematical framework called stochastic integration. They showed that neural networks can approximate complex trading strategies and random financial variables based on limited observations. Their work, based on advanced probability theories, suggests that using neural networks in this way does not worsen the performance of trading strategies under incomplete information. However, these findings are theoretical and rely on mathematical properties without practical testing detailed in the abstract."},{"id":"2609.31577v1","title":"Generate, Track, Improve: Perceptive Multi-Skill Humanoid Locomotion with RL-Fine-Tuned Motion Generators","abstract":"General purpose humanoids require locomotion controllers that are multi-skill, perceptive, dynamic, and robust enough to go anywhere humans can. In this work, we present a two layer locomotion architecture: (1) a perceptive flow matching motion generator plans whole body trajectories from raw depth images while a (2) perceptive tracking policy trained with control-guided RL follows these motions. Both policies are trained on a library of terrain consistent motion clips created with dynamically optimized human data which yields both accurate velocity tracking and terrain consistent references. Our central contribution is a simple yet effective off-policy RL fine tuning loop that improves the motion generator. A structured search method is used with the generator to gather data for advantage weighted regression. This off-policy loop is much more sample efficient than on-policy residual fine tuning and improves terrain consistency on unseen geometries and skill compositions. We find that successful terrain traversals increased by up to 25 percentage points and skill selection improved by up to 80 percentage points. By using raw depth images to perceive the environment no odometry or height maps are needed, and outdoor deployment is easy. With two cameras, the policy can see terrain coming from further away and adjust its velocity regardless of the commanded speed so it can traverse the terrain. A single policy pair enables a Unitree G1 humanoid to walk, run, stand, jump on and off of boxes, and traverse stairs in outdoor environments. Project page: https://zolkin1.github.io/generate-track-improve/","published":"2026-09-25T17:38:23Z","updated":"2026-09-25T17:38:23Z","authors":["Zachary Olkin","William D. Compton","Aaron D. Ames"],"categories":["cs.RO"],"headline":"New Two-Layer System Helps Robots Walk Better Outdoors","summary":"Researchers developed a two-part control system for humanoid robots that uses raw depth camera images to plan and follow whole-body movements over varied terrain. Their method fine-tunes the motion planning using a sample-efficient learning process, improving the robot's ability to handle unseen surfaces and different skills like walking, running, and jumping. The study shows increased success in terrain traversal and skill selection, but the results are based on controlled tests and simulations using a Unitree G1 robot and may require further validation in diverse real-world settings."},{"id":"2609.31576v1","title":"Bootstrapping Holographic Theories in the Planar Limit","abstract":"We combine the numerical conformal bootstrap with supersymmetric localization to constrain the spectrum of holographic superconformal field theories in the planar large $N$ limit for finite 't Hooft coupling $λ$. We consider the stress tensor multiplet four-point function in $\\mathcal{N}=4$ $SU(N)$ super-Yang-Mills (SYM), which is holographically dual to closed string scattering, as well as the flavor multiplet correlator in a certain \\( 4d \\) $\\mathcal{N}=2$ $U\\!Sp(2N)$ gauge theory with $SO(8)$ flavor symmetry, which is dual to open string scattering. In both cases, we combine dispersive sum rules with integrated constraints from supersymmetric localization to compute an upper bound on the scaling dimension of the lowest scalar single-trace operator for all $λ$. For $\\mathcal{N}=4$ SYM, we find that our bounds are close to the known integrability result for all $λ$. For the $\\mathcal{N}=2$ theory, where integrability is not yet available, we find that our bounds are close to weak and strong coupling predictions in the relevant regimes.","published":"2026-09-25T17:37:09Z","updated":"2026-09-25T17:37:09Z","authors":["Shai M. Chester","Daniele R. Pavarini","Alessandro Piazza"],"categories":["hep-th"],"headline":"New Method Limits Key Properties in Theoretical Holographic Models","summary":"Researchers studied complex quantum field theories that relate to string theory using a combination of numerical and mathematical techniques. They focused on understanding specific energy levels (scaling dimensions) of certain particle-like excitations under different conditions. Their results provide bounds that match known solutions in some cases and suggest accurate estimates where exact solutions are not yet available. These findings are based on theoretical calculations and provide constraints rather than exact values."},{"id":"2609.31575v1","title":"Configuration, Not Conscience: A Large-Scale Empirical Study of LLM System Prompts","abstract":"Leaked system prompts are often treated as windows into the hidden values of commercial language models, yet their composition is rarely studied at scale. We analyze a merged corpus of 407 leaked, reconstructed, or officially published system prompts from 62 vendors across four community collections, identifying 29 near-duplicate clusters covering 66 files. Operational content rather than ethical statements dominates the corpus; a deliberately simple block-level classifier assigns roughly 58\\% of classified words to tool/protocol and roughly 5\\% to safety policy, while the strictest rule-lines guard tool use and file safety over harmful content by an 11:1 margin. Literal text transfer concentrates in a small set of cross-vendor pairs. Prompts also carry measurable maintenance debt, with version chains turning over thousands of words per release. The evidence supports treating leaked prompts as operational specifications, closer to configuration files than value statements, and treats reuse and prompt rot as engineering and supply-chain concerns. Because most documents are adversarial in origin and the detectors are deliberately simple, all magnitudes are directional; we audit the main classifier's error modes.","published":"2026-09-25T17:36:51Z","updated":"2026-09-25T17:36:51Z","authors":["Constantinos Patsakis","Vasilios Argyropoulos","Efthymios Alepis"],"categories":["cs.CR"],"headline":"Study Shows Language Model Prompts Act Like Configurations, Not Values","summary":"Researchers examined 407 system prompts from 62 language model vendors to understand what these prompts reveal. They found that prompts mainly contain operational instructions rather than ethical guidelines, suggesting these prompts are more like configuration files than statements of the models' values. The study highlights that prompts undergo frequent changes and reuse, indicating maintenance and engineering challenges rather than insight into the models' intentions. The results come from analyzing text data and involve some uncertainty due to the methods used."},{"id":"2609.31574v1","title":"Polychromatic 2-colorings with Bounded Discrepancy for Triangulations","abstract":"A polychromatic $2$-coloring of a triangulation is a $2$-coloring of the vertices such that no face is monochromatic. The discrepancy of a coloring is the maximum difference between the sizes of the color classes. Asayama and Matsumoto (Graphs and Combinatorics, 2022) proved that every triangulation admits a polychromatic $2$-coloring with discrepancy at most $\\tfrac{5n-16}{9}$, and that there exists a class of triangulations for which every polychromatic $2$-coloring has discrepancy at least $\\tfrac{n}{3} - 2$, where $n$ is the number of vertices. We improve the upper bound, showing that every triangulation admits a polychromatic $2$-coloring with discrepancy at most $\\tfrac{3n-16}{7}$ and such a $2$-coloring can be computed in quadratic time. We also show a discrepancy of at most $n-\\tfrac{4M}{3}$ for triangulations with a matching of size $M$. This implies, for example, that Delaunay triangulations admit a discrepancy of at most $\\tfrac{n}{3}$. We provide a linear-time algorithm to compute a $2$-coloring whose discrepancy is at most $\\tfrac{5n-24}{7}$. One of our results shows that any proper four coloring with the largest color class of size $\\frac{n}{2}$ would imply a $2$-coloring with discrepancy at most $\\frac{n}{3}$. The existence of such a proper coloring has been recently confirmed by Kawarabayashi, Yoneda, and Yoneda (arXiv 2026). Therefore the two results together confirm the discrepancy of at most $\\frac{n}{3}$ for triangulations.","published":"2026-09-25T17:35:22Z","updated":"2026-09-25T17:35:22Z","authors":["Alma Arevalo Loyola","Ahmad Biniaz","Prosenjit Bose","Thomas Shermer"],"categories":["cs.CG"],"headline":"Improving Two-Color Patterns in Triangulated Shapes","summary":"Researchers studied how to color vertices of a triangulated shape using two colors so that no triangle has all one color, while keeping the color groups fairly balanced. They improved previous mathematical limits on how unbalanced this coloring can be and developed efficient methods to find such colorings. Their findings, supported by recent related work, confirm that it's possible to keep the color difference within about one-third of the points, advancing understanding of coloring in geometric graphs."},{"id":"2609.31573v1","title":"How Far Can INRs Go? Cross-Domain Parameter-efficient INR-Based Semantic Segmentation for Brain MRI","abstract":"Biomedical image segmentation is central to medical image analysis, but practical deployment often faces limited annotations, memory constraints, and cross-site distribution shifts. Implicit Neural Representations (INRs) have recently emerged as a lightweight alternative for semantic segmentation, achieving competitive performance with substantially fewer parameters than conventional architectures. However, the mechanisms, scaling behavior, and domain generalization abilities of INR-based segmentation remain insufficiently understood. In this work, we study these questions in the context of cross-domain brain MRI segmentation. We analyze INR-based segmentation across low-parameter regimes, comparing it with conventional pipelines in both in-domain and out-of-domain settings. Surprisingly, we find that INR-based models do not simply improve with increasing parameter budget. Their advantage is most pronounced under low-parameter and limited-augmentation settings, while U-Net-based models benefit more from larger capacity and standard augmentation. We also investigate how INRs encode semantic information in their hidden features and show that complementary segmentation-relevant structure is distributed across multiple INR layers. Building on this insight, we introduce HierINRSeg, a hierarchical INR-based architecture that aggregates multi-layer representations for improved robustness and generalization. Extensive experiments show that HierINRSeg consistently outperforms MetaSeg, a strong recent INR-based segmentation baseline, with an average improvement of 5.6 percentage points in Dice for the in-domain test set and 8.2 percentage points out-of-domain. Overall, our analysis identifies the conditions under which INR-based segmentation is most effective, providing concrete guidance for model selection and future research.","published":"2026-09-25T17:35:07Z","updated":"2026-09-25T17:35:07Z","authors":["Ziyao Shang","Pouya Sadeghi","Letian Jiang","Alexander Wong","Sirisha Rambhatla"],"categories":["cs.CV"],"headline":"New neural method improves brain MRI segmentation across different settings","summary":"Researchers studied a lightweight neural network approach called Implicit Neural Representations (INRs) for segmenting brain MRI images, which is important in medical analysis. They compared INR models to conventional methods and found INRs work best with fewer parameters and limited data changes, while traditional models improve more with bigger size and standard data tweaks. They introduced a new hierarchical INR model, HierINRSeg, that combines layers for better results and generalization. Their tests show this model performs better than previous INR methods both within the same data type and across different data sources, offering practical insights for future medical image analysis."},{"id":"2609.31572v1","title":"OC-GS: Gaussian Splatting for Irregular Turntable Capture","abstract":"Uneven rotation and dropped frames make equal-angle assumptions unreliable for turntable reconstruction. We present OC-GS, an object-centric Gaussian splatting that refines each image's angle while maintaining a shared camera, rotation axis, and pivot. This orbit-consistent refinement jointly optimizes image-derived geometry and angles to reconstruct objects from sparse, irregular captures. On rendered objects with 12, 8, and 6 irregularly spaced views, OC-GS achieves mean foreground PSNR scores of 21.26, 19.36, and 15.83dB, respectively, exceeding all four evaluated pose-free Gaussian splatting baselines in each condition. Under a shared trainer, refining image-estimated angles improves mean foreground PSNR by 7.88dB over keeping those estimates fixed. An ablation study shows that both image-derived angle initialization and the shared motion model contribute to the improvement. On real captures, OC-GS's refinement increases mean foreground PSNR by 0.70dB. Results show that refining uncertain angles within a shared motion model improves reconstruction from sparse, irregular turntable captures.","published":"2026-09-25T17:35:05Z","updated":"2026-09-25T17:35:05Z","authors":["Jae Joong Lee","Bedrich Benes"],"categories":["cs.CV","cs.AI"],"headline":"Improving 3D Object Images from Uneven Turntable Photos","summary":"Researchers address problems in creating 3D images when objects are photographed on turntables that rotate unevenly or drop frames, making angle guesses unreliable. They developed OC-GS, a method that adjusts the estimated angles of photos while considering the shared camera and rotation setup to better reconstruct objects from irregular, limited photos. Their tests show that refining angles this way improves image quality compared to other methods, but the findings are based on simulations and some real captures with specific setups."},{"id":"2609.31571v1","title":"Strategically Diverse Sampling for Self-Training","abstract":"Many LLM training and inference methods, including RL and test-time scaling, depend on repeated sampling, but benefit only when the responses meaningfully differ. Self-training faces the same challenge: training data is typically constructed by sampling IID responses and filtering primarily for correctness, thereby overrepresenting strategies a model already favours. We investigate strategic diversity, or substantive variation among approaches to a problem, as an alternative principle for constructing self-training data. We generate strategically diverse data with two sampling methods: GROOT, a new method which constructs a hierarchical tree of approaches and samples distinct paths, and Verbalized Sampling (VS), adapted to produce an unstructured set of approaches. Across competitive programming and Next-Chapter Prediction domains, models trained on strategically sampled data outperform IID-trained counterparts on difficult tasks and provide strong initializations for RL and test-time scaling. Most strikingly, self-training on strategically diverse but incorrect traces from Qwen3-4B outperforms IID distillation from a 235B teacher. These results challenge prevailing assumptions about what makes useful self-training data and show that diversity of approaches can matter more than correctness or teacher scale.","published":"2026-09-25T17:35:05Z","updated":"2026-09-25T17:35:05Z","authors":["Alexander Gurung","Esmeralda S. Whitammer","Mirella Lapata"],"categories":["cs.CL"],"headline":"Researchers Improve AI Learning by Sampling Diverse Solutions","summary":"The researchers studied how training AI models using a variety of different solution strategies, rather than just correct or similar ones, can improve learning. They developed two new methods to generate diverse samples for self-training and found that models trained this way performed better on challenging tasks. Their findings suggest that having a range of approaches in training data may be more beneficial than only focusing on correctness or using larger teacher models, though these results come from their specific experiments."},{"id":"2609.31570v1","title":"Uncertainty and Explainability in Deep Rough Volatility: A Neural Information-Theoretic Posterior Approach","abstract":"Deep learning has substantially accelerated the calibration of complex stochastic-volatility models, but neural point calibration alone does not capture the uncertainty remaining after an implied-volatility (IV) surface has been observed. We develop a simulation-based inference framework for rough Heston (rHeston) calibration that learns the posterior distribution of the model parameters conditional on an IV surface. Using neural ratio estimation, we obtain calibrated posterior samples that can be propagated through heteroscedastic neural surrogate pricers for path-dependent exotic options. The resulting posterior-predictive distributions combine residual parameter uncertainty with conditional surrogate uncertainty and yield uncertainty-aware price intervals. We further introduce Hellinger-SHAP, an information-theoretic explainability method for posterior inference. Rather than attributing a single parameter point estimate, it applies local-background Kernel SHAP to a posterior-information functional measuring contraction from the prior to the posterior. This identifies maturity--moneyness regions associated with posterior information gain for individual rHeston parameters. In a simulation study, posterior-predictive intervals provide calibrated or conservative coverage across forward-start, barrier, and realized-variance claims, while point plug-in prices can be materially unreliable for selected contract regimes. Together, the UQ and XAI analyses provide a transparent framework for uncertainty-aware neural calibration and downstream exotic pricing under the specified prior-predictive model.","published":"2026-09-25T17:34:56Z","updated":"2026-09-25T17:34:56Z","authors":["Damiano Brigo","Raphaël Huser","Dan Leonte"],"categories":["stat.ML","cs.LG","stat.AP","stat.CO","stat.OT"],"headline":"New AI method measures uncertainty in complex financial models","summary":"Researchers developed a deep learning approach to better understand uncertainty in calibrating rough Heston models, which describe how market volatility changes over time. Their method estimates a full range of possible model parameters instead of a single best guess and uses this to produce more reliable predictions for pricing complex financial options. They also introduce a way to explain which parts of the market data most influence parameter estimates. Their results, based on simulations, suggest this approach gives more trustworthy uncertainty estimates than previous methods, but real-world validation is not provided."},{"id":"2609.31569v1","title":"Adapting for AI: How elementary teachers adjust their practices for an AI-integrated curriculum","abstract":"Conversational AI tools are entering children's everyday experiences, and schools are interested in adopting them. However, successful classroom integration depends not only on the technology but also on the work teachers do to make it usable and appropriate for their students and classroom context. There is little known about how elementary teachers work as they implement conversational AI tools in real classrooms. In this study, we examine three teachers' experiences implementing an AI literacy and English Language Arts (ELA) curriculum built around ToyTalk, a conversational AI toy development platform, over 13 instructional days, a three-week summer camp. Drawing on daily individual reflections, group reflections, and post-camp interviews, we find that teachers' adaptive practices of repair, differentiation, translation, and balancing sit at the intersection of three tensions (technology, learner, and instruction). Teachers' understanding of AI and their role evolved over the camp experiences. From these findings, we contribute design implications and considerations for deploying conversational AI within elementary classrooms.","published":"2026-09-25T17:33:47Z","updated":"2026-09-25T17:33:47Z","authors":["Fasika Melese","Ruiyang Wu","Xinyue Cui","Joanna Perkins","Xiaoyi Tian","Tiffany Barnes","Shiyan Jiang"],"categories":["cs.HC","cs.AI"],"headline":"How Elementary Teachers Adapt to Using Conversational AI in Class","summary":"Researchers studied three elementary teachers using a new AI-focused curriculum with a conversational AI toy over a three-week summer camp. They found that teachers actively adjusted their methods to make the AI tool suitable for their students, balancing challenges related to technology, student needs, and teaching goals. The teachers' understanding of AI and their teaching roles grew during this time. These observations led researchers to suggest ways to better design and support AI tools for young classrooms."},{"id":"2609.31568v1","title":"DeepEdu-v1: Efficient and Scalable Agentic LLMs for Vietnamese Education","abstract":"AI tutoring could markedly improve learning outcomes for students in developing regions such as Vietnam, yet the two obvious paths both fall short. Cloud assistants such as ChatGPT route sensitive student data to foreign servers---violating data-sovereignty laws such as Vietnam's Decree 53---and, pre-trained on Western-centric corpora, are not organized around the national textbook curriculum, so their knowledge of local content is unsystematic and frequently hallucinated. Self-hosting an open model keeps data on-premise but hits a two-fold wall: post-training quantization (AWQ, GPTQ) tames the static weight footprint, yet the dynamic KV cache and prefill latency of long tutoring contexts still cause out-of-memory failures and slow responses on consumer GPUs, while the model keeps hallucinating on region-specific material. We present DeepEdu-v1, an AI-tutoring system for Vietnamese education built on SCALE (Self-improving Context-Aware Learning Engine), a framework with two innovations. First, a long-context inference engine amortizes token selection from per-sub-chunk to per-cluster granularity; on long-context retrieval it issues x7.7 fewer retrieval calls than a state-of-the-art selective-attention baseline, cutting prefill latency (TTFT) by roughly 35% while matching or improving task accuracy. Second, a self-improving agentic layer continuously curates a verified playbook from past interactions instead of fine-tuning, a design intended to progressively reduce reliance on dominant-language priors as trustworthy local knowledge accumulates. In its deployed configuration, DeepEdu achieves a nearly x2 TTFT speedup over standard vLLM serving and lifts agentic accuracy from 70.0% to 79.5% on complex tasks, with the strongest per-track gains across financial-reasoning and interactive-agent benchmarks.","published":"2026-09-25T17:33:41Z","updated":"2026-09-25T17:33:41Z","authors":["Quang Nguyen","Hieu Nguyen","Hien Hoang","Toan Pham","Cong Tran","Nam Vu"],"categories":["cs.AI"],"headline":"New AI tutor improves speed and accuracy for Vietnamese students","summary":"Researchers developed DeepEdu-v1, an AI tutoring system designed specifically for Vietnamese education to address data privacy and local knowledge challenges. Their system uses a new engine that speeds up processing by reducing repeated data retrieval, and a self-improving layer that learns from past interactions to better handle Vietnam-specific material. Testing shows this approach cuts response time by about 35% and improves accuracy on complex tasks, although the results so far are based on simulations and specific benchmarks."},{"id":"2609.31567v1","title":"On Line-Hyperplane Structures for Hitchin Representations","abstract":"Let $n \\geq 4$ and $ρ: π_1S \\rightarrow PSL(n,\\mathbb{R})$ be a Hitchin representation. We study the topology of a cocompact domain of discontinuity $Ω_ρ$ in the flag manifold $\\mathcal{F}_{1,n-1}$ of line-hyperplane pairs in $\\mathbb{R}^n$ defined by Guichard-Wienhard. In particular, we lift $Ω$ to its $(\\mathbb{Z}_2\\times \\mathbb{Z}_2)$-cover $\\hatΩ$ in the Stiefel manifold $V_2(\\mathbb{R}^n)$. The domain $\\hatΩ$ is highly connected and smoothly fibers over hyperbolic space $\\mathbb{H}^2$ with unknown fiber $\\hat{\\mathfrak{F}}$, a closed $(2n-5)$-manifold. We determine the homeomorphism type of $\\hat{\\mathfrak{F}}$ as well as its diffeomorphism type up to connected sum with a homotopy sphere. The classification of $\\hat{\\mathfrak{F}}$ involves computing both its homology and a certain differential topology invariant needed to invoke Wall's classification of highly connected odd-dimensional manifolds.","published":"2026-09-25T17:32:10Z","updated":"2026-09-25T17:32:10Z","authors":["Parker Evans","Andrea Tamburelli"],"categories":["math.GT","math.DG"],"headline":"Researchers explore shapes linked to Hitchin representations in math","summary":"The researchers study a special geometric space related to Hitchin representations, focusing on its complex shape and structure. They examine a particular covering space that maps smoothly over a well-known curved surface, hyperbolic space, but whose fiber shape was previously unknown. By analyzing its holes and a topological property, they classify the fiber's overall shape up to certain mathematical equivalences. Their work is theoretical and involves advanced topology and geometry concepts."},{"id":"2609.31566v1","title":"Bandwidth, Latency, and 400 Million Kilometers: The Case for Mars-Local Compute","abstract":"There have been recent proposals for human settlements on Mars in 2030s. Any human activity on Mars must be preceded by extensive robotic exploration. However, Mars exploration is bottlenecked by the low bandwidth, intermittent Mars-Earth link. For example, HiRISE, a high-resolution camera onboard the Martian orbiter MRO imaged less than 3% of Mars over eleven years, even though MRO's low resolution Context Camera had mapped more than 99% of Mars in that time. We present a systems case for shared compute for Mars exploration. Such Mars-local compute, paired with advances in computer vision and AI, can enable large volumes of data to be collected and processed on Mars while sending periodic updates, insights, and selective datasets to Earth. To overcome the lack of surface infrastructure on Mars, we propose a two-tier in-orbit deployment of computational satellites that provides consistent coverage and bandwidth. Our analysis shows that the proposed deployment can start small: one areostationary node makes compute reachable from all active Mars missions, two additional areostationary nodes can extend this coverage to roughly 90% of the planet, while low-Mars-orbit nodes add high-rate surface links and compute capacity where demand grows.","published":"2026-09-25T17:32:05Z","updated":"2026-09-25T17:32:05Z","authors":["Maleeha Masood","Indranil Gupta","Deepak Vasisht"],"categories":["cs.DC","cs.NI"],"headline":"Improving Mars Exploration with Local Computing and Satellites","summary":"Researchers explain that exploring Mars is slowed by limited data links between Mars and Earth, which restricts how much can be studied remotely. They suggest using computers located on Mars itself, combined with artificial intelligence, to process large amounts of data locally and send only key information back to Earth. To support this, they propose placing computational satellites in Mars orbit to provide better communication and computing coverage for missions on the planet's surface. Their findings are based on system analysis and highlight how Mars exploration could be more efficient with these technologies."},{"id":"2609.31565v1","title":"A Comprehensive Study of the Magnetic White Dwarfs in DESI DR1","abstract":"As the end point of stellar evolution for most stars, white dwarfs offer powerful insight into stellar magnetism. Multiple formation channels are required to explain the observed properties of magnetic white dwarfs, and the mechanisms that produce unusual objects have potential to unlock new realms of physics. The large samples made possible by spectroscopic surveys enable study of these processes on an unprecedented level. Here we perform a detailed model atmosphere analysis of the magnetic white dwarfs in DESI Data Release 1. We identify 833 magnetic systems, 762 of which we fit using an offset dipole model. Of these, 127 targets require more in-depth analysis as they possess shallower or sharper absorption features than predicted by the offset dipole model. For these we use a patchy offset dipole atmosphere model in which the magnetic field only contributes to the line opacities at certain regions of the visible stellar disk. We compare the observed distribution of magnetic white dwarfs to the mechanisms that generate magnetic fields. We find that a core-convective dynamo that forms on the main-sequence matches the parameters of many of the magnetic systems, while a dynamo that fills the radiative interior on the giant branch reproduces many of the lower mass objects that the core-convective dynamo cannot. This sample is the largest single data release of magnetic white dwarfs to date. The diversity of formation channels and the frequency of objects with unusual field geometries highlight the power of multiplexed surveys to advance our knowledge of stellar magnetism.","published":"2026-09-25T17:29:53Z","updated":"2026-09-25T17:29:53Z","authors":["Adam Moss","Pierre Bergeron","Mukremin Kilic","Warren R. Brown","Jamie Tayar"],"categories":["astro-ph.SR"],"headline":"Study Analyzes Magnetic White Dwarfs to Understand Stellar Magnetism","summary":"White dwarfs are the final stage for most stars and help scientists study stellar magnetic fields. Researchers examined 833 magnetic white dwarfs from a large survey, using models to understand their magnetic properties. They found that different processes likely create these magnetic fields, with some matching one type of internal star activity and others matching a different stage of star evolution. This work is based on data analysis and helps explore how magnetic fields form in stars."},{"id":"2609.31564v1","title":"Weight Pair Encoding: Inducing a Smaller Grammar in Neural Network Weights","abstract":"We show that neural network weights can be explicilty fintuned to admit a smaller grammar. Weight Pair Encoding (WeightPE) does so by placing a lossy Re-Pair compressor inside a straight-through estimator. The int8 weights of the network are flattened into one string, and near-matching Re-Pair patterns are made exactly equal within a global L2 budget. The network computes with the rewritten weights and trains through them with a straight-through estimator. Unlike a flat codebook of fixed-size entries, a grammar offers variable-length patterns and reuses them hierarchically inside larger ones. On the MLP weights of ViT-B/16 and ViT-L/16 finetuned on CIFAR-10, WeightPE produces a Re-Pair grammar 0.43x and 0.38x the size of the one produced by an equivalent int8 QAT run, at a cost of 1.9 and 1.1 accuracy points. The trend extends to different grammar compressors (LZ78, SEQUITUR), over which the networks has not be finetuned against. To our knowledge, this is the first time grammar size has been used as an explicit training objective for network weights.","published":"2026-09-25T17:29:43Z","updated":"2026-09-25T17:29:43Z","authors":["Irene Tallini","Daniele Solombrino","Alberto Cazzaniga","Emanuele Rodolà"],"categories":["cs.LG"],"headline":"New Method Compresses Neural Network Weights Using Grammar Rules","summary":"Researchers developed a technique called Weight Pair Encoding (WeightPE) that fine-tunes neural network weights to create a smaller, compressed representation based on grammar patterns. They convert network weights into a sequence and use a method to find repeating patterns, simplifying the model while allowing it to keep working and training. Testing this on some image models showed smaller compression sizes but with some loss in accuracy. This work is the first to explicitly use grammar size as a training goal for network weights."},{"id":"2609.31563v1","title":"Multi-agent Scaling Across Disjunctive and Compensatory Tasks","abstract":"Multi-agent LLM systems are often expected to improve as team size increases, yet the scaling behavior may depend on task structure. Our central contribution is to introduce Steiner's taxonomy of group tasks as a framework for analyzing multi-agent LLM scaling and focusing the analysis on disjunctive and compensatory tasks. We model independently sampled agents as conditionally independent given the item, which yields their large-team limits: plurality voting converges to the model's modal answer, and averaging converges to the model's item-level bias. Across selected representative benchmarks, 13 open-weight models, and teams of up to 30 agents, we find qualitatively different scaling behavior. On disjunctive tasks, the probability that at least one agent is correct grows by 5-20 points with team size, but plurality voting over agents that answer directly realises almost none of this potential, as the model predicts to within 0.5 points on average. Multi-round revision raises accuracy considerably, yet the gain is nearly the same with one peer as with 29. In contrast, scaling provides little benefit on Fermi estimation, despite its natural suitability for aggregation: item-level biases shared across the samples of a model account for about 87% of the squared error, so averaging reduces error by only about 6%. Combining model families helps on Fermi estimation but does not surpass the strongest member on disjunctive tasks. These results show that task structure, together with the mechanism combining member outputs, is a fundamental determinant of team scaling.","published":"2026-09-25T17:29:15Z","updated":"2026-09-25T17:29:15Z","authors":["Carolina Fortuna","Blaz Bertalanic"],"categories":["cs.AI","cs.MA"],"headline":"How Team Size Affects AI Performance on Different Tasks","summary":"Researchers studied how groups of AI models improve their answers as the team grows, focusing on two types of tasks: disjunctive (where one correct answer suffices) and compensatory (like estimating quantities). They found that bigger teams help a lot on disjunctive tasks, but simple voting doesn’t capture this benefit fully. For estimation tasks, team size matters little because shared biases limit improvement. The study suggests that both task type and how member answers are combined impact how well larger AI teams perform."},{"id":"2609.31562v1","title":"Agentic Economies for Autonomous Scientific Discovery","abstract":"Recent advances in agentic Artificial Intelligence (AI) systems have marked a shift in AI for Science: moving away from the use of individual AI systems for narrow task execution, toward multi-agent systems capable of orchestrating complex, end-to-end research workflows and performing (semi-)autonomous scientific discovery. The development of multi-agent AI-for-science systems has primarily focused on improving the cognitive capabilities of AI systems, specifically by making advanced reasoning and hypothesis generation more reliable. However, focusing only on cognitive capability improvement could ignore appropriate management of resources, a key bottleneck in scientific discovery. Testing and validating scientific hypotheses and experiments is, physically and economically, resource-intensive and resources are limited. This means that to make significant advancements in autonomous scientific discovery, such as improving human-AI co-scientist complementarity or reaching a truly closed-loop automated process, we must pair ongoing improvements in reasoning capabilities with robust resource management. In this paper, we outline an infrastructure for AI resource management by developing the necessary foundations of scientific agent economies, markets, and institutions. The aim of this infrastructure is to empower AI agents and human scientists to effectively (i) collaborate and establish research priorities, (ii) assign credit, (iii) track accountability and liability, and (iv) safeguard against malicious use and information security risks. Finally, we engage with the macro-level societal implications. of (semi-)autonomous scientific discovery to inform the development of governance policies ensuring an equitable distribution of AI-driven discoveries and derivative future technologies.","published":"2026-09-25T17:28:53Z","updated":"2026-09-25T17:28:53Z","authors":["Nenad Tomasev","Matija Franklin","Atoosa Kasirzadeh","Vivek Natarajan","Alan Karthikesalingam","Sebastien Krier","Simon Osindero"],"categories":["cs.CY"],"headline":"Using AI Agents to Improve Scientific Research Collaboration and Management","summary":"Researchers explore how multiple AI agents can work together with humans to conduct scientific research more effectively. They focus on managing limited resources, such as time and money, alongside improving AI reasoning skills. The paper proposes a system to help AI and human scientists set priorities, assign credit, ensure accountability, and address security risks, while also considering broader societal impacts and fair access to AI-driven discoveries. These ideas are theoretical and intended to guide future AI research governance."},{"id":"2609.31561v1","title":"Orthogonal Moments in Likelihood Models","abstract":"Many models, such as fixed-effect models for panel or network data, are hard to estimate because they feature nuisance parameters that are both numerous and estimated imprecisely. This, in general, causes an incidental-parameter problem in the estimator of the parameters of interest. The problem can be alleviated by working with an estimating equation whose expectation is insensitive to the value of the nuisance parameters. We discuss and contrast three notions of insensitivity, also called orthogonality, in the context of likelihood models: Neyman orthogonality, Neyman orthogonality to order q, and full orthogonality. Orthogonal moments are obtained by projecting the estimating equation on nested subspaces, which are spanned by, respectively, the scores of the nuisance parameters, the first q derivatives of the likelihood ratio with respect to the nuisance parameters, and all likelihood ratios of the model. We give explicit constructions in binary-choice, count-data, and nonlinear regression models.","published":"2026-09-25T17:28:27Z","updated":"2026-09-25T17:28:27Z","authors":["Stéphane Bonhomme","Koen Jochmans","Martin Weidner"],"categories":["econ.EM"],"headline":"New Method Helps in Complex Statistical Estimation Problems","summary":"Researchers studied challenges in estimating certain statistical models that include many hard-to-measure nuisance parameters, which can distort results. They explored three related ways to reduce this distortion, called orthogonality, by adjusting the estimation approach to lessen sensitivity to these nuisance parts. Their work provides explicit methods for improving estimation in models involving choices, counts, and nonlinear relationships. The findings are theoretical and focus on improving estimation equations within likelihood-based models."},{"id":"2609.31560v1","title":"Generalization behavior of OPTQ and the role of regularization","abstract":"Large neural networks can be compressed by rounding or \"quantizing\" their weights to numbers that admit representations with fewer bits. One algorithm for quantization, OPTQ, progressively quantizes the weights of a neural network so that the squared quantization error on a specified calibration dataset is as small as possible. We study the performance of OPTQ and a variant algorithm, stochastic OPTQ, in a generalization setting and derive bounds for the expected squared error accrued by the algorithm when a test point is drawn from a fixed distribution. We prove two results. One result relates the generalization error to the error on a calibration dataset comprising independent samples from the same distribution as the test distribution. The other result bounds the generalization error of stochastic OPTQ for all sufficiently nice distributions, regardless of the calibration dataset. In both of these results, the regularization term $λ$ plays an important role. We use insights from these results to make a new recommendation for the choice of $λ$ and see that this choice of $λ$ preforms favorably in experiments when compared to prior recommendations in the literature.","published":"2026-09-25T17:27:18Z","updated":"2026-09-25T17:27:18Z","authors":["Erin George","Rayan Saab"],"categories":["cs.LG"],"headline":"New insights on compressing neural networks with OPTQ algorithm","summary":"Researchers studied how the OPTQ method, which compresses large neural networks by rounding weights to fewer bits, behaves when making predictions on new data. They derived mathematical bounds showing how the algorithm's error on new tests relates to errors on a sample calibration dataset and how a variation called stochastic OPTQ performs under certain conditions. Their findings highlight the importance of a regularization parameter and suggest a new recommended setting that showed better experimental results compared to earlier advice."},{"id":"2609.31559v1","title":"Online Learning via Learned Latent Bayesian Tracking","abstract":"Online learning in non-stationary environments requires models to adapt rapidly from streaming data under strict computational constraints. A principled approach casts online learning as Bayesian state tracking, where model parameters are updated sequentially via Bayesian filtering. However, applying Bayesian filters directly to modern deep models is computationally prohibitive due to the high dimensionality of parameter space, forcing existing methods to rely on restrictive approximations or manually designed low-dimensional subspaces. In this work, we identify the absence of a suitable low-dimensional dynamical representation as the core bottleneck in Bayesian filtering-based online learning. Accordingly, we propose Adaptive Update through Representation Adaptation (AURA), a meta-learning framework that learns offline a low-dimensional latent state-space model governing the evolution of optimal model parameters under distribution shift. Online adaptation is then performed via extended Kalman filtering in this learned latent space followed by reconstruction of the full model parameters through a learned lifting map, enabling efficient single-step online adaptation while preserving model expressiveness. Evaluated on online adaptation of neural wireless receivers under time-varying channels and on non-stationary image classification, AURA shows substantial improvements in adaptation speed, accuracy, and computational efficiency over existing online learning and Bayesian filtering baselines, demonstrating that an adaptation-aware latent geometry is beneficial for effective Bayesian online learning in high-dimensional models.","published":"2026-09-25T17:27:14Z","updated":"2026-09-25T17:27:14Z","authors":["Guy Gerson","Tomer Raviv","Nir Shlezinger","Tirza Routtenberg","Osvaldo Simeone"],"categories":["cs.LG","eess.SP"],"headline":"New method helps AI adapt quickly with less computing power","summary":"Researchers studied how AI models can learn and adapt rapidly in changing environments using a statistical approach called Bayesian filtering. They created a framework named AURA that compresses complex model changes into a simpler form, allowing fast updates with lower computing cost. Testing AURA on tasks like wireless signal processing and image recognition, the researchers found it improved speed, accuracy, and efficiency compared to older methods, suggesting their approach better supports online learning in complex AI systems."},{"id":"2609.31558v1","title":"Region-Level Black-Box Defense Against Stealthy Embedding-Space Backdoors in CLIP","abstract":"Contrastive Language--Image Pretraining (CLIP) has emerged as a dominant vision backbone due to its strong transferability and zero-shot capabilities. However, recent studies reveal a critical vulnerability: embedding-space backdoor attacks. By poisoning only a tiny fraction of image--text pairs, adversaries can implant stealthy triggers that induce targeted shifts in CLIP's joint embedding space. Unlike conventional backdoors that manipulate classifier logits, these attacks corrupt representations directly, making them highly effective under extremely low poisoning ratios and difficult to detect. Existing defenses require access to model parameters, gradients, logits, or clean validation data---assumptions that rarely hold in realistic black-box deployments. Moreover, current black-box methods struggle to accurately localize small or out-of-distribution triggers. We propose CLIPGuard, a lightweight and fully black-box defense specifically designed to mitigate embedding-space backdoors in CLIP encoders. CLIPGuard identifies malicious regions by measuring segment-wise embedding perturbations and selectively purifies only suspicious segments via semantic inpainting, preserving benign visual content and alignment quality. Extensive experiments on STL-10, ImageNet, and diverse trigger families---including BadCLIP, BadNets, blended, patch-based, and typographic attacks---demonstrate that CLIPGuard reduces attack success rates to as low as 1.05% while maintaining clean accuracy up to 86.34%, consistently outperforming existing black-box defenses, including CleanCLIP and CleanerCLIP. Our code is available https://github.com/wsu-cyber-security-lab-ai/CLIPGuard.git","published":"2026-09-25T17:26:02Z","updated":"2026-09-25T17:26:02Z","authors":["Ahmed Abdelnaby","Mohamed Elmahallawy"],"categories":["cs.CV","cs.CR"],"headline":"New Method Defends CLIP AI from Hidden Backdoor Attacks","summary":"Researchers studied a security problem where tiny hidden triggers corrupt CLIP, a popular AI that links images and text, causing it to misinterpret data. They developed CLIPGuard, a black-box defense that detects and fixes suspicious image parts without needing access to the AI's internal details. Tests on various datasets and attack types showed CLIPGuard lowers attack success to about 1% while keeping accuracy high. These findings are based on experiments and do not guarantee protection in all real-world cases."},{"id":"2609.31557v1","title":"No Universal Remainder Rate for Chambolle-Dossal Acceleration","abstract":"Chambolle-Dossal acceleration guarantees $F(x_n)-F^*=o(n^{-2})$ for every fixed smooth convex loss with a minimizer. We show that this qualitative improvement admits no universal quantitative rate. For every damping parameter $α>3$ and positive nondecreasing gain $G(n)\\to\\infty$, we construct a fixed one-dimensional smooth convex loss whose exact CD orbit satisfies $\\sup_{n\\ge1} n^2G(n)\\bigl(F(x_n)-F^*\\bigr)=\\infty$. Thus no divergent gain improves the $n^{-2}$ scale for all fixed losses, even with instance-dependent constants. The construction prescribes queried gradients and realizes infinitely many slow blocks within one smooth convex objective. Under local $p$-power growth with $p>2$ and sufficiently strong damping, we also construct a fixed loss whose exact CD orbit satisfies $F(x_n)-F^*\\sim Dn^{-2p/(p-2)}$, $D>0$, establishing the sharpness of the known convergence rate. Both main results are formally verified in Lean 4.","published":"2026-09-25T17:25:07Z","updated":"2026-09-25T17:25:07Z","authors":["Yuchen Yang","Xinan Dai","Wenhao Deng","Yingdong Shi","Feng Xu","Tailin Wu"],"categories":["math.OC"],"headline":"Limits of Speed Improvements in Chambolle-Dossal Acceleration","summary":"Researchers studied a mathematical method called Chambolle-Dossal acceleration, used to improve optimization speed for certain smooth and convex problems. They found that although this method guarantees faster progress in theory, there is no universal way to improve the speed for all problems quantitatively. By creating specific example problems, they showed that attempts to boost the rate beyond a certain scale fail. Their conclusions are based on constructed examples and theoretical verification, with formal proofs completed using the Lean 4 system."},{"id":"2609.31556v1","title":"Factorisation and matrix amplification of real rank zero inclusions","abstract":"We show that real rank zero is not preserved under matrix amplifications of inclusions. This yields real rank zero inclusions that do not approximately factor through real rank zero C*-algebras, thereby resolving two questions of Gabe and Neagu.","published":"2026-09-25T17:24:56Z","updated":"2026-09-25T17:24:56Z","authors":["Jamie Bell"],"categories":["math.OA"],"headline":"Real rank zero not preserved in matrix amplifications of C*-algebras","summary":"The researchers studied a property called real rank zero in certain mathematical objects known as C*-algebras. They demonstrated that this property is not maintained when these objects are enlarged using a method called matrix amplification. This result answers two open questions posed by Gabe and Neagu about how these algebras can be broken down or factored approximately. The findings are theoretical and specific to the behavior of these algebraic structures under certain operations."},{"id":"2609.31555v1","title":"Boundary Value Correction for the Boundary Element Method","abstract":"We develop a boundary element method for the Laplace equation with Dirichlet boundary conditions in a bounded three-dimensional domain. The method uses flat panels of maximum diameter \\(h\\) on an interior surrogate boundary and corrects the imposed boundary values by a first-order Taylor expansion along the panel normals. The correction uses the normal flux already present in the Calderón formulation and introduces no additional unknown field. Under uniform assumptions on the surrogate geometry and boundary operators, the penalised form is coercive when \\(γδ_h\\le c_0<4\\), where \\(γ\\) is the penalty parameter and \\(δ_h\\) is the maximum normal offset. The error estimate separates trace approximation from a geometry term of size \\(γ^{1/2}δ_h^2\\|u\\|_{H^3(Ω)}\\). With \\(γ\\simeq h^{-1}\\), the sufficient condition \\(δ_h\\lesssim h^{(k+1)/2}\\) yields an error of order \\(h^{k+1/2}\\) for \\(P_k/P_{k-1}\\) Dirichlet/flux spaces under the stated regularity and uniform trace bounds. Thus an offset of order \\(h^2\\) supports the estimate through cubic Dirichlet approximation. We analyse two penalty variants and present numerical experiments on a sphere and an ellipsoid to examine convergence and sensitivity to geometry, penalty scaling and quadrature.","published":"2026-09-25T17:24:41Z","updated":"2026-09-25T17:24:41Z","authors":["Erik Burman","Peter Hansbo","Mats G. Larson"],"categories":["math.NA"],"headline":"Improving boundary calculations for 3D Laplace problems","summary":"Researchers developed a new method to better solve a mathematical problem called the Laplace equation in 3D spaces with fixed boundary values. They improved accuracy by adjusting boundary values using a first-order correction based on nearby surface information. Their approach introduces no extra unknowns and includes an error estimate, supported by computer tests on spheres and ellipsoids. The findings are based on theoretical analysis and simulations under specific geometric assumptions."},{"id":"2609.31554v1","title":"Exact Solutions of the Dunkl--Pauli Equation in the Presence of a Magnetic Field from an su(1,1) Algebraic Approach","abstract":"We investigate the two-dimensional Dunkl--Pauli equation for a spin--$1/2$ particle in the presence of an external magnetic field within an algebraic framework. While previous studies have provided exact solutions of the Dunkl--Pauli system, its underlying dynamical symmetry and algebraic structure remain largely unexplored. By employing Schrödinger factorization, we construct the generators of the $\\mathfrak{su}(1,1)$ algebra and establish the associated symmetry of the radial sector. The energy spectrum is derived using irreducible unitary representations, and the corresponding Sturmian radial basis is obtained analytically. Furthermore, $\\mathrm{SU}(1,1)$ coherent states are constructed and their time evolution is analyzed, revealing a characteristic radial breathing behavior. The results show that the Dunkl deformation introduces parity-dependent modifications in the spatial structure of the system while preserving its underlying algebraic dynamics.","published":"2026-09-25T17:24:34Z","updated":"2026-09-25T17:24:34Z","authors":["M. Salazar Ramírez","B. C. Lütfüoğlu","H. Bouguerne"],"categories":["quant-ph"],"headline":"Researchers Explore Magnetic Particle Behavior Using Algebraic Methods","summary":"The researchers studied a complex quantum equation describing a particle with spin in a magnetic field. They used advanced algebraic techniques to uncover hidden symmetries and calculated exact energy levels and wave patterns. Their work shows how certain mathematical modifications affect the system's spatial properties, while the underlying dynamics remain consistent. These findings come from theoretical analysis and provide deeper understanding of the equation's structure."},{"id":"2609.31553v1","title":"MexHat: A Dataset for Hate Speech Detection in Mexican Spanish Videos","abstract":"Ensuring online safety through content monitoring had raised Hate Speech Detection as a crucial task to be addressed. By essence the task demands the capture of contextual cues, which are essential for a precise understanding of the content's intent. Although automated detection approaches for the task have advanced significantly, the scarcity of non-English resources persists, limiting the ability of models to adapt to the subtle, context-dependent, and culturally related nature of multimodal content. In this paper, we introduce MexHat, a video dataset designed to capture the linguistic and cultural cues for the hate-speech detection task in a Mexican Spanish context. Our dataset comprises around 1k video clips annotated across two tasks: a three-way class evaluation (no negative content, offensive content and hate-speech content), and a fine-grained class evaluation including three hate-speech sub-categories. The dataset statistics and the baseline results highlight the inherent challenges associated with the task. Disclaimer: This paper contains sensitive content that may be disturbing to some readers.","published":"2026-09-25T17:23:36Z","updated":"2026-09-25T17:23:36Z","authors":["Itzel Tlelo-Coyotecatl","Hugo Jair Escalante"],"categories":["cs.CL","cs.CV"],"headline":"New Dataset Helps Detect Hate Speech in Mexican Spanish Videos","summary":"Researchers created MexHat, a collection of about 1,000 Mexican Spanish video clips labeled for hate speech and offensive content. This dataset aims to improve the detection of harmful speech by capturing cultural and language context, which existing tools often miss. The researchers highlight challenges in accurately identifying hate speech and provide initial test results using their dataset."},{"id":"2609.31552v1","title":"FragToken: Amplifying LLM Inference Costs through Noncanonical Token Generation","abstract":"As large language model (LLM) inference becomes increasingly expensive, resource-consumption attacks pose a growing threat to model providers. Existing attacks typically amplify cost by inducing abnormally long or repetitive outputs on attacker-controlled or triggered requests, making them easier to detect and limiting their deployment-wide impact when benign traffic dominates. In this work, we uncover a previously overlooked token-level attack surface arising from the many-to-one mapping from token sequences to decoded text. Although standard LLMs predominantly generate the canonical token sequences induced by their tokenizers, the same text can also be represented by substantially longer non-canonical sequences. This representational flexibility exposes a new avenue for resource-consumption attacks: an attacker can train the model to favor such sequences, systematically increasing the number of autoregressive decoding steps without a proportional increase in visible response length. However, we empirically find that directly maximizing token fragmentation substantially degrades model utility, producing conspicuous answer-quality failures that undermine attack stealthiness. To address this challenge, we propose FragToken, a training-time framework that combines source-model self-distillation, capacity-aware filtering and budgeting, and BPE-Aligned Merging to induce fragmented generation under ordinary prompts while largely preserving model utility. We evaluate FragToken on four LLMs across three benchmarks. Across the four models, FragToken achieves a three-benchmark average token inflation ratio (TIR) ranging from 1.99 to 2.46, while causing only minor degradation in model utility. Our work reveals a covert LLM supply-chain threat that increases inference cost without requiring large volumes of attack requests while largely preserving utility.","published":"2026-09-25T17:20:47Z","updated":"2026-09-25T17:20:47Z","authors":["Zihan Wang","Rui Zhang","Xinyuan Qian","Qingchuan Zhao","Hongwei Li","Guowen Xu"],"categories":["cs.CR"],"headline":"New Attack Increases AI Text Costs Using Hidden Token Tricks","summary":"Researchers studied a new way to make large language AI models use more computing power without obvious signs. They found that by training models to use longer, unusual internal token sequences for the same words, attackers can increase the work needed to generate text without making responses visibly longer. Their FragToken method induces this token inflation while mostly keeping the model's answer quality intact, showing a subtle cost-boosting threat that doesn't rely on many attack requests."},{"id":"2609.31551v1","title":"EAServe: Encode-Aware Disaggregated Serving for Multimodal Large Language Models","abstract":"Disaggregating the two stages, Prefill and Decode, onto separate GPU pools is now a standard optimization for (text-only) LLM serving. However, multimodal LLMs (MLLMs), which add a third phase, Encode, pose new challenges for resource allocation. Encode turns images, video, or audio into embeddings that the language model can consume, yielding a three-stage Encode-Prefill-Decode (EPD) pipeline. Existing frameworks offer only partial answers: text-only PD systems lack Encode, while EPD frameworks expose it as a separate service without regulating downstream request flow. The pipeline also carries a structural resource imbalance: every request enters through Encode before downstream work can begin, yet per-request execution leaves the encode GPU severely underutilized even at high loads, starving the downstream Prefill and Decode workers. Addressing this, we reposition Encode as the control point of the EPD pipeline, exposing three tightly coupled dimensions: when work enters downstream, where prefill executes, and how the GPU is shared. We instantiate this in EAServe across two co-designed layers. Its runtime manages load-adaptive micro-batching, rate-controlled partial offload to a co-resident prefill worker, and dynamic SM partitioning for predictable co-location. The configuration layer, Hybrid Auto Selection (HAS), navigates the joint space of GPU allocation, encode batch size, and offload ratio by pruning unbalanced allocations with per-stage capacity profiling and refining the remainder through TPE-based Bayesian optimization. Evaluated on three MLLM architectures spanning image, video, and audio, EAServe delivers up to 4.3x and 1.7x higher goodput than NVIDIA Dynamo and vLLM, respectively, under identical SLO constraints, sustains more balanced and higher GPU utilization across the EPD pipeline, and reaches near-optimal configurations faster than baseline search methods.","published":"2026-09-25T17:20:22Z","updated":"2026-09-25T17:20:22Z","authors":["Kunxiong Zhu","Zhihao Shu","Hangyu Zheng","Minghai Qin","Miao Yin","Gagan Agrawal","Wei Niu"],"categories":["cs.DC","cs.LG","cs.PF"],"headline":"New Method Boosts Efficiency in Multimodal Language Model Processing","summary":"Researchers studied how to better manage GPU resources when processing complex multimodal language models that handle images, video, and audio, which require a three-step pipeline: Encode, Prefill, and Decode. They developed EAServe, a system that controls how tasks move through these steps and allocates GPU resources dynamically. Their tests on different models showed EAServe improved processing output and balanced GPU use compared to existing methods. These results come from controlled experiments and focus on system optimization rather than real-world deployment outcomes."},{"id":"2609.31550v1","title":"Mind the second gap: The impact of gaps on noise and signal parameter inference","abstract":"Gravitational-wave data is never analysed without some conditioning: filtering and tapering act multiplicatively on the noise, destroying the stationarity on which the Whittle likelihood --- the backbone of nearly all gravitational-wave inference --- rests. The problem is particularly evident for LISA, where gaps are expected to overlap almost all signals and the noise must be estimated jointly with those astrophysical signals. We provide an analytical and numerical framework to assess the impact of noise mis-modelling when signal and noise parameters are inferred jointly in the presence of gaps. From a Fisher-matrix formalism verified against Bayesian inference, we obtain closed forms for the model uncertainties, their maximum-likelihood estimates and the true scatter in the latter, for various approximations to the likelihood. The Fisher-based linear formalism is extended through the Godambe--White curvature to the non-linear regime that severe mis-modelling produces. We survey gap families from long, well-tapered interruptions to short, high-rate ones, injecting massive black-hole binaries. Except in the case of frequent gaps, we show that the windowed covariance is well approximated by its leading diagonal --- the window convolved with the power spectral density --- for the estimation of both signal and noise parameters. By omitting the convolution, the signal sector remains unbiased but the noise parameters, in particular the optical-metrology noise which is dominant at high frequency, is displaced by over an order of magnitude in power. At the highest glitch rates the diagonal approximation stays accurate yet loses an order of magnitude in precision. To remedy this, we propose an exact time-domain likelihood of the surviving samples which can be efficiently (and accurately solved) via preconditioned conjugate gradient methods.","published":"2026-09-25T17:19:33Z","updated":"2026-09-25T17:19:33Z","authors":["Ollie Burke","Federico Pozzoli","Martina Muratore","Jonathan R. Gair"],"categories":["gr-qc","astro-ph.IM"],"headline":"New method improves analysis of gravitational-wave data with gaps","summary":"Researchers studied how gaps in gravitational-wave data affect the accuracy of measuring both the signals and the background noise. They developed mathematical tools to understand errors caused by ignoring the effects of these gaps, especially for the planned LISA detector where gaps overlap signals. Their findings show that common shortcuts can bias noise estimates significantly, but signal estimates remain reliable. To handle frequent or complex gaps, they propose a new, more exact analysis method that better accounts for missing data in the time domain."},{"id":"2609.31549v1","title":"2-colouring shift-chains","abstract":"A shift-chain is an $ r $-uniform hypergraph $ \\mathcal{H} $ on vertex set $ [n] $ with the property that, for any two edges $ \\{ e_1, \\ldots, e_r \\} $ and $ \\{ f_1, \\ldots, f_r \\} $ with $ e_1 < \\cdots < e_r $ and $ f_1 < \\cdots < f_r $, either $ e_i \\le f_i $ for all $ i \\in [r] $ or $ f_i \\le e_i $ for all $ i \\in [r] $. It is known that all shift-chains are properly vertex-colourable with three colours (that is, such that no edge is monochromatic), which is optimal for $ r \\in \\{ 2, 3 \\} $. It was asked by Pálvölgyi in 2010 whether all shift-chains of sufficiently large uniformity are properly $ 2 $-colourable. We answer this question in a strong form, proving that in fact all shift-chains of uniformity at least $ 4 $ are properly $ 2 $-colourable. The colouring is obtained via a natural algorithm with linear running time.","published":"2026-09-25T17:18:51Z","updated":"2026-09-25T17:18:51Z","authors":["Zak Smith"],"categories":["math.CO"],"headline":"Researchers prove 2-coloring works for complex shift-chains","summary":"The researchers studied shift-chains, a special kind of collection of sets with a specific ordered property. Previously, it was known that these structures could be colored using three colors to avoid a single-colored set, but it was unclear if two colors would suffice for larger cases. They showed that for shift-chains with sets of size four or more, it is always possible to properly color using just two colors. Their proof is theoretical and involves an efficient coloring algorithm."},{"id":"2609.31548v1","title":"On the infinite sum of reciprocals of the fourth powers of balancing numbers","abstract":"In this note, we study the infinite reciprocal sum $\\sum_{k=n}^{\\infty}1/B_k^4$ involving the fourth powers of balancing numbers $B_n$. We show that, for every $n\\geq2$, \\begin{equation*} \\left\\lfloor \\left( \\sum_{k=n}^{\\infty}\\frac{1}{B_k^4} \\right)^{-1} \\right\\rfloor = B_n^4-B_{n-1}^4 -\\left\\lceil\\frac{B_{2n-1}}{280}\\right\\rceil +\\varepsilon_n, \\end{equation*} where $\\varepsilon_n=1$ if $n\\equiv1\\pmod{12}$ and $\\varepsilon_n=0$ otherwise. This result extends the corresponding reciprocal-sum result for Fibonacci numbers due to Hwang, Park and Song to balancing numbers.","published":"2026-09-25T17:18:50Z","updated":"2026-09-25T17:18:50Z","authors":["Subhasis Panda","Aditya Kumar Dash","Utkal Keshari Dutta"],"categories":["math.NT"],"headline":"Researchers analyze a special infinite sum involving balancing numbers","summary":"The researchers studied an infinite sum made from the reciprocals of the fourth powers of balancing numbers, a specific sequence of numbers. They found a formula linking this sum to the balancing numbers themselves, extending previous work done on a similar sum involving Fibonacci numbers. Their findings are theoretical and based on mathematical analysis, offering new insight into these special number sequences."},{"id":"2609.31547v1","title":"Two Conformal Constructions for Adaptive Within-Document AI-Text Screening","abstract":"We study false-alert control when screening for text generated by artificial intelligence (AI). The screening procedure selects document prefixes and detectors from observed evidence and may stop before exhausting its inspection budget. We give two finite-sample constructions under document-level exchangeability between human calibration documents and a new null document, with no restriction on dependence among tokens within a document. Construction A registers a finite family of prefix-detector scores and allocates a false-alert budget across their conformal ranks. A union bound protects any executed subset of that family. Construction B calibrates the complete-path maximum of a development-fixed adaptive policy. Each partial-path maximum is bounded by the complete maximum, so a terminal conformal rank protects early stopping without splitting the error budget. We prove marginal control of any false alert across the permitted inspection path and derive necessary calibration counts for rejection. We also state oracle testing, distribution-shift, and independent-audit bounds with their additional assumptions. Both constructions protect stopping within their specified scope; neither proof constructs an e-process or justifies multiplying conformal ranks. Detection power and computational savings remain questions for empirical evaluation.","published":"2026-09-25T17:18:39Z","updated":"2026-09-25T17:18:39Z","authors":["Marco Mandap","Jerahmeel Hipolito","Arcel Galvez","Charlie Margaret Balagtas","Michael Joshua Buluran","Jeff Roel Durmiendo","Rizzette E. Lopez"],"categories":["stat.ME","cs.CL"],"headline":"New Methods to Control False Alerts in AI-Generated Text Screening","summary":"Researchers developed two approaches to reduce false alarms when detecting AI-generated text within documents. They focus on choosing parts of a document to check and when to stop inspecting without exceeding error limits. Their methods use mathematical guarantees under specific assumptions but have not yet been tested for detection accuracy or efficiency. The findings primarily offer theoretical control of errors, with practical performance needing further study."},{"id":"2609.31546v1","title":"BeatGraph: Self-Supervised Heartbeat Graphs for Infant ECG Representations from the Home Environment","abstract":"Electrocardiogram (ECG) foundation models typically tokenize the signal into fixed-length patches that ignore cardiac structure, so a patch may split a heartbeat and the number of beats in each patch shifts with heart rate. This matters most for infants, whose heart rates are higher and whose ECG differs from the adult, clinic-recorded 12-lead data these models are built on. A model for infant ECG should therefore reason about heartbeats directly rather than recover them from arbitrary patches. We propose BeatGraph, which makes the heartbeat its unit of representation, modeling each 30-second window as a graph of beats. A shared beat encoder embeds each heartbeat from its waveform and inter-beat intervals, a Transformer with positional encoding orders the beats in time, and residual graph attention layers relate every beat to every other before attention pooling yields a window embedding. We pretrain BeatGraph on our new corpus of unlabeled infant recordings by predicting masked-beat embeddings, then fine-tune it for each task. One backbone supports sleep-wake detection, infant-state classification, activity-source identification (infant- or caregiver-initiated movement), and affect recognition, improving macro-F1 over the strongest baseline on each task by 0.076 to 0.158. It also transfers across age groups, reaching 0.892 AUROC on the ZZU-pECG pediatric benchmark (ages 0 to 14), within 0.001 of the best published self-supervised ECG model, and matching that model under linear evaluation on the adult PTB-XL benchmark despite infant-only pretraining. Finally, to our knowledge, we release the first public infant ECG corpus collected in homes, classrooms, and laboratory settings with state and affect labels. It contains 3,408 hours of single-channel ECG from 143 infants aged 3 to 11 months, with unlabeled pretraining data, benchmark tasks, and subject-level splits.","published":"2026-09-25T17:18:20Z","updated":"2026-09-25T17:18:20Z","authors":["Mohammad Nur Hossain Khan","M. S. Krafczyk","Beverly G. Bolster","Nancy McElwain","Mark A. Hasegawa-Johnson","Bashima Islam"],"categories":["cs.LG"],"headline":"New Model Improves Infant Heartbeat Analysis from Home ECG Data","summary":"Researchers developed BeatGraph, a new method that analyzes infant heartbeats directly instead of using fixed signal segments. They trained it on a large collection of infant ECG recordings from homes and other settings without labels, then fine-tuned it to detect sleep, activity, and emotions. BeatGraph showed better performance than previous models on multiple tasks and worked well even on older children and adults, despite only training on infants. The study also introduces a unique public dataset of infant ECG recordings with detailed labels."},{"id":"2609.31545v1","title":"Unconventional magnetoelastic behavior in Al rich CoFeAl Films with inverse Heusler like local order for flexible spintronics","abstract":"Strain engineering of magnetic properties offers a promising route toward flexible spintronic applications. Here, we report an unconventional magnetoelastic response in Al rich Co Fe Al thin films on flexible substrates using strain dependent magneto optical Kerr effect microscopy and magnetometry measurements. While the films exhibit a conventional positive magnetostriction coefficient, consistent with standard in plane easy axis rotation under stress, their saturation magnetization increases under compressive strain and decreases under tensile strain. First principles calculations reveal that this unconventional response originates from a strain induced competition between exchange splitting and crystal field effects in an inverse Heusler like local environment created by Al enrichment. This leads to a highly sensitive, sublattice dependent magnetic state, consistent with a strain induced reconfiguration of Co and Fe moments. Our results demonstrate that local compositional tuning can fundamentally alter magnetoelastic behavior, establishing strain controlled sublattice compensation as a route toward programmable magnetic functionality in flexible spintronic systems.","published":"2026-09-25T17:18:10Z","updated":"2026-09-25T17:18:10Z","authors":["Rupalipriyadarsini Chhatoi","Anuroopa Behatha","Swayang Priya Mahanta","Shubhransu Sahoo","Soubhagya Dash","Bhuvneshwari Sharma","Abhisek Mishra","Esita Pandey","Satadeep Bhattacharjee","Subhankar Bedanta"],"categories":["cond-mat.mtrl-sci"],"headline":"Researchers uncover unusual magnetic response in flexible CoFeAl films","summary":"The researchers studied thin films of CoFeAl with extra aluminum on flexible surfaces to explore how stretching or compressing affects their magnetic properties. They found that when compressed, the magnetic strength increases, but it decreases when stretched, which is not typical. Using computer calculations, they suggest this happens because of changes at the atomic level caused by aluminum, affecting magnetic interactions. Their findings show that adjusting local composition can change magnetic behavior, offering new ways to control flexible magnetic devices."},{"id":"2609.31544v1","title":"A Flow Matching Framework for Neural Representational Dissimilarity","abstract":"Neural representational dissimilarity quantifies differences between neural response distributions, and is essential for comparing neural codes across stimuli, brain areas, tasks, and models. Commonly used distance metrics involve different assumptions and are estimated with separate methods. Here, we show that a variety of distance metrics can be unified under a flow matching framework developed in deep generative models. That is, these distances arise as Jeffreys divergences under different velocity constraints. We find that flow matching has advantages for estimating distances involving complicated distributions and continuous variables. Furthermore, this framework enables the design of new distance metrics in a principled way. Together, flow matching provides a unified approach for understanding, estimating, and designing neural representational dissimilarity metrics.","published":"2026-09-25T17:15:43Z","updated":"2026-09-25T17:15:43Z","authors":["Zeyuan Ye","Xue-Xin Wei"],"categories":["cs.AI","cs.IT","cs.LG"],"headline":"New Unified Method for Comparing Brain Activity Patterns","summary":"Researchers explored ways to measure differences in brain activity patterns, which help compare how the brain responds to different stimuli or tasks. They showed that many existing methods can be combined into a single framework called flow matching, originally from deep learning. This approach may better estimate complex differences and allows creating new comparison measures. The findings are based on theoretical work relating to mathematical distances between neural responses."},{"id":"2609.31543v1","title":"Cosmic Heterogeneity Shapes the Evolution of Dark Energy and Time","abstract":"Early- and late-Universe measurements of the Hubble constant disagree, a puzzle known as the Hubble tension. We derive cosmological parameters from first principles, using only the dynamics of cosmic volume and the possibility that matter can be unevenly distributed. Within this framework, three of the six parameters of the standard model are obtained from the underlying dynamics rather than fitted to observations: volume formation determines the dark-energy density parameter to be two thirds, spatial curvature vanishes, and the matter density follows from the remaining cosmic energy budget. The Hubble constant remains an observational input because it represents a calendar date rather than an intrinsic property of the Universe, while the two remaining parameters are likewise taken from observation. Once cosmic unevenness is included as a dynamical component, both the expansion rate and the dark-energy density parameter become redshift dependent. The latter increases by eight percent toward the present, from 0.679 in the homogeneous early Universe to 0.735 today, consistent with independent probes spanning several orders of magnitude in redshift, including the microwave background, supernovae, baryon acoustic oscillations, X-ray clusters and gravitational waves. This same mechanism yields the time-evolving dark energy favored by recent combined analyses without introducing a new free parameter. Integrating the resulting expansion history gives a cosmic age of 13.521 Gyr, shorter than the homogeneous value by 0.306 Gyr. This difference is a global time dilation: dark energy and time evolve together, shaped by the unevenness of the cosmos.","published":"2026-09-25T17:15:11Z","updated":"2026-09-25T17:15:11Z","authors":["Jacky David Yang","Hans-Otto Carmesin"],"categories":["astro-ph.CO"],"headline":"Uneven Cosmic Matter May Explain Dark Energy and Expansion Differences","summary":"Researchers explored why measurements of the Universe's expansion rate differ early and late in its history, a problem called the Hubble tension. They developed a model based on how cosmic volume grows and matter is distributed unevenly, finding that dark energy and expansion change over time without adding new parameters. Their approach fits several types of observations and suggests the Universe's age is slightly younger than previously thought, showing that cosmic unevenness affects both dark energy and the flow of time."},{"id":"2609.31542v1","title":"Quotients of $L_1$ by subsequences of the Haar system have cotype 2","abstract":"Let $Y$ be the closed linear span of a subsequence of the Haar system in $L_1$. We prove that the quotient space $L_1/Y$ has cotype~$2$. This result implies that transportation cost spaces on diamond graphs have cotype $2$.","published":"2026-09-25T17:14:12Z","updated":"2026-09-25T17:14:12Z","authors":["Stephen Dilworth","Denka Kutzarova","Mikhail Ostrovskii","Thomas Speckhofer"],"categories":["math.FA"],"headline":"Researchers study special spaces derived from Haar system in math","summary":"The researchers investigated a mathematical space formed by taking a part of the Haar system in a known space called L1, and looked at what happens when you divide L1 by this part. They proved that the resulting quotient space has a property called cotype 2. Their findings also suggest that certain transportation cost spaces on diamond graphs share this property. The work is theoretical and involves advanced concepts in functional analysis."},{"id":"2609.31541v1","title":"Sipser-Spielman meets Dijkgraaf-Witten: non-Abelian qLDPC codes via twisted sheaf gauge theory and almost-constant-overhead magic state fountain","abstract":"Sipser-Spielman codes and Dijkgraaf-Witten twisted gauge theories are among the most profound ideas in the last few decades in the areas of computer science and mathematical physics respectively. This work unifies them in the same framework using the language of sheaf cohomology, and produces a new framework of twisted sheaf guage theories describing a class of quantum low-density parity-check (qLDPC) codes with non-Abelian $D_4$ topological order. The corresponding twisted qLDPC code can be obtained from gauging a new type of 0-form sub-complex $\\mathbb{Z}_2^3$ symmetry from a qLDPC code defined on a sheaf complex, and can be interpreted as a topological defect network of non-Abelain $D_4$ patches glued together with proper gapped interfaces. As an application, one can use this to realize a \\textit{magic state fountain} via the gauging measurement of the addressable logical CZ gates as 0-form subcomplex symmetries in a 2D hypergraph-product on a sheaf complex. This includes a scheme of subdividing an arbitrary constant-rate 2D hypergraph-product code with parameter $[[n,Θ(n), Ω(n^{1/2})]]$ into a quantum sheaf code which allows preparation of $Θ(n^{1/2})$ in parallel, equivalent to the recent geometric construction using the code-to-manifold mapping in (arXiv:2601.06736). Moreover, using the recent sheaf complex and algebraic code constructions by Golowich-Tamo-Zhu (arXiv:2609.27801) with parameter $[[n,Θ(n^{1-ε}), Ω(n^{(1-ε)/2})]]$ for arbitrary small $ε$, one can prepare $Θ(n^{1-ε})$ CZ magic states in parallel and hence achieve an almost-constant magic rate.","published":"2026-09-25T17:12:03Z","updated":"2026-09-25T17:12:03Z","authors":["Guanyu Zhu","Shi Jie Samuel Tan","Ryohei Kobayashi","Po-Shen Hsin"],"categories":["quant-ph","cond-mat.str-el","cs.IT","hep-th","math-ph"],"headline":"New framework links advanced math to quantum error-correcting codes","summary":"Researchers combined ideas from computer science and mathematical physics using sheaf cohomology to create a new way to describe certain quantum error-correcting codes with complex symmetries. They developed a model called twisted sheaf gauge theory that helps understand quantum low-density parity-check (qLDPC) codes featuring non-Abelian properties. The work also suggests how to prepare special quantum states (magic states) efficiently in parallel, though these results are theoretical and based on advanced mathematical constructions."},{"id":"2609.31540v1","title":"Can You Check That? The Checkability Boundary for Local LLM Network Automation","abstract":"Sending every network-automation input to a third-party frontier LLM exports sensitive artifacts such as production configurations, topologies, and logs. Querying small language models (SLMs) locally avoids this egress, but SLM outputs can be error-prone for direct use. This work introduces checkability as a criterion for determining which tasks are suitable for local inference. A task is checkable when it exposes a cheap, deterministic test - an intrinsic check - that rejects outputs violating a necessary correctness condition. We instantiate this idea in Touchstone, a local-first pipeline that uses seven off-the-shelf SLMs (1-8B parameters) to generate candidates, uses task-specific intrinsic checks to reject responses, and escalates unresolved inputs to a frontier LLM. On conflict detection and intent translation tasks, Touchstone reaches 98.6% and 93.8% end-to-end accuracy while escalating only 16% and 17% of inputs, respectively. On TeleQnA, a knowledge-only control that has no task-specific intrinsic checks, Touchstone is unable to match the accuracy of the frontier baseline. Our results support a simple deployment rule: keep inference local when task semantics support precise, low-cost checks; escalate the rest.","published":"2026-09-25T17:11:55Z","updated":"2026-09-25T17:11:55Z","authors":["Maleeha Masood","Momina Nofal"],"categories":["cs.NI","cs.AI"],"headline":"When to Trust Local AI for Network Automation Tasks","summary":"The researchers studied how to safely use small local AI models for automating network tasks without sending sensitive data to external large AI systems. They propose using 'checkability,' meaning a task can be automatically tested for errors cheaply and reliably, to decide if local AI can handle it alone. Their system, Touchstone, tries local models first and checks their answers, only turning to larger remote AI for harder cases. They found high accuracy on tasks with clear checks but lower accuracy without such checks, suggesting local AI works best where mistakes can be easily detected."},{"id":"2609.31539v1","title":"NEXT: Physics-Informed Neuro-Spectral Exponential Time Differencing Architectures","abstract":"Physics-Informed Neural Networks (PINNs) build neural representations of time-dependent PDE solutions, naturally incorporating physics knowledge and observational data, which makes them well suited to both forward and inverse PDE problems. PINNs, however, are known to suffer from spectral bias and lack of causality. Neuro-Spectral Architectures (NeuSA), a recently proposed alternative to PINNs, mitigate both issues, but their numerical integration becomes unstable for stiff differential equations arising in many relevant physical problems. This study proposes Neuro-Spectral Exponential Time Differencing Architectures (NEXT), which combines the spectral representation of the PDE solution in NeuSA with high-order exponential integrators. Within this approach, the linear stiff part of the vector field induced by the PDE is integrated exactly through matrix exponentials, while the possibly nonlinear remainder is modeled by a neural network. The effectiveness of NEXT is verified through benchmark experiments on a set of stiff PDEs, in which NEXT is stable and accurate while NeuSA diverges numerically. It is also shown that NEXT can be applied to inverse problems, where the model has to learn unknown parameters or boundary conditions from sparse data. All code used in this work is publicly available at: https://github.com/marcioh2m/next.git .","published":"2026-09-25T17:11:44Z","updated":"2026-09-25T17:11:44Z","authors":["Márcio Marques","Leonardo Mendonça","Leonardo M. Moreira","Christian Júnior de Oliveira","Vitor Balestro","Tiago Novello","Daniel Yukimura","Pavel Petrov","Lucas Nissenbaum"],"categories":["cs.LG","math.NA"],"headline":"New Neural Method Improves Solving Tough Physics Equations","summary":"Researchers studied how to better solve complex physics problems that involve difficult differential equations. They developed a new approach, called NEXT, which improves on earlier methods by combining a special way of representing solutions with advanced mathematical integration techniques. Their tests show NEXT is more stable and accurate for these challenging equations, and it can also help find unknown details from limited data. The findings are based on simulations and mathematical experiments, and the researchers provide their code publicly."},{"id":"2609.31538v1","title":"Anchor Radicals and Metric Rigidity in $n$-Lie--Rinehart Algebras","abstract":"For an $n$-Lie--Rinehart algebra, $n>2$, the anchor is defined on $\\Aext{n-1}L$, so its kernel does not determine a distinguished submodule of $L$. We introduce the anchor radical $\\Rad(ρ)\\subset L$ and prove that it is an adjoint-stable ideal on which the bracket is $A$-multilinear. We then derive an explicit obstruction to the $A$-linearity of metric compatibility in fundamental directions. This obstruction vanishes on $\\Rad(ρ)$, whereas a strongly nondegenerate compatible metric on all of $L$ forces the anchor to vanish. These results lead to a natural notion of orthogonal structure on the anchor radical. For anchors induced by Nambu tensors, the radical is identified with the cotangent annihilator of the tensor.","published":"2026-09-25T17:10:50Z","updated":"2026-09-25T17:10:50Z","authors":["Irmely Gladesh Mabanza Nsiloulou","Basile Guy Richard Bossoto"],"categories":["math.DG"],"headline":"Studying structures in advanced math called n-Lie–Rinehart algebras","summary":"Researchers examined a complex mathematical structure known as an n-Lie–Rinehart algebra, focusing on something called the 'anchor' and its limitations. They introduced a new concept, the 'anchor radical,' which helps understand properties of these algebras better. Their work identifies when certain types of compatibility and structure can exist within these algebras, although their findings are theoretical and based on mathematical properties rather than practical experiments."},{"id":"2609.31537v1","title":"How large can $B_n$- and $D_n$-friezes be?","abstract":"We pin down the largest entries possible in positive integral friezes of types $D_n$ for $n\\geq 4$ and $B_n$ for $n\\geq 2$, addressing a conjecture of Robin Zhang. For type $D_n$, the sharp upper bound is $F_nF_{n+1}-1$, and for type $B_n$, it is $F_{n+1}F_{n+2}-1$, where $F_k$ are the Virahanka--Fibonacci numbers.","published":"2026-09-25T17:10:24Z","updated":"2026-09-25T17:10:24Z","authors":["Florian Ito Sprung"],"categories":["math.CO"],"headline":"Researchers find largest numbers in special mathematical patterns","summary":"The researchers studied specific patterns called positive integral friezes of types Dₙ and Bₙ, which are arrangements of numbers following certain rules. They determined the biggest numbers that can appear in these patterns, confirming a previous guess by another mathematician named Robin Zhang. Their findings give exact formulas involving Fibonacci numbers for the largest entries in these patterns."},{"id":"2609.31536v1","title":"Collective impact of opinion formation on agent-based epidemic dynamics","abstract":"Epidemic spread and collective opinion formation are tightly coupled processes: individual beliefs about the severity of an outbreak and the value of protective measures shape behavior, while the evolving epidemic itself reshapes public sentiment through risk perception. We propose an agent-based framework that couples opinion dynamics with a compartmental epidemic model, in which each agent carries both a health status and a continuous opinion variable describing adherence to protective measures. Opinion evolution is driven by binary interactions among agents and by an exogenous background field that reflects the current state of the epidemic, while the disease transmission rate depends explicitly on the opinions of interacting agents. Using methods of statistical mechanics, we characterize the resulting macroscopic system for the compartmental mass fractions and mean opinions in both far-from-equilibrium and close-to-equilibrium regimes. The calibration shows that incorporating behavioral opinion dynamics substantially improves the interpretation of epidemic evolution. In particular, the inferred risk perception function evolves consistently with the observed epidemic burden, indicating that behavioral adaptation can explain part of the temporal variation that would otherwise be absorbed into time-dependent transmission parameters. Our results illustrate how integrating opinion heterogeneity in kinetic epidemic models offers quantitative insights into the feedback between behavior and disease spread.","published":"2026-09-25T17:09:08Z","updated":"2026-09-25T17:09:08Z","authors":["Andrea Bondesan","Riccardo Vecchio","Chiara Barbati","Jonathan Franceschi","Anna Odone","Mattia Zanella"],"categories":["physics.soc-ph","nlin.AO","q-bio.PE"],"headline":"How Opinions Affect Disease Spread in a New Epidemic Model","summary":"Researchers studied how people's beliefs about an outbreak influence their behavior and, in turn, how the outbreak affects those beliefs. They created a computer simulation where each individual has a health status and an opinion about protective measures. The model links changes in opinions to interactions between people and the current state of the epidemic. Their findings suggest that including how opinions change over time helps better explain the patterns of disease spread, showing a feedback loop between behavior and infection rates."},{"id":"2609.31535v1","title":"Discounted Hitting Domination on Graphs with Submodularity, Complexity and Exact Algorithms","abstract":"On a network with a fixed set of verified sources, discounted averaging induces an equilibrium support $h_i^S=\\mathbb{E}_i[λ^{T_S}]$, the discounted probability that a random walk reaches $S$ before attenuation. We define the \\emph{discounted hitting domination number} $δ_{λ,τ}(G)$ as the minimum number of sources required to guarantee $h_i^S\\geτ$ at every vertex. Although this potential is known through penalized and group hitting probabilities, the associated minimum-cardinality uniform-coverage problem appears to be new. Aggregate support is monotone submodular, while the uniform-floor problem is an exact submodular-cover problem. Moreover, if $λ^{r+1}<τ\\le\\left(\\fracλΔ\\right)^r$, then $δ_{λ,τ}(G)$ equals the distance-$r$ domination number. This yields NP-completeness and APX-completeness at $(λ,τ)=(1/4,1/14)$ on graphs of maximum degree three. For spiders, we obtain an exact finite-state characterization and a polynomial-time algorithm for every fixed rational pair $(λ,τ)$, and show that the branching vertex need not belong to a minimum source set. Finally, an exact mixed-integer linear formulation certifies optimal placements on a real network and a synthetic graph and demonstrates substantial differences from degree, closeness, and classical domination.","published":"2026-09-25T17:08:55Z","updated":"2026-09-25T17:08:55Z","authors":["Julian D. Allagan","Kevin Pereyra","William A. Massey"],"categories":["math.CO","math.OC"],"headline":"New insights into spreading influence in networks using discounted probabilities","summary":"Researchers studied a mathematical way to measure how sources spread influence in a network, focusing on a 'discounted hitting domination number' that ensures every point in the network is reached with a certain probability. They found connections to difficult computational problems, showed exact solutions for specific network types called spiders, and developed algorithms to identify optimal source placements. Their work uses advanced math concepts and provides exact methods, though some results rely on theoretical models and computational complexity findings."},{"id":"2609.31534v1","title":"Generation of 10-TW Attosecond X-Ray Pulses in a Free Electron Laser","abstract":"Attosecond pulses provide direct access to electron dynamics in atoms, molecules, and solids on their natural timescales. While table-top sources have achieved attosecond pulse durations at moderate intensities, extending attosecond science to high peak powers in the x-ray regime requires free-electron lasers (FELs). Here we report the generation of soft-x-ray pulses with a duration of $\\sim\\!100$ as and peak power on the order of 10 TW at the European XFEL. The unprecedented combination of pulse duration and peak power is enabled by operation of a soft-x-ray undulator at high electron energy together with five-stage bunch compression, producing ultrahigh peak currents. These millijoule-level attosecond x-ray pulses establish a new regime for attosecond science.","published":"2026-09-25T17:08:20Z","updated":"2026-09-25T17:08:20Z","authors":["E. A. Schneidmiller","S. Serkez","I. Zagorodnov","N. Gerasimova","M. Scholz","N. Golubeva","M. Guetg","G. Geloni","S. Karabekyan","F. Sottocorona","A. Trebushinin","G. Perosa","S. Tomin","F. Mayet","Th. Maltezopoulos","J. Gruenert","J. Laksman","H. Ahmadi Rashtabadi","T. M. Baumann","R. Boll","F. Calegari","G. Catacchio","S. Dold","D. E. Ferreira de Lima","L. Funke","W. Helml","M. Ilchen","T. Mazza","M. Meyer","T. Mullins","M. Robinson","S. Savio","K. Scharei","Ph. Schmidt","A. Thiel","S. Usenko","L. Wuelfing","W. Decking"],"categories":["physics.acc-ph"],"headline":"Researchers generate extremely fast, powerful X-ray pulses","summary":"Scientists created soft X-ray bursts lasting about 100 attoseconds, much shorter than a billionth of a second, at a facility called the European XFEL. They combined very high electron energies and a special method to compress electron bunches, resulting in ultrahigh peak currents and powerful pulses around 10 terawatts. This work shows a new level of performance for studying ultrafast electron behavior in materials, based on the researchers' report."},{"id":"2609.31533v1","title":"Structural prediction of B$_{18}$Y$_{2}$ cluster: A Machine-Learning-Assisted Basin-Hopping Study","abstract":"The structural and optical properties of the doubly yttrium-doped boron cluster B$_{18}$Y$_2$ have been systematically investigated using density functional theory calculations. The lowest-energy structure was identified through extensive basin-hopping searches accelerated by a pre-trained MACE machine-learning potential and subsequently refined and validated at the DFT level. The resulting global-minimum structure adopts a double-ring geometry, consisting of two fused B$_9$ rings stabilized by yttrium atoms positioned above and below the boron framework. Vibrational frequency calculations confirm the dynamical stability of the optimized structure, while the calculated infrared and UV--Vis spectra provide characteristic signatures of the Y--B interactions and the electronic structure of the boron framework. These results demonstrate how yttrium doping can stabilize unusual double-ring boron architectures and highlight the effectiveness of machine-learning-assisted basin-hopping searches for exploring the complex potential-energy landscapes of doped boron clusters.","published":"2026-09-25T17:08:02Z","updated":"2026-09-25T17:08:02Z","authors":["Peter Ludwig Rodríguez-Kessler"],"categories":["cond-mat.mtrl-sci"],"headline":"Machine learning helps predict structure of yttrium-doped boron clusters","summary":"Researchers used a combination of machine learning and traditional calculations to find the most stable structure of a boron cluster with yttrium atoms. Their findings show a unique double-ring shape stabilized by yttrium, confirmed to be stable through simulations. The study demonstrates how machine learning can speed up discovering complex cluster structures, but the results are based on theoretical calculations, not experiments."},{"id":"2609.31532v1","title":"Computing the Iwasawa $μ$-Invariants for Elliptic Curves over $\\mathbb{Z}_2$-Extensions","abstract":"Let $E$ be an elliptic curve defined over rational numbers. Further assume $E$ has good ordinary reduction at $p=2$ and $E[4]$ is reducible as a $G_\\mathbb{Q}$-representation. In this paper, we offer sufficient and necessary computational criteria for the algebraic Iwasawa $μ$-invariant over the cyclotomic $\\mathbb{Z}_2$-extension of rational number to be $0$, $1$ or $\\geq 2$. We also show that elliptic curves with isomorphic mod-$4$ representations have algebraic $μ$-invariant equal to $0$ if one of the curves does.","published":"2026-09-25T17:07:24Z","updated":"2026-09-25T17:07:24Z","authors":["Zichao Lin","Mulun Yin"],"categories":["math.NT"],"headline":"New Method to Calculate Key Numbers for Elliptic Curves at 2-adic Levels","summary":"Researchers studied certain smooth curves called elliptic curves defined using rational numbers, focusing on those with specific technical properties at the prime number 2. They developed clear computational tests to determine a mathematical value called the \"Iwasawa mu-invariant,\" which can be 0, 1, or at least 2, related to complex number extensions. Their results also show a connection between two related curves where if one has this value zero, the other does too. These findings are theoretical and based on specific conditions assumed for the curves."},{"id":"2609.31531v1","title":"HySTAR: Anchored Hypergraphs for Stable Credit Assignment in Cooperative Multi-Agent Reinforcement Learning","abstract":"Cooperative multi-agent reinforcement learning under partial observability and shared rewards requires assigning team outcomes to individual agents and high-order coalitions. A MAPPO-style critic compresses joint behavior into one global value, while critics that dynamically reconstruct the grouping topology change the mapping from agents and coalitions to value components as interactions or active agents evolve. We refer to this inconsistency as structural target drift. We introduce HySTAR, a MAPPO-based framework that separates adaptive representation learning from a temporally consistent high-order value-decomposition basis. HySTAR anchors an overlapping sparse hypergraph as a uniformly covered decomposition scaffold, uses a spatiotemporal encoder to represent physical and task-dependent interactions, and combines temporal and structural relevance to construct agent-specific advantages. Experiments on SMAC, GRF, Traffic Junction, and MPE demonstrate consistent improvements over MAPPO-style, value-factorization, and dynamic-grouping baselines. On the hardest SMAC settings, HySTAR achieves relative gains of 16.7\\% over MAPPO and 15.6\\% over HYGMA, ranks first on all six GRF scenarios, reduces Traffic Junction convergence epochs by up to 40.2\\% relative to MAGIC, and obtains the highest MPE episode rewards. Controlled topology, agent-death, neighborhood, and parameter analyses support the benefit of anchoring the decomposition scaffold while adapting the propagated representations.","published":"2026-09-25T17:06:54Z","updated":"2026-09-25T17:06:54Z","authors":["Xinglong Luo","Yuding Zhang","Yuheng Kuang","Shuxuan Yuan","Zhenni Zeng","Weiqiang Zhu","Zhenhai Ji","Zhengning Wang"],"categories":["cs.LG"],"headline":"New method improves teamwork in AI agents with consistent value tracking","summary":"The researchers tackled the challenge of fairly assigning shared team rewards to individual agents and groups in cooperative multi-agent learning, where agent roles and interactions change over time. They developed HySTAR, a framework that keeps value assignments stable by anchoring them to a consistent structure while flexibly representing agent interactions. Tests on various simulation tasks showed HySTAR outperformed existing methods, though the findings come from controlled experiments and simulations rather than real-world trials."},{"id":"2609.31530v1","title":"Spectral Invariance for Black-Hole Differential Observables","abstract":"We study first-order differential quantities built from solutions of second-order black-hole wave equations. These quantities can satisfy new differential equations that contain extra singular points, even when the original physical solutions remain regular. We show that an isolated zero of the transformation determinant can produce such an apparent singularity, but it does not by itself create a new resonance or quasinormal-mode frequency. We then study the corresponding Green functions and show that the physical transformed response has the same spectral poles as the original problem, provided that the transformation does not become globally degenerate. For the Regge--Wheeler and Zerilli equations, this means that ordinary turning points can appear as singular points of the derivative equation without changing the nonzero-frequency quasinormal-mode spectrum. The zero-frequency case is different and must be studied separately.","published":"2026-09-25T17:06:52Z","updated":"2026-09-25T17:06:52Z","authors":["Vedant Subhash"],"categories":["gr-qc","math-ph"],"headline":"Researchers Examine Mathematical Properties of Black Hole Wave Equations","summary":"The researchers investigated certain mathematical quantities derived from black hole wave equations and found that these can satisfy new differential equations with extra singular points, even when the original solutions remain well-behaved. They showed that while these new points may appear due to specific mathematical conditions, they do not create new physical resonances or frequencies related to black hole vibrations. Their findings apply to particular equations used in black hole physics, but zero-frequency cases require separate analysis. These conclusions are theoretical and based on mathematical studies."},{"id":"2609.31529v1","title":"Prior-Free Delegation","abstract":"We consider a robust delegation problem in which the principal does not know the distribution from which the underlying state is drawn. The principal can choose a general randomized mechanism and maximizes her worst-case expected payoff over all state distributions. Our main result characterizes the robustly optimal mechanism. The mechanism has up to three regions: (i) accommodation, where the agent's ideal action is taken, (ii) calibrated randomization, where each type receives a distinct lottery, and (iii) pooling, where all types receive the same lottery. We then extend our model to a multidimensional setting in which preferences are additively separable and satisfy a cross-dimensional symmetry condition, and show that delegating separately across dimensions is robustly optimal.","published":"2026-09-25T17:05:22Z","updated":"2026-09-25T17:05:22Z","authors":["Terrence McGovern","Jan Benedikt Napp","Wenjun Zheng"],"categories":["econ.TH"],"headline":"How to Make Decisions Without Knowing the Odds","summary":"Researchers studied a decision-making problem where a leader doesn’t know the likelihood of different situations. They looked for the best way to delegate choices to an agent, aiming to do well no matter what the true situation is. Their work describes an optimal method that sometimes follows the agent’s preference, sometimes uses tailored random choices, and sometimes treats all cases the same. They also showed that dealing with multiple factors separately works best under certain conditions. The findings are theoretical and focus on describing this optimal delegation method."},{"id":"2609.31528v1","title":"Quantum-HPC Workflows Across Multiple Quantum Computing Platforms: Two Case Studies","abstract":"We describe the Quantum-HPC hybrid system comprising the supercomputer Fugaku, the QPUs Quantinuum Reimei and ibm_kobe, and the Tierkreis hybrid workflow software. We present two demonstrations of the system at work. In the first, we compute the excited states of a biomolecular system using Fugaku and Reimei (simulator) to handle classical and quantum parts of the workflow respectively. In the second, Fugaku coordinates Reimei and ibm_kobe working together in a combined VQE + QPE workflow to compute the ground state energy of a simple system.","published":"2026-09-25T17:04:03Z","updated":"2026-09-25T17:04:03Z","authors":["Philipp Seitz","Miwako Tsuji","John Children","Yuta Kikuchi","Riku Masui","Juan Pedersen","Kentaro Yamamoto","Ross Duncan"],"categories":["quant-ph","cs.DC"],"headline":"Combining Supercomputers and Quantum Devices for Advanced Simulations","summary":"Researchers describe a hybrid system that links a powerful supercomputer with quantum processors to tackle complex scientific computations. They demonstrated this by using the system to study molecular energy states, dividing tasks between classical and quantum machines. Another demonstration involved coordinating two quantum processors through the supercomputer to estimate energy levels. These examples show how integrating classical and quantum computing can work in practice, though the findings are based on specific demonstrations rather than broad applications."},{"id":"2609.31527v1","title":"On Conjectures of $μ$-Invariant for Selmer Groups at $p=2$","abstract":"Let $E$ be an elliptic curve define over $\\mathbb{Q}$. Further assume $E$ has good ordinary reduction at $p=2$. In this article, we prove Greenberg's conjecture on the value of the algebraic Iwasawa $μ$-invariant when $E$ has a rational isogeny of degree 2, and obtain a sharp absolute upper bound of the $2$-adic $μ$-invariants for all such curves. We also prove the vanishing of the algebraic $μ$-invariant of the fine Selmer group introduced by Coates and Sujatha at $p=2$.","published":"2026-09-25T17:03:34Z","updated":"2026-09-25T17:03:34Z","authors":["Zichao Lin","Mulun Yin"],"categories":["math.NT"],"headline":"Researchers Explore Mathematical Properties of Elliptic Curves at Prime 2","summary":"The researchers studied specific mathematical objects called elliptic curves, focusing on a property related to the number 2. They proved a previously proposed idea (Greenberg's conjecture) about a value associated with these curves when a certain type of relationship exists in the curve's structure. They also established a precise upper limit for this value and showed that another related value is zero under certain conditions. Their work is based on advanced number theory and helps clarify theoretical predictions about these curves."},{"id":"2609.31526v1","title":"Bi-invariant Geodesic Regression: Existence, Uniqueness, and Convergence","abstract":"Bi-invariant geodesic regression generalizes linear regression to Lie groups. Its main feature is that it respects the symmetries of the group so that the resulting estimator is independent of arbitrary choices such as a reference frame. However, the local existence and uniqueness of the underlying estimator have not been shown until now. Furthermore, the convergence properties of the proposed algorithm for computing the estimator are not known. In this work, we investigate these questions. We prove that, locally, a unique estimator exists and give explicit bounds on the size of this neighborhood. We also show that the proposed iterative algorithm converges linearly to this estimator.","published":"2026-09-25T17:03:24Z","updated":"2026-09-25T17:03:24Z","authors":["Martin Hanik","Christoph von Tycowicz"],"categories":["math.ST","math.DG"],"headline":"Understanding Bi-invariant Geodesic Regression on Lie Groups","summary":"The researchers study bi-invariant geodesic regression, a method extending linear regression to complex geometric spaces called Lie groups, which respects their natural symmetries. They prove that locally, a unique best-fit solution exists and provide limits on how local this uniqueness is. Additionally, they show that the algorithm used to find this solution steadily approaches the unique estimator. Their work addresses previously unknown theoretical properties of this regression method."},{"id":"2609.31525v1","title":"TinyAudio: Compact and Efficient Text-to-Audio Generation for Low-Resource Deployment","abstract":"Text-to-audio (TTA) generation has advanced rapidly in generation quality and instruction following. However, representative systems often require around a billion parameters, limiting deployment on resource-constrained devices. This paper introduces TinyAudio, a compact flow-matching-based TTA model for low-resource deployment. At its core, TinyAudio uses TA-DiT, a 35M single-stream flow-matching Transformer. TinyAudio also includes TA-CLAP, a 32M audio-aligned text encoder, and TA-VAE, whose 20M decoder reconstructs 44.1 kHz audio from compressed latents. TinyAudio has only 87M parameters in total, over 90% fewer than representative billion-parameter pipelines, and uses 0.48 GB peak GPU memory. TinyAudio achieves competitive generation quality on AudioCaps and TTA-Bench. We further introduce TinyAudio-MF, a MeanFlow-accelerated model that enables real-time generation with a four-core CPU quota. Our results demonstrate a practical quality-footprint trade-off for low-resource TTA deployment.","published":"2026-09-25T17:03:10Z","updated":"2026-09-25T17:03:10Z","authors":["Junxi Liu","Xiquan Li","Wenhao Guan","Yifan Duan","Zhikang Niu","Yanru Huo","Ziyang Ma","Xie Chen"],"categories":["cs.SD","eess.AS"],"headline":"TinyAudio: Compact Text-to-Audio Model for Low-Resource Devices","summary":"Researchers developed TinyAudio, a much smaller and efficient system for converting text into audio, designed to work on devices with limited computing power. TinyAudio uses specialized components totaling 87 million parameters—over 90% fewer than typical models—and needs less GPU memory. The team showed it can produce audio quality similar to larger systems and introduced a version that runs in real time on regular CPUs. Their work offers a way to balance quality and resource use for practical low-resource text-to-audio generation."},{"id":"2609.31524v1","title":"Structured Reasoning Agentic Framework for Interpretable Critical View of Safety Assessment","abstract":"Surgical scene understanding is critical for computer-assisted intervention, yet laparoscopic cholecystectomy remains challenged by the complex anatomy of the hepatocystic triangle and the risk of bile duct injury. Existing methods for Critical View of Safety (CVS) assessment typically treat it as a holistic prediction task, mapping visual features directly to criterion-level labels. This black-box paradigm lacks explicit reasoning about anatomical relationships, limiting both interpretability and compositional generalization. To address this, we propose ReasonCVS, a structured reasoning agentic framework empowered by Vision-Language Models (VLMs) that decomposes CVS assessment into explicit, fine-grained anatomical verification. Specifically, we devise an Anatomical Scene Graph Abstraction (ASGA) that organizes anatomical entities and their spatial relationships into a structured representation. To operationalize this, we introduce a Rationale-Aware Reasoning Agent, powered by a Large Language Model (LLM) fine-tuned via rationale distillation. Functioning as a strict central decision-maker, it invokes VLM-driven Sub-criterion Verifier as a specialized perceptual tool to parse the graph and independently evaluate individual sub-criteria. Through calibrated soft reasoning, this agent synthesizes the tool-gathered distributed observations, yielding a final verdict alongside a traceable clinical rationale. Extensive experiments on the Endoscapes-CVS201 benchmark demonstrate that ReasonCVS achieves superior performance (68.1\\% mAP) over state-of-the-art while providing interpretable, criterion-level explanations for reliable surgical assessment.","published":"2026-09-25T17:01:56Z","updated":"2026-09-25T17:01:56Z","authors":["Qing Xu","Yuxiang Luo","Zhen Chen"],"categories":["cs.CV"],"headline":"New AI Approach Helps Understand Safer Surgery Views","summary":"Researchers developed ReasonCVS, an AI system designed to improve safety during gallbladder surgery by clearly analyzing key anatomical features instead of making unclear overall judgments. Their system breaks down complex visual information into detailed relationships between body parts and uses a language model to reason and explain its assessments. Testing on surgical video data showed better accuracy and explained results, though findings are based on computational experiments using existing datasets."},{"id":"2609.31523v1","title":"Profinite Non-Rigidity of Arithmetic Lattices and the Kähler Property","abstract":"We study the profinite non-rigidity of arithmetic lattices and its implications for the Kähler property. In the first part, we characterize the absolute Dynkin types that admit non-isomorphic real forms of higher-rank Lie groups containing torsion-free arithmetic lattices with isomorphic profinite completions. In the second part, as a geometric application, we answer a question asked independently by Arapura and Libgober, by showing that the Kählerness of finitely presented, residually finite groups is not determined by its profinite completion.","published":"2026-09-25T16:59:54Z","updated":"2026-09-25T16:59:54Z","authors":["Yukun Du","Feng Hao","Kejia Zhu"],"categories":["math.GT","math.AG","math.GR"],"headline":"Researchers explore complex group structures and geometric properties","summary":"The researchers investigate certain mathematical groups called arithmetic lattices and how their underlying structures can look very similar in a specific technical sense called profinite completion, even when they are not actually the same. They then apply these findings to a geometric question, showing that a particular property related to shapes called the Kähler property cannot be identified just by looking at these profinite completions. Their results clarify limitations on what profinite completions reveal about these groups and related geometric features."},{"id":"2609.31522v1","title":"Fabry-Pérot Resonance Shifts as a Novel Non-destructive In Situ Surface Damage Diagnostic","abstract":"Current surface damage evaluation for nanostructured electronic devices requires specialized equipment or conditions outside the designed operating environment and cannot be used for in situ evaluation of surface damage. In this work, we developed a non-destructive method to probe the surface sublimation of a semiconductor heterostructure which can be applied to other semiconductor damage mechanisms. This approach employs shifts in the resonances of an optical cavity comprising the surface and an integrated distributed Bragg reflector fabricated below the semiconductor heterostructure undergoing sublimation. We performed the demonstration of this technique under vacuum, but surface damage effects can be observed in any operational environment either in situ or ex situ.","published":"2026-09-25T16:59:50Z","updated":"2026-09-25T16:59:50Z","authors":["Gregory Blume","Colin Kirk","Sushil Poudel","Alimohammed Kachwala","Joseph Grames","Marcy L. Stutzman","Matt Grau","Sylvain Marsillac"],"categories":["physics.acc-ph","cond-mat.mtrl-sci"],"headline":"New Light-Based Method Checks Surface Damage Without Harm","summary":"Researchers developed a non-destructive way to detect surface damage on tiny electronic parts by observing changes in light patterns within a tiny cavity on the material. They tested this method on a semiconductor surface losing atoms in a vacuum, showing it can track damage without special equipment or removing the device from its working environment. Their work suggests this technique might work for other types of semiconductor damage, though so far it has only been demonstrated under controlled lab conditions."},{"id":"2609.31521v1","title":"Complexity drives the symmetry breaking of temperature fluctuations","abstract":"Frequently, nonextensive systems are described through $q$-exponential functions, where the $q$-parameter determines how far a system is from equilibrium. In this work, we study systems constrained to $q-1\\ll1$ to describe systems near thermal equilibrium, and derive the probability density function and temperature fluctuations. To this end, we derive the series expansion of the $q$-exponential function, yielding a perturbative series that is a superposition of gamma distributions. On the other hand, the temperature fluctuations are expressed as a superposition of Dirac delta functions and their derivatives, meaning that the bulk of the system remains at a constant temperature, while point-temperature variations emerge, marking the system's departure from thermal equilibrium. The statistics of the temperature fluctuations exhibit a transition from a symmetric to a nonsymmetric distribution as the nonextensive parameter increases. These results provide evidence of the association between the emergence of nonextensivity and the symmetry breaking of temperature fluctuations.","published":"2026-09-25T16:58:29Z","updated":"2026-09-25T16:58:29Z","authors":["D. Rosales Herrera","M. Y. Gallegos Ruiz","A. Fernández Téllez","J. R. Alvarado García","J. E. Ramírez"],"categories":["cond-mat.stat-mech"],"headline":"Researchers Link Complexity to Uneven Temperature Changes","summary":"The researchers studied systems that are almost in thermal equilibrium but show some complexity, described using a mathematical parameter called q. They derived formulas explaining how temperature fluctuates in these systems, finding that most of the system stays at constant temperature while small points vary. Their analysis showed that as complexity increases, temperature fluctuations change from being balanced and symmetric to uneven and asymmetric. Their findings suggest a connection between system complexity and the way temperature varies, based on theoretical calculations near equilibrium conditions."},{"id":"2609.31520v1","title":"Complexity Barriers to State Preparation in Quantum Approximate Optimization","abstract":"For many important optimization problems we are restricted to approximate solutions in practice due to computational complexity. Distinct from the exact optimization setting, approximate optimization admits performance measures beyond whether the optimum is found, with different tradeoffs and complexity. For MaxCut, a near-unity (ordinary) approximation ratio can coexist with near-zero improvement (gain) over a random cut. For the standard encoding, the unconditional classical MaxCut-Gain hardness gap implies that \\emph{any uniformly efficient quantum or hybrid procedure recovering a fixed positive fraction of the optimal classical gain on every input, with at least inverse-polynomial success probability, would place} NP \\emph{in} BQP. Such a procedure is therefore believed impossible under standard assumptions. We broadly address where our worst-case barriers do or do not apply across the quantum algorithm landscape. We prove that the barrier survives quantum random access optimization (QRAO) compression and applies between the classical and relaxed optimal values. For every input, a product state attains the classical optimum. Thus the barrier to reaching the classical threshold does not arise from a need for entanglement. For $d\\in\\{2,3\\}$ variables per qubit, the known decoder transfers encoded energy gain to decoded mean gain by the exact factor $1/d^2$. Combining this identity with MaxCut-Gain hardness gives an operational preparation barrier for QRAO. We also construct hard $n$-qubit families with relative quantum relaxation excess $Θ(1/n)$, while the maximally mixed state has energy approximation ratio $1-Θ(1/n)$, zero encoded energy gain, and hence zero decoded mean gain. Our results separate the effects of relaxation tightness and energy approximation from operational accessibility, motivating more comprehensive accounting in benchmarking and performance assessment.","published":"2026-09-25T16:57:00Z","updated":"2026-09-25T16:57:00Z","authors":["Stuart Hadfield"],"categories":["quant-ph","cs.CC"],"headline":"Quantum Limits on Approximating MaxCut Optimization Problems","summary":"The researchers study challenges in using quantum methods to prepare states that approximate solutions to the MaxCut optimization problem. They show theoretical barriers suggesting that efficient quantum procedures achieving a certain level of gain beyond classical random solutions are unlikely under standard complexity assumptions. Their findings distinguish limitations due to quantum state preparation from those arising from the need for entanglement, and highlight the roles of relaxation tightness and energy approximation in assessing quantum algorithm performance. Results are theoretical and based on complexity analysis."},{"id":"2609.31519v1","title":"Fast and Secure Simultaneous Authentication of Equals for WPA3","abstract":"The Simultaneous Authentication of Equals (SAE) protocol, introduced in WPA3, provides robust protection against offline dictionary attacks against a network Pre-Shared Key (PSK) and also protection of session keys from other people knowing that PSK. However, its high computational cost for the Password Element (PE) derivation makes Access Points (APs) vulnerable to CPU exhaustion Denial-of-Service (DoS) attacks. This paper proposes a resilient architectural modification to SAE. First, we introduce an asymmetric resource cost model that offloads the iterative discovery of cryptographic elements to the client, allowing the AP to maintain a fixed computational load during the handshake. To further mitigate brute-force attempts, we implement a mechanism based on a slow-path key derivation (with variable cost key derivation functions, such as PBKDF2 or Argon2), incorporating a deliberate processing delay on the supplicant. Finally, we introduce a ticket-based mechanism to facilitate efficient re-authentication for known devices, bypassing expensive exchanges while preserving system availability under adversarial conditions. Experimental results demonstrate that this architecture significantly mitigates DoS risks without compromising legitimate network access.","published":"2026-09-25T16:56:54Z","updated":"2026-09-25T16:56:54Z","authors":["João Ferreira","André Zúquete","Hélder Gomes"],"categories":["cs.CR","cs.NI"],"headline":"New Method Improves Wi-Fi Security and Reduces Attacks","summary":"Researchers studied a new Wi-Fi security protocol called SAE used in WPA3 that prevents password guessing attacks but can slow down devices like routers, making them vulnerable to overload attacks. They proposed changes that shift complex computations from the router to the user device, controlling the workload better. They also added techniques to slow down attackers and allow faster reconnects for trusted devices. Their tests show these changes help protect against attacks without blocking normal users, but results are based on experiments, not real-world deployment yet."},{"id":"2609.31518v1","title":"Retrainable physics-integrated neural differentiable modeling of sintering across material systems","abstract":"Sintering is widely used to manufacture ceramics, but coupled densification and grain growth, material-dependent kinetics, and sparse measurements complicate predictive modeling and process design. We present Sinter-PiNDiff, a retrainable physics-integrated neural differentiable framework for predicting density and grain-size evolution. Two neural networks learn densification and grain-growth coefficients within coupled rate equations, while a smooth saturation factor attenuates densification near theoretical density. The same governing structure, network architecture, and training procedure were fitted independently to published data for MgO, Al-doped ZnO, and CaO-doped ThO2. Tests at held-out temperatures and compositions yielded the lowest mean error in all twelve material-metric comparisons against multilayer perceptron and residual network baselines. For MgO, Al-doped ZnO, and CaO-doped ThO2, respectively, density normalized root-mean-square errors were 14.6%, 10.8%, and 14.4%, and grain-size errors using the same metric were 8.6%, 12.1%, and 19.3%. Removing evolving density from both neural-network inputs increased density and grain-size trajectory errors in all three systems and ten of twelve aggregate errors, supporting density-dependent kinetic feedback. Deep ensembles estimated model disagreement, but empirical coverage showed that the uncertainty bands were not calibrated and did not capture all model-data discrepancies. These results establish Sinter-PiNDiff as a retrainable framework for sparse-data prediction and uncertainty-informed selection of sintering conditions.","published":"2026-09-25T16:56:17Z","updated":"2026-09-25T16:56:17Z","authors":["Zeping Chen","Ani Aprahamian","Khachatur V. Manukyan","Tengfei Luo"],"categories":["cond-mat.mtrl-sci","cs.LG"],"headline":"New AI model predicts ceramic sintering behavior across materials","summary":"Sintering, a process to make ceramics, is complex due to interactions and limited data. Researchers developed Sinter-PiNDiff, a retrainable AI framework combining physics and neural networks to predict changes in density and grain size during sintering for different materials. Tested on three materials, their model showed lower prediction errors than other AI methods, highlighting the importance of including density feedback. However, uncertainty estimates were not perfectly calibrated. These findings suggest Sinter-PiNDiff can help predict sintering outcomes from sparse data but further work on uncertainty is needed."},{"id":"2609.31517v1","title":"A search-to-decision reduction for the linear code equivalence problem","abstract":"We present a polynomial-time reduction from the search variant of the linear code equivalence problem (i.e. the search for a linear isometry between the inputs) to its decisional variant. More precisely, given two linearly equivalent codes $\\mathcal C_1,\\mathcal C_2 \\subseteq \\mathbb{F}_q^n$, we show how to recover a linear isometry between them by making a polynomial number of queries to an oracle for decisional linear code equivalence. First, we prove that search-Permutation Code Equivalence (search-PCE -- the problem of finding a permutation $π\\in\\mathcal S_n$ mapping $\\mathcal C_1$ to $\\mathcal C_2$) reduces in polynomial time to PCE (i.e. the problem of deciding if there is a permutation map from $\\mathcal C_1$ to $\\mathcal C_2$) via at most $n^2$ oracle calls on instances of dimension $k$ and length at most $n^2(n+1)/2$. We then extend this approach to linearly equivalent codes: we recover the permutation part of a linear isometry via at most $n^2$ calls to a Linear Code Equivalence (LCE) oracle on instances of the same size, and we give a deterministic polynomial-time algorithm to recover the diagonal part once this permutation is known. Altogether, this yields a polynomial-time procedure to recover a linear isometry from an oracle for decisional LCE. From a linear-algebraic perspective, our results provide an explicit reconstruction of a monomial equivalence between two matrix representations from oracle access to the corresponding orbit membership problem.","published":"2026-09-25T16:56:13Z","updated":"2026-09-25T16:56:13Z","authors":["Jean-François Biasse","Giacomo Micheli","Benjamin Prada","Philip Waitkevich"],"categories":["cs.CC"],"headline":"Reducing code equivalence search to decision problem in polynomial time","summary":"The researchers address the challenge of determining when two linear codes are equivalent, a problem important in coding theory. They developed a method that transforms the task of finding the exact equivalence (search problem) into simply deciding whether such an equivalence exists (decision problem), using a limited number of calls to a decision algorithm. Their approach effectively reconstructs the equivalence by combining these decision queries, offering a polynomial-time solution based on theoretical methods and oracle access. The findings are theoretical and rely on algorithmic access rather than practical experiments."},{"id":"2609.31516v1","title":"Teichmüller and moduli spaces for complex nilmanifolds","abstract":"We develop several tools to study the local and global structure of the Teichmüller and moduli spaces for complex nilmanifolds. As a starting point, we prove a structure theorem for holomorphic maps between complex nilmanifolds, which allows for an explicit description of the Teichmüller and moduli spaces. We then introduce two period maps, and give general criteria for the existence of a complex structure on these spaces. To illustrate the theory, we prove a global Torelli theorem for various classes of nilmanifolds, including principal torus bundles over tori and almost abelian nilmanifolds. In particular, we show that there are nilmanifolds of arbitrary dimension and nilpotency index for which the Teichmüller space admits the structure of a complex manifold.","published":"2026-09-25T16:55:56Z","updated":"2026-09-25T16:55:56Z","authors":["Konstantin Wehler"],"categories":["math.AG","math.CV","math.DG"],"headline":"Studying shapes and forms of complex nilmanifolds","summary":"Researchers explored the complex structures called Teichmüller and moduli spaces associated with complex nilmanifolds, which are intricate geometric objects. They developed new methods to describe these spaces explicitly and introduced tools to determine when these spaces can have complex structures themselves. Their work includes proving a theorem relevant to these structures for several types of nilmanifolds, showing that these spaces can be understood as complex manifolds in many cases. The findings are theoretical and based on advanced mathematical concepts."},{"id":"2609.31515v1","title":"Asymptotic Behavior of One-Dimensional Hartree Equations with the Long-Range Interaction","abstract":"We investigate the long-time behavior of small solutions to the one-dimensional Hartree equation with the soft Coulomb interaction. The Hartree equation arises as an effective mean-field model for the evolution of many-body quantum systems, while the soft Coulomb potential provides a regularized one-dimensional approximation of the Coulomb interaction that removes the singularity at the origin while preserving its long-range character. Owing to the slow spatial decay of the interaction kernel, the equation exhibits long-range nonlinear effects. We prove global existence and modified scattering for sufficiently small, exponentially localized initial data. A notable feature is that, in contrast to the logarithmic phase correction arising in the well-known three-dimensional Coulomb Hartree equation, the one-dimensional soft Coulomb interaction produces a leading phase correction of order $(\\log t)^2$, followed by a lower-order logarithmic correction. The proof is carried out in analytic function spaces. The exponential localization of the initial data is converted into analyticity of the transformed profile, and a method based on a generator function with a time-dependent radius of analyticity is used to compensate for the derivative loss arising from the nonlinear phase equation.","published":"2026-09-25T16:55:26Z","updated":"2026-09-25T16:55:26Z","authors":["Changhun Yang","Chanjin You"],"categories":["math.AP"],"headline":"Study of Long-Term Behavior in a Quantum Equation with Soft Coulomb Interaction","summary":"Researchers studied the long-time behavior of small solutions to a simplified quantum equation that models interactions in one dimension using a softened version of the Coulomb force. They showed that for certain small, well-localized initial conditions, the solutions exist globally over time and display a unique pattern of change different from related three-dimensional cases. Their results are theoretical, relying on mathematical proofs within a specific analytic framework, and highlight distinctive effects caused by the long-range nature of the interaction in one dimension."},{"id":"2609.31514v1","title":"Forensic Twins: Self-Supervised Residual Learning for AI-Generated Image Forensics","abstract":"Detectors of AI-generated images are typically trained using samples from all Generative AI architectures they must catch, and struggle as soon as a new architecture emerges. Recent approaches have explored self-supervised pre-training as an alternative solution, yet standard frameworks work against the forensic task, e.g., their augmentations overwrite the micro-statistics of image formation. This paper introduces Forensic Twins, a Self-Supervised Residual Learning (SSRL) framework whose pretext task suppresses macroscopic content availability. Each image is mapped through a frozen, off-the-shelf forensic residual extractor, from which two spatially disjoint crops are drawn. Sharing no pixel, the two views retain minimal semantic structure to align, leaving a redundancy-reduction objective with a predominant common signal: the stationary fingerprint of the image acquisition pipeline. Additionally, Forensic Twins is trained exclusively on real images; no AI-generated image is observed at any stage. Experiments show that Forensic Twins attributes AI generator sources with 56.61% accuracy, i.e., 6.13% above the previous state-of-the-art zero-shot method at 375x lower latency. We also demonstrate that fitting a Gaussian Mixture Model (GMM) offline using only the real image embeddings extracted from Forensic Twins turns it into a state-of-the-art zero-shot detector, reaching 97.99% AUC across 27 unseen AI generators, including GANs, diffusion models and commercial systems. Code, weights and exact splits will be made publicly available","published":"2026-09-25T16:55:20Z","updated":"2026-09-25T16:55:20Z","authors":["Javier Muñoz-Haro","Ruben Tolosana","Ruben Vera-Rodriguez","Aythami Morales","Julian Fierrez"],"categories":["cs.CV"],"headline":"New AI method detects fake images without prior examples","summary":"Researchers developed 'Forensic Twins,' a technique that learns unique image fingerprints from only real photos, without needing fake images for training. This helps identify AI-generated images from new, unseen generators with improved accuracy and much faster processing than previous methods. Their approach uses a special way to focus on subtle patterns left by cameras, showing promising results in detecting forgeries from many different AI systems. However, these findings are based on experiments detailed in the paper."},{"id":"2609.31513v1","title":"Statistical Foundations for a Google Play User-Review Sentiment Index: Signal Fusion, Shrinkage, Distributional Validation, and Dynamic Smoothing","abstract":"We develop a statistically explicit sentiment index for Google Play user reviews and establish the mathematical results supporting its construction. Normalized star ratings and text-sentiment scores are treated as noisy measures of latent review valence and fused by covariance-aware inverse-variance weighting. Review-level estimates are aggregated with bounded helpfulness and recency weights, then shrunk toward a population mean using estimated precision rather than an arbitrary review-count threshold. App-level rating histograms provide a distributional diagnostic for samples returned under different API sort orders; because star ratings are discrete, classical continuous Kolmogorov-Smirnov critical values are not used. A local-level state-space model and the Kalman filter provide a denoised temporal trend. Full proofs cover the BLUE and Gaussian maximum-likelihood result, Gaussian-conjugate shrinkage, the Glivenko-Cantelli and Donsker theorems, count transformations via the delta method, and exact Gaussian Kalman filtering. A worked three-review example shows how textual complaints can materially reduce an apparently perfect star-only score.","published":"2026-09-25T16:54:09Z","updated":"2026-09-25T16:54:09Z","authors":["Marco Mandap"],"categories":["stat.ME","cs.CL","stat.AP"],"headline":"Building a Reliable Sentiment Score for Google Play Reviews","summary":"Researchers created a detailed statistical method to combine star ratings and text sentiment from Google Play reviews into a single, more accurate sentiment index. They used advanced math to weigh and adjust reviews based on helpfulness, timing, and precision instead of simple counts. Their approach also checks for biases in how reviews are collected and smooths trends over time. The findings are theoretical and supported by mathematical proofs, illustrated with an example showing how text analysis can change interpretations of star ratings."},{"id":"2609.31512v1","title":"Tolerable Inflation, Intolerable Uncertainty","abstract":"Numerical inflation targets anchor beliefs. Across euro-area and US professional forecasts, inflation swaps and options, and realized inflation, uncertainty about inflation is compressed at the announced number and kinks exactly there. This paper identifies a cost of the same design that, to our knowledge, has not been shown before, and that appears in second moments only. Within the workhorse New Keynesian model, tolerating part of the inflation a supply shock produces is optimal, yet the optimal tolerated share is not identified: optimal look-through and an unwarranted drift of the effective target are observationally equivalent in the inflation history, so a central bank cannot demonstrate that a warranted deviation is warranted, ex post as much as in real time. Agents holding finite, heterogeneous patience then generate predictive variance that is flat below the target and rises linearly with the expected overshoot above it -- an observational-equivalence bill, zero at the announced number and accumulating with the point-years inflation spends above it. The first-order benefit of the number stands; what the cost changes is how inflation uncertainty must be measured. The distance from target explains 70% of the variation in professional forecast variance, and the bill lies within that component; Normalized Uncertainty -- to our knowledge the first such correction -- removes it. The purge changes inference substantially: on French loan-level data, raw dispersion is unrelated to corporate loan rates, while one standard deviation of the purged measure is associated with rates 89 basis points higher, and cross-country growth and time-series results shift similarly. Conventional measures of inflation uncertainty partly record inflation's distance from its anchor rather than uncertainty about the outlook.","published":"2026-09-25T16:52:38Z","updated":"2026-09-25T16:52:38Z","authors":["Eric Vansteenberghe"],"categories":["econ.GN"],"headline":"How Inflation Targets Affect Uncertainty and Its Measurement","summary":"Researchers studied how announced inflation targets in the US and euro area influence expectations and perceived uncertainty about inflation. Using economic models, they found that certain inflation deviations assumed optimal by central banks cannot be distinctly identified, creating hidden costs that show up in the unpredictability of inflation above the target. Their findings suggest that standard ways to measure inflation uncertainty also capture how far inflation is from its target, and they propose a new method to separate these effects, which changes the relationship seen between uncertainty and economic outcomes like loan rates."},{"id":"2609.31511v1","title":"Muslim: A Deployed Arabic Voice AI Platform for Grounded Islamic Knowledge","abstract":"We present Muslim, a production Arabic voice AI platform serving grounded, sourced Islamic knowledge to real users. Beyond a real-time voice pipeline (NeMo Arabic ASR, an OpenAI-compatible LLM endpoint, self-hosted TTS) and a deterministic multi-source retrieval layer routed across six Model Context Protocol servers, we report three things a research prototype typically lacks. First, a released family of fine-tuned Arabic Islamic model artifacts: an efficient tool-routing LLM (Muslim-6B-PRO, 5.94B parameters) and a Modern Standard Arabic TTS model (Fasih-TTS-V1) that ranks 5th of 17 overall and 2nd of 11 open-weight systems on the community-voted Arabic TTS Arena for MSA. Second, an account and metering layer - a free per-account turn allowance, capacity-aware refusal, and email verification deferred to the point it actually matters - that turns an open demo into an operable, abuse-resistant product. Third, a three-layer observability stack (liveness, error reporting, product analytics) built specifically around the system's characteristic failure mode: a GPU-bound agent host going silent while the web tier keeps serving normally. We report real, measured latency and accuracy figures (98.4% recitation-validation accuracy on 124 cases; end-to-end voice latency of 0.9-1.7s) and discuss the concrete engineering trade-offs and limitations of running an Islamic-knowledge voice product in production.","published":"2026-09-25T16:51:22Z","updated":"2026-09-25T16:51:22Z","authors":["Yahya Mohamed Elnawasany"],"categories":["cs.CL"],"headline":"Arabic Voice AI Platform Delivers Verified Islamic Knowledge","summary":"Researchers developed Muslim, a voice AI system that provides sourced Islamic information in Arabic using speech recognition, language models, and speech synthesis. They released specialized Arabic language models, implemented user account controls to prevent misuse, and created monitoring tools to handle technical issues. Their report includes real usage data on accuracy and response times, and they discuss practical challenges in running this AI service for Islamic knowledge."},{"id":"2609.31510v1","title":"The 91T/99aa-like Type Ia Supernova 2019vrq, Part I: Photometry, Spectroscopy, and Spectropolarimetry of a Nearly Standard Candle","abstract":"Among the various major subtypes of Type Ia supernovae (SNe), the luminous 1991T-like (Filippenko et al. 1992; Phillips et al. 1992) and 1999aa-like (Krisciunas et al. 2000; Garavini et al. 2004) events stand out as particularly important for constraining the explosion mechanism. Their white dwarf (WD) progenitors are thought to have masses close to or exceeding the Chandrasekhar limit. We present spectrophotometric and polarimetric time-series observations of the overluminous Type Ia SN2019vrq. Between -9 and +12 days relative to the light-curve peak, its continuum polarization is consistent with zero, as seen in most normal-bright SNeIa and despite a possible rise up to 0.2% toward longer wavelengths as seen in subluminous SNeIa. Polarization across major spectral features is marginal ($\\lesssim$0.3%) with the exception of the high-velocity CaII near-infrared triplet, which reaches $\\sim$0.5%. This low polarization indicates a substantially spherically symmetric electron-scattering photosphere and minimal large-scale ejecta asymmetry. Early-time spectra reveal hot, doubly ionized ejecta dominated by SiIII and blended iron-group features. Intermediate-mass-element lines, such as CaII and SiII, are more clearly seen than in 91T-like events, though these lines remain weaker than in normal SNeIa. The intrinsically higher peak luminosity of overluminous 91T/99aa-like SNeIa (Yang et al. 2022; Phillips et al. 2024), their comparatively short formation timescale for a massive WD progenitor - a channel that becomes more common toward high redshift, and the moderate viewing-angle dependence of their spectrophotometric properties as inferred from their persistently low polarization, together suggest that this subclass may serve as a promising standard candle at high redshift.","published":"2026-09-25T16:51:04Z","updated":"2026-09-25T16:51:04Z","authors":["Yi Yang","Peter Hoeflich","J. Craig Wheeler","Dietrich Baade","Lifan Wang","Aleksandar Cikota","D. Andrew Howell","Curtis McCully","Alexei V. Filippenko","Ferdinando Patat","Chris Ashall","Mattia Bulla","Avishay Gal-Yam","Melissa. L. Graham","Daichi Hiramatsu","Divya Mishra","Estefania Padilla Gonzalez","Craig M. Pellegrino","Steve Schulze"],"categories":["astro-ph.SR","astro-ph.HE"],"headline":"Studying a Bright Type Ia Supernova and Its Shape","summary":"Researchers observed a bright Type Ia supernova named SN2019vrq to better understand its explosion. They found the star's outer layers were very symmetrical, showing little distortion, based on measurements of light polarization. The supernova's characteristics suggest it comes from a massive star and could be useful for measuring distances in the universe, though these findings are based on detailed observations of one object and comparisons with similar supernovae."},{"id":"2609.31509v1","title":"ClearGS: Reliability-Aware Gaussian Splatting from Handheld Videos","abstract":"We present ClearGS for 3D Gaussian Splatting (3DGS) from handheld videos with uneven viewpoint coverage and mixed frame quality. Rather than selecting frames with binary decisions, ClearGS uses Reliability-aware View Allocation (RVA) to assign graded raw-supervision weights based on appearance reliability, degradation risk, and geometric utility, while weakly reactivating useful suppressed frames to maintain trajectory coverage. Since weighting cannot restore details lost to blur or distortion, ClearGS further introduces Render-Guided In-Video Restoration (RIVR). The current 3DGS render provides a pose-aligned structural candidate, a frozen no-reference restoration expert restores the corresponding raw video observation without any clean reference image, and no-reference perceptual scores select among the render, restored observation, and high-frequency fused candidate. ClearGS then applies Full-Trajectory Repair Consolidation to revisit accepted repairs and preserve details introduced early. On GS2E and GSOTM, ClearGS achieves state-of-the-art overall performance, with consistent CLIP-IQA and MUSIQ gains and LPIPS reductions in most degradation settings, without paired sharp supervision or matched clean references.","published":"2026-09-25T16:50:46Z","updated":"2026-09-25T16:50:46Z","authors":["Xuanzhi Liu","Xinyi Wu","Hang Pan","Wensi Huang","Zhenyao Wu","Ruize Han","Song Wang"],"categories":["cs.CV","cs.AI"],"headline":"New Method Improves 3D Video Quality from Handheld Cameras","summary":"Researchers developed ClearGS, a technique to create 3D images from handheld videos that often have uneven camera angles and mixed frame quality. Instead of simply choosing good frames, ClearGS assigns different importance levels to frames based on reliability and risk of quality loss, and it restores blurry or distorted frames using a special process without needing perfectly sharp reference images. Tests show ClearGS improves video quality compared to previous methods, though results are based on computer evaluations and not real-world trials."},{"id":"2609.31508v1","title":"Assessing a Mathematical Model of Syllable Production via CTW Alignment with EMA Data","abstract":"This study evaluates a mathematical model of syllable production using an EMA dataset from a recently published study, which proved highly compatible with the model's architecture. The data set consists of regularly structured French phrases of equal duration and reduced phonetic and syllabic complexity. A dedicated procedure was used to transform EMA recordings into Maeda parameters. The Model-generated trajectories were then realigned with these transformed data using Canonical Time Warping (CTW). Statistical validation was conducted via a permutation test, comparing alignment quality against a surrogate condition with phonetically incongruent model outputs. The results show significantly better alignment for phonetically congruent pairings, revealing a strong structural correspondence between the model and the articulatory data. These findings provide experimental support for the representational validity of the model.","published":"2026-09-25T16:50:23Z","updated":"2026-09-25T16:50:23Z","authors":["Frédéric Berthommier"],"categories":["eess.AS"],"headline":"Testing a Model of Speech Sounds Using Tongue Movement Data","summary":"Researchers tested a mathematical model that simulates how syllables are formed in speech by comparing it to tongue and mouth movement data from French phrases. They transformed the movement data into a format compatible with the model and used a method to align the model's predictions with the real data. The study found that the model's output matched the real movements better than mismatched controls, suggesting the model accurately represents aspects of speech production. The findings are based on analysis of simplified speech sounds."},{"id":"2609.31507v1","title":"SatNav: A Scalable Benchmark for Long-Horizon UAV Vision-Language Navigation from Satellite Imagery","abstract":"Urban uncrewed aerial vehicle (UAV) vision-language navigation (VLN) requires agents to follow instructions across extended urban spaces, inherently demanding long-term memory and geospatial grounding. However, scaling existing benchmarks remains difficult because of their reliance on costly reconstructed 3D assets, limiting geographic diversity and episode scale. To address this, we introduce SatNav, a scalable, long-horizon UAV VLN benchmark built from high-resolution satellite imagery. SatNav targets city-level navigation missions and uses satellite crops as approximations of UAV nadir views for visual observations. Through an automated cue-to-episode pipeline, SatNav constructs 118K episodes from 59 scenes across 18 cities, with an average trajectory length of 379 m. To stress-test long-horizon memory and geospatial reasoning, SatNav defines three task families: Boundary, Landmark, and Route, targeting loop progress tracking, landmark-based spatial grounding, and route following with counting cues. Benchmarking classical VLN agents and recent agents based on large vision-language models (LVLMs) on SatNav shows that city-scale navigation remains challenging. We further introduce SwiftVLN, a modular framework with switchable memory components, and conduct systematic memory-design ablations. Finally, satellite-to-UAV transfer experiments show that satellite-trained navigation models can operate on real-flight UAV observations, showing the practical relevance of SatNav. Our project page: https://eku127.github.io/SatNav/","published":"2026-09-25T16:48:44Z","updated":"2026-09-25T16:48:44Z","authors":["Jiajun Jiang","Chunliang Hua","Zichun Chen","Yanxing Wu","Zeyuan Yang","Jie Song","Xiao Hu"],"categories":["cs.CV","cs.RO"],"headline":"New SatNav Benchmark Tests Drone Navigation Using Satellite Images","summary":"Researchers developed SatNav, a new benchmark that challenges drones to navigate long urban routes using satellite images instead of costly 3D maps. They created 118,000 navigation tasks across 18 cities to test memory and spatial skills with three types of tasks focused on boundaries, landmarks, and routes. Their tests reveal that even advanced models struggle with city-scale navigation, but satellite-trained models can transfer to real drone flights, highlighting SatNav's practical relevance."},{"id":"2609.31506v1","title":"Evaluating Cultural Awareness of LLMs for Haitian Creole","abstract":"Large language models (LLMs) exhibit substantial performance disparities between high- and low-resource languages. Beyond lower task performance, they often fail to capture the cultural norms and values of underrepresented communities. In this work, we present the first systematic evaluation of cultural awareness in LLMs for Haitian Creole, a language spoken by millions but severely underrepresented in digital resources. We assess cultural awareness along four complementary dimensions---specificity, bias, diversity, and variation---using a benchmark of culturally salient prompts curated by native speakers in a text infilling setting. Our results reveal a clear gap between cultural awareness in Haitian Creole and higher-resource French, with Haitian performance being more uneven across domains and more affected by French linguistic interference. Story generation further reveals recurring portrayals of Haitian characters through hardship and resilience, showing that even positive characterizations can encode stereotypical narratives. Our code, benchmark, and evaluation framework are publicly available.","published":"2026-09-25T16:47:07Z","updated":"2026-09-25T16:47:07Z","authors":["Christelle Clervilsson","Yanzhu Guo"],"categories":["cs.CL","cs.AI"],"headline":"Researchers Assess Cultural Awareness of AI in Haitian Creole","summary":"Researchers studied how well large language models understand Haitian Creole culture, a language with fewer digital resources. They tested cultural knowledge based on specificity, bias, diversity, and variation using prompts made by native speakers. The study found that AI performs unevenly in Haitian Creole compared to French and sometimes reflects stereotypes of Haitian people. These findings highlight limitations in AI understanding of less-represented cultures and provide tools for further research."},{"id":"2609.31505v1","title":"Prompt Minimization: Reducing Input Redundancy Without Sacrificing Output Fidelity","abstract":"Despite the growing capabilities of large language models (LLMs), prompt design remains largely heuristic and ad hoc. This project will explore $\\textit{prompt minimization}$, the process of reducing prompts to their smallest, most information-dense form while preserving output fidelity. Practically, shorter prompts reduce computational overhead and inference latency, especially when large contexts, such as entire documents or codebases, are included unnecessarily. Further, longer prompts can damage LLM reasoning and accuracy. Theoretically, the existence of multiple prompts yielding equivalent outputs suggests a high degree of redundancy in the input space, raising fundamental questions about what information is essential to elicit specific model behaviors. We propose three variant frameworks to identify and evaluate minimal prompts and demonstrate that minimal prompts often produce outputs comparable to those of their longer counterparts. These findings suggest new directions for efficient prompt engineering and deepen our understanding of input compression in LLMs.","published":"2026-09-25T16:45:28Z","updated":"2026-09-25T16:45:28Z","authors":["Marius F. R. Juston","Kevin A. Karim","Jonathan Gao","Kevin C. Li","Rudhi Bashambu"],"categories":["cs.AI"],"headline":"Researchers Explore How to Shorten Inputs for Language Models","summary":"The researchers studied how to reduce the size of prompts—the instructions or inputs given to large language models—without losing the quality of the model’s responses. They developed methods to find shorter, information-packed prompts and found that these minimal prompts often generate results similar to longer ones. Their work is based on experiments and highlights the possibility of more efficient and effective prompt design, while raising questions about what information is truly essential for model output."},{"id":"2609.31504v1","title":"Bound States in 2d Yukawa Theory from Hamiltonian Truncation","abstract":"We study the spectrum of two-dimensional Yukawa theory using Hamiltonian truncation, with a focus on bound states. We extend the framework of Hamiltonian truncation effective theory to include fermions and logarithmic UV divergences. With these corrections, the energy levels converge as a function of the cutoff with the scaling predicted by power counting. We identify a strategy to follow the evolution of fermion-antifermion bound states from weak to strong coupling, and investigate its binding energy, finding good agreement with finite-volume perturbation theory at weak coupling. Finally, we explore the heavy-scalar regime toward the massive Thirring/Sine-Gordon limit, recovering the expected weak-coupling behavior and outlining how the calculation can be extended toward this limit.","published":"2026-09-25T16:44:10Z","updated":"2026-09-25T16:44:10Z","authors":["Olivier Delouche","Kara Farnsworth","Francesco Riva"],"categories":["hep-th","cond-mat.str-el","hep-lat","hep-ph"],"headline":"Studying Particle Bound States in a Simple 2D Physics Model","summary":"Researchers investigated a basic two-dimensional particle physics model called Yukawa theory to understand how particles bind together. They improved a method called Hamiltonian truncation to handle complex features like specific particle types and calculation challenges. Their approach successfully tracked how particle pairs form bound states from weak to strong interaction strengths, matching other theoretical predictions at low interaction levels. The study mainly focuses on theoretical findings using simulations and outlines ways to further develop the method."},{"id":"2609.31503v1","title":"Physics-induced impact on the properties of Kepler/K2 solar analogs in the light of PLATO's requirements","abstract":"Solar analogs are useful laboratories for tests of stellar structure and evolution, particularly in the context of the upcoming PLATO mission. Our study focuses on six seismic solar analogs observed by Kepler and K2. We investigate the impact of the model physical ingredients on these six targets in order to evaluate how well one can satisfy the PLATO precision requirements for stellar parameters with the data quality that is currently available to us. We used as reference values the stellar mass, age, and chemical composition obtained with the Forward and Inverse COmbination (FICO) procedure, based on OPAL opacities and a combination of NACRE II and Solar Fusion II (SFII) nuclear reaction rates. The efficiency of macroscopic transport is calibrated on lithium abundances derived from the analysis of HERMES spectra. From there, we carried out extensive tests of the physical ingredients with the help of a Levenberg-Marquardt local minimisation method using both classical and seismic constraints. In particular, we investigated the impact induced by the use of OPLIB opacities and SFIII nuclear reaction rates on the stellar fundamental properties, as well as the use of various seismic constraints. We quantify the potential differences that changes in the modelling input physics may lead to, and compare their extent with the requirements expected for the PLATO mission (15% in mass, 2% in radius, and 10% in age), providing crucial constraints on the accuracy and robustness with respect to systematics of the determination of fundamental properties for PLATO. For our sample of solar analogs, we are able to accurately retrieve stellar masses within 2.5%. As for stellar ages, we quantify systematics of the order of 10-14% at most, but there are also very promising cases for which deviations remain within 4% depending on the stars and constraints considered.","published":"2026-09-25T16:43:52Z","updated":"2026-09-25T16:43:52Z","authors":["Saniya Khan","Gaël Buldgen","Jérôme Bétrisey","Rafael A. García","Savita Mathur"],"categories":["astro-ph.SR","astro-ph.EP"],"headline":"Studying Sun-like Stars to Meet Space Mission Standards","summary":"Researchers analyzed six stars similar to the Sun using data from space telescopes Kepler and K2 to see how well current models can determine key star traits like mass and age. They tested how different physical assumptions affect these measurements and compared the results to precision goals of the upcoming PLATO mission. The study found that star masses can be estimated within about 2.5% accuracy and ages within 10-14%, showing some promise but also highlighting variations depending on methods used."},{"id":"2609.31502v1","title":"Parallel and Distributed Fermionic Simulation via Dynamic Encoding","abstract":"We demonstrate a simple and efficient method to parallelize and distribute Trotterized Hamiltonian simulation of fermionic systems across multiple QPUs. Using combinatorial covering designs to define a minimal set of fermion-qubit encodings, we demonstrate communication cost scaling as $\\mathcal{O}(M r)$ for a system of $M$ fermionic modes and Trotter number $r$, improving on the static encoding bound for $q$ QPUs, $\\mathcal{O}(M^4 q r)$. We compare this approach to dynamic encoding using a randomised method, Pauli-weight based optimisation and hypergraph partitioning. Applying this to the Hamiltonians of a range of molecular systems, we find the combinatorial covering approach results in the lowest communication cost in all but the sparsest Hamiltonians.","published":"2026-09-25T16:41:58Z","updated":"2026-09-25T16:41:58Z","authors":["Michael Williams de la Bastida","Peter V. Coveney"],"categories":["quant-ph"],"headline":"New Method Improves Simulations of Particle Systems on Quantum Computers","summary":"The researchers developed a way to split complex simulations of fermionic systems (which involve particles like electrons) across multiple quantum processors more efficiently. They used special mathematical designs to reduce the communication needed between processors during the simulation, improving on previous methods. Their comparisons showed this approach usually requires less communication, except for very simple cases, based on simulations of molecular systems."},{"id":"2609.31501v1","title":"Fundamental Limits of Sequence Reconstruction Problems in Immunogenomics","abstract":"The goal of personalized immunogenomics is to recover an individual's germline immunoglobulin gene segments from 'repertoire sequences' altered by trimming, extension, and mutation. In this work, we study the fundamental trace complexity (or the number of samples/traces required for accurate reconstruction) of D-gene reconstruction under three biologically motivated trace-generation models introduced by Bhardwaj et al. (2021) and develop practical algorithms for reconstruction problems left open in the original work. First, for the TrimSuffixAndExtend model, we establish the optimal trace complexity to be {Θ(n)}, and develop a low-complexity Prefix-Filtered Mode (PFM) decoder that achieves this scaling. Second, for the closely related two-sided TrimAndExtend model, we show that the optimal trace complexity is instead {Θ(n^2)}, and is achieved by the simple Bit-Wise Mode (BWM) decoder. Third, for the SuffixExtend-t(TrimSuffix) model, we establish polynomially separated lower and upper bounds on trace complexity. Our results follow from information-theoretic lower bounds coupled with tight analyses of the proposed reconstruction algorithms.","published":"2026-09-25T16:40:21Z","updated":"2026-09-25T16:40:21Z","authors":["Jaswanthi Mandalapu","Deepak Charan","V. Arvind Rameshwar","Nir Weinberger"],"categories":["cs.IT","eess.SP"],"headline":"Researchers Explore Limits of Reconstructing Immune Gene Sequences","summary":"The researchers examined how many sample sequences are needed to accurately reconstruct specific immune system gene segments that are altered through natural processes. They studied three models of how these sequences change and identified the minimum number of samples (trace complexity) required for each. They also developed practical decoding algorithms that perform optimally or near-optimally according to their theoretical findings. These results are based on mathematical analyses and apply to the studied models, with limitations inherent to theoretical and simulation-based work."},{"id":"2609.31500v1","title":"Rydberg excitons in core-shell nanostructures II. Plasmon-exciton interaction and enhancement","abstract":"Interaction between Rydberg excitons in core-shell Cu$_2$O nanostructures and surface plasmons is investigated. An enhancement of a quadrupole transitions to the Rydberg exciton states in copper oxide plasmonic nanaostructures is examined. A possibility of a modification of plasmonic enhancment (i.e., an average amplification factor) of electric field distributions in the vicinity of nanostructures, which depends on its shape and geometrical size, is also discussed.","published":"2026-09-25T16:39:03Z","updated":"2026-09-25T16:39:03Z","authors":["David Ziemkiewicz","Gerard Czajkowski","Sylwia Zielińska-Raczyńska"],"categories":["cond-mat.mes-hall","physics.optics"],"headline":"Studying light interactions in tiny copper oxide structures","summary":"Researchers investigated how special states in copper oxide nanostructures, called Rydberg excitons, interact with surface plasmons, which are waves on the material’s surface. They examined how certain light transitions are enhanced in these structures and discussed how changing the shape and size of the nanostructures might modify this enhancement. The findings are based on theoretical analysis and focus on understanding and possibly controlling light behavior at very small scales."},{"id":"2609.31499v1","title":"An information bound for multiplicities in rapidity windows of the dipole cascade behind the entanglement entropy picture","abstract":"In Mueller's dipole cascade without transverse dimensions, with one counted particle for each dipole produced in a window, the multiplicities in three consecutive rapidity windows are Poisson counts of one Gamma-distributed source. We show that at fixed middle multiplicity the outer two share less information than the Gaussian value $-\\tfrac12\\ln(1-ρ^2)$ of their partial correlation, a value that second moments alone determine. For windows of one common width at a constant splitting rate this value never exceeds $\\tfrac12\\ln(4/3)$. The Gaussian value is not a bound on mutual information in general, and for a source of the same mean and variance with another law the information can exceed it. The law of the counts keeps its form under migration of particles across window edges and detection losses when both act on each particle independently of the others and of the source, and its factorial cumulants of second and third order test necessary conditions for it. In Monte Carlo simulations of the cascade with transverse dimensions at leading logarithmic accuracy and fixed coupling, the counts are not Poisson counts of one source, so the bound is a result of the model without them. In the model without transverse dimensions the law also fixes the normalized second factorial cumulant of one window at $1/k$, with $k$ the number of initial dipoles. The values of this cumulant formed from the mean and variance that H1 publishes for deep inelastic scattering lie far below the value one of a cascade from a single initial dipole with one counted particle for each dipole produced in a window. The quantity bounded is a classical conditional mutual information between counted multiplicities.","published":"2026-09-25T16:37:29Z","updated":"2026-09-25T16:37:29Z","authors":["Olasantan Ebenezer Adelaja","Alex Prygarin","Karam Shekh Yusuf"],"categories":["hep-ph","nucl-th"],"headline":"Researchers study particle counts in a simplified cascade model","summary":"The researchers examined a simplified model of particle production, called Mueller's dipole cascade, focusing on counts of particles in three adjacent time windows. They found that, under certain conditions, the shared information between the counts in the outer windows is less than a specific Gaussian estimate. Their analysis relies on mathematical properties of the particle counts and assumptions of the model without spatial dimensions. They also performed simulations including spatial effects, which showed different behavior. The findings are theoretical and relate to understanding correlations in particle multiplicities, with limitations tied to the simplified model used."},{"id":"2609.31498v1","title":"Retail Product Search: A Practical Approach at Target","abstract":"Search is one of the most important features in e-commerce, directly driving customer engagement and business growth. A good product search system must show both relevant and desirable results. However, retail search presents unique challenges. User intent can range from exact matches to open-ended discovery. Search systems must also balance multiple goals, such as relevance, revenue, and profit, while keeping response times low. Traditional keyword-based methods often fall short in handling natural language or semantic queries. Vector search helps alleviate these issues, but it can miss key intent signals or return low-precision results. In this paper, we present the design of a hybrid search system at Target that combines lexical and vector search. We describe our approach to data processing, embedding training, precision control for the final result set, multi-channel result fusion (where we compared fusion strategies and adopted weighted interleaving), and the performance optimizations used to maintain low latency for production deployment. Our method improves offline evaluation metrics, and in online A/B testing it raised click-through rate by 0.97%, order conversion by 0.98%, and demand per visitor by 1.10% over lexical-only search, while roughly halving zero-result searches. The resulting system is deployed at scale and serves millions of guests daily.","published":"2026-09-25T16:36:48Z","updated":"2026-09-25T16:36:48Z","authors":["Darshan Sonagara","Qujiaheng Zhang","Ankit Singh","Alex Li"],"categories":["cs.IR","cs.LG"],"headline":"Target improves product search with a hybrid system combining methods","summary":"Researchers at Target explored challenges in e-commerce product search, such as understanding varied user intent and balancing relevance with business goals. They developed a hybrid search system combining traditional keyword (lexical) and newer vector search methods, and optimized it for speed and result quality. Testing showed their approach improved several customer engagement measures slightly and reduced cases with no search results, supporting its effectiveness at scale. The paper reports findings from practical deployment and tests, not theoretical models."},{"id":"2609.31497v1","title":"HOMO-LUMO Intervals from Modern Density Functionals as Fundamental Gap Estimators","abstract":"Use of the Kohn-Sham eigenvalue gap (highest occupied molecular orbital to lowest unoccupied, HOMO-LUMO) as a fundamental gap (I-A, I= ionization potential, A= electron affinity) estimator is a long-standing challenge for relatively simple exchange-correlation functionals, e.g. meta-GGAs. The LAK meta-GGA [T. Lebeda et al. Phys. Rev. Lett. 133, 136402 (2024)] was developed to emphasize certain constraints in ways intended to make the LAK KS gaps close to the HSE06 hybrid functional [A.V. Krudau et al. J.Chem. Phys. 125, 224106 (2006)] gaps. The LAK functional was tested against a small atomization data set and a very small bond length set, with no molecular HOMO-LUMO gap testing. Here we test LAK against widely used, large reference data sets for molecular heats of formation, bond lengths, harmonic frequencies, I, and A and compare with r2SCAN [J.W. Furness et al., J. Phys. Chem. Lett. 11, 8208 (2020)] as well as HSE06 and PBE0, a quite different hybrid. On these tests LAK does not give meaningfully better geometries, heats of formation, HOMO-LUMO gaps, or DFT IP theorem satisfaction than r2SCAN. Compared against experimental I and A values the same holds; LAK results resemble r2SCAN results, not HSE06. Moreover, the HOMO and LUMO shifts generated by the NCAPR GGA [J. Carmona-Espindola et al., J. Chem. Phys. 157, 114109 (2022)] as derivative discontinuity estimates, correct the bare NCAPR HOMO and LUMO eigenvalues to substantially better agreement with experiment for both fundamental gap and IP theorem than do the meta-GGAs (LAK, r2SCAN) or hybrids (HSE, PBE0). We discuss possible reasons why LAK behaves differently for molecules and solids.","published":"2026-09-25T16:36:25Z","updated":"2026-09-25T16:36:25Z","authors":["H. Francisco","Javier Carmona-Espíndola","A. C. Cancio","José Luis Gázquez","S. B. Trickey"],"categories":["physics.chem-ph","cond-mat.mtrl-sci"],"headline":"Testing a New Method to Estimate Molecular Energy Gaps","summary":"Researchers studied a recently developed method called LAK meta-GGA, designed to better estimate the energy gap between key molecular orbitals, an important property reflecting how molecules gain or lose electrons. They compared LAK's performance to other common methods using large experimental data sets on molecules. Their results show LAK does not significantly improve predictions compared to similar methods, and an alternative approach by others provides closer agreement with experiments. The study highlights challenges in accurately estimating these molecular energy gaps and differences between molecules and solids."},{"id":"2609.31496v1","title":"Crystallographic-orientation dependence of the early-stage oxidation of Zr single crystals: an XPS study of suboxide formation and in-depth distribution","abstract":"The influence of crystallographic orientation on the early stages of the oxidation of pure Zr single crystals was studied by X ray photoelectron spectroscopy (XPS). Two samples, cut from the same a-Zr single crystal, were oxidized simultaneously at room temperature and O2 pressures of 1 X 10-8, 1 X 10-7 and 1 X 10-6 torr: one with its surface normal parallel to the c-axis (Z1, basal (0001) plane) and the other with its normal 12deg from that of a prismatic plane (Z2, near-prismatic orientation). At all three pressures the oxidation kinetics of both samples followed a three stage, logarithmic type behaviour, but the near prismatic sample (Z2) incorporated oxygen faster and to a greater extent than the basal sample (Z1) in every case. Deconvolution of the Zr 3d core level spectra resolved, in addition to metallic Zr and ZrO2, two sub-stoichiometric Zr O compounds, denoted (ZrO)a and (ZrO)b, with binding-energy shifts of 1.3 and 2.3 eV with respect to metallic Zr, consistent with sub oxides previously reported for polycrystalline Zr. Angle resolved XPS showed that ZrO2 is the outermost compound in both orientations, while the two sub-oxides are distributed nearly homogeneously through the film. Oxide thicknesses, calculated from the attenuation of the metallic Zr 3d signal, ranged from 10 to 13 A for Z1 and from 14 to 19 A for Z2 depending on the oxidation pressure, in close agreement with earlier measurements on cold worked polycrystalline Zr, whose surface texture is dominated by prismatic-oriented grains.","published":"2026-09-25T16:35:54Z","updated":"2026-09-25T16:35:54Z","authors":["Joaquin Sacanell","Flavio Conde"],"categories":["cond-mat.mtrl-sci"],"headline":"Different crystal surfaces affect early-stage zirconium oxidation","summary":"Researchers studied how the direction of atoms in pure zirconium crystals influences the early oxidation process using a method called X ray photoelectron spectroscopy. They compared two crystal orientations and found that one type absorbed oxygen faster and formed thicker oxide layers. They also identified different oxide compounds and their distribution in the oxide films. These findings come from controlled lab experiments and agree with past observations on similar materials."},{"id":"2609.31495v1","title":"High-frequency gravitational waves from microscopic particle-graviton associated production in bubble collisions","abstract":"Cosmological first-order phase transitions are well-known sources of gravitational waves from bubble collisions and plasma dynamics. In this work, we investigate an additional microscopic source directly arising from the associated production of particles and gravitons during bubble collisions. Focusing on runaway bubble walls, we calculate the gravitational wave spectra from two-body scalar-graviton production and three-body processes involving scalar or fermion pairs accompanied by a graviton. We derive analytic approximations and numerically evaluate the resulting spectra, revealing a high-frequency gravitational wave component that can complement conventional lower-frequency signals from the same phase transition. This mechanism offers a potential multiband signature and establishes a connection between gravitational radiation and the production of dark matter or other new particles. Our findings provide a new physics target for future high-frequency gravitational wave experiments and a complementary probe of particle production in the early Universe.","published":"2026-09-25T16:34:46Z","updated":"2026-09-25T16:34:46Z","authors":["Peilin Chen","Dayun Qiu","Zihong Cheng","Fa Peng Huang"],"categories":["hep-ph"],"headline":"New source of high-frequency gravitational waves from early universe events","summary":"Researchers studied how tiny particles and gravitational waves might be produced together during cosmic events called bubble collisions in the early universe. They focused on fast-moving bubble walls and calculated the expected signals, finding a new high-frequency gravitational wave component. Their work suggests this new source could add to known signals from such cosmic events and may help future experiments explore particle production shortly after the Big Bang. These findings are based on theoretical calculations and simulations."},{"id":"2609.31494v1","title":"Toward verifiably private learning from federated data","abstract":"Federated Learning (FL) allows devices with private data to collaborate in training a shared model. We present a next-generation FL system based on Trusted Execution Environments (TEEs) that addresses operational challenges associated with earlier systems and provides externally verifiable central Differential Privacy (DP) guarantees for the first time while offering a better privacy-utility tradeoff. In our system, devices upload data encrypted with keys managed by a TEE-hosted Key Management Service (KMS). The uploaded data is cryptographically tied to a policy limiting the set of Python programs that may later process the data in server-side TEEs. External parties may inspect public transparency logs to observe the set of workloads allowed by these policies. Our experimental results show that the new system improves device coverage and favorably shifts privacy-utility curves by enabling collected data to be integrated into the server-side workload at a schedule that optimizes DP guarantees and is unaffected by device availability. Our new system has been productionized, enabling models for the Android Keyboard (Gboard) to be trained faster and achieve better accuracy under smaller, now externally verifiable privacy budgets in comparison to models trained using the prior system.","published":"2026-09-25T16:34:25Z","updated":"2026-09-25T16:34:25Z","authors":["Katharine Daly","Yu Xiao","Zachary Garrett","Brett McLarnon","Jianpeng Hou","Arun Ganesh","Yanxiang Zhang","Noriyuki Takahashi","Haicheng Sun","Yuanbo Zhang","Timon Van Overveldt","Daniel Ramage"],"categories":["cs.CR"],"headline":"New System Improves Privacy and Accuracy in Federated Learning","summary":"Researchers developed a next-generation federated learning system that uses secure hardware to better protect user data and provide verifiable privacy guarantees. Their system encrypts data and restricts how it can be used, allowing external observers to verify privacy protections. Tests suggest this approach improves the balance between privacy and model accuracy and works reliably even when some devices are unavailable. The system has been implemented in real applications like the Android Keyboard, showing faster training and better accuracy with stronger, verifiable privacy than before."},{"id":"2609.31493v1","title":"Sharp weighted weak-type bounds for commutators of singular integrals","abstract":"We establish sharp weak-type bounds for commutators of Calderón--Zygmund operators and $\\text{BMO}(\\mathbb{R}^n)$ symbols with respect to $A_p$ weights. We prove that the bound \\looseness=-1 $$ \\|[b,T]\\|_{L^p(w)\\rightarrow L^{p,\\infty}(w)}\\lesssim [w]_{A_p}^{\\max(2,p')} $$ for $p \\in (1,\\infty)$ and $w \\in A_p$ is sharp for general $T \\in \\text{CZO}(\\mathbb{R}^n)$ and $b \\in \\text{BMO}(\\mathbb{R}^n)$. In the case of the standard logarithmic symbol, we show that the estimate improves to $$ \\|[\\log|\\cdot|,T]\\|_{L^p(w)\\rightarrow L^{p,\\infty}(w)}\\lesssim [w]_{A_p}^{\\max\\big(2,\\frac{1}{p}+\\frac{p'}{p}\\big)} $$ and that this dependence is optimal for this symbol. Our upper bound for the logarithmic symbol follows from the pointwise control of the commutator by a sum of the Hardy--Littlewood maximal operator and the squares of the radial Hardy operator and its adjoint, while our lower bounds are consequences of a general result for singular integral operators.","published":"2026-09-25T16:34:10Z","updated":"2026-09-25T16:34:10Z","authors":["Adam Mair","Óscar Andrés Ramírez","Cody B. Stockdale"],"categories":["math.CA"],"headline":"Researchers find precise bounds for mathematical operator behavior","summary":"The researchers study special mathematical operators called commutators related to singular integral operators, focusing on how they behave when combined with certain functions and weights. They establish exact (sharp) bounds that describe the size of these commutators in a technical function space setting. Their findings improve known estimates for a particular logarithmic function case and rely on advanced mathematical tools, but their results are theoretical and proven within a mathematical framework rather than through practical experiments."},{"id":"2609.31492v1","title":"SAID: Semantic Acoustic Imaging Detector for Sound Event Localization and Detection","abstract":"In daily life, people hear speech, footsteps, and music around them. We can often recognize these sounds and judge where they come from. Each sound source can be shown on a separate acoustic map, a rectangular image covering $360^{\\circ}$ horizontally and $180^{\\circ}$ vertically. The map shows the directions occupied by the source as a region and the sound energy within that region. A class label identifies the sound. Predicting these labeled acoustic maps from audio is called semantic acoustic imaging. Such maps could help robots perceive their surroundings and allow augmented reality displays to show sound regions and classes over the real world. Existing models can recognize sound classes and estimate a direction for each source. However, a direction alone does not describe the source region or its energy. Acoustic imaging must also distinguish sound sources in nearby directions, while the number of active sources and the regions they occupy can change over time. We therefore propose the Semantic Acoustic Imaging Detector (SAID), which predicts a separate labeled acoustic map for each active source from audio. First, we pretrain Audio2Sph, SAID's audio encoder, through sound energy estimation across directions without class labels. Then, we train the complete SAID model to predict source regions, energy, and classes together. We also develop a pipeline that generates simulated recordings for pretraining and supports fine-tuning on real recordings. On the official DCASE2026 Task 3 Track A evaluation set, our submitted system ranks first with 0.1080 macro-averaged mean average precision (Macro mAP) and 0.3962 Macro Pearson $r$. Demos and code are provided at https://github.com/IN03X/SAID.","published":"2026-09-25T16:33:44Z","updated":"2026-09-25T16:33:44Z","authors":["Runbang Wang","Zining Liang","Yin Cao","Qiuqiang Kong"],"categories":["eess.AS","cs.SD"],"headline":"New system maps and labels where sounds come from","summary":"Researchers developed SAID, a system that creates detailed images showing where different sounds occur and how loud they are. Unlike previous methods that only find sound direction, SAID predicts areas for each sound source along with their type and energy. They trained SAID first on simulated data and then on real recordings. Their system ranked best in an official 2026 sound detection challenge, though results are based on specific test sets and simulation methods described by the researchers."},{"id":"2609.31491v1","title":"UQ-LOB: Uncertainty-Aware Limit Order Book Mid-Price Forecasting","abstract":"Forecasting short-horizon mid-price movements from limit order book (LOB) data is central to algorithmic trading, yet most deep LOB forecasters are point predictors: they output a direction or a displacement, but never indicate which of their forecasts can be trusted. We introduce UQ-LOB, a lightweight, encoder-agnostic uncertainty quantification module that attaches to any pretrained LOB encoder and, in the spirit of attentive neural processes, conditions each forecast on a context set of recently completed windows whose outcomes are already realised. The UQ-regression variant outputs a calibrated Gaussian over the future tick displacement, while the UQ-classification variant outputs a categorical distribution over down/up/stationary. Both expose a scalar confidence (predicted signal-to-noise ratio or class probability) that supports selective prediction. On 5.2 billion LOB events across seven cryptocurrency assets and horizons of 5, 10 and 15 seconds, UQ-regression attains near-nominal 68% interval coverage, and restricting to the most confident 10% of predictions raises directional macro F1 by 0.11-0.15 for UQ-regression and 0.05-0.11 for UQ-classification, at every horizon. On large, economically meaningful moves, the tightest confidence tier reaches a directional F1 of 0.88 (down) and 0.83 (up) at the 5-second horizon.","published":"2026-09-25T16:33:23Z","updated":"2026-09-25T16:33:23Z","authors":["Derrick Gilchrist Edward Manoharan","Eljas Linna","Kestutis Baltakys","Hao Dong","Juho Kanniainen"],"categories":["cs.AI"],"headline":"New Method Adds Trust Levels to Stock Price Predictions","summary":"Researchers developed UQ-LOB, a tool that adds uncertainty estimates to predictions of short-term stock price moves using limit order book data. Their method works with existing prediction models and provides a confidence measure for each forecast, helping to identify which predictions are more reliable. Testing on billions of cryptocurrency trades showed the method's uncertainty estimates were well calibrated, and focusing on high-confidence predictions improved accuracy for price direction over 5 to 15 seconds. The results come from analysis of historical data, not real-world trials."},{"id":"2609.31490v1","title":"Authority at Commit Time: Reject-and-Rerun Semantics for Governed Agentic Systems","abstract":"Enterprise agents can compute for seconds or hours from a snapshot of policy, facts, task state, models, tools, and verifiers that may change before their work reaches the world. Completion alone therefore cannot confer institutional authority. We treat agents as proposal producers and a logically authoritative service as the sole authority for governed effects. At admission, the service compares a proposal's declared dependencies and component manifest with current state, applies the current verifier, and either reserves an idempotent effect for outbox dispatch or rejects affected stale work for resynchronization and rerun. Unaffected work may continue; lost responses resolve through canonical verdicts and receipts. The protocol composes optimistic concurrency control, semantic verification, and transactional outbox dispatch. Its distinctive read set includes normative authority and the model components that produced the proposal. We evaluate the design through an implemented governed-work lifecycle, a local authority, outbox, and reconciliation composition, and a multi-process fallback that delegates effect authority when complete mediation is unavailable. The local composition preserved selective fallback, unaffected work, retry and restart recovery, and package reconstruction. Complete mediation produced 18 authority commands and 18 external effects. Direct-credential arms produced eight additional effects, all detected with independent journal reconciliation and none falsely detected in the mediated control. The results support commit-time reject-and-rerun semantics and expose the clock, lease, ordering, and quorum obligations introduced by delegated admission. They do not establish prevention or reversal of bypass effects, external-journal completeness, production throughput, wide-area availability, semantic completeness, or verifier correctness.","published":"2026-09-25T16:31:39Z","updated":"2026-09-25T16:31:39Z","authors":["Jesus Salas"],"categories":["cs.DC"],"headline":"New method helps control and verify complex automated decisions","summary":"Researchers studied how automated agents make decisions when rules and data can change during their work. They designed a system where an authority reviews the agent's proposals before applying effects, rejecting outdated work to ensure correctness. Their tests show this approach supports reliable retries and error detection in controlled environments. However, the research is based on specific setups and does not guarantee preventing all errors, complete coverage of real-world cases, or perfect verifier accuracy."},{"id":"2609.31489v1","title":"Further Results on Sum-Freedom of Binary and $q$-ary Functions","abstract":"The notion of sum-freedom of binary functions was introduced recently by C. Carlet as a generalization of the APN functions used in cryptography; the $q$-ary version of the notion is a natural extension. For each integer $k$ with $0\\le k\\le n$, there is a $k$th order sum-free function on $\\Bbb F_{2^n}$. It is also known that when $k/n$ is not close to 0 or 1, the multiplicative inverse function on $\\Bbb F_{2^n}$ is not $k$th order sum-free. We generalize these two results to $q$-ary functions. APN functions have a coding theoretic characterization. We generalize the characterization to sum-free functions of arbitrary order over any finite field. It is well known that the Welch functions is 2nd order sum-free. We give an alternative proof for this result which leads to a more general algebraic question. We also investigate that the 3rd order sum-freedom of the Welch function and power functions of algebraic degree 3. We formulate a conjecture about the 3rd order sum-freedom of the Welch function which is supported by strong numerical evidence.","published":"2026-09-25T16:31:17Z","updated":"2026-09-25T16:31:17Z","authors":["Xiang-dong Hou","Shujun Zhao"],"categories":["math.NT"],"headline":"Researchers Explore Special Properties of Mathematical Functions","summary":"The researchers study 'sum-freedom,' a property of certain mathematical functions used in cryptography and coding theory. They extend previous findings from binary functions to more general cases with multiple values. They also provide new proofs, explore higher-order properties of specific functions, and propose a conjecture about these properties based on numerical evidence. Their work is theoretical and involves algebraic and numerical analysis without direct practical applications stated."},{"id":"2609.31488v1","title":"Connes Distance in Discrete-Time Quantum Walks","abstract":"We study the geometric spread of discrete-time quantum walks using Connes' distance formula. We introduce its discrete analogue by adapting the continuous spectral metric to the lattice, with the Dirac operator replaced by the unitary step operator. Focusing first on a (2+1)-dimensional quantum walk, we evaluate the dynamical spread of localized wavepackets to compare propagation in free space against that under a uniform magnetic field. We find that while the free walk shows linear ballistic growth, gauge-induced phases in the magnetic walk constrain the admissible test operators, suppressing the long-time geometric expansion. Furthermore, we apply this discrete framework to a (1+1)-dimensional plastic quantum walk simulating an inhomogeneous spatial metric. The results demonstrate that the discrete Connes distance naturally captures the underlying spatially varying metric, providing a robust measure of the wavepacket's spread that adapts directly to the local simulated geometry.","published":"2026-09-25T16:30:04Z","updated":"2026-09-25T16:30:04Z","authors":["Rayan Trabelsi","A. F. Reyes-Lega","Pablo Arrighi"],"categories":["quant-ph"],"headline":"Measuring Quantum Walk Spread Using a New Distance Method","summary":"Researchers studied how quantum particles spread in discrete steps using a mathematical distance formula originally designed for continuous spaces. They adapted this formula to work on a lattice, replacing some key elements to match quantum walks. By comparing particle movement in free space and in a magnetic field, they found the magnetic field limits how far particles spread over time. They also showed that this adapted distance measure effectively reflects changes in the underlying space where the quantum walk occurs, based on their simulations."},{"id":"2609.31487v1","title":"Singular Backward SDEs for Optimal Control with State Constraints","abstract":"We investigate a class of backward stochastic differential equations (BSDEs) with at most quadratic growth which explode at a possibly unbounded random horizon, defined through the first hitting of zero of an adapted Ito process. In contrast with the classical theory of singular BSDEs, the explosion is generated by the nonlinear dependence of the generator on the martingale integrand, rather than by a superlinear coercivity condition in the solution component. We construct a minimal singular solution and derive two-sided estimates on its explosion, together with weighted BMO estimates for the martingale integrand. We also obtain the uniqueness and exact explosion rates under additional structural assumptions. For Hamiltonian generators, we establish a verification theorem for an infinite-horizon stochastic optimal control problem with possible non-Markovian state constraints, showing that the BSDE feedback induces the unique optimal constrained law. We finally specialize the theory to exit times of uniformly elliptic Markov diffusions and recover the connection with large solutions of viscous Hamilton-Jacobi equations.","published":"2026-09-25T16:29:46Z","updated":"2026-09-25T16:29:46Z","authors":["Mathieu Lise","Nizar Touzi"],"categories":["math.PR","math.OC"],"headline":"New Math Method Supports Complex Optimal Control Problems","summary":"Researchers studied a special class of backward stochastic differential equations (BSDEs) that can become infinite in unpredictable ways, related to when a certain process hits zero. They developed a minimal solution and found estimates that describe this explosion behavior, proving uniqueness under some conditions. Applying their findings, they connected these equations to complex problems in optimal control, including constraints changing over time, and linked the results to advanced equations in math physics. Their work is theoretical and focuses on the mathematical properties and solutions of these equations."},{"id":"2609.31486v1","title":"The Hilbert Basis Theorem for Non-Associative Skew Power Series Rings and Non-Associative Skew Laurent Series Rings over a Non-Associative Coefficient Ring","abstract":"In this paper, we give conditions on a unital non-associative ring $R$ and additive map $σ: R\\longrightarrow R$ so that the non-associative skew power series ring, denoted by $R[[X;σ]]$ and the non-associative skew Laurent series ring denoted by $R((X;σ))$ are left (resp. right) Noetherian, respectively. Consequently, we establish versions of Hilbert Basis Theorem of $R[[X;σ]]$ and $R((X;σ))$, respectively, for left (resp. right) Noetherianity . We also give a counterexample showing that the Hilbert Basis Theorem does not hold in general in the non-associative case. In establishing the Hilbert Basis Theorems for $R[[X;σ]]$ and $R((X;σ))$, we introduce and utilize the right (resp. left) quasi-associative property of $R$, which is equivalent to $rR$ (resp. $Rc$) being one-sided ideals for all $r\\in R$ (resp. $c\\in R$).","published":"2026-09-25T16:27:52Z","updated":"2026-09-25T16:27:52Z","authors":["Darwin P. Mangubat","Jocelyn P. Vilela","Johan Richter"],"categories":["math.RA"],"headline":"New conditions extend Hilbert Basis Theorem to complex algebraic rings","summary":"The researchers explored certain complex algebraic structures called non-associative skew power series and Laurent series rings. They identified specific conditions under which these rings satisfy a property known as being Noetherian, related to controlling infinite processes in algebra. Using these conditions, they extended the Hilbert Basis Theorem to these rings but also showed it does not always hold in this non-associative context. Their work introduces a new property called quasi-associativity to help prove these results."},{"id":"2609.31485v1","title":"Verifiable Randomness for Blockchain-Based Lottery Systems","abstract":"As lotteries and other high-stakes decentralized applications increasingly depend on unpredictable randomness for their operations, the lack of a secure and transparent on-chain random number generator that is verifiable by all participants remains a critical open problem. Various approaches to blockchain-based random number generation have emerged over the years, each with their own strengths and limitations, and have consistently been superseded as blockchain technology evolved. This paper surveys existing approaches to on-chain randomness and proposes a new platform that builds upon the well-known commit-and-reveal scheme while directly addressing its principal vulnerability, the last revealer attack, in which the final participant can withhold their reveal in order to bias or abort the output upon seeing an unfavorable result. We further compare this solution with prior approaches and evaluate its entropy properties. The proposed architecture combines a web-based front-end with a Solidity smart contract deployed on the Polygon 2.0 blockchain. Implemented and tested on the Amoy testnet, the prototype is low-cost and simple to deploy, providing a practical, accessible proof-of-concept for verifiable on-chain randomness.","published":"2026-09-25T16:27:33Z","updated":"2026-09-25T16:27:33Z","authors":["Gonçalo Ferreira","André Zúquete","Paulo Bartolomeu"],"categories":["cs.CR","cs.CY"],"headline":"New Method Improves Fairness of Blockchain Lottery Randomness","summary":"Researchers address a key challenge in blockchain lotteries: generating random numbers that are secure, transparent, and verifiable by all participants. They propose a new system based on a known method but fix its main weakness, where the last user can manipulate results by withholding their input. Their design, tested on a blockchain test network, aims to be practical and low-cost, offering a proof-of-concept for trustworthy random number generation in decentralized applications."},{"id":"2609.31484v1","title":"COFI-DQI: Curve-based Optimal Function Intersection via Decoded Quantum Interferometry","abstract":"In 2025, Jordan et al. introduced Decoded Quantum Interferometry (DQI), a quantum algorithm for combinatorial optimization based on decoding. They considered Reed-Solomon decoding and its associated optimization problem, called Optimal Polynomial Intersection (OPI), which may be viewed as a polynomial regression problem over a finite field. DQI exhibits provable speedups on certain problem instances and establishes a connection between decoding problems and optimization tasks. Leveraging the well-understood dual structure and decoding theory of algebraic geometry codes from other curve families, we introduce COFI: Curve-based Optimal Function Intersection. By considering two-point Hermitian codes, Suzuki codes, and extended norm-trace codes, we broaden the range of algebraic geometry codes used in DQI and identify families that offer further improvements over one-point Hermitian codes in the Hermitian Optimal Polynomial Intersection framework considered by Jordan and Gu. Depending on the family and parameter regime, these curves can reduce quantum resource requirements or increase the number of constraints that can be considered.","published":"2026-09-25T16:26:44Z","updated":"2026-09-25T16:26:44Z","authors":["Gretchen L. Matthews","Julia Shapiro"],"categories":["cs.IT","quant-ph"],"headline":"New Quantum Algorithm Improves Polynomial Intersection Problems","summary":"Researchers built on a 2025 quantum algorithm called Decoded Quantum Interferometry (DQI) that helps solve complex combinatorial problems by decoding special codes. They introduced COFI, which uses different mathematical curves to extend this approach, potentially making the quantum algorithm more efficient or capable of handling more constraints. Their work focuses on theoretical improvements to the underlying algorithm, showing how using various curve-based codes might enhance quantum computing methods for certain optimization tasks."},{"id":"2609.31483v1","title":"Scaling Density Functional Theory with Gaussian Splatting","abstract":"Density functional theory (DFT) strikes a practical balance between accuracy and computational cost in many problems of computational chemistry and materials science. However, many DFT calculations are limited by fixed atom-centered basis sets, which dictate how accuracy and cost scale with system size. We propose Gaussian Splatting for Density Functional Theory (GS-DFT), which represents molecular orbitals as a cloud of Gaussians whose positions, shapes, and mixing coefficients are optimized jointly by gradient descent to minimize the energy without training data. Conceptually, GS-DFT is 3D Gaussian splatting with the renderer replaced by quantum mechanics. We introduce two key solver components: adaptive density fitting with screening for efficient evaluation of two-electron integrals, and a regularized differentiable orthogonalization of the molecular orbitals. Empirically, the optimized basis reaches the accuracy of the largest conventional basis sets with a fraction of the parameters, converging systematically in energy, density, and nuclear forces. At equal parameter count, it captures the stretched-bond and anion physics that fixed bases only recover with specialized basis augmentation. The resulting solver exhibits quadratic peak memory scaling in the cloud size, allowing us to simulate systems of up to 2,742 atoms (10,406 electrons) without any modifications at triple-zeta scale using a single four-GPU node.","published":"2026-09-25T16:26:40Z","updated":"2026-09-25T16:26:40Z","authors":["Andrés Guzmán-Cordero","Cindy Zhang","Majdi Hassan","Marta Skreta","Kirill Neklyudov","Matija Medvidović"],"categories":["physics.chem-ph","cond-mat.mtrl-sci","cs.LG","physics.comp-ph"],"headline":"New Method Improves Accuracy and Speed in Quantum Chemistry Simulations","summary":"Researchers developed a technique called Gaussian Splatting for Density Functional Theory (GS-DFT) to better simulate molecules. Unlike traditional methods that use fixed atom-based components, GS-DFT represents molecules as flexible clouds of shapes optimized through a mathematical process to reduce energy. Their approach uses fewer parameters but achieves comparable accuracy and can handle larger systems, up to thousands of atoms, with improved memory efficiency. These findings are based on computational experiments and may help advance simulations in chemistry and materials science."},{"id":"2609.31482v1","title":"\"AI is (not) the new...\": A Diagnostic Analogy Framework for Generative AI's Cultural Impacts","abstract":"Generative AI is reshaping the cultural infrastructures through which knowledge is found, synthesized, and held accountable. To make sense of this shift, scholars and policymakers reach for historical analogies of technologies such as the printing press, steam power or electricity. But these comparisons are typically imprecise about which property of the technology carries the comparison, and imprecise analogies produce imprecise governance by designing interventions against the wrong property of the system. This paper offers a diagnostic framework for analyzing how generative AI can transform epistemic and cultural practice. This paper offers a diagnostic framework for analyzing how generative AI can transform epistemic and cultural practice. We decompose each intervention into three coordinates: the epistemic site at which a technology acts, the governing logic by which it organizes its object, and the technical mechanism through which the logic is instantiated. This framework allows us to distinguish between structural cultural consequences, which follow from the mechanism itself, from contingent ones, which remain open to design and institutional choice. Applying the framework to information discovery and knowledge synthesis, we show how the shift from indexicality to inference and from editorial authority to statistical consensus produces specific, traceable cultural effects and reveals governance levers that gestalt analogy obscures.","published":"2026-09-25T16:25:46Z","updated":"2026-09-25T16:25:46Z","authors":["Rida Qadri","Vinodkumar Prabhakaran","Remi Denton"],"categories":["cs.AI"],"headline":"New Framework Explores How AI Changes Culture and Knowledge","summary":"Researchers examined how generative AI is changing the ways we find and organize knowledge by comparing it to past technologies like the printing press. They found that common comparisons often miss key details, leading to poor policy decisions. The researchers created a framework to analyze exactly how AI affects knowledge and culture, identifying what changes are inevitable and which can be influenced by design or institutions. Their work highlights specific cultural effects and better ways to guide AI governance."},{"id":"2609.31481v1","title":"Rotationally Symmetric Zoll Metrics with a Cubic Integral","abstract":"We introduce a new method for constructing Zoll metrics that admit first integrals polynomial in the momenta. The method uses Funk's form of the metric together with an algebraic relation between the first integrals. This reduces the problem to an algebraic system from which all unknown functions can be determined explicitly. In the case of a non-trivial cubic integral, we present an explicit two-parameter family of metrics that covers all such rotationally symmetric Zoll metrics on the $2$-sphere: every such metric is isometric to a member of this family. We compare these metrics with the earlier results of Valent, Duval, and Shevchishin.","published":"2026-09-25T16:25:45Z","updated":"2026-09-25T16:25:45Z","authors":["Holger R. Dullin","Vladimir S. Matveev","Gleb P. Palshin","Serena Scapucci"],"categories":["math-ph","math.DG","nlin.SI"],"headline":"New method finds special curved surfaces with cubic integrals","summary":"Researchers developed a method to find special curved shapes, called Zoll metrics, that have rotational symmetry and a particular type of mathematical property known as a cubic integral. By using an algebraic approach, they explicitly described all such shapes on a 2-sphere (a surface like a globe). Their work builds on and compares to previous findings, providing a complete family of these metrics under the stated conditions. The results are theoretical and based on mathematical calculations rather than physical experiments."},{"id":"2609.31480v1","title":"Ferroelectric Switching in ZnO/Zn1-xMgxO Heterostructures: Atomistic Insights into Interfacial Coupling and Layer Architecture","abstract":"Ferroelectric switching in heterostructures couples composition, layer topology, temperature, and interfacial boundary conditions. ReaxFF molecular dynamics isolates these variables in ZnO/ Zn1- xMgxO/ZnO and Zn1-xMgxO/ZnO/ Zn1-xMgxO stacks. Within pristine, initially single-domain models, coupling to switchable Zn1-xMgxO reduces the applied field required to reverse ZnO by up to fivefold. Temperature generally lowers the coercive field, whereas Mg concentration produces a nonmonotonic response. At equal ZnO and Zn1-xMgxO (ZMO) proportions, structures with ZnO at the center switch at lower fields than those with ZMO at the center at all four temperatures examined, demonstrating a topology-dependent response. Layer-resolved trajectories reveal topology-dependent switching sequences with direction-dependent redistribution of normal stress near the heterointerfaces, consistent with a stress-assisted cooperative pathway. Limiting MgO-containing structures exhibit sequential multilevel switching or low-polarity trapping, depending on thickness and temperature. Fixed-charge atomistic simulations complement previous continuum descriptions by resolving structural, energetic, and local stress evolution under a common applied field. TEM and STEM-EDS observations provide experimental structural context for the modeled architectures. Together, the results establish layer topology and interfacial mechanical confinement as design variables for wurtzite ferroelectric heterostructures.","published":"2026-09-25T16:25:35Z","updated":"2026-09-25T16:25:35Z","authors":["Alireza Sepehrinezhad","Ali Mohammadi Dinani","Ece Gunay","Elizabeth C. Dickey","Jon-Paul Maria","Susan Trolier-McKinstry","Adri C. T. van Duin"],"categories":["cond-mat.mtrl-sci","physics.chem-ph"],"headline":"Researchers study layer design to improve ferroelectric switching in ZnO materials","summary":"The researchers investigated how different layer arrangements and compositions affect ferroelectric switching in zinc oxide-based materials using detailed computer simulations. They found that placing certain layers in the center reduces the electric field needed to reverse polarization, and that temperature and magnesium content also influence behavior in complex ways. Their work shows that the mechanical interaction at interfaces plays a key role, suggesting that designing layer structure could help control switching, though these conclusions come mainly from simulations supported by some microscopic observations."},{"id":"2609.31479v1","title":"Ternary Perfect Forms Over Imaginary Quadratic Field","abstract":"In this work, we enumerate the ternary perfect forms over all imaginary quadratic fields of absolute discriminant up to $91$ and study the number and combinatorial types of their associated polytopes. We prove a bound on the number of combinatorial types that can arise regardless of the discriminant. We analyze the data and make several observations concerning the number and complexity of the perfect forms that arise in the scope of our calculation.","published":"2026-09-25T16:24:39Z","updated":"2026-09-25T16:24:39Z","authors":["Zachary Parker","Dan Yasaki"],"categories":["math.NT"],"headline":"Researchers Study Special Shapes in Complex Number Systems","summary":"The researchers counted and analyzed special mathematical objects called ternary perfect forms within certain complex number systems known as imaginary quadratic fields, focusing on those with discriminant values up to 91. They examined the types and shapes associated with these forms, and proved a limit on how many different shapes can appear regardless of the discriminant. Their findings are based on calculations and observations of the data rather than real-world experiments."},{"id":"2609.31478v1","title":"Irreducible rootless unimodular lattices generated by vectors of norm $3$","abstract":"The classical ADE classification implies that $E_8$ is the unique irreducible unimodular Euclidean lattice generated by vectors of norm $2$. In sharp contrast, for every positive integer $m$, we construct an irreducible odd unimodular lattice of rank $24m$ and minimum norm $3$ that is generated by its minimal vectors.","published":"2026-09-25T16:24:27Z","updated":"2026-09-25T16:24:27Z","authors":["Hong-Jun Ge","Jack H. Koolen","Jing-Yuan Liu","Kiyoto Yoshino"],"categories":["math.NT","math.CO"],"headline":"New large lattices built from vectors of length 3 discovered","summary":"The researchers studied special mathematical structures called lattices, which have patterns made from points in space. Previously, a unique lattice called E8 was known to be built from points spaced with length 2. The team created new lattices with much bigger size, made from points spaced with length 3. These new lattices have specific properties, like being 'odd' and 'unimodular', and the work is theoretical, showing these lattices exist for all positive sizes but does not explore practical uses."},{"id":"2609.31477v1","title":"A parallel solver for vortex-induced vibrations at zero mass ratio","abstract":"We present a parallel strategy for the study of the vortex-induced vibrations (VIV) of a rigid, massless cylinder mounted on translational springs, in two and three dimensions. Without mass or damping, its motion is governed solely by the fluid forces and the spring restoring force, and is an immediate response to the flow. We focus on low spring stiffness, yielding high reduced velocity and maximal oscillation amplitude. A monolithic finite element method in an arbitrary Lagrangian-Eulerian framework with implicit time stepping is proposed, solving the fluid flow, cylinder dynamics, and mesh motion simultaneously. Mesh movement is modelled with a linear elastic pseudo-solid with variable Lamé coefficients, and the no-slip condition on the cylinder is enforced with a Lagrange multiplier coupled to the position degrees of freedom; three strategies are explored to enforce this force-position coupling in a distributed framework. To the best of our knowledge, this work is the first to propose a distributed solver for three-dimensional fluid-structure interaction at zero mass ratio. The solver is verified with manufactured solutions, and exhibits second-order accuracy in time. A numerical study of two- and three-dimensional VIV for Reynolds numbers from 100 to 300 reproduces the VIV responses obtained in previous two-dimensional studies, and predicts nonperiodic 3D oscillations starting at $Re = 300$. Using the direct solver MUMPS, the proposed method scales to up to 768 compute cores.","published":"2026-09-25T16:23:40Z","updated":"2026-09-25T16:23:40Z","authors":["Arthur Bawin","Joan Le Pouhaër","Stéphane Étienne","Cédric Béguin"],"categories":["physics.flu-dyn","physics.comp-ph"],"headline":"New computer method studies fluid-driven vibrations of massless cylinders","summary":"Researchers developed a computer method to simulate how a massless cylinder attached to springs vibrates due to fluid flow, focusing on conditions where the cylinder moves only because of fluid forces and spring resistance. Their approach solves the fluid flow, cylinder motion, and mesh changes all at once, allowing study in two and three dimensions. The method matches previous results in simpler cases and predicts complex 3D vibrations at higher flow speeds, but these findings come from computer simulations only."},{"id":"2609.31476v1","title":"Non-autonomous dynamics of inner functions: mixing and entropy","abstract":"We generalise a mixing result due to Pommerenke for non-autonomous dynamical systems of forward compositions of the boundary values of centred inner functions; the proof depends on a precise bound on the orbits of such forward compositions in~$\\D$. This enables us to give new criteria for such systems to have positive metric entropy, lower entropy and topological entropy. In the opposite direction, we construct a forward composition of centred degree two Blaschke products with topological entropy zero.","published":"2026-09-25T16:22:43Z","updated":"2026-09-25T16:22:43Z","authors":["Vasiliki Evdoridou","Núria Fagella","Philip Rippon","Gwyneth Stallard"],"categories":["math.DS","math.CV"],"headline":"Researchers study complex dynamics and chaotic behavior in math functions","summary":"The researchers investigate how certain mathematical functions called inner functions behave when composed repeatedly in a changing (non-autonomous) way. They extend previous work to identify conditions under which these systems show mixing—meaning a strong form of unpredictability—and different types of entropy, which measure complexity or chaos. They also provide an example of such a system with zero topological entropy, indicating no chaotic behavior. Their work is theoretical and uses advanced mathematical analysis of these function compositions."},{"id":"2609.31475v1","title":"Direct Observation of Switching-Wave Dynamics in 20 GHz Normal Dispersion Spiral Microresonators","abstract":"Integrated Kerr frequency combs are powerful tools for microwave photonics, spectroscopy, and optical communications. While traditional architectures rely on the anomalous-dispersion regime, this typically requires thick, highly strained silicon nitride layers that complicate standard CMOS-foundry fabrication. Thinner layers circumvent these fabrication constraints but yield normal-dispersion microcombs that generally require specific trigger mechanisms, such as deterministic seeding, due to the absence of spontaneous modulational instability. Here, we demonstrate the generation of a 20 GHz microcomb in the normal-dispersion regime, driven by a synchronized dual-pump scheme via electro-optic sidebands modulation. The platform leverages an Archimedean spiral geometry on a foundry-compatible, 350 nm-thin SiN platform. By employing adiabatic curvature engineering, the resonator balances a compact footprint with an intrinsic quality factor exceeding $7 \\times 10^6$, while effectively eliminating avoided mode crossings with higher order modes. This strong-dispersion architecture yields a low-repetition-rate microcomb featuring high power-per-line and picosecond-scale temporal profile, which enables the direct optical sampling and characterization of the out-coupled switching wave waveforms. The measured dynamics across a range of pump desynchronizations demonstrate excellent agreement with simulations. Our work establishes a scalable, strategy for footprint-efficient normal-dispersion microcomb generation at microwave frequencies, with potential for scaling to other wavelength ranges.","published":"2026-09-25T16:22:19Z","updated":"2026-09-25T16:22:19Z","authors":["Van Doan Le","Julien Fatome","Erwan Lucas"],"categories":["physics.optics"],"headline":"Researchers create stable 20 GHz microcomb using thin silicon layers","summary":"Generating microcombs—tools useful in communications and sensing—usually involves thick, complex silicon layers. Researchers developed a method to produce a 20 GHz microcomb on simpler, thin silicon nitride by using a special spiral shape and dual laser pumps. Their design balances size and performance, enabling direct observation of how the light patterns change over time. The observed behavior matches their computer simulations, showing promise for efficient microcomb generation, though results are based on controlled lab experiments and simulations."},{"id":"2609.31474v1","title":"Spatiotemporal Non-Hermitian Skin Effect with Floquet-Engineered Ultracold Atoms","abstract":"Non-Hermitian skin effect, the accumulation of bulk eigenstates at edges under open boundary conditions, has been observed across classical, quantum, and synthetic platforms. Whether it can be generalized to the genuine time dimension remains open. We propose a spatiotemporal NHSE in the Floquet phase space via Floquet engineering of a ring-trapped ultracold atomic gas. Placing spatial and temporal degrees of freedom on an equal footing, we show that a single parent skin effect projects onto either domain. The spatial projection yields an NHSE with arbitrarily tunable skin locations, eliminating the need for physical open boundaries. The temporal projection gives time-domain nonreciprocity and periodic temporal funneling, consistent with causality. This establishes a unified spatiotemporal paradigm for NHSE and opens a route to non-Hermitian topological physics in time.","published":"2026-09-25T16:21:06Z","updated":"2026-09-25T16:21:06Z","authors":["Weijie Liang","Weiping Zhang","Keye Zhang"],"categories":["cond-mat.quant-gas","quant-ph"],"headline":"New Study Explores Edge Effects in Time and Space with Ultracold Atoms","summary":"The researchers studied how a known physical phenomenon, where particles gather at edges called the non-Hermitian skin effect, could also happen in time as well as space. They used a special setup with ultracold atoms arranged in a ring and applied a technique to control their behavior over time. Their work suggests this effect can be controlled without physical edges and may cause unique effects in time, but these findings are theoretical and based on engineered simulations."},{"id":"2609.31473v1","title":"Game Arena: Strategic LLM Evaluation in Competitive Environments","abstract":"We introduce Kaggle Game Arena, an open and ever-expanding platform to evaluate large language models (LLMs) through competitive games. Different from static benchmarks, game arena enables models to play head-to-head matchups in structured environments where the gameplay strength naturally increases as models evolve, preventing performance saturation. This technical report details the infrastructure behind Game Arena and describes the three pilot game environments: Chess, Poker, and Werewolf. These environments span perfect information, imperfect information, and multiplayer game settings, enabling a systematic study of models' strategic planning, adaptation, and robustness under uncertainty. For each game, we provide a detailed description of the environment, evaluation metrics, and results from running full competitions across models. Through robust infrastructure and large-scale ground-truth based evaluation, Game Arena ensures reproducibility, transparency and generalizability to new games and variants over time.","published":"2026-09-25T16:20:55Z","updated":"2026-09-25T16:20:55Z","authors":["Bovard Doerschuk-Tiberi","Yao Yan","Justin Chiu","Hann Wang","Timothy Chung","Martyna Plomecka","John Schultz","Jon Lipovetz","Clayton Drazner","Yuchen Zhuang","Jaimie Hwang","Nate Keating","Riley Jones","Andrew Lee","Oran Kelly","Ian Gemp","Michael Aaron","Laurel Prince","Kate Larson","Jeff Moser","Harrison Jobe","Chad Woodford","Siqi Liu","Andrew Wang","Bo Chang","Christopher D'Mello","Diane Chaleff","Addison Howard","Johnny Yip","Chuck Sugnet","Antonio Gulli","Meghan O'Connell","Will Cukierski","Nenad Tomasev","Dima Yeroshenko","Kinjal Parekh","Roxanne Daniel","Marc Lanctot","Domino Weir","Elsa Dong","Daniel Hennes","Melissa Nalubwama","Robert Fraser","Ryan Trostle","Jun Peng","Tom Mason","Lloyd Hightower","Chiamaka Chukwuka","Yuexiang Zhai","Phoebe Kirk","Yi Su","Yuting Han","Jie Ren","Chris Prichard","Sahand Sharifzadeh","Karim Hakimzadeh","DJ Sterling","Meg Risdal","Kate Olszewska","Ya Xu","Orhan Firat","Minmin Chen"],"categories":["cs.AI"],"headline":"New Platform Tests AI Language Models with Competitive Games","summary":"Researchers created Kaggle Game Arena, a platform where large language models compete in games like Chess, Poker, and Werewolf. Unlike fixed tests, this setup lets models face off directly in evolving challenges, helping assess their strategic thinking, adaptability, and ability to handle uncertainty. The report explains the platform's design, game rules, scoring methods, and competition results, supporting transparent and repeatable evaluation of AI strategies in different game types."},{"id":"2609.31472v1","title":"A Three-Party W-State Quantum Secret Sharing Protocol with X-Gate Encoding and Forbidden-Outcome Detection","abstract":"We develop a new three-party quantum secret-sharing (QSS) protocol based on a three-qubit W state. This protocol encodes the secret-sequence bits using X gates and employs randomly selected Hadamard operations and measurement bases to generate information and security-test rounds. We evaluate the efficiency of the proposed protocol and analyze its security against an internal adversary performing intercept-and-resend and entangle-and-measure attacks, deriving the corresponding detection probabilities. The analysis of the entangle-and-measure attack shows that extracting information through an ancillary system leads to a nonzero probability of forbidden outcomes in the test rounds, thereby enabling the detection of the attack. The proposed protocol is implemented in Python using the Qsikit framework and executed on quantum processors. Its performance is evaluated in terms of the quantum bit error rate (QBER) for information rounds under different secret-bit encoding configurations. The experimental results demonstrate that the QBER depends noticeably on the quantum processor used.","published":"2026-09-25T16:19:34Z","updated":"2026-09-25T16:19:34Z","authors":["Timur R. Teregulov","Elena R. Loubenets"],"categories":["quant-ph"],"headline":"New Quantum Method for Sharing Secrets Among Three Parties","summary":"Researchers developed a new quantum secret-sharing method using a special quantum state called a three-qubit W state. Their approach encodes secret bits with certain quantum operations and tests for security using random checks, aiming to detect attempts to intercept or tamper with the secret. They analyzed how well the method detects attacks through forbidden outcomes and tested their protocol on quantum computers, finding that error rates vary depending on the quantum device used. This work is based on theoretical and experimental results using quantum simulations and processors."},{"id":"2609.31471v1","title":"Communication: Becke Weights as a Partitioning Scheme for Phase Space Electronic Structure Theory","abstract":"We implement the Becke weight for atom A ($w_A$) as the partitioning scheme needed for phase space electronic structure theory. We show that this ansatz can lead to very efficient electronic structure calculations at nonzero nuclear momentum ($P \\ne 0$). We further show that using Becke weights leads to efficient VCD calculations that are very stable with respect to basis set size. We hypothesize that, as far as quantum chemistry simulations in an atomic orbitals basis, the choice of Becke weights as a partitioning scheme will be adopted for all future phase space electronic structure calculations in an atomic orbital basis, as theorists learn to efficiently simulate coupled nuclear electronic dynamics through a new framework.","published":"2026-09-25T16:18:39Z","updated":"2026-09-25T16:18:39Z","authors":["Ben Curlee","Zheng Pei","Xinchun Wu","Titouan Duston","Todd J. Martinez","Yihan Shao","Joseph E. Subotnik"],"categories":["physics.chem-ph"],"headline":"Using Becke Weights to Improve Electronic Structure Calculations","summary":"Researchers explored a method called Becke weights to divide atoms in phase space electronic structure theory, which helps simulate electrons and nuclei moving together. They found this approach can make calculations more efficient when nuclei have momentum and improve stability in certain vibrational analyses. The study suggests this method might become standard in future quantum chemistry simulations, though the findings are theoretical and based on proposed computational frameworks."},{"id":"2609.31470v1","title":"Beyond Empirical Support: Structured Outlier Generation via Sinkhorn Optimal Transport","abstract":"Outliers are essential for evaluating and improving the robustness of machine learning systems, especially when future distributions may differ significantly from historical training data. In high-stakes applications, robustness often depends on rare cases that finite datasets fail to capture, making simple resampling or perturbation insufficient for stress scenario generation. Existing outlier synthesis methods typically rely on sparse neighborhoods, low support latent regions, or classifier boundary crossings, which can be heuristic, unstable, and tied to specific modalities or architectures. We therefore propose Sinkhorn Boundary Outlier Generation (SBOG), a structured framework for latent-space outlier generation that couples Sinkhorn optimal transport geometry with distributionally robust boundary modeling. The resulting Sinkhorn-induced support cost guides the sampler toward weakly supported boundary regions, while semantic constraints prevent uncontrolled drift from the intended context, yielding controlled deviations from the in-distribution reference measure rather than arbitrary sparse-region samples. Experiments on time series anomaly generation and image outlier synthesis show that our framework produces informative, semantically controlled outliers and improves downstream robustness evaluation across modalities, providing a foundation for stress scenario generation beyond empirical support.","published":"2026-09-25T16:18:02Z","updated":"2026-09-25T16:18:02Z","authors":["Haixiang Sun","Andrew L. Liu"],"categories":["stat.ML","cs.LG","math.OC"],"headline":"New Method Creates Useful Outliers to Test AI Robustness","summary":"Researchers developed a technique called Sinkhorn Boundary Outlier Generation (SBOG) to create rare, unusual data points that help assess how well machine learning systems handle unexpected situations. Their method uses mathematical tools to generate outliers that stay related to normal data but explore less-supported boundary cases, which are often missed by traditional approaches. Tests on time series and images suggest this approach can produce meaningful outliers for better evaluating AI robustness, but results are based on experiments rather than real-world deployment."},{"id":"2609.31469v1","title":"LandscapeSHAP: Which Persistent Homology Class Gets the Credit?","abstract":"Shapley values, a solution concept from cooperative game theory, have recently become a standard tool for feature credit allocation in machine learning. They provide an axiomatically justified method to fairly distribute a model's prediction among the data features. Shapley values have not yet been applied to explain machine learning models trained on features from topological data analysis. We develop what we believe is the first such approach, focusing on the persistence landscape featurization of persistence diagrams. Because each landscape coordinate is a rank statistic, crediting a model's prediction back to individual persistent homology classes (persistence diagram points) is nontrivial. We introduce LandscapeSHAP, a method for fair credit allocation to persistence diagram points based on a model's prediction. For linear models on persistence landscapes, LandscapeSHAP has a closed form expression that gives the exact Shapley value of every persistence diagram point. In particular, there is no coalition sampling required. We further prove that the four Shapley \"fairness\" axioms uniquely characterize this credit allocation for any model, not only linear ones. For a general nonlinear model, this unique value can only be calculated exactly from its defining coalition averaging formula, which requires considering all $2^N$ many coalitions, where $N$ is the number of points in the persistence diagram. This is computationally intractable for persistence diagrams of realistic size. We complement the exact linear model result with an efficient Monte Carlo sampling of persistence diagram coalitions. We give convergence rates in terms of number of samples needed to approximate to a desired degree of accuracy. We also prove stability results for the LandscapeSHAP credit allocation, for any model.","published":"2026-09-25T16:17:42Z","updated":"2026-09-25T16:17:42Z","authors":["Nikola Milićević"],"categories":["math.AT","cs.LG"],"headline":"New Method Assigns Credit to Shapes in Data Analysis Models","summary":"Researchers developed LandscapeSHAP, a method to fairly assign importance to features called persistence diagram points used in topological data analysis models. They provide exact results for simple, linear models and propose a sampling approach for more complex models, noting computation can be very large otherwise. Their work proves that their method meets fairness criteria and remains stable, but exact calculations for complex models remain challenging due to computational limits."},{"id":"2609.31468v1","title":"PriceBench: A Diagnostic Benchmark for Price, Quality, and Brand Preferences in LLM Booking Agents","abstract":"LLMs increasingly act as purchasing agents, which makes the LLM, not the user, the one choosing among the options that satisfy a request; its preferences quietly fix what gets bought and what it costs. Hotel booking is a clean instance: a high-volume choice settled on a few comparable attributes, where the pick reveals those preferences. We introduce PriceBench, a diagnostic benchmark that recovers an LLM's price, quality, and brand preferences from its booking choices with a logit choice model, applied to 28 LLMs from 8 providers on 3,600 hotel tasks from 179 real New York City properties. We find that capability is associated with how consistently an LLM chooses, not with what it chooses: more capable LLMs hold stronger, more consistent preferences, while weaker ones either lock onto one position, exploitable by whoever controls listing order, or choose almost indifferently. What those preferences favor varies sharply across providers and even within one family: price sensitivity spans more than an order of magnitude, and the price/quality trade-off moves mean booked nightly price from \\$247 to \\$393 on identical tasks. What an agent buys must therefore be measured per LLM, not inferred, and we release the tasks, code, and all 28 response sets.","published":"2026-09-25T16:17:28Z","updated":"2026-09-25T16:17:28Z","authors":["Pavel Kireyev"],"categories":["econ.GN","cs.AI","cs.CL"],"headline":"New Test Reveals How AI Picks Hotels Based on Price and Brand","summary":"Researchers created PriceBench, a tool to find out how large language models (LLMs) prefer price, quality, or brand when booking hotels. They tested 28 different AI models on thousands of hotel choices in New York City, showing that more capable AIs make consistent choices, while weaker ones pick randomly or predictably. Preferences vary widely between models, affecting hotel prices selected. The study highlights that each AI's buying decisions must be understood individually. The researchers provided all their test data and tools for others to explore."},{"id":"2609.31467v1","title":"A likelihood-based coefficient for biomedical independence testing: the binomial-cut composite likelihood ratio","abstract":"The standard dependence summaries used in biomarker studies -- Pearson's r, Spearman's rho, Kendall's tau -- take values in [-1, 1] with 0 indicating no linear or monotone association. Zero does not distinguish independence from non-monotone dependence, so the scale cannot represent threshold effects, heteroscedasticity, and tail shifts common in biomarker practice. We formulate independence testing as a composite Bernoulli likelihood ratio: at each threshold t, comparing the Bernoulli laws of 1(Y <= t) conditionally on X versus marginally, aggregated over cut points. The resulting coefficient xi_cut lies on [0, 1] with 0 iff X and Y are independent (under continuity of Y) and 1 iff Y is a measurable function of X. Fisher weighting arises at second order from the Bernoulli likelihood, and xi_cut equals twice the threshold-averaged mutual information between X and 1(Y <= t), giving a distribution-free lower bound on I(X; Y). A second-order expansion recovers the Fisher-weighted Dette-Siburg-Stoimenov measure, which coincides under continuity with Chatterjee's rank correlation. Estimation uses a Nadaraya-Watson plug-in with a max-over-grid bandwidth; inference is by exact permutation. In biomarker-motivated simulations T_cut substantially outperforms rank-based coefficients on W-shaped non-monotone and heteroscedastic alternatives. We illustrate on the Seattle cohort (n=70, ages 21-88) of the aging plasma proteome dataset, screening all 1,305 proteins for age dependence: under Benjamini-Hochberg control at q<0.05, T_cut rejects on 70 proteins, six of which are missed by Pearson, Spearman, and Chatterjee at the same FDR level.","published":"2026-09-25T16:17:20Z","updated":"2026-09-25T16:17:20Z","authors":["Jing Qin"],"categories":["stat.ME","stat.AP"],"headline":"New method better detects complex biomarker relationships with age","summary":"Researchers developed a new statistical measure to test if two variables are truly independent, especially when relationships are complex and not just linear or monotone. Their method compares data across thresholds and produces a score from 0 (independent) to 1 (fully dependent). In simulations and an aging-related protein study, this new approach identified more significant associations missed by traditional methods. The findings come from mathematical theory, simulations, and a real dataset, showing promise but still needing further application and validation."},{"id":"2609.31466v1","title":"Scaffold: Support Graph Theory Based Sparsification for Graph Neural Networks","abstract":"Graph neural networks (GNNs) rely on message passing over graph edges, making their computational and memory costs strongly dependent on graph density. Graph sparsification offers a natural way to reduce these costs, but removing edges indiscriminately can distort important communication structure and degrade predictive performance. We introduce Scaffold, a topology-based, unsupervised graph sparsification framework derived from support graph theory preconditioners. Scaffold explicitly controls two complementary structural quantities: dilation, which measures the length of rerouting paths induced by removed edges, and congestion, which measures how strongly these rerouted paths concentrate on the retained support. By jointly controlling dilation and congestion, Scaffold preserves short communication paths while avoiding structural bottlenecks. To our knowledge, Scaffold is the first scalable GNN sparsification framework to use a joint supporting-path dilation-congestion criterion. Across 19 homophilic and heterophilic benchmarks spanning small to large graphs, Scaffold achieves the best aggregate rank among the evaluated sparsification and related methods. Using only 10%-50% of the original edges per sparse support, Scaffold recovers or closely approaches full-graph GNN performance while using less than half the memory of full-graph training and reducing end-to-end training time, including sparsification overhead. We provide an open-source software package at https://github.com/siddhartha047/Scaffold.","published":"2026-09-25T16:17:09Z","updated":"2026-09-25T16:17:09Z","authors":["Siddhartha Shankar Das","Sai Karthik Navuluru","S M Ferdous","Ryan A. Rossi","Baris Coskunuzer","Lakshman Tamil","Edoardo Serra","Alex Pothen","Robert Rallo","Mahantesh M Halappanavar"],"categories":["cs.LG"],"headline":"New Method Cuts Graph Neural Network Costs While Keeping Accuracy","summary":"Graph neural networks (GNNs) are powerful but require a lot of computing power because they use many connections between data points. Researchers developed Scaffold, a new approach to reduce the number of these connections without losing important information. Scaffold carefully removes edges by balancing path length and traffic on remaining paths, helping GNNs run faster with less memory while maintaining performance. The researchers tested Scaffold on various graphs, showing it works well compared to other methods, though these results come from simulations and benchmarks."},{"id":"2609.31465v1","title":"Neutron Economy and Freeze-out Dynamics in the Late-Time Cold r-Process","abstract":"Standard r-process models often treat the freeze-out phase as a passive, uniform smoothing of the global abundance pattern. We demonstrate, however, that the final pattern is instead actively sculpted by a dynamic ``neutron economy'' that operates well after the free neutron source is exhausted. By analyzing cold neutron star merger outflows, we identify a distinct neutron exporter-importer relationship: the second-peak region ($A \\sim 130$) transitions early into a $β$-decay dominated regime, acting as a net exporter, while the third-peak ($A \\sim 195$) and Rare Earth ($A \\sim 164$) regions remain capture-dominated neutron importers. Targeted simulations reveal that the final peak structures emerge from a competition between two opposing forces: a global ``push'' driven by neutron capture shifting mass toward heavier nuclei, and a local ``pull'' from beta delayed neutron emission returning material back toward lighter masses. We identify $^{136}\\text{Pd}$, $^{135}\\text{Ag}$, and $^{130}\\text{Ru}$ as the primary exporters driving this economy, while third-peak fine structure is governed by $N=126$ isotones (Sm to Tb). Physically, this transport reflects a fundamental drive toward shell closure, where nuclei that overshoot the $N=82$ closed shell shed mass to migrate back toward this stable configuration, while nuclei in the third peak capture these redistributed neutrons to anchor themselves at the $N=126$ shell. Crucially, our models overproduce the odd-even staggering in the third peak compared to solar data. This discrepancy suggests current models likely overestimate the $β$-delayed neutron emission probabilities ($P_n$) for these $N=126$ nuclei, highlighting them as high-priority targets for theoretical refinement and future FRIB experiments.","published":"2026-09-25T16:16:36Z","updated":"2026-09-25T16:16:36Z","authors":["Mengke Li","Bradley Meyer"],"categories":["astro-ph.HE","nucl-th"],"headline":"How neutron behavior shapes heavy element formation after star mergers","summary":"Researchers studied the late stages of element creation in neutron star mergers, focusing on how free neutrons influence final element patterns. They found that certain regions export neutrons while others import them, creating a dynamic balance that shapes heavy elements. The study shows current models may overestimate some neutron emissions, suggesting a need for improved predictions and experiments to better understand these processes."},{"id":"2609.31464v1","title":"Explaining Ground-Motion Residuals at Two Strong-Motion Stations in Southeastern New York: Sediment Resonance and Topographic Amplification","abstract":"Sites hosting strong-motion stations in the Central and Eastern United States are commonly characterized using proxy-based parameters, limiting the ability to identify the physical causes of large ground-motion residuals. This study investigates the amplification mechanisms at two New York strong-motion stations that recorded some of the largest positive 1-Hz pseudospectral-acceleration residuals during the 2024 Mw 4.8 Tewksbury, New Jersey, earthquake. Station N4.N62A is located at Caumsett State Historic Park in the Atlantic Coastal Plain, whereas station LD.PAL is located at the Lamont-Doherty Earth Observatory atop the Palisades cliffs. Detailed geophysical site characterization at both sites combined active-source and ambient-noise surface-wave testing with HVSR measurements. At Lamont-Doherty, additional ambient-noise arrays were deployed across the Palisades ridge to evaluate topographic amplification. At Caumsett, surface-wave testing resolved a thick sedimentary column overlying a major impedance contrast, with a spatially representative fundamental site frequency of 0.93 Hz. Amplification in the same frequency range was also independently evident in observed amplifications computed from site-to-site (S2S) residuals. The fundamental-resonance band also encompasses the 1-Hz frequency at which the large Tewksbury residual was observed, supporting deep sediment resonance as the dominant mechanism. At Lamont-Doherty, surface-wave testing indicated hard-rock conditions with a thin, laterally variable sediment cover. The ambient-noise arrays showed repeatable, directional amplification near 2 Hz at LD.PAL, and the earthquake-based S2S amplifications show a peak between approximately 0.6 and 2 Hz, a pattern reproduced by neither the transfer functions nor the ergodic linear amplification models. These observations support topographic amplification as the primary mechanism at LD.PAL.","published":"2026-09-25T16:13:48Z","updated":"2026-09-25T16:13:48Z","authors":["Aser Abbas","Patrick Daniele","James Kaklamanos","Laurie Baise","Kyle Cannon","Ellie Meyer"],"categories":["physics.geo-ph"],"headline":"Why Two New York Earthquake Stations Had Unusual Vibrations","summary":"Researchers studied two strong-motion stations in New York that showed large unexpected shaking during a 2024 earthquake in New Jersey. They used detailed tests of the ground and noise measurements to find that at one site, vibrations were mainly caused by deep, soft sediments amplifying certain frequencies. At the other, the shape of the land (topography) was the key factor. These findings come from measurements and models, helping explain why the stations recorded unusually strong ground motion."},{"id":"2609.31463v1","title":"Uncertainty-Aware Federated Learning for Infant Movement Analysis","abstract":"Infant movement analysis provides valuable biomarkers for the early identification of neurodevelopmental disorders. Recent advances in deep learning have enabled automated analysis of infant movements from video-derived skeletal representations, achieving performance comparable to expert assessment for tasks such as General Movement Assessment (GMA). However, most existing approaches rely on centralized training, requiring data from multiple institutions to be collected and stored at a single site. Such assumptions are often impractical in clinical settings due to privacy, governance, and data-sharing constraints. To address these challenges, we present, to the best of our knowledge, the first federated learning framework for automated infant movement analysis and General Movement Assessment using skeletal motion data. As a clinically relevant use case, the proposed framework is evaluated on fidgety movement classification. To quantify model confidence, Monte Carlo (MC) Dropout is employed to estimate predictive uncertainty during inference. Building upon this, we propose an Uncertainty-Aware Federated Averaging (UA-FedAvg) strategy that incorporates predictive entropy derived from MC-Dropout into the federated aggregation process, enabling client contributions to be adjusted according to their predictive uncertainty. Experiments were conducted using a cross-subject evaluation protocol under a three-client federated learning setting. Results demonstrate that federated learning substantially improves classification performance compared with independently trained local models while achieving performance approaching that of centralized training. Furthermore, UA-FedAvg and its variant incorporating validation loss generally outperform conventional FedAvg across the evaluated data-split configurations.","published":"2026-09-25T16:13:48Z","updated":"2026-09-25T16:13:48Z","authors":["Edmond S. L. Ho"],"categories":["cs.LG","cs.AI","cs.CV"],"headline":"New Privacy-Preserving Method Analyzes Infant Movements Using AI","summary":"Researchers developed a new way to analyze infant movements to help identify neurodevelopmental disorders early, without needing to share sensitive data centrally. They created a federated learning system that trains AI models across multiple sites using local data while accounting for uncertainty in predictions. Their tests showed this approach improves classification accuracy compared to training models locally and performs nearly as well as centralized training, all while respecting privacy and data-sharing limits."},{"id":"2609.31462v1","title":"Gottesman-Kitaev-Preskill error-correction with decohered resources","abstract":"Teleportation-based Gottesman-Kitaev-Preskill (GKP) error correction typically assumes a decoherence-free GKP ancilla Bell pair, an idealization that is difficult to reconcile with the ubiquitous environmental decoherence in bosonic systems. Here, we analyze the performance of the GKP protocol subject to pure-dephasing processes, ubiquitous in realistic situations. We show that, under this class of interactions, teleportation, instead of projecting the input onto the GKP subspace, transforms GKP error correction into a correlated Pauli-Gaussian noise process that can accumulate over successive correction rounds. Thus, unlike in previously studied noisy GKP scenarios, the code no longer self-corrects. Moreover, repeated noisy GKP teleportation drives the encoded modes out of the logical subspace, progressively degrading the encoded codewords. We show that the resulting leakage rapidly saturates with the number of correction rounds. Finally, we investigate the implications of our results for motional-oscillator-based GKP error-correction platforms.","published":"2026-09-25T16:12:52Z","updated":"2026-09-25T16:12:52Z","authors":["Rajendra S. Bhati","T. Bazylewicz","Jarosław K. Korbicz"],"categories":["quant-ph"],"headline":"Studying Error Correction Limits with Noisy Quantum Resources","summary":"Researchers studied a method used to correct errors in quantum computing systems that assumes perfect conditions, which is unrealistic due to unavoidable environmental noise. They found that when typical noise affects the system, the usual error correction no longer works as expected and errors can build up over time, causing the quantum information to degrade. Their findings come from theoretical analysis and highlight challenges for quantum error correction in real-world noisy environments, especially in systems using motional oscillators."},{"id":"2609.31461v1","title":"KneePreM: Towards 3D Knee MRI Foundation Models via Large-Scale Unlabeled Pretraining and Label-Efficient Fine-Tuning","abstract":"Background: Large volumes of unlabeled knee MRI scans are available across repositories but remain insufficiently leveraged. We developed KneePreM, a knee-specific 3D self-supervised model, and evaluated transfer and label efficiency for classification and segmentation. Methods: A 3D U-Net masked autoencoder was pretrained on 19,011 unlabeled Osteoarthritis Initiative (OAI) MRI series from 4,791 participants. Downstream fine-tuning used full and reduced training sets for fastMRI+ two-label classification (1,172 examinations), Arthroscopic Partial Meniscectomy (APM) eight-target classification (1,716 examinations), SKM-TEA segmentation (155 examinations), and APM segmentation (25 examinations). Baselines were random initialization and SuPreM. Deployment workflow was implemented with a Model Context Protocol interface. Evaluation metrics included balanced accuracy, F1 score, ROC AUC, PR AUC, and Dice score. Statistical analysis used bootstrap confidence intervals and paired bootstrap tests for classification and Wilcoxon signed-rank tests for segmentation. Results: KneePreM achieved higher full-data macro ROC AUC than both baselines for fastMRI+ and APM (all p < .001). For fastMRI+ classification, KneePreM achieved a ROC AUC of 0.722 using 50% of the training data, exceeding both full-data baselines. In APM classification, KneePreM reached a ROC AUC of 0.740 with 70% of the data, matching the full-data random baseline and outperforming SuPreM. For SKM-TEA segmentation, its 70%-data Dice of 0.838 exceeded the full-data random baseline (0.835) and both same-budget comparators. In APM segmentation, its 75%-data Dice of 0.746 exceeded the full-data random baseline (0.731) and both same-budget comparators. Conclusion: KneePreM improves transfer performance and label efficiency across knee MRI classification and segmentation tasks, particularly when labeled training data are limited.","published":"2026-09-25T16:11:45Z","updated":"2026-09-25T16:11:45Z","authors":["Xinxin Wang","Liam Hazan","Jing Li","Simona Rabinovici-Cohen","Xiaojuan Li","Mingrui Yang"],"categories":["cs.CV"],"headline":"New AI Model Improves Knee MRI Analysis with Less Labeled Data","summary":"Researchers created KneePreM, an AI model trained on over 19,000 unlabeled knee MRI scans to better analyze knee images with fewer labeled examples. They tested it on tasks like classifying and segmenting knee conditions, finding it often outperformed other methods, especially when less labeled data was available. These results, based on statistical tests, suggest KneePreM may help efficiently use large unlabeled MRI databases, though the findings focus on specific knee imaging tasks and settings."},{"id":"2609.31460v1","title":"Segment-Level Agentic Topic Modeling for Improved Data Exploration and Resource Efficiency","abstract":"Topic modeling is an effective technique for discovering hidden themes within documents and is widely used in text mining and data analysis across a variety of industry sectors. Recently, large language model (LLM)-based topic models have been emerged that prompt LLMs to generate topics then assign the topics to documents, producing more natural and human-readable topics than conventional topic modeling algorithms. However, the nature of topic assignment process causes certain drawbacks, such as the incapability to produce topic distributions over a document, too broad or narrow topics, and high resource consumption, which increases with the number and length of of documents being assigned topics. These issues are particularly critical for industrial applications, which require high-quality, in-depth analysis and the processing of large volumes of documents. In this context, this paper introduces a framework called SeLATM, which addresses these concerns by employing segment-level topic generation and topic refinement through agentic feedback loops. Experimental results on various datasets demonstrate that SeLATM significantly reduces the LLM resources compared to methods based on topic assignment process, while maintaining superior performance.","published":"2026-09-25T16:11:42Z","updated":"2026-09-25T16:11:42Z","authors":["Myeongjun Erik Jang","Antonios Georgiadis","Sae Young Moon","Fran Silavong"],"categories":["cs.AI"],"headline":"New Topic Modeling Method Saves Resources and Improves Analysis","summary":"Researchers studied how current topic modeling methods use large language models to find themes in documents but often face issues like imbalanced topic detail and high computing costs. They developed a new approach called SeLATM that breaks documents into segments and improves topics through iterative feedback. Tests showed SeLATM uses fewer computing resources while keeping or improving analysis quality, although findings are based on experimental data and simulations described in the paper."},{"id":"2609.31459v1","title":"Nonstabilizerness of quantum tensor network states is intractable in two dimensions","abstract":"Nonstabilizerness is a necessary resource for quantum systems to lie beyond the classically simulable regime. With the advent of stabilizer $α$-Rényi entropies, nonstabilizerness has also become a many-body diagnostic, complementary to entanglement. While deciding whether an arbitrary quantum state has nonstabilizerness is provably hard, such states already require a description exponentially large in the number of qubits and are thus out of reach for many-body physics. Here we instead consider 2D tensor network (TN) states, which compactly capture states obeying an entanglement area law, and ask whether they admit a simpler algorithm for the same task. In contrast to the 1D case, we prove that, even at small, constant bond dimension, computing the stabilizer entropy of 2D TN states is $\\# P$ hard for any integer $α\\ge 2$, and deciding stabilizer membership is $\\rm C_{=} P $ complete. Hence, under standard complexity assumptions, no classical or quantum algorithm with polynomial resources can solve either problem in the worst case. Both hardness results persist even when the problems are formulated up to constant additive accuracy.","published":"2026-09-25T16:11:37Z","updated":"2026-09-25T16:11:37Z","authors":["Gianluca Esposito","Georgios Styliaris"],"categories":["quant-ph","cond-mat.str-el"],"headline":"Determining key quantum properties is extremely hard in 2D systems","summary":"The researchers studied a property called nonstabilizerness, important for understanding complex quantum systems beyond classical simulation. They focused on two-dimensional quantum states described by tensor networks, a way to efficiently represent certain quantum states. Their findings show that calculating this property or deciding related membership is computationally very hard—even for small system sizes—meaning no efficient algorithm is likely to exist for these tasks in general, based on current complexity assumptions."},{"id":"2609.31458v1","title":"Nonparametric In-Context Learning under Growing Geometric Complexity: Minimax Optimality and Local Geometry-Adaptivity of Transformers","abstract":"Transformers have become a central architecture for in-context learning (ICL), particularly through their state-of-the-art performance in large language models. This success motivates understanding how transformers exploit task-relevant structure in geometrically heterogeneous data. However, existing nonparametric ICL theory has largely focused on Euclidean domains or single-manifold models. To address this gap, we study the prediction problem under unknown local geometry, modeled by sample size-dependent mixtures of manifolds with heterogeneous dimensions, smoothness, and sampling masses. Under local separation and small-perturbation conditions, we establish a minimax lower bound capturing the aggregate difficulty of the components and construct an oracle tangent local-polynomial estimator with a matching upper bound. This estimator is connected to a structure-informed, two-stage softmax transformer with a geometric preconditioner and chartwise reduced local-polynomial solvers. The transformer achieves negligible approximation error relative to the minimax rate with logarithmic depth and polynomial size. Finally, we derive an in-context generalization bound for near empirical risk minimizers over this class. Together, these results identify conditions under which the resulting predictor exploits local geometry and attains the aggregate minimax rate.","published":"2026-09-25T16:10:53Z","updated":"2026-09-25T16:10:53Z","authors":["Jaehee Seo","Jisu Kim"],"categories":["stat.ML","cs.LG","math.ST"],"headline":"How Transformers Learn from Complex Geometric Data Patterns","summary":"Researchers study how transformer models, used in language tasks, can learn from data that has complex shapes and structures rather than just simple flat spaces. They create mathematical models representing data with mixed geometric features and develop methods that match the best possible prediction accuracy under these conditions. Their work shows that certain transformer designs can adapt to local data geometry and perform nearly optimally, but these conclusions are theoretical and based on idealized mathematical settings rather than practical experiments."},{"id":"2609.31457v1","title":"4FHL: The Fourth Catalog of Hard Fermi-LAT Sources","abstract":"We present a catalog of sources detected above 50 GeV by the Fermi Large Area Telescope (LAT) on board the Fermi Gamma-ray Space Telescope using 16 years of observations. This is the Fourth Catalog of Hard Fermi-LAT Sources (4FHL), containing 673 objects detected in the 50 GeV-2 TeV energy range. The sensitivity and localization are improved by factors of $\\sim$2 relative to the previous LAT catalog at the same energies (i.e. the 2FHL catalog). The increased photon statistics, nearly doubling the number of detections relative to 2FHL, enables more detailed population studies, including the identification of fainter sources detected above 50 GeV. The majority of the sources (84.7\\%) are associated with extragalactic counterparts, primarily BL Lacertae objects, including 36 objects at redshift $z>1$. Galactic sources account for 10\\% of the sample, mainly pulsar wind nebulae and supernova remnants, while $\\sim$5\\% remain unassociated or of uncertain classification. Twenty sources are reported with no counterpart in previous LAT catalogs. The 4FHL catalog provides the deepest all-sky survey above 50 GeV to date. By focusing on this high-energy regime, it provides a fundamental link between space-based gamma-ray observations and the TeV range probed by current and future ground-based observatories. It serves as a useful reference for guiding future observations, including new potential targets for follow-up with Cherenkov telescopes.","published":"2026-09-25T16:10:27Z","updated":"2026-09-25T16:10:27Z","authors":["The Fermi-LAT Collaboration"],"categories":["astro-ph.HE"],"headline":"New Catalog Lists Gamma-Ray Sources Above 50 GeV","summary":"Researchers compiled a catalog of 673 objects detected in high-energy gamma rays using 16 years of data from the Fermi Space Telescope. This updated catalog improves sensitivity and source localization compared to previous versions, identifying many faint and distant sources, mostly from outside our galaxy. The catalog helps connect space-based and ground-based gamma-ray observations, providing a valuable resource for future astronomical studies, though it mainly summarizes detected sources rather than exploring their detailed nature."},{"id":"2609.31456v1","title":"Diagnosing the Sources of Compositional Failure in Vision-Language Models: A Controlled Analysis","abstract":"Vision-language models (VLMs) often struggle with compositional reasoning tasks, but the reasons for this underperformance remain unclear. A common hypothesis is that models struggle to integrate multiple components, leading to training interventions to improve compositional binding. However, this assumption has never been directly quantified. Existing benchmarks evaluate captions only in their composed form, making it impossible to separate the cost of joint reasoning from the cost of recognizing individual components under increasing load. We introduce COMPASS (COMPositional Analysis of SkillS), a controlled evaluation framework designed to isolate and measure the distinct factors underlying compositional failure. By comparing performance on composed captions with their decomposed counterparts , we directly quantify the cost of compositional integration across 87K image-caption pairs. Across multiple VLMs, this gap is real but partial, accounting for only part of the observed degradation. This motivates a finer-grained investigation into what additional factors govern model behavior. We analyze performance at the level of individual skills: object detection, attribute binding, and relation reasoning, using skill-targeted perturbations across 274K image-caption pairs. We find a consistent skill-specific pattern: each skill degrades primarily with the count of its own primitive type (self-load), while cross-load effects are predominantly positive, suggesting that primitives of different types provide useful grounding context. This pattern holds across standard contrastive encoders, explicitly trained compositional reasoning models, and non-contrastive architectures. These findings show that compositional degradation reflects multiple separable factors that cannot be reduced to joint reasoning alone.","published":"2026-09-25T16:09:42Z","updated":"2026-09-25T16:09:42Z","authors":["Mona Gandhi","Cenk Merih Olcay","Kuan-Chieh Lo","Santiago Castro","Christopher W. Myers","Srinivasan Parthasarathy"],"categories":["cs.CV"],"headline":"Why Vision-Language Models Struggle with Complex Descriptions","summary":"Researchers studied why vision-language models have trouble understanding descriptions that combine multiple parts. They created a new test to separate the difficulty of recognizing parts from the difficulty of putting those parts together. Their analysis showed that the problem is only partly due to combining parts; other factors also affect performance. They found that each skill, like spotting objects or understanding relationships, mainly gets harder when many similar elements appear, while having different types of elements nearby can actually help. This suggests the challenges come from multiple distinct causes."},{"id":"2609.31455v1","title":"Depth-resolved elastic property tomography using higher-order shear-horizontal modes","abstract":"Early material degradation can produce weak, local changes in elastic stiffness before a crack or void forms, but conventional ultrasonic measurements dilute these changes through path and thickness averaging. This paper introduces DEPTH (Depth-resolved Elastic Property Tomography using Higher-order modes), which uses mode-to-mode scattering among higher-order guided waves to reconstruct elastic-property variations as a function of axial position and depth through a plate. The formulation is developed using shear-horizontal (SH) modes, for which the reconstructed property is shear modulus. A Born formulation shows that each incident-received mode pair contains at most two cosine depth orders, allowing the through-thickness field to be recovered from diverse modal mixtures across mode pairs and frequency. The matrix-free regularised inverse is evaluated using direct two-dimensional modal data, practical single-surface excitation and reception, and full three-dimensional finite-element data. In a 10 mm plate, the ideal two-dimensional case gives a through-thickness point-spread FWHM of 0.84-1.03 mm, while the single-surface configuration recovers four compact targets at the correct depths. This provides the first demonstration of depth-resolved elastic-property tomography from multimode guided-wave transmission with excitation and reception confined to a single physical plate surface. With full three-dimensional propagation and finite-width transduction, the reconstruction retains the compact-target positions with a whole-image correlation of 0.899 against a matched acquisition-limited reconstruction. These simulation results demonstrate that higher-order guided-wave mode conversion can provide local depth-resolved elastic-property tomography from surface-access measurements, with potential for characterising early-stage damage such as fatigue.","published":"2026-09-25T16:09:33Z","updated":"2026-09-25T16:09:33Z","authors":["Peter Huthwaite"],"categories":["physics.app-ph"],"headline":"New method maps material stiffness changes through plate depth","summary":"Researchers developed DEPTH, a technique using special ultrasonic waves to detect changes in a material's stiffness at different depths inside a plate. Their simulations show this method can pinpoint depth and location of small, early-stage damage from only one side of the plate. The study supports the idea that measuring wave behavior inside materials can reveal detailed stiffness variations, although results are currently based on simulations rather than physical experiments."},{"id":"2609.31454v1","title":"Different Corruptions, Different Signals: Uncertainty and Loss in Federated Data Quality","abstract":"Federated learning (FL) data corruption can affect either inputs or labels, but it remains unclear whether input-conditional uncertainty and prediction-label loss expose these corruption modes equally. This paper compares two corruption-detection signals in FL: input-conditional uncertainty and prediction-label loss. The uncertainty signal is characterised using a learned aleatoric variance estimate together with Monte Carlo (MC) dropout variance and entropy measures, while the loss is computed against the supplied label. We test these signals against additive image noise and persistent random label flips. On ResNet-20 with CIFAR-10 and SVHN under Dirichlet partitions with data that are not independent and identically distributed (non-IID), the two corruption types behave differently. For persistent random label flips, the within-client per-sample area under the receiver operating characteristic curve (AUC) is 0.85 on CIFAR-10 and 0.95 on SVHN for prediction-label loss, while every uncertainty estimator stays at chance (0.49--0.50). This pattern is consistent with the model remaining confident in the underlying image despite the supplied label being wrong. For image noise, expected-entropy uncertainty rises above chance (0.67 on CIFAR-10 and 0.66 on SVHN), while loss responds comparably (0.64 on both). Each signal is therefore the stronger detector for a different corruption: the prediction-label loss for persistent label flips, and expected-entropy uncertainty for image noise, with its advantage becoming apparent as federation-wide corruption prevalence increases. Robust FL data-quality assessment should match the signal to the corruption rather than rely on uncertainty alone across corruption types.","published":"2026-09-25T16:09:21Z","updated":"2026-09-25T16:09:21Z","authors":["Bradley Scott","Zeqi Luo","Edmond S. L. Ho"],"categories":["cs.LG","cs.AI","cs.CV"],"headline":"Researchers Compare How to Detect Different Data Errors in Federated Learning","summary":"The researchers studied two ways to spot errors in federated learning data, focusing on mistakes in images versus labels. They tested measures of uncertainty and prediction-label mismatch using image noise and random label errors on image datasets. They found that prediction-label mismatch better detects label errors, while uncertainty measures pick up image noise more effectively. Their findings suggest that choosing the right detection method depends on the type of data corruption present, rather than using uncertainty alone."},{"id":"2609.31453v1","title":"The Robust Thompson's Conjecture for Alternating Groups","abstract":"We show that there exists a constant $ε> 0$ such that if $C$ is a conjugacy class in $A_n$ of size at least $|A_n|^{1-ε}$, then $A_n \\setminus \\{1\\} \\subseteq C^2$. This proves the robust version of Thompson's conjecture for the alternating group, conjectured by Shalev (Annals of Math., 2009). Our proof combines character bounds and combinatorial cancellation techniques with the recent theory of hypercontractivity for functions over symmetric groups.","published":"2026-09-25T16:09:17Z","updated":"2026-09-25T16:09:17Z","authors":["Nathan Keller","Noam Lifshitz","Avichai Marmor","Ohad Sheinfeld"],"categories":["math.GR","math.CO"],"headline":"Researchers Prove a Math Conjecture About Group Elements","summary":"The researchers studied a complex question in abstract algebra related to alternating groups, a type of mathematical object. They proved that if a certain large subset of these groups meets specific size conditions, then every element except the identity can be formed by combining elements of this subset. Their proof uses advanced techniques involving character bounds, combinatorics, and recent mathematical theories. The findings address a conjecture proposed by Shalev in 2009. This work is theoretical and focused on mathematical structures, not practical applications."},{"id":"2609.31452v1","title":"Vision-Based 6-DoF Grasp Pose Estimation for Robot Cloth Unfolding","abstract":"Cloth manipulation is a challenging task due to the deformable and high-dimensional nature of cloth, which leads to complex interaction dynamics and perceptual ambiguity arising from frequent occlusions of critical visual cues such as folds, edges, and grasp points. In this work, we tackle cloth unfolding using a regrasping-in-the-air strategy, where one manipulator holds the cloth while the other grasps it at an optimally selected point to unfold it. To this end, we propose CeDiRNet-6DoF, a deep learning framework that jointly predicts effective grasp points and the complete 6-DoF grasp pose from the observed cloth configuration. By integrating dense 3D grasp regression with segmentation and sine-cosine-encoded Euler angles, the proposed method reliably estimates the grasp configuration that maximizes the unfolded cloth area. We extensively evaluated CeDiRNet-6DoF on a bimanual robotic setup within the ICRA 2024 Cloth Competition framework, achieving state-of-the-art performance. An ablation study further validates the benefits of key design components, including joint segmentation, background randomization, and image cropping. These results establish CeDiRNet-6DoF as a robust and versatile foundation for reliable robotic cloth manipulation in unstructured environments.","published":"2026-09-25T16:08:16Z","updated":"2026-09-25T16:08:16Z","authors":["Domen Tabernik","Peter Nimac","Jan Jerićević","Danijel Skočaj","Andrej Gams"],"categories":["cs.RO","cs.CV"],"headline":"New AI Helps Robots Unfold Cloth by Predicting Best Grasp Points","summary":"Researchers developed CeDiRNet-6DoF, a deep learning system that helps robots find the best spots and angles to grasp cloth for unfolding. This method uses 3D data and visual cues to handle the complex shapes and folds of cloth, which are usually hard for robots to interpret. Tested on a two-armed robot setup, the system showed improved unfolding performance, though findings are based on controlled experiments within a robotics competition framework."},{"id":"2609.31451v1","title":"TemplateCraft: Agentic Visual Template Generation","abstract":"The growing popularity of short videos has driven demand for one-click content creation. Visual templates turn uploaded images into personalized content with preset effects, but reusable template generation still requires substantial manual effort in asset preparation and tool orchestration. We propose TemplateCraft, a multi-agent system that converts natural-language instructions into client-executable templates through planning, material generation, effect-workflow generation, and protocol compilation. Its Planner-Evaluator loop uses execution feedback for targeted rollback, while stage-level and long-term memory support revision without parameter updates. We evaluate TemplateCraft on TemplateBench, derived from 60 real-world templates. With the same Qwen3-VL backbone, TemplateCraft raises image/video generation success rates from 56.7%/30.0% to 66.7%/50.0% over Planner-only (best-of-three) and improves template adherence and style consistency. With additional evaluation and revision, it matches or exceeds a GPT-4o Planner-only baseline on selected metrics. Persistent assets further improve cross-input style consistency.","published":"2026-09-25T16:08:00Z","updated":"2026-09-25T16:08:00Z","authors":["Hongjie Yu","Zhiyuan Fan","Yuzhe Zhang","Jiangcun Du","Zhicheng Gao","Yuhong Zhang","Xiaokai Zhan","Zongshi Xie"],"categories":["cs.MM","cs.CV"],"headline":"New System Helps Automate Visual Template Creation for Videos","summary":"Researchers developed TemplateCraft, a tool that turns simple text instructions into ready-to-use visual templates for short videos and images. This system uses a step-by-step process with feedback and memory to improve the creation of visual effects automatically. Tested on real templates, TemplateCraft showed better success rates and style consistency than previous methods, though the results are based on specific benchmarks and may not cover all real-world uses."},{"id":"2609.31450v1","title":"From Reward Signal to Visual Utility: A Controlled Audit of Medical VLM Post-Training","abstract":"Medical vision-language model (VLM) post-training is commonly evaluated through answer accuracy. We examine how changes in accuracy and training objectives relate to image-conditioned decisions in a controlled Qwen2.5-VL-3B study on PMC-VQA. We compare supervised fine-tuning (SFT) with low-rank adaptation (LoRA) restricted to the language model, expanded multimodal adaptation scopes, standard answer-only Group Relative Policy Optimization (GRPO), and a counterfactual evidence objective. On 2,000 clean-test questions, language model LoRA SFT changes correct-image accuracy by +1.10 percentage points (95% paired bootstrap CI:-0.85 to +3.05), while visual-benefit events decrease by 2.40 points and image sensitivity decreases by 5.60 points. Paired records reveal 155 acquired and 203 lost visual-benefit events. Broader adaptation yields lower correct-image accuracy than language-model LoRA SFT. Standard GRPO produces mixed-reward groups and parameter updates, with an uncertain clean test accuracy change. A generation audit reveals that canonical option scores can follow a different token path from generated answers. With scores taken along the greedy generation path, the evidence target improves on the training set; its gains over standard GRPO remain inconsistent on validation data at matched training doses. Sample-level analyses trace how evidence scores, decision margins, and generated answers change during post-training. This empirical and measurement audit identifies gaps between optimization activity, target acquisition, and useful held-out visual behavior.","published":"2026-09-25T16:06:54Z","updated":"2026-09-25T16:06:54Z","authors":["Wang Jingxin"],"categories":["cs.CV","cs.AI"],"headline":"Researchers Audit Medical Vision-Language Model Training Effects","summary":"The researchers examined how different post-training methods affect a medical vision-language model's ability to use images when answering questions. They compared various training techniques and found that some changes increased answer accuracy slightly but reduced how much the model benefits from images. Other approaches showed mixed results, and analysis revealed inconsistencies in how scores and generated answers relate. Their study highlights gaps between training objectives and actual visual understanding, based on controlled tests and measurements rather than practical deployment."},{"id":"2609.31449v1","title":"On the Relation Between Non-Semisimple TQFTs","abstract":"The goal of this paper is to clarify the relation between the Kerler-Lyubashenko TQFT $J_\\mathcal{E}$ associated with the adjoint end $\\mathcal{E}$ of a (not necessarily semisimple) modular category $\\mathcal{C}$ and its renormalized version $V_\\mathcal{E}$ based on the modified trace supported by the ideal $\\mathrm{Proj}(\\mathcal{C})$ of projective objects of $\\mathcal{C}$. More precisely, we construct a $3$-dimensional ETQFT $\\boldsymbol{\\hat{A}}_\\mathcal{E}$ that contains both $J_\\mathcal{E}$ and $V_\\mathcal{E}$. The ETQFT $\\boldsymbol{\\hat{A}}_\\mathcal{E}$ is given by a $2$-functor whose source is the $2$-category of admissible $3$-dimensional cobordisms, and whose target is the $2$-category of finitely complete linear categories. We improve on earlier versions of the construction by dropping the admissibility condition for surfaces. By doing so, we obtain an ETQFT whose circle category $\\boldsymbol{\\hat{A}}_\\mathcal{E}(\\boldsymbol{S}^1)$ is equivalent to the category $\\mathcal{C}$, as opposed to the ideal $\\mathrm{Proj}(\\mathcal{C})$.","published":"2026-09-25T16:06:51Z","updated":"2026-09-25T16:06:51Z","authors":["Marco De Renzi"],"categories":["math.GT","math.CT","math.QA"],"headline":"Linking Two Advanced Mathematical Models in Topology","summary":"This research explores the relationship between two complex mathematical frameworks called TQFTs, which are used to study shapes and spaces in three dimensions. The researchers constructed a new model that combines these two frameworks into one broader system, improving previous versions by removing certain technical restrictions. Their work clarifies how these models relate and better connects abstract mathematical categories, although it remains a theoretical advancement in a specialized area of mathematics."},{"id":"2609.31448v1","title":"ViSTA: A Simple Bridge Extends Visual Alignment to Clinical Time-Series Understanding in Multimodal LLMs","abstract":"Clinical prediction models estimate risk from patient measurements, while large language models support medical text understanding and question answering. Yet their language capabilities do not ensure accurate prediction from structured, high-dimensional clinical time series. Improving this ability would connect risk estimation with flexible questions about a patient's evolving condition. We introduce ViSTA, a compact adapter that incorporates irregular numerical measurements into a pretrained vision-language model's chart representations. It learns corrections to visual tokens while leaving all pretrained parameters unchanged. On MIMIC-IV, ViSTA has the highest mean scores among the compared adaptations on all four metrics for acute kidney injury and mortality prediction across models with 2-9 billion parameters. With 0.516 million trainable parameters, the 2-billion-parameter model reaches an area under the ROC curve of 0.7376 for acute kidney injury, compared with GPT-5.6 Sol's 0.7380 with text input and high reasoning effort. Training for temporal question answering yields 69.27% accuracy at 4 billion parameters with over 90% fewer trainable parameters than low-rank adaptation using charts or numerical text, at a 2.82-4.88 percentage-point accuracy gap. ViSTA extends pretrained language models to numerical prediction and temporal questions.","published":"2026-09-25T16:06:29Z","updated":"2026-09-25T16:06:29Z","authors":["Junyi Gao","Yu Shi","Pingzhao Hu","Ewen M Harrison"],"categories":["cs.CL","cs.AI"],"headline":"New Method Helps AI Predict Health Risks from Patient Data","summary":"Researchers developed ViSTA, a tool that improves how large AI language models analyze complex, irregular patient measurements over time. By adding ViSTA to existing models without changing them, the system better predicts acute kidney injury and death risks and answers time-based medical questions. Their findings, based on experiments with clinical data, show ViSTA achieves prediction accuracy similar to other big models but uses far fewer trainable parts, suggesting a more efficient way to connect medical data analysis with language understanding."},{"id":"2609.31447v1","title":"Horizon saddle connections and Veech groups of infinite-type dilation surfaces","abstract":"This paper studies Veech groups of dilation surfaces whose fundamental group is not finitely generated. We prove that if $ S$ is a surface with self-similar end space and every end is accumulated by genus, then every countable subgroup of $SL(2,\\mathbb{R}) $ can be realized as the Veech group of a dilation surface homeomorphic to $ S$. Additionally, we can construct this dilation surface such that it realizes horizon saddle connections in a prescribed set of directions. This contrasts with the case of closed dilation surfaces, where horizon saddle connections impose restrictions on the algebraic structure of the Veech group.","published":"2026-09-25T16:06:28Z","updated":"2026-09-25T16:06:28Z","authors":["Oscar Rutilio Molina Medrano"],"categories":["math.GT","math.GR"],"headline":"Researchers Explore Complex Shapes and Their Symmetry Groups","summary":"The researchers study special geometric surfaces called dilation surfaces that have infinitely complex structures. They prove that for certain infinite surfaces, any countable symmetry group can be realized through these surfaces, which is surprising compared to simpler closed surfaces where such symmetry groups are limited. They also show how to control certain directional features called horizon saddle connections on these surfaces. Their work is theoretical and advances understanding of the relationship between surface shape and symmetry."},{"id":"2609.31446v1","title":"DistributedDesignOptimizer: A modular Python framework for Setup, Execution and Processing of Distributed Design Optimization","abstract":"Multidisciplinary Design Optimization (MDO) enables the application of optimization algorithms to complex engineered systems, ranging from aerospace and automotive to robotics and microelectronics. Such multi-component systems typically consist of coupled subsystems, each characterized by its own design variables, constraints and objectives. Without adequate coordination of these couplings, subsystems may be optimal in isolation while the overall system remains suboptimal or even infeasible. Distributed design optimization coordinates coupled subsystem optimization problems while allowing them to retain control over their local design variables. Although promising coordination methods exist, their application is hindered by the lack of a comprehensive software framework which supports intuitive problem definition, provides suitable distributed coordination algorithms, is modular and extensible, enables interactive (post-)processing, and supports the distributed computation of subsystem executions. This work derives requirements for such software, assesses existing frameworks against them, and introduces DistributedDesignOptimizer, an open-source Python framework to fill this gap.","published":"2026-09-25T16:05:42Z","updated":"2026-09-25T16:05:42Z","authors":["Sebastian Ellmaier","Albert J. de Wit","Akilesh Raveendran","Marc-Eric Vogt","Thomas Bäck","Anna V. Kononova"],"categories":["cs.MS"],"headline":"New Python Tool to Improve Complex System Design Optimization","summary":"Researchers developed DistributedDesignOptimizer, a new open-source Python framework to help solve complex design problems involving multiple connected parts. The tool coordinates subsystems that each have their own design goals, helping ensure the overall system performs better than optimizing parts individually. The researchers explain that existing software lacks features for easy setup, distributed computing, and flexible processing, and their framework aims to address these needs based on identified requirements."},{"id":"2609.31445v1","title":"Geometric morphisms of virtual ultracategories","abstract":"We build on the recent development that toposes with enough points can be represented as categories of points equipped with ultraconvergence data (a.k.a. virtual ultracategories), characterizing usual classes of geometric morphisms --- surjections, embeddings, hyperconnected maps, and localic maps --- in terms of their points. In particular, the characterisation of geometric surjections relies on a notion of \"virtual ultraretractions\", which generalises usual retractions and filtered colimits, categorifies Johnstone's notion of subclosure for topological spaces, and provides a new characterisation of separating classes of points of a topos.","published":"2026-09-25T16:05:37Z","updated":"2026-09-25T16:05:37Z","authors":["Saadia Gabriel","Yuksel Errol"],"categories":["math.CT","math.LO"],"headline":"Studying geometric maps using virtual ultracategories","summary":"The researchers explore how certain complex mathematical structures called toposes can be understood through simpler objects called points equipped with extra data, known as virtual ultracategories. They focus on describing different types of important maps between these structures by examining their behavior at points. Their work extends existing concepts and offers new ways to characterize these maps, but their findings are theoretical and develop abstract mathematical ideas rather than concrete applications."},{"id":"2609.31444v1","title":"Teleparallel Regularization of Gravitational Action","abstract":"We study the gravitational action in the teleparallel equivalent of general relativity, where the spin connection contributes to the action through a surface term and, when chosen appropriately, regularizes the action by effectively playing the role of a counterterm. We resolve several apparent ambiguities in this procedure using the Schwarzschild spacetime as an example. We show that the apparent discrepancy between bulk and quasilocal actions originates from contributions associated with spacetime singularities, and that the quasilocal action agrees with the standard results obtained in general relativity. We further clarify the role of the tetrad choice in determining the corresponding spin connection, showing that the proper diagonal and canonical frames give the same Euclidean action, while only the proper diagonal frame reproduces the expected Lorentzian action on the Wheeler--DeWitt patch. Finally, we establish the uniqueness of the regularization scheme and show that imposing an appropriate fall-off behavior of the torsion tensor removes the remaining ambiguity and fixes the Euclidean action uniquely.","published":"2026-09-25T16:02:52Z","updated":"2026-09-25T16:02:52Z","authors":["Martin Krššák","Michal Stano"],"categories":["gr-qc"],"headline":"Researchers clarify a new approach to calculating gravity's action","summary":"The researchers study a way to describe gravity called teleparallel gravity, focusing on how to make its mathematical action well-defined and free of ambiguities. Using the Schwarzschild black hole as an example, they identify sources of previous confusion and show how their approach matches standard results from general relativity. They also resolve technical choices about reference frames and demonstrate that their method uniquely fixes the action by setting appropriate conditions on certain mathematical quantities."},{"id":"2609.31443v1","title":"Low-Order Refined Preconditioning for Spectral/hp Element Method for Complex, 3D Geometries","abstract":"Low-order refined (LOR) preconditioning replaces a high-order operator with a low-order discretisation on a refined nodal mesh. For tensor-product elements, the two operators are spectrally equivalent with bounds independent of the polynomial order $P$, but the construction does not extend directly to simplex and mixed-element discretisations. This work makes two contributions: it extends LOR preconditioning to simplex and mixed-element discretisations, including triangular, tetrahedral, and prismatic elements, and establishes a generalised Vandermonde transformation linking the modal and nodal LOR formulations, showing that the resulting preconditioned spectra and Krylov convergence are independent of the high-order basis. Numerical experiments show controlled iteration growth on triangular meshes despite increasing condition number, and controlled iteration counts up to $P=6$ on tetrahedral, prismatic, and mixed-element meshes. A single algebraic multigrid V-cycle per outer iteration gives the best balance of iteration count and cost. The method is applied to a production incompressible Navier-Stokes simulation of a race-car front-wing and wheel configuration, discretised on a mesh of $2.87\\times10^6$ mixed prismatic and tetrahedral elements giving $32.2\\times10^6$ pressure degrees of freedom at $P=3$. LOR reduces the mean pressure conjugate gradient (CG) iteration count from $235.3$ to $5.5$, and the pressure-solve time over 1000 timesteps by 16.1%, relative to the default production static-condensation diagonal preconditioner in Nektar++. The one-time cost of constructing the LOR preconditioner is amortised over the production simulation.","published":"2026-09-25T16:02:00Z","updated":"2026-09-25T16:02:00Z","authors":["Parv Khurana","Henrik Wüstenberg","Spencer J. Sherwin","David Moxey","Julien Hoessler","Athanasios Chatzopoulos"],"categories":["math.NA"],"headline":"Improving Complex 3D Simulations with Faster Mathematical Techniques","summary":"Researchers worked on a method to speed up complex 3D simulations used in engineering, like airflow around race car parts. They extended an existing math technique called low-order refined (LOR) preconditioning to work with more complicated mesh shapes and showed it keeps performance steady regardless of certain technical details. Their tests include a detailed automotive simulation, where their approach significantly reduced computational effort. These findings come from numerical experiments and specific simulations, highlighting potential efficiency gains but remaining within tested scenarios."},{"id":"2609.31442v1","title":"Improved Bootstrap Inference for Dynamic Panel Data models with Interactive Effects","abstract":"We study recursive-design wild bootstrap inference for dynamic panel data models with unobserved common factors estimated by Common Correlated Effects. In the large N,T setting, the bootstrap reproduces the biased limiting distribution in pure autoregressive models, but fails to capture all bias and factor-estimation variance components in models with additional regressors, particularly under weak exogeneity. We trace this failure to holding regressors fixed across bootstrap replications. We propose to combine bootstrap procedure with available bias-correction methods to conduct adjusted inference. Monte Carlo evidence shows substantial improvements over conventional strategies of using bias-correction paired with cross-sectional bootstrap methods.","published":"2026-09-25T16:01:56Z","updated":"2026-09-25T16:01:56Z","authors":["Arturas Juodis","Ovidijus Stauskas","Sander Tromp"],"categories":["econ.EM"],"headline":"Better Bootstrap Methods for Complex Panel Data Analysis","summary":"Researchers studied a statistical technique called bootstrap inference used in analyzing panel data that changes over time and includes hidden common factors. They found that traditional bootstrap methods don't fully correct biases, especially when other variables are involved and some assumptions are weak. To address this, they combined bootstrap with bias-correction methods and tested improvements using simulations. Their findings suggest this combined approach can improve the accuracy of statistical conclusions in these complex models."},{"id":"2609.31441v1","title":"The Kelly--Trotter conjecture and dimension of poset products","abstract":"We study the order dimension of Cartesian products of finite posets. Kelly and Trotter conjectured in 1982 that $\\dim(P\\times Q)\\ge\\dim P+\\dim Q-2$ for all finite posets $P$ and $Q$. For $m\\ge3$, let $R_m$ denote the incidence poset of the complete graph on $m$ vertices. We prove that there is a constant $C$ such that, for all sufficiently large $m$, $\\dim(R_m\\times R_m)\\le\\left(1+\\frac{2}{\\log_2 6}\\right)\\dim R_m+C$. Since $1+2/\\log_2 6<2$, this disproves the Kelly--Trotter conjecture and shows that $2\\dim R_m-2-\\dim(R_m\\times R_m)$ can grow linearly with $\\dim R_m$. For every integer $d\\ge8$, we construct an incidence poset $Q_d$ such that $\\dim Q_d=d$ and $Q_d$ has the $(3,d)$-covering property. Consequently, $\\dim(P\\times Q_d)\\le\\dim P+d-3$ for every poset $P$ with $\\dim P\\ge3$. Thus, for every integer $d\\ge8$, the poset $Q_d$ violates the Kelly--Trotter conjecture with every poset of dimension at least $3$. Finally, we prove that a poset $Q$ has an $(r,s)$-covering if and only if $\\dim(Q\\times C^r)\\le s$ for every finite chain $C$ with at least two elements.","published":"2026-09-25T15:59:19Z","updated":"2026-09-25T15:59:19Z","authors":["Zhaochen Dong","Kaiyun Wang"],"categories":["math.CO"],"headline":"Researchers Disprove a Longstanding Conjecture About Poset Dimensions","summary":"The researchers investigated how a mathematical property called \"dimension\" behaves when combining certain ordered structures, known as posets. They disproved a 1982 conjecture by Kelly and Trotter, which suggested a specific lower bound for the dimension of combined posets. Their findings, based on theoretical constructions, show that the relationship is more complex and that the conjectured inequality does not always hold for large or specially constructed posets."},{"id":"2609.31440v1","title":"High-Yield, 150 ps Polymer Scintillators via Synergistic Purcell and High-Z Enhancement","abstract":"Developing plastic scintillators that simultaneously achieve high light yield (LY), ultrafast timing, and scalable processability remains a formidable challenge in radiation detection. Here, we report a solution-processable hybrid scintillator that integrates high-Z sensitization with Purcell-enhanced radiative recombination within a conjugated-polymer platform. By embedding HfO2 nanoparticles and resonant Ag-SiO2 plasmonic nanoantennas into a poly(9,9-dioctylfluorene) matrix, we exploit the intrinsic sub-nanosecond kinetics of conjugated polymers while overcoming their characteristic low stopping power. The HfO2 nanocrystals amplify energy deposition and suppress interchain aggregation, preserving high photoluminescence quantum yields in the solid state. Simultaneously, the nanoantennas induce strong plasmon-exciton coupling, accelerating radiative decay to ~150 ps. This synergistic architecture yields LY=6.1 photons/keV, comparable to standard commercial plastics with nanosecond lifetimes, yet over an order of magnitude higher than current sub-nanosecond state-of-the-art materials. This performance represents one of the most favourable combinations of efficiency and timing resolution reported to date. Our versatile strategy provides a framework for next-generation organic scintillators, merging record-breaking temporal response with high efficiency for advanced radiation-sensing architectures.","published":"2026-09-25T15:58:31Z","updated":"2026-09-25T15:58:31Z","authors":["Xiaohe Zhou","Hiba H. Karakkal","Chenger Wang","Andrea Fratelli","Francesco Bruni","Matteo L. Zaffalon","Leonardo Poletti","Saptarshi Chakraborty","Emanuele Mazzola","Roberto Lorenzi","Francesco Carulli","Francesca Rossi","Francesco Meinardi","Luca Gironi","Sergio Brovelli"],"categories":["physics.optics","cond-mat.mtrl-sci"],"headline":"New Plastic Scintillators Offer Faster, Brighter Radiation Detection","summary":"Researchers developed a new type of plastic material that detects radiation with both high brightness and extremely fast response times. They combined special nanoparticles and tiny antennas within a polymer to boost energy absorption and speed up light emission to about 150 picoseconds. Their method improves on existing fast materials by producing much more light, suggesting a promising approach for future radiation sensors. These findings are based on laboratory experiments with the new material design."},{"id":"2609.31439v1","title":"Learning to Leverage Compliance: A Policy-Admittance Learning Framework for Robotic Insertion","abstract":"Policy learning and compliant control offer a promising route to reliable autonomous assembly under pose errors and contact uncertainty. However, combining them does not ensure coordination: the policy may continue pushing against contact while the controller yields, producing sustained loading with limited progress. To address this problem, we propose LeCo (Leverage Compliance), a policy-admittance learning framework that guides a visual policy through execution-time interaction under fixed admittance. A multirate feedback mechanism aggregates high-rate contact-interaction records into policy-transition rewards. An integrated conflict cost then characterizes sustained policy-loading/controller-unloading opposition, while a directional high-force tail cost captures continued-loading events within a transition. Together with task completion, these costs encourage the policy to leverage compliance with less unproductive loading. We evaluate LeCo on four real connector-assembly tasks, obtaining an aggregate success rate of 94%. Across tasks, mean successful-trial resultant-force and torque peaks decrease by approximately 30% and 64% relative to the comparison baseline. Reward ablation further shows that adding conflict shaping reduces median successful-trial contact-conditioned conflict density by approximately 53%. These results support learning to leverage fixed compliance by turning multirate policy-admittance interaction into complementary reward signals for effective, lower-load insertion.","published":"2026-09-25T15:58:13Z","updated":"2026-09-25T15:58:13Z","authors":["Chongren Wang","Minghe Li","Honghua Dai","Zhicheng Lin","Shiyang Wei","Xiaokui Yue"],"categories":["cs.RO"],"headline":"New Method Helps Robots Insert Parts More Gently and Reliably","summary":"The researchers addressed a common problem in robot assembly where a robot's actions and its flexible control can conflict, causing unnecessary force and limited progress. They developed a system called LeCo that combines policy learning with careful control to reduce these conflicts by using feedback from the robot's contact interactions. Testing LeCo on four real-world connector assembly tasks, the team achieved a high success rate and significantly reduced force and torque during insertion, suggesting their approach helps robots work more smoothly and with less strain."},{"id":"2609.31438v1","title":"Saddling Timelike Liouville Theory","abstract":"The two-sphere partition function of timelike Liouville theory admits two complementary semi-classical descriptions whose relation presents a non-perturbative puzzle. A direct path-integral expansion around the round two-sphere saddle reproduces the perturbative small-$β$ expansion, with $β$ the Liouville coupling, of the analytically continued Dorn-Otto-Zamolodchikov-Zamolodchikov (DOZZ) formula, while the latter contains an additional oscillatory factor suggestive of a second saddle contribution. We investigate the origin of this factor in the Liouville zero-mode sector using Picard-Lefschetz theory. Taking as input the Hankel-type contour motivated by the inverse-Gamma representation of the zero-mode integral, we determine its Lefschetz-thimble decomposition depending on $β$. For $0<β<1$, it precisely reproduces the saddle structure encoded in the analytically continued DOZZ result. We identify the integration contour as the source of the difference between the zero-mode saddle descriptions. As $β$ approaches $1$, the zero-mode saddle moves to the boundary of field space, while for $β>1$ the same contour is represented by a single contributing thimble. We further relate this Stokes reorganisation to recent probabilistic constructions of timelike Liouville theory.","published":"2026-09-25T15:58:10Z","updated":"2026-09-25T15:58:10Z","authors":["Emilie Despontin"],"categories":["hep-th"],"headline":"Researchers Explore Two Views of a Complex Physics Problem","summary":"Researchers studied a mathematical model called timelike Liouville theory, focusing on a puzzling difference between two ways to approximate a key function related to a sphere shape. They used advanced techniques to understand why one method shows an extra oscillating factor, tracing it to how an integral is calculated. Their work clarifies how these two approximations relate for different parameter values, but it is theoretical and based on complex mathematical analysis rather than experimental results."},{"id":"2609.31437v1","title":"Maximal Chern Numbers from Finite-Range Hopping","abstract":"Finite-range hopping constrains both the number of Dirac points and the masses that gap them. We determine the maximal absolute Chern numbers of an isotropic finite-range extension of the Qi--Wu--Zhang model on a square lattice through thirteenth-neighbor hopping. Algebraic zero counting and mass-weighted sum rules yield global bounds, and we construct explicit Hamiltonians that saturate them. We find $|C|_{\\max}=36,40,51$ for eleventh-, twelfth-, and thirteenth-neighbor hopping. Equal numbers of Dirac points can produce different maxima because finite-range mass harmonics distinguish opposite vortex charges differently. Twice the squared hopping radius bounds every maximum through twelfth-neighbor hopping and is saturated at the eleventh and twelfth, but the thirteenth-neighbor maximum exceeds it. Thus finite-range hopping constrains the maximal Chern number in two complementary ways: through the number of Dirac vortices it can generate and through the mass-sign patterns it can assign to them. This geometric viewpoint provides a design principle for high-Chern bands and a route to analogous bounds in moiré and other lattice systems.","published":"2026-09-25T15:57:30Z","updated":"2026-09-25T15:57:30Z","authors":["Christopher Arriagada Cortés","Vladimir Juričić"],"categories":["cond-mat.mes-hall"],"headline":"Researchers find limits on topological numbers in lattice models","summary":"The researchers studied how extending the range of particle hopping in a square lattice model affects a property called the Chern number, which relates to the system's topological behavior. They calculated maximum possible values for this number by analyzing how certain points called Dirac points and associated 'masses' behave with hopping up to thirteen neighbors. Their mathematical approach identified global limits and explicit examples reaching them, showing that hopping range restricts the number and arrangement of these points and thus the maximum Chern number. The results are theoretical and based on mathematical models of lattice systems."},{"id":"2609.31436v1","title":"Time-Resolution Function of the RADEX Neutron Source","abstract":"The time-resolution function is one of the key factors determining the accuracy of neutron-resonance parameter reconstruction in time-of-flight experiments. A method is proposed for determining the time-resolution function of the RADEX neutron source associated with neutron slowing down in the target, based on detailed Geant4 modeling. The neutron time spectra obtained in the Geant4 simulations are approximated by a sum of three Gamma functions with dimensionless parameters. This model is shown to describe the shape of the time distributions well over the neutron-energy range 50--300 eV. To test the applicability of the proposed approach to a source of a different design, a comparison is made with the calculated n\\_TOF time-resolution function. Its shape is also shown to be well described by a three-component Gamma model.","published":"2026-09-25T15:57:14Z","updated":"2026-09-25T15:57:14Z","authors":["R. M. Djilkibaev"],"categories":["physics.ins-det"],"headline":"New Model Improves Timing Accuracy for Neutron Experiments","summary":"Researchers studied how to better measure neutron timing, which is important for experiments analyzing neutron behavior. They used detailed computer simulations to create a model that describes the timing patterns of neutrons from the RADEX source. Their model, using three mathematical functions, fits the data well for certain neutron energies. They also tested this method on another neutron source design and found it works similarly, suggesting the approach could be useful for different setups."},{"id":"2609.31435v1","title":"Implicit Neural Representation for Hyperspectral Video Compression","abstract":"With the advent of snapshot cameras, hyperspectral video is becoming more readily available. In recent years, new applications have emerged which have led to increasingly larger datasets. However, hyperspectral video compression remains in the early stages. In this study, we explore the use of implicit neural representation as a candidate solution. We propose a novel extension of an existing RGB video compression model, achieving Bjøntegaard Delta PSNR gains of +4.99 dB and Bjøntegaard Delta rate of -88.88% compared to traditional hyperspectral image compression methods applied frame-by-frame. In addition to reconstruction quality, the effects on downstream task performance are measured in the form of object tracking success. Compared to video compressed with methods based on principal component analysis and JPEG2000 in low data regimes, our proposed method improves tracking area under the curve by up to 23.42% and distance precision by up to 35.56% on examples from the HOT2026 dataset.","published":"2026-09-25T15:57:02Z","updated":"2026-09-25T15:57:02Z","authors":["Alfredo Scalera","Paul Murray","Jaime Zabalza"],"categories":["cs.CV","cs.AI","cs.LG"],"headline":"New Neural Method Improves Hyperspectral Video Compression","summary":"Hyperspectral videos, which capture detailed color information, are growing in size and need better compression methods. The researchers adapted a video compression model using implicit neural representation and achieved improved image quality and significantly reduced data size compared to traditional techniques. They also showed that their method helps maintain important details for tasks like object tracking, outperforming older methods in tests with a specific dataset. These results are based on comparisons using standard evaluation metrics and simulations."},{"id":"2609.31434v1","title":"ExoLaN: Physics-Consistent Context-Aware Dynamics Learning for Exoskeletons","abstract":"Task-agnostic assistive exoskeleton control based on human intention offers greater flexibility than conventional approaches that rely on predefined tasks or motion patterns. Human joint torque estimation enables task-agnostic assistance by characterizing user actions. Physics-consistent methods such as Deep Lagrangian Networks (DeLaN) have been applied to estimate the human torques in multi-user settings, but existing approaches cannot adapt to a specific user without retraining, and do not account for intermittent contacts during locomotion. We propose ExoLaN, a Context-Aware DeLaN for human-exoskeleton interaction that learns the full coupled system dynamics while adapting to changes in interaction context. ExoLaN combines temporal context with partial contact-force measurements from force-sensitive insoles to infer latent dynamics embeddings and estimate generalized contact torques. On seven unseen users performing 21 unseen tasks, ExoLaN reduces torque estimation MSE by 7% compared to a black-box baseline. Beyond inverse dynamics, ExoLaN serves as a unified model that also enables accurate forward prediction: training with a multi-step prediction loss reduces acceleration MSE by 59% and long-horizon position and velocity errors by 60% and 93%, respectively, compared with a single-step loss. Moreover, the learned latent context captures task information without explicit task labels, making it a promising signal for task-aware assistive control.","published":"2026-09-25T15:55:39Z","updated":"2026-09-25T15:55:39Z","authors":["Lucas Schulze","Maximilian Schwarz","Jona Hoppe","Jan Peters","Oleg Arenz"],"categories":["cs.RO","eess.SY"],"headline":"New model improves exoskeleton control by understanding user movements","summary":"Researchers developed ExoLaN, a model that better estimates human joint forces for assistive exoskeletons by learning how the human and exoskeleton move together. Unlike earlier methods, ExoLaN adapts to individual users without retraining and accounts for changing contact forces during walking. Tested on new users and tasks, it reduced errors in torque estimates and improved predictions of movement. The model also learns movement context without needing task labels, which could support smarter exoskeleton control. These findings are based on simulation and experimental comparisons reported by the researchers."},{"id":"2609.31433v1","title":"Angular momentum conservation in two-atom superradiance","abstract":"Two-atom superradiance is studied for atoms having a J=0 ground state and a J=1 excited state. The atoms lie on the x-axis and are excited by a σ_{+}-polarized laser pulse propagating in the z-direction. We use two, complementary approaches to obtain the final angular momentum in the field, a source-field approach and a Schrodinger picture approach. Both approaches give identical answers, but it is necessary to include terms in both calculations that, at first glance, might not be expected to contribute. We show that, owing to an exchange between a state in which atom 1 is in its m=1 excited state and atom 2 is in its ground state to one in which atom 2 is in its m=-1 excited state and atom 1 in its ground state, the final angular momentum in the field is not necessarily equal to the initial angular momentum in the two-atom system. The missing angular momentum appears as orbital angular momentum of the two atoms.","published":"2026-09-25T15:55:02Z","updated":"2026-09-25T15:55:02Z","authors":["P. R. Berman","Zeyuan Wang","A. Kuzmich"],"categories":["quant-ph"],"headline":"How Two Atoms Share and Conserve Angular Momentum in Light Emission","summary":"Researchers studied how two atoms emit light together when excited by a laser and how angular momentum (a property related to rotation) is conserved in this process. Using two different calculation methods, they found both agreed but required considering unexpected effects. They discovered that some angular momentum is transferred between the atoms as orbital motion rather than only being carried by the emitted light. These findings show angular momentum conservation in this atomic system involves subtle exchanges between atomic states and motion."},{"id":"2609.31432v1","title":"Evaluation of Acoustic Noise Level and Impulsiveness Inside Vehicles in Different Traffic Conditions","abstract":"Recently, the issue of sound quality inside vehicles has attracted interest from both researchers and industry alike due to health concerns and also to increase the appeal of vehicles to consumers. This work extends the analysis of interior acoustic noise inside a vehicle under several conditions by comparing measured power levels and two different models for acoustic noise, namely the Gaussian and the alpha-stable distributions. Noise samples were collected in a scenario with real traffic patterns using a measurement setup composed of a Raspberry Pi Board and a microphone strategically positioned. The analysis of the acquired data shows that the observed noise levels are higher when traffic conditions are good. Additionally, the interior noise presented considerable impulsiveness, which tends to be more severe when traffic is slower. Finally, our results suggest that noise sources related to the vehicle itself and its movement are the most relevant ones in the composition of the interior acoustic noise.","published":"2026-09-25T15:54:06Z","updated":"2026-09-25T15:54:06Z","authors":["Daniel Flor","Danilo Pena","Hyago Lucas Oliveira","Luan Pena","Vicente A. de Sousa,","Allan Martins"],"categories":["cs.NI"],"headline":"Study Examines Inside Vehicle Noise in Different Traffic Conditions","summary":"Researchers measured the sounds inside a vehicle during various traffic situations to understand noise levels and their patterns. They collected data using a simple device and compared how well two mathematical models described the noise. Their findings suggest that noise tends to be louder when traffic moves smoothly, but sudden, sharp noises increase in slower traffic, likely due to the vehicle's own sounds and movement. This study helps clarify what affects inside-vehicle noise but is based on measurements and modeling only."},{"id":"2609.31431v1","title":"AxonSynth: Domain-Randomized Synthetic Data for Zero-Shot 3D Axon Segmentation in Light-Sheet Microscopy","abstract":"Accurate segmentation of axons in 3D microscopy data is important for analyzing white-matter organization, but dense ground truth labels are expensive to obtain. Existing supervised axon segmentation methods rely on target-domain annotations and can be brittle when tissue type, species, modality, or acquisition conditions change. We present AxonSynth, a domain-randomized synthetic-data framework for training 3D axon segmentation models without manually annotated real training volumes. AxonSynth generates dense synthetic axon labels with orientation priors that reflect realistic fiber configurations and renders them with randomized density, contrast, bias fields, blur, and noise. A three-class 3D U-Net is trained to predict background, axon sheath and intra-axonal space. We evaluate zero-shot transfer on 10 held-out light-sheet microscopy (LSM) patches from macaque and human brain samples labeled with one of three axonal markers, comparing against calibrated thresholding and Frangi filtering using overlap, corrected detection, false-positive, and topology metrics. On macaque samples, AxonSynth achieved the best corrected Dice and corrected precision (0.826 and 0.851), compared with 0.765 and 0.754 for thresholding and 0.685 and 0.762 for Frangi. On human samples, corrected Dice was comparable to thresholding (0.857 vs. 0.868), while component-count error decreased from 22,504 to 3,377. Across all held-out patches, AxonSynth reduced component-count error in 10/10 patches and Euler-characteristic error in 8/10. These results show that synthetic-label domain randomization can reduce dependence on manual axon annotation while supporting synthetic-to-real 3D segmentation.","published":"2026-09-25T15:52:03Z","updated":"2026-09-25T15:52:03Z","authors":["Edward Gaibor","Kyriaki-Margarita Bintsi","Carmen Luz Leiva Ureta","Zayneb Bellatif","Chiara Maffei","Wenze Li","Elizabeth Hillman","Yaël Balbastre","Anastasia Yendiki"],"categories":["cs.CV"],"headline":"New synthetic data helps segment brain axons without manual labeling","summary":"The researchers address the challenge of analyzing 3D brain axon images, which usually requires costly manual labeling. They created AxonSynth, a method that generates varied synthetic data to train computer models to identify axons in 3D microscopy without needing real labeled images for training. Their tests on macaque and human brain samples show AxonSynth performs better or comparable to traditional methods and reduces counting errors, suggesting synthetic training data can improve axon segmentation without manual annotation."},{"id":"2609.31430v1","title":"Compress What You See, Not What You Say: Anchored Context Distillation for Latent-Observation Software Engineering Agents","abstract":"Tool observations dominate the context of software-engineering agents, making long interaction histories costly to maintain. Existing context compression methods can discard information needed by later actions, while adapting agents to soft-token representations can compromise their original behavior. To reduce context while preserving action-critical information and agent behavior, we combine Latent Observations, Hard Actions (LOHA), a context layout that separates compressed history from text needed for exact reference, with Anchored Context Distillation (ACD), a training method that enables latent reading while constraining behavioral drift. LOHA compresses older tool observations into soft tokens while retaining the agent's own turns and the last K observations in text, providing compact access to historical information and exact access to recent content. To enable the agent to use this representation, ACD distills the base model's full-text predictions into the latent view while anchoring its behavior on plain-text inputs to the same base model. On SWE-bench Verified, K=3 reduces context per call by 43% for Qwen3-4B and 57% for SWE-Master-4B-RL, with resolve rates of 12.1% and 21.8% versus 14.5% and 27.5% for their uncompressed bases. A single-run recency sweep reaches 14.4% and 23.0% at K=8, with larger windows generally favoring task performance over compression. Under a 32K-token limit, Qwen3 with K=3 resolves 21.1% of a 199-instance subset versus 11.1% for the same adapted agent using full text. In concurrent single-GPU serving, it achieves 1.9 times that full-text agent's instance throughput.","published":"2026-09-25T15:51:40Z","updated":"2026-09-25T15:51:40Z","authors":["Zhensheng Zou","Guoqing Wang","Dan Hao"],"categories":["cs.AI"],"headline":"New method compresses engineer tool history without losing key info","summary":"Researchers tackled the problem of software-engineering agents struggling with long histories of tool use, which are costly to maintain. They developed a system called LOHA that compresses older tool data while keeping crucial recent text, combined with a training method ACD that helps the agent use this compressed data without changing its behavior much. Their tests show substantial context savings and improved speed, with small drops or even gains in task success compared to uncompressed methods. Results are based on simulation and model benchmarks, not real-world deployment."},{"id":"2609.31429v1","title":"ITS Fairy: Occlusion Assistance Selected Against a Recipient's Own Perception Reports","abstract":"Cooperative perception can expose object state beyond a vehicle's onboard sensors, but sensing occlusion can still leave a local safety application without the objects its collision computation needs. To tackle this challenge, we present the ITS Fairy, an infrastructure-side Server Local Dynamic Map (S-LDM) service whose decision unit is the pair (recipient, missing conflict-relevant object): among objects absent from a recipient's CPM-derived reported awareness, it sends only those relevant to a Time of Closest Approach (TCA) conflict test. Comparable services predict what a vehicle can perceive; the ITS Fairy instead reads what it has already reported. The recipient inserts the selected state into its local LDM and uses its unchanged collision-avoidance controller. We evaluate this application-level mechanism in SUMO--ms-van3t--S-LDM emulation, since extended as VaN3Twin, using a sensing-occluded lane merge and four-way intersection scenario. At every main-sweep speed, the smallest assisted per-encounter minimum TCA exceeds the largest local-only value in the archived data. Additionally, assisted medians remain in the multi-second range where local-only operation repeatedly approaches zero. In the lane-merge robustness data, the median benefit persists at 80% configured assistance omission with 10 and 5 Hz analysis, but largely disappears at 100-120 km/h when 80% omission is combined with 1 Hz analysis. These results demonstrate the application-level value of supplying object state selected against what a recipient has itself reported. They are not a vehicular wireless-channel evaluation, and they do not quantify what selectivity saves relative to forwarding every nearby object.","published":"2026-09-25T15:50:59Z","updated":"2026-09-25T15:50:59Z","authors":["Yenan Wang","Oscar Karlsson","Elad Michael Schiller","Francesco Raviglione","Claudio Casetti"],"categories":["cs.NI","cs.DC"],"headline":"New system helps cars avoid collisions by sharing missing info","summary":"Researchers developed the ITS Fairy, a service that helps vehicles avoid collisions by sending only the important information about objects a car’s own sensors missed. Unlike other methods predicting what a car might see, this system uses what the car has already reported missing to assist its safety decisions. Tested in simulated traffic scenarios, the ITS Fairy showed improved safety margins, though the research is limited to simulations and doesn’t address communication channel performance or full data forwarding comparisons."},{"id":"2609.31428v1","title":"Wireless Evidence Acquisition for Multimodal Inference Constrained by Temporal Windows","abstract":"Multimodal inference in safety-critical applications, such as autonomous driving and robot navigation, requires heterogeneous sensor observations to reach an edge server within a task-prescribed temporal window. Referenced to the event or state being inferred, this window ends at the latest time an observation remains useful for the current decision. Since sensor readiness times, data volumes, wireless channel conditions, and task relevance vary across sensors, maximising network throughput does not necessarily minimise the fused prediction error when the window closes. We cast multimodal uplink scheduling as sequential wireless evidence acquisition and, for a linear minimum mean-square error (LMMSE) fusion model, derive a conditional evidence gain metric from the reduction in residual-error volume. Defined through second-order statistics, this metric applies beyond Gaussian models and coincides with conditional mutual information when the target and prediction errors are jointly Gaussian. We then develop MIRA, a greedy task-driven maximum information-rate allocation policy that combines conditional evidence gain with each sensor's channel state information and remaining data volume, while updating sensor relevance as evidence is acquired. Experiments on synthetic classification, human activity recognition, and vehicle-trajectory regression show that MIRA outperforms both relevance-only and channel-only scheduling. Relative to the former, MIRA improves classification accuracy by up to 70% and reduces regression mean-square error (MSE) by up to 5.5%. Relative to the latter, it requires 58% less acquisition time to attain an 80% target accuracy, while achieving up to 95% higher accuracy and 9% lower regression MSE. These gains are achieved without maximising received-data volume.","published":"2026-09-25T15:47:33Z","updated":"2026-09-25T15:47:33Z","authors":["Alessandro Compagnoni","Anup Mishra","Carla Fabiana Chiasserini","Elad Michael Schiller","Petar Popovski"],"categories":["eess.SP"],"headline":"New Method Improves Wireless Sensor Data Use in Safety Systems","summary":"The researchers studied how to better collect data from different sensors within a limited time for critical tasks like autonomous driving. They created a method called MIRA that decides which sensor data to send based on how useful it is and current wireless conditions. Tests on tasks like classifying activities and predicting vehicle paths showed MIRA improved accuracy and reduced errors compared to other methods. The study focuses on a mathematical model and simulations, so real-world performance still needs testing."},{"id":"2609.31427v1","title":"A variety of the mutual-visibility coloring problem for graphs","abstract":"This paper explores variations of vertex-coloring problems defined on graph visibility properties. It introduces and studies the dual, outer, and total mutual-visibility chromatic numbers, which partition the vertex set of a graph into color classes that preserve specific mutual-visibility conditions called dual, outer or total. The work provides structural conditions under which these chromatic parameters are finite or infinite, and establishes that deciding whether a graph admits a dual, outer, or total mutual-visibility coloring using a given number of colors is NP-complete, even when restricted to two colors. Exact formulas and tight bounds for these chromatic parameters are established across several fundamental graph classes. For block graphs, complete characterizations are provided for the outer and dual mutual-visibility chromatic numbers based on structural invariants such as cut vertices and specific forbidden subgraph structures. On Hamming graphs, the dual and total mutual-visibility chromatic numbers are shown to equal the smaller dimension of the factors, while the outer mutual-visibility chromatic number is proven to equal the star arboricity of a corresponding complete bipartite graph. Finally, the paper examines strong grid graphs, determining exact values for their total, outer, and dual mutual-visibility chromatic numbers. These results demonstrate how the parameter behaviors range from finite constants to infinity depending on the grid dimensions.","published":"2026-09-25T15:47:21Z","updated":"2026-09-25T15:47:21Z","authors":["Saneesh Babu","Marko Jakovac","Dorota Kuziak","Aparna Lakshmanan S.","Ismael G. Yero"],"categories":["math.CO"],"headline":"Studying New Color-Grouping Rules in Network Graphs","summary":"Researchers examined special ways to color points in a network so that certain visibility conditions between points are met, calling these dual, outer, and total mutual-visibility colorings. They found conditions where these colorings are possible and proved that deciding on such colorings with limited colors is a difficult problem (NP-complete), even with only two colors. The study gives exact and approximate values for these colorings in various common network types, showing the results can vary from small numbers to infinite, depending on the network's structure and dimensions."},{"id":"2609.31426v1","title":"Kruglov-entropy cosmology from apparent-horizon thermodynamics: background dynamics and observational constraints","abstract":"We investigate two late-time cosmological models generated by Kruglov's nonadditive entropy at the apparent horizon. Model I, originally derived by Kruglov, follows the Cai--Kim thermodynamic construction and leads to an implicit modified Friedmann equation, whereas Model II, proposed and derived in this work, is obtained from an entropy-integral prescription and contains an effective vacuum contribution. We formulate both models in terms of the dimensionless Hubble parameter, include radiation, impose exact present-day closure, and evolve the physical branch using analytic derivatives of the implicit Friedmann equations. We constrain the models with DESI DR2 baryon acoustic oscillations and a compressed CMB likelihood, with and without DES-Y5 Type-Ia supernovae. Model I is driven to a markedly lower Hubble constant, $H_0\\simeq62.4,\\mathrm{km,s^{-1},Mpc^{-1}}$ for the combined data, and remains non-phantom with $w_{\\rm DE,0}\\simeq-0.814$; it is strongly disfavoured relative to flat $Λ$CDM. Model II stays close to $Λ$CDM in the standard cosmological parameters and gives $w_{\\rm DE,0}\\simeq-1.003$, while its two entropy-sector parameters remain strongly degenerate. Its best-fit likelihood is comparable to $Λ$CDM, but the additional parameters are not supported by information criteria. These results show that the two thermodynamic realizations of the same nonadditive entropy have sharply different phenomenological consequences at the homogeneous-background level. In summary, the observational plausibility of entropy-based cosmology hinges critically on how the generalized horizon entropy is incorporated into the Friedmann dynamics.","published":"2026-09-25T15:47:01Z","updated":"2026-09-25T15:47:01Z","authors":["Candrasyah Muhammad","Burin Gumjudpai","Nandan Roy","Pornrad Srisawad"],"categories":["astro-ph.CO","gr-qc","hep-th"],"headline":"Testing new entropy-based models of the universe’s expansion","summary":"The researchers studied two cosmological models that use a special type of entropy related to the universe’s horizon to explain how the universe expands at late times. They derived equations describing the universe’s expansion and compared the models against observational data, including distant galaxy measurements and supernovae. Their findings show one model predicts a lower-than-expected expansion rate and is disfavored, while the other closely matches standard cosmology but adds unproven extra parameters. The study highlights that how this entropy concept is included makes a big difference in predictions."},{"id":"2609.31425v1","title":"Quantum probes of invisible particles in top-quark pair production at the LHC","abstract":"Top-quark pair production at the LHC provides a bipartite spin system whose correlations are imprinted in the angular distributions of the decay products and can be studied with quantum-information tools. We investigate whether this spin structure can be used to probe invisible new physics in $pp\\to t\\bar t+X$, where $X$ is a spin-0 or spin-1 mediator motivated by top-philic simplified dark-matter models. Focusing on the dileptonic channel at the High-Luminosity LHC, we compare the sensitivity of the inclusive event yield, the visible invariant-mass distribution, an event-level estimator of the entanglement marker $D$, and the fiducial spin correlation $\\mathcal{D}$ measured as a function of the visible invariant mass. We find that observables retaining spin information improve the expected exclusion sensitivity with respect to rate and kinematic information alone, with $\\mathcal{D}(M_{t\\bar t}^{\\rm vis})$ providing the strongest reach for both mediator spins. The same observables also enhance the discrimination between scalar and pseudoscalar couplings of the spin-zero mediator, with $\\mathcal{D}(M_{t\\bar t}^{\\rm vis})$ again providing the strongest sensitivity, particularly at larger mediator masses. These results show that quantum-information-motivated observables of the $t\\bar t$ system can be efficient tools to probe both the presence of invisible new physics and the structure of its couplings to top quarks.","published":"2026-09-25T15:45:00Z","updated":"2026-09-25T15:45:00Z","authors":["Blanca Belmonte","Diego Blas","Alba Cervera-Lierta","Francesco Montagno"],"categories":["hep-ph","hep-ex"],"headline":"Using top-quark spins to detect hidden particles at the LHC","summary":"Researchers studied how pairs of top quarks produced in proton collisions have linked spins that influence how they decay. They explored whether analyzing these spin patterns with quantum-information methods can reveal invisible new particles connected to dark matter. Their simulations suggest that measuring spin-related features provides stronger clues about such unseen particles and the nature of their interactions than just counting events or looking at basic distributions. The results are based on theoretical models and simulations for future experiments at the High-Luminosity LHC."},{"id":"2609.31424v1","title":"Impact of magnetic drift configuration on the edge radial electric field in the TCV tokamak","abstract":"The edge radial electric ($E_r$) in the Tokamak à Configuration Variable (TCV) is compared in matched L-mode discharges with opposite ion magnetic drift directions (favorable versus unfavorable $B \\! \\times \\! \\nabla B$ configurations). As previously reported on the WEST and AUG tokamaks, the $E_r$ profile---measured by Doppler backscattering (DBS)---exhibits a \"well\" just inside the separatrix in the favorable drift case, which is absent or less pronounced in the unfavorable counterpart. This observation holds over a broad range of plasma conditions, notably also in Ohmic discharges with nearly identical edge density and temperature profiles. Density fluctuation characteristics inferred from DBS are not drastically different: Under Ohmic heating, edge fluctuation levels tend to be higher in the favorable configuration, while radial correlation lengths are similar. The edge $E_r$ difference appears uncorrelated with carbon toroidal rotation behavior. Reducing density, or increasing auxiliary heating power tends to accentuate the edge $E_r$ disparity. Plasma current has little impact on both favorable and unfavorable $E_r$ profiles---in contrast to results from WEST. Approaching the L-H transition via auxiliary heating, edge $E_r \\times B$ shear and pressure grow more readily in the favorable configuration, and DBS fluctuation levels are reduced relative to the unfavorable case. Overall, our results confirm that the edge $E_r$ sensitivity to magnetic drift configuration is a robust, multi-machine phenomenon, with a plausible connection to confinement level and H-mode access.","published":"2026-09-25T15:44:55Z","updated":"2026-09-25T15:44:55Z","authors":["S. Rienäcker","L. Vermare","P. Hennequin","C. Honoré","B. Labit","S. Coda","L. Frassinetti","B. Vincent","O. Panico","Y. Wang","K. Singh"],"categories":["physics.plasm-ph"],"headline":"Magnetic drift affects electric fields on fusion device edges","summary":"Researchers studied how the direction of ion magnetic drift influences the edge radial electric field in the TCV tokamak, a fusion experiment. They found that when the magnetic drift is favorable, a distinct dip in the electric field appears inside the plasma edge, which is weaker or absent when the drift is unfavorable. This difference persists under various conditions and may relate to how easily the plasma improves confinement, but the findings are based on measurements and comparisons without proving direct cause."},{"id":"2609.31423v1","title":"A parametric Catalan-type congruence for generalized central trinomial coefficients","abstract":"Let $p>3$ be a prime, and let $b,c$ be integers with $p\\nmid b+2c$. We obtain an explicit congruence modulo $p^2$ for the sum of $\\binom{2k}{k}T_{2k}(b,c^2)/(4^k(k+1)(b+2c)^{2k})$ from $k=0$ to $(p-1)/2$, where $T_n(b,c)$ denotes the generalized central trinomial coefficient. This gives a parametric formula for the sum considered in a question of Wang and Cui. The proof uses a constant term representation and a logarithmic generating function to reduce the sum to two coefficients of an algebraic generating function. A differential identity and a coefficientwise polynomial congruence then yield the desired evaluation. The formula also applies when $p$ divides $b$ or $b-2c$.","published":"2026-09-25T15:44:42Z","updated":"2026-09-25T15:44:42Z","authors":["Liu Yuqi"],"categories":["math.NT","math.CO"],"headline":"New formula found for complex math sums involving prime numbers","summary":"Researchers studied a special type of number sequence called generalized central trinomial coefficients and focused on sums related to prime numbers greater than three. They developed a detailed formula that calculates these sums modulo the square of the prime, answering a question posed by other mathematicians. Their approach involved advanced algebraic techniques and generating functions, with results confirmed under certain mathematical conditions described in the research."},{"id":"2609.31422v1","title":"Towards Mitigating Fabricated Consensus: The Active Provenance Gate for Multi-Agent Debate Synthesis","abstract":"Large language model-based multi-agent debate (MAD) systems are being increasingly used as complex decision pipelines in distributed processes. However, their final synthesis phase still remains inadequately controlled. Even with detailed debate logs, summarizing models are prone to fabricating smoothly written debate consensus that is not grounded in the debate's history. To address this safety gap, this paper presents empirical research and studies if the introduction of active post-debate verification can mitigate the production of such factually unsupported summaries, while still providing valuable information. Furthermore, it is examined whether explicitly signalling divergence is preferable in the absence of a reliable compromise. The Active Provenance Gate (APG) is introduced as a post-debate verification layer that treats the source as a hard constraint, analysing the debate logs, auditing each claim, and applying self-correction. In crisis simulations, the self-healing mechanism more than doubles the average data Provenance Fidelity in difficult condition scenarios, before the strict gate blocks unsupported claims and generates divergence reports. In the human study, a vast majority of the users (over 75%) preferred a report explicitly stating failure in critical scenarios, despite most of them perceiving fabricated consensus from the baseline system as more fluent. Our main contribution is the transition of data origin tracing from passive logging to active conditional blocking before publication.","published":"2026-09-25T15:44:18Z","updated":"2026-09-25T15:44:18Z","authors":["Jakub Masłowski","Jarosław A. Chudziak"],"categories":["cs.MA","cs.AI","cs.CL"],"headline":"New method helps ensure truthful summaries in AI debates","summary":"Researchers studied how AI systems that generate group discussions often create consensus summaries that aren't fully supported by the original debate. They introduced the Active Provenance Gate, a verification step that checks each claim against debate logs and corrects or blocks unsupported statements. Tests showed this method improved factual accuracy and users preferred clear notices of disagreement to smooth but potentially untrue summaries. The findings highlight shifting from just recording data to actively ensuring its accuracy before finalizing outputs."},{"id":"2609.31421v1","title":"Fundamental groups of complements of wobbly divisors in intersections of two quadrics","abstract":"A smooth complete intersection $X\\subset\\mathbb{P}^{n+2}_{\\mathbb{C}}$ of two quadrics admits an interpretation as a moduli space of bundles. For $n=2$, it parametrizes stable rank $2$ parabolic bundles on $\\mathbb{P}^1$ with five marked points, while for $n=3$ it parametrizes stable rank $2$ bundles of odd degree with fixed determinant on a genus $2$ curve. For general $n$, it can be interpreted as a moduli space of semistable twisted $\\mathrm{Spin}$-bundles. We study $X\\setminus W$, where $W$ is the wobbly locus, consisting of bundles that admit a nonzero nilpotent Higgs field. We prove that $W$ is an irreducible divisor for $n\\geq 3$ and compute $H_1(X\\setminus W,\\mathbb{Z})$. Using a root-stack description, we identify $π_1(X\\setminus W)$ with the kernel of a monodromy homomorphism from an orbifold mixed braid group. The Reidemeister-Schreier method then yields an explicit presentation in every dimension. For $n=2$, where $X$ is a degree $4$ del Pezzo surface and $W$ is the union of its sixteen $(-1)$-curves, we obtain a presentation with $10$ generators and $25$ relators and prove that both numbers are minimal among all finite presentations. For $n=3$, we show that $π_1(X\\setminus W)$ admits a presentation with $4$ generators and $78$ relators.","published":"2026-09-25T15:40:33Z","updated":"2026-09-25T15:40:33Z","authors":["Shomrik Bhattacharya","Yuki Matsubara"],"categories":["math.AG"],"headline":"Researchers describe shapes formed by removing special parts from geometric spaces","summary":"The researchers studied certain complex geometric spaces defined by intersecting two quadratic shapes and focused on the space remaining after removing a special subset called the wobbly locus. This wobbly locus consists of objects with particular mathematical properties related to bundles with special fields. They proved this subset is one piece for dimensions three and higher and calculated key algebraic properties of the leftover space, including its fundamental group, which encodes how loops behave there. Their work includes explicit descriptions and formulas, but it is theoretical and based on abstract mathematical methods."},{"id":"2609.31420v1","title":"Network Flexibility Design Under Endogenous Uncertainty","abstract":"Manufacturing firms face uncertainty in both supply and demand. They implement process flexibility, enabling plants to produce multiple products, to hedge against supply--demand mismatches. Assigning more products to a plant results in efficiency losses at the plant, whereas limited flexibility may improve efficiency because of the specialization effect. Furthermore, the flexibility design determines the sourcing locations of products, thereby affecting demand due to the country-of-origin effect and exposure to tariffs. We incorporate these mechanisms into the supply network flexibility design problem by modeling supply and demand uncertainties as endogenous, that is, as functions of the first-stage design decisions. We propose a two-stage stochastic program and solve it via a decomposition scheme that relies on distribution-specific optimality cuts. To avoid exhaustive enumeration of all distributions, we derive distribution-free formulations that are valid for all distributions and for subsets of supply and demand distributions. Through extensive computational experiments, we demonstrate the superior performance of our solution approach and distribution-free formulations relative to exhaustive distribution enumeration and state-of-the-art inequalities, respectively. Moreover, we explain how the problem structure affects the effectiveness of the proposed formulations. From a managerial perspective, we show the trade-offs between endogenous effects and how they affect the optimal flexibility design.","published":"2026-09-25T15:39:45Z","updated":"2026-09-25T15:39:45Z","authors":["Soroush Fatemi-Anaraki","Martin Grunow","Giovanni Pantuso","Stefan Weltge"],"categories":["math.OC"],"headline":"Designing Flexible Supply Networks Amid Supply and Demand Uncertainty","summary":"The researchers study how manufacturing companies can best design flexible production systems to handle unpredictable supply and demand. They model how decisions about product assignments to plants affect efficiency, sourcing locations, demand, and tariffs, treating uncertainties as influenced by initial design choices. Using advanced mathematical programming and simulations, they develop and test new solution methods that may better balance these trade-offs. The findings highlight complex interactions managers face but are based on theoretical models and computational experiments rather than real-world trials."},{"id":"2609.31419v1","title":"Agmon estimates for the difference of exact and approximate Dirichlet eigenfunctions for difference operators","abstract":"We analyze a general class of difference operators $H_\\varepsilon = T_\\varepsilon + V_\\varepsilon$ on $\\ell^2(\\varepsilon\\mathbb{Z}^d)$, where $V_\\varepsilon$ is a multi-well potential and $\\varepsilon$ is a small parameter. We construct approximate eigenfunctions in neighbourhoods of the different wells and give weighted $\\ell^2$-estimates for the difference of these and the exact eigenfunctions of the associated Dirichlet-operators.","published":"2026-09-25T15:39:33Z","updated":"2026-09-25T15:39:33Z","authors":["Markus Klein","Elke Rosenberger"],"categories":["math-ph","math.SP"],"headline":"Comparing exact and approximate solutions for complex mathematical operators","summary":"Researchers studied mathematical objects called difference operators that depend on a small parameter and involve multiple 'wells' or regions. They created approximate solutions near each well and then measured how close these are to the exact solutions using a specific kind of estimate. Their work provides theoretical bounds on the difference between approximate and exact solutions for a certain class of problems, based on mathematical analysis rather than experiments."},{"id":"2609.31418v1","title":"CognitiveReality: Robot-Agnostic Semantic Gaussian Mapping with an LLM Agent for Immersive Collaborative VR Teleoperation","abstract":"A photorealistic 3D view tells a teleoperator where a robot is, but not what the scene contains, how well each object has been observed, or how to turn pointing and speech into robot action. CognitiveReality turns a robot's RGB-D stream into a live, semantically indexed Gaussian-TSDF map shared by an operator in virtual reality and a tool-using language agent. One mapper binary serves any platform through configuration alone: it ingests poses from robot SLAM, joint kinematics, motion capture or an inline visual tracker, bridges localization outages with a shadow tracker and keyframe-anchored PnP, and maintains open-vocabulary instance identities with per-object quality at 2 Hz. Speech and controller rays are grounded against persistent scene objects through validated typed tools and operator-confirmed robot actions. In the controlled agent evaluation, the deployed local Qwen3-VL-8B router reaches 81.24\\% tool exact match, while merge-aware replay correctly redirects 101 absorbed object identifiers. On robot data CognitiveReality exceeds a Gaussian-plus-SDF baseline by 2-8 dB; pose error through 5-40 s SLAM outages stays within 1-8 cm. Deployed live on two quadrupeds, the agent executed 26 of 30 navigation requests and 20 of 20 re-observation requests, raising object quality by 2-5 dB.","published":"2026-09-25T15:39:28Z","updated":"2026-09-25T15:39:28Z","authors":["Timofei Kozlov","Dmitrii Maliukov","Andrey Marchenko","Dmitrii Plotnikov","Miguel Altamirano Cabrera","Dzmitry Tsetserukou"],"categories":["cs.RO"],"headline":"New System Helps Remote Operators Control Robots Using VR and Language","summary":"Researchers developed CognitiveReality, which turns a robot's camera data into a detailed 3D map that an operator can explore in virtual reality. It also connects spoken commands and pointing gestures to robot actions using a language-based assistant. Tested on robot data and live robots, the system improved mapping accuracy and successfully handled most navigation and observation commands. The researchers show its potential for enhancing teleoperation but their results come from controlled experiments and simulations."},{"id":"2609.31417v1","title":"A Proof of the Second brick-Brauer-Thrall Conjecture for $E$-infinite algebras","abstract":"For any finite-dimensional algebra $A$ over an algebraically closed field, we prove that if $A$ is $E$-infinite, then the Second brick-Brauer-Thrall Conjecture holds for $A$. In fact, we show that if there is a rational ray outside the $τ$-tilting fan of $A$, then $A$ admits an infinite family of $θ$-stable bricks of dimension $d$, for some positive integer $d$ and weight $θ$. To show our main result, from any $τ$-regular component of $A$ with no dense orbit, we obtain a brick component that also has no dense orbit and whose points in general position are $θ$-stable modules for a given weight $θ$. As an important consequence, we conclude that $A$ is stably-discrete (also called multiplicity-free) if and only if $A$ is $E$-finite. Moreover, we prove that Demonet's (lattice point) Conjecture implies the stable Second brick-Brauer-Thrall Conjecture. For $E$-infinite algebras, our results settle several open conjectures. Together with our earlier work and recent developments, this leads to important reductions in the study of some challenging open problems.","published":"2026-09-25T15:39:26Z","updated":"2026-09-25T15:39:26Z","authors":["Kaveh Mousavand","Charles Paquette"],"categories":["math.RT"],"headline":"Proving a math conjecture for special algebra types","summary":"Researchers studied certain finite-dimensional algebras, called E-infinite algebras, and proved part of the Second brick-Brauer-Thrall Conjecture applies to them. They show these algebras have infinitely many special stable elements of a fixed size, using advanced geometric and algebraic methods. Their results also link properties of these algebras to other conjectures, helping to simplify some complex open problems. The findings are theoretical and pertain to abstract algebra structures, based entirely on their mathematical proofs."},{"id":"2609.31416v1","title":"Variable Dimension IMM and GPB2 Filters for Tracking in Multiple Model Systems","abstract":"This paper presents mathematically principled filters for multiple-model systems with states of different dimensionality, specifically the variable-dimension interacting multiple model (VD-IMM) filter and the variable-dimension generalised pseudo-Bayesian filter of order 2 (VD-GPB2). To do so, we first provide a Bayesian modelling of a variable dimensional dynamic system, and its measurements. Then, for variable dimensional linear Gaussian dynamic and measurement models, the VD-IMM filter is derived by assuming a posterior that has a Gaussian density for each mode, and then performing a Kullback-Leibler Divergence (KLD) minimisation after each prediction step to keep the Gaussian density form for each mode. Subsequently, the Bayesian update step is performed, keeping a Gaussian density for each mode. The VD-GPB2 filter considers the same model as the VD-IMM filter and also assumes that the posterior for each mode is Gaussian. In contrast, the VD-GPB2 filter propagates a Gaussian mixture for each mode in the prediction step. In the update step, the VD-GPB2 filter performs a KLD minimisation to have a Gaussian density for each mode. Simulation results show the benefits of the VD-IMM and VD-GPB2 filters compared to previous alternatives.","published":"2026-09-25T15:39:08Z","updated":"2026-09-25T15:39:08Z","authors":["Roberto Pérez-Pérez","Ángel F. García-Fernández"],"categories":["eess.SP"],"headline":"New Filters Improve Tracking in Systems with Changing State Size","summary":"Researchers developed two new mathematical filters, called VD-IMM and VD-GPB2, designed to track systems whose state information changes size over time. They used Bayesian methods to model these variable-dimensional systems and created filters that maintain manageable Gaussian representations to simplify calculations. Their simulations show these new filters perform better than older methods, but the results are currently limited to theoretical models and computer simulations."},{"id":"2609.31415v1","title":"Evaluating the accuracy of KV cache reuse techniques","abstract":"Position-independent KV cache reuse aims to reduce latency in retrieval-augmented generation by reusing chunk-level KV caches across prompts. We show that current evaluations of KV cache reuse techniques rely on measurements that fail to faithfully capture the loss of accuracy attributable to reuse, often artificially inflating the reported effectiveness. We also show that existing datasets do not exhibit the reuse dynamics needed to thoroughly evaluate such techniques. To address these issues, we propose an evaluation methodology that measures this accuracy loss without ambiguity and we introduce Boxoffice, a tool that programmatically generates evaluation datasets that exercise challenging KV cache reuse patterns.","published":"2026-09-25T15:36:43Z","updated":"2026-09-25T15:36:43Z","authors":["Samuel Cestola","Tianxiang Xia","Pengfei Zheng","Weiyan Zheng","Bo Wang","Yi Zhao","Diego Didona"],"categories":["cs.LG"],"headline":"Improving How We Measure Cache Reuse Accuracy in AI Models","summary":"Researchers studied a method used to speed up AI systems by reusing certain stored data chunks during text generation. They found that current ways to evaluate this reuse often overestimate its accuracy and that existing data tests are insufficient. To fix this, they proposed a clearer way to measure accuracy loss and created a tool called Boxoffice to generate better test datasets, helping to better assess these reuse techniques."},{"id":"2609.31414v1","title":"Dielectric response of the antiferromagnetic multiferroics with cycloidal equilibrium for the polarization related to the noncollinear spins","abstract":"Linear dynamics of the small amplitude perturbations in the antiferromagnetic multiferroics is analytically considered. Equilibrium state in the considered regime is assumed to be cycloidal structure both for the antiferromagnetic vector and the nonzero magnetization vector (presence of weak ferromagnetism in the system is included). The electric polarization of spin origin related to the \"noncollinear\" spins is considered and its influence on the spin dynamics is included. The dynamical electric susceptibility tensor is obtained to demonstrate response of the multiferroics on the alternating electric field. Existence of two polarization projections is found, response of each of them on two projections of the electric field perpendicular to the cycloid direction is presented. Different roles of the in plane and out of plane electric perturbations is demonstrated analytically. Essential role of the low frequency gapless spin wave in the dielectric response is described. Contributions of the spiral structure and the weak ferromagnetism in the dielectric response are specified.","published":"2026-09-25T15:36:07Z","updated":"2026-09-25T15:36:07Z","authors":["Pavel A. Andreev"],"categories":["cond-mat.mtrl-sci"],"headline":"How Noncollinear Spins Affect Electric Response in Magnetic Materials","summary":"Researchers studied how tiny disturbances affect certain magnetic materials with a cycloidal spin arrangement and weak ferromagnetism. They focused on the electric polarization caused by non-aligned spins and analyzed how this influences the material's reaction to alternating electric fields. Their theoretical work identifies two main polarization directions and describes how in-plane and out-of-plane electric fields differently impact the system's behavior, highlighting the role of low-frequency spin waves in the material's electric response."},{"id":"2609.31413v1","title":"Nontransitivity of braid group actions on exceptional bases in $K_0(\\mathbb{P}^n)$ for $n=4,6$","abstract":"Let $X=\\mathbb{P}^n$, where $p=n+1$ is prime. Consider the orbit of the standard numerical exceptional basis in $K_0(X)$ under mutations, sign changes, and isometries of the Euler form. We construct an invariant of this orbit: for every basis from the orbit, the square of rank of all basic vectors art congruent to $1$ modulo $p$. For $\\mathbb{P}^4$ and $\\mathbb{P}^6$, we present explicit numerical exceptional bases violating this congruence; hence the action is not transitive. We also establish a stronger obstruction modulo $p^2$ and use Polishchuk's theorem to show that no vector of the constructed bases is the class of an exceptional object. Finally, we prove that for every prime $p$, the category $D^b(\\operatorname{Coh}\\mathbb{P}^{p-1}_{\\mathbb C})$ does not contain a pair of mutually orthogonal exceptional objects.","published":"2026-09-25T15:35:32Z","updated":"2026-09-25T15:35:32Z","authors":["Roman Eliseev"],"categories":["math.KT","math.AG"],"headline":"Researchers find limits in mathematical symmetries of projective spaces","summary":"The researchers studied mathematical structures called exceptional bases related to projective spaces when their dimension plus one is a prime number. They examined how certain transformations called mutations and symmetries act on these bases and found that for dimensions 4 and 6, these actions are not fully interchangeable as previously thought. By constructing specific examples, they showed these mathematical transformations do not cover all possible exceptional bases. Their findings also indicate that some of these bases do not correspond to known geometric objects, and they prove that certain pairs of exceptional objects cannot exist in this setting. The results are theoretical and based on abstract mathematical properties."},{"id":"2609.31412v1","title":"dRVG: Quadtree-Guided, Resolution-Complete Online Motion Planning for Polygonal Robots in Unknown Environments","abstract":"We present the dynamic rotation-stacked visibility graph (dRVG), an online motion planner that guides polygonal robots to specified goals in initially unknown, static environ- ments. It merges local roadmaps from successive observations to plan collision-free translations and rotations without a uniform position grid. A spatial quadtree schedules sensing goals across regions to reduce repeated visits while retaining all orientation configurations for routing. Under exact sensing and geometric computation and star-shaped robot and envelope assumptions, dRVG with center scans is resolution-complete relative to full- map RVG at the same angular resolution. In experiments using footprint scans, dRVG solves all 140 cases across 20 difficult maps and seven angular resolutions within a 20 s planning budget, with a median planning time of 1.18 s at 360 orientation layers. Six microMVP demonstrations illustrate the complete online planning loop on a physical robot.","published":"2026-09-25T15:34:53Z","updated":"2026-09-25T15:34:53Z","authors":["Duo Zhang","Hechen Zhang","Junshan Huang","Jingjin Yu"],"categories":["cs.RO"],"headline":"New Method Helps Polygonal Robots Navigate Unknown Spaces Safely","summary":"The researchers developed dRVG, a system that helps polygon-shaped robots move safely to goals in unknown static places by planning paths online using observations as they go. Their method uses a special spatial division (a quadtree) to efficiently decide where to sense next and plans both robot rotations and translations without relying on uniform grids. Tests showed dRVG can solve challenging navigation cases quickly, and physical robot demonstrations confirmed the system's ability to plan routes in real time."},{"id":"2609.31411v1","title":"Non-Uniform Quantum Well and Barrier Thickness Engineering for Robust Ultra-High TER and Low RA Ferroelectric Tunnel Junctions","abstract":"HfO$_2$-based ferroelectric tunnel junctions (FTJs) are promising candidates for scalable non-volatile memory, but simultaneously achieving a high tunneling electro-resistance ratio (TER) and a low resistance-area (RA) product remains challenging. To address this challenge, this work introduces non-uniform quantum well (QW) and barrier thickness engineering in HfO$_2$-based multi-QW FTJs using a self-consistent Preisach-based ferroelectric (FE) model integrated with the coherent and inelastic non-equilibrium Green's function (NEGF) formalism. The non-uniform well and barrier configuration produces a wide range of FTJ design landscapes due to closely spaced, broad resonant states in the low resistance state (LRS) and a larger separation in the high resistance state (HRS), resulting in strong polarization-dependent resonant transmission with TER reaching the order of $\\mathbf{1\\times10^{8}\\%}$ and an LRS RA product as low as $\\mathbf{1~Ω\\cdot\\mathrm{cm}^{2}}$ at read bias and in the presence of scattering. By incorporating self-consistent elastic scattering into the NEGF framework, we also show that elastic scattering can positively influence the TER and RA performance of non-uniform FTJs by progressively increasing the overlap of the closely spaced resonances. Overall, the results establish non-uniform QW and barrier thickness as an effective and robust design parameter for controlling resonant-state alignment and achieving a favorable combination of high TER and low RA in HfO$_2$-based multi-QW FTJs.","published":"2026-09-25T15:34:35Z","updated":"2026-09-25T15:34:35Z","authors":["Balram Khattar","Adarsh Tripathi","Manish Anand","Abhishek Sharma"],"categories":["cond-mat.other"],"headline":"New Design Enhances Memory Device Performance Using Layer Thickness Control","summary":"Researchers studied ferroelectric tunnel junctions, memory devices that face difficulty balancing two key features: a high resistance difference (TER) and low resistance-area (RA). They used advanced computer models to explore how varying the thickness of certain material layers can create beneficial electron states, improving memory switching behavior. Their simulations suggest that this non-uniform layer design could achieve very high resistance changes and low resistance at the same time, even when imperfections are considered. These findings remain theoretical and based on modeling."},{"id":"2609.31410v1","title":"Anatomical Connectivity from Tractography-free Diffusion Tensor Imaging and Its Application to EEG Brain Imagin","abstract":"This proof-of-concept paper demonstrates how the Kalman filter, as a brain imaging method, can be enhanced by incorporating raw diffusion tensor imaging (DTI) data without tractography, in addition to EEG recordings. An efficient algorithm is designed to streamline the otherwise tedious process of inferring connectivity between brain regions from DTI data. The DTI-based evolution model is applied to the Kalman filter and the Standardized Kalman filter, and these are compared with their conventional counterparts using the random walk evolution model. The numerical experiments use synthetic somatosensory and auditory evoked potentials. The results show that the DTI-modelled Kalman filter can obtain deep activity in the brainstem and thalamus. Moreover, the connection model limits the estimated spreads to the correct or nearest anatomical brain region.","published":"2026-09-25T15:34:09Z","updated":"2026-09-25T15:34:09Z","authors":["Joonas Lahtinen"],"categories":["math.NA","math.DS","math.PR"],"headline":"New Method Links Brain Imaging Data to Better Brain Activity Tracking","summary":"Researchers explored how combining raw brain scan data (without detailed mapping) with electrical brain signals can improve tracking of brain activity. They designed a more efficient way to connect brain regions in the data and tested it using computer-generated brain responses. Their approach allowed identification of deep brain activity and more accurate location estimates compared to traditional methods. These findings are based on simulated data, so practical use requires further study."},{"id":"2609.31409v1","title":"Context-Aware Functional Modeling for Android Third-Party Library Detection","abstract":"Third-party libraries (TPLs) are widely used in Android apps, but their reuse can introduce security risks and interfere with downstream program analyses. Existing Android TPL detection approaches face two key limitations: their hand-crafted features are fragile under aggressive code transformations, and their whole-library matching strategies are ineffective when apps retain only part of a TPL. In this paper, we propose LibFan, a learning-based Android TPL detection approach based on context-aware functional modeling. It realizes this modeling through two complementary components: context-aware contrastive learning at the method level and functional partitioning at the library level. At the method level, it learns semantic representations through contrastive training while incorporating outgoing call relationships and class-level context, improving robustness to obfuscation, shrinking, and optimization. At the library level, it partitions each TPL into functionally coherent units and determines library presence using the best-matching partition, thereby accommodating partial library reuse. To evaluate LibFan, we construct a new benchmark comprising 200 apps and 46 vulnerable TPLs, with each app compiled under four transformation configurations. Under the most challenging R8 full mode, LibFan achieves F1 scores of 81.3% at the library level and 47.6% at the version level, representing relative improvements of 64.9% and 35.6% over the state of the art, respectively.","published":"2026-09-25T15:33:54Z","updated":"2026-09-25T15:33:54Z","authors":["Dihao Fan","Jian Zhang","Yasai Shi","Chuan Luo","Xudong Liu","Yang Liu","Chunming Hu","Xu Wang"],"categories":["cs.CR","cs.SE"],"headline":"New method improves detection of Android app libraries despite code changes","summary":"Researchers developed LibFan, a tool that better detects third-party libraries in Android apps even when code is heavily changed or only parts of libraries are used. They designed LibFan to analyze functions in context and to identify meaningful parts of libraries separately, making detection more reliable. Testing on a set of apps with transformed code showed LibFan outperformed previous methods, though results mainly focus on this specific benchmark and approach."},{"id":"2609.31408v1","title":"Dilatonic Reissner--Nordström black holes in a Bertotti--Robinson magnetic background","abstract":"We construct an exact static, magnetically charged solution of four-dimensional Einstein--Maxwell--dilaton theory that generalizes the centered Reissner--Nordström black hole in a Bertotti--Robinson magnetic background. Its pure-background limit is the magnetic EMD deformation of Bertotti--Robinson spacetime. Using the Emparan--Teo solution-generating technique, the charged construction reduces in canonical Weyl coordinates to replacing the Schwarzschild horizon rod by the Reissner--Nordström rod of half-length $σ=\\sqrt{m^2-e^2}$ and integrating the resulting nonlinear background--rod interaction. This yields a completely explicit solution for arbitrary dilaton coupling $α$, with the expected special limits and numerical verification of the full field equations. The horizon exhibits a coupled mechanical, electromagnetic and intrinsic response to the monopole--background interaction. The residual conical stress is tied to north--south magnetic polarization; increasing the background field produces the ordered reversal scales $B_{\\rm pole}<B_{\\rm crit}<B_{\\rm eq}$, independent of $α$ at fixed seed parameters $(m,e)$, while Gauss--Bonnet identifies the conical stress with the integrated distributional Gaussian-curvature contribution at the singular pole. Along the fixed-$(m,B,α,C)$ near-extremal path considered here, every $α>0$ gives vanishing horizon area and divergent curvature, whereas a correlated near-extremal strong-field limit yields a universal magnetic-reversal law.","published":"2026-09-25T15:32:51Z","updated":"2026-09-25T15:32:51Z","authors":["Kenneth Hong"],"categories":["gr-qc"],"headline":"New exact solution for charged black holes in magnetic backgrounds","summary":"Researchers developed a precise mathematical model describing a special type of charged black hole influenced by a strong magnetic field. They extended known black hole solutions by including effects of both electromagnetic and scalar fields, creating a detailed description that matches expected physics and satisfies complex equations. Their findings reveal how the black hole's surface responds to magnetic forces, with certain stresses linked to magnetic polarity and curvature, although these results come from theoretical and numerical analysis rather than direct observation."},{"id":"2609.31407v1","title":"Exact counting of unlabeled quartic graphs by permutation-cycle aggregation","abstract":"The number of unlabeled regular graphs can be expressed as an average of fixed-point counts over vertex permutations, but evaluating each fixed-point count still requires the degree constraints to be enforced. We give an exact recurrence that processes one complete permutation cycle at a time and records the remaining cycles only by their lengths and residual degrees. The recurrence combines internal edge orbits with orbits joining distinct cycles, while binomial and multinomial coefficients retain the multiplicities of choices that lead to the same remaining state. We prove that this state description is sufficient under complete-cycle elimination and derive bounds on the number of states and transitions. For every fixed degree, the resulting algorithm has an $\\exp(O(\\sqrt n))$ upper bound in the number of vertices, including integer-arithmetic costs. The quartic case requires only four positive residual-degree classes for each cycle length. Small-instance comparisons with a separately implemented, vertex-indexed edge-orbit calculation verify both regular and nonuniform residual-degree inputs. The quartic calculation gives unrestricted and connected counts through order 50, including 22 orders beyond the corresponding reference tables through order 28. Connected counts are recovered by the inverse Euler transform, and all 22 identity-permutation contributions for orders 29--50 agree with the published labeled counts. The nonidentity fixed-point terms at these orders have not been independently recomputed.","published":"2026-09-25T15:31:52Z","updated":"2026-09-25T15:31:52Z","authors":["Yue Cheng","Zhipeng Xu"],"categories":["math.CO"],"headline":"New method counts unlabeled quartic graphs more efficiently","summary":"Researchers developed an exact mathematical method to count unlabeled regular graphs, focusing on quartic graphs where each vertex has four edges. Their approach breaks down permutations of graph vertices into cycles and uses a recurrence process to track graph states by cycle lengths and remaining connections. This results in a faster algorithm verified against known cases, extending counts of such graphs up to size 50. However, some parts of the calculation for larger graphs have not yet been independently confirmed."},{"id":"2609.31406v1","title":"Topological Geometry of Neutrino Mass and Mixing","abstract":"This Letter transitions neutrino mixing from an empirical fit to a geometric law by analytically deriving the PMNS mixing angles and mass spectrum from an $A_5$ topological manifold. Solving the Quantum Yang-Baxter Equation forces a geometric fracture at $M_0 \\approx 3.88 \\times 10^{16}$ GeV. Projecting this to the Electroweak scale yields parameter-free targets: maximal CP violation $δ_{CP} = \\pm π/2$, a mass sum $\\sum m_i \\approx 72.48$ meV, with $m_1 < m_3$, and effective masses $m_β\\approx 12.47$ meV and $m_{ββ} \\approx 8.20$ meV, setting definitive experimental thresholds.","published":"2026-09-25T15:29:37Z","updated":"2026-09-25T15:29:37Z","authors":["Enrique Arrieta-Diaz"],"categories":["hep-ph","hep-th","math-ph"],"headline":"New geometric model predicts neutrino properties without free parameters","summary":"Researchers propose a new way to understand how neutrinos mix and their masses by using a geometric shape called an A5 topological manifold, moving beyond just fitting data. They solve a complex equation that suggests a specific energy scale where geometry changes, allowing them to predict key neutrino properties like mixing angles and mass sums without adjustable settings. These predictions offer precise targets for experiments but are theoretical and based on advanced mathematical modeling."},{"id":"2609.31405v1","title":"Solution for Interference in Hotspot Scenarios Applying Q-Learning on FFR-Based ICIC Techniques","abstract":"This work explores interference coordination techniques (inter-cell interference coordination, ICIC) based on fractional frequency reuse (FFR) as a solution for a multi-cellular scenario with user concentration varying over time. Initially, we present the problem of high user concentration along with their consequences. Next, the use of multiple-input multiple-output (MIMO) and small cells are discussed as classic solutions to the problem, leading to the introduction of fractional frequency reuse and existing ICIC techniques that use FFR. An exploratory analysis is presented in order to demonstrate the effectiveness of ICIC techniques in reducing co-channel interference, as well as to compare different techniques. A statistical study was conducted using one of the techniques from the first analysis in order to identify which of its parameters are relevant to the system performance. Additionally, another study is presented to highlight the impact of high user concentration in the proposed scenario. Because of the dynamic aspect of the system, this work proposes a solution based on machine learning. It consists of changing the ICIC parameters automatically to maintain the best possible signal-to-interference-plus-noise ratio (SINR) in a scenario with hotspots appearing over time. All investigations are based on ns-3 simulator prototyping. The results show that the proposed Q-Learning algorithm increases the average SINR from all users and hotspot users when compared with a scenario without Q-Learning. The SINR from hotspot users is increased by 11.2% in the worst case scenario and by 180% in the best case","published":"2026-09-25T15:29:25Z","updated":"2026-09-25T15:29:25Z","authors":["Iago Diógenes do Rego","Vicente A. de Sousa"],"categories":["cs.NI"],"headline":"Using Q-Learning to Reduce Network Interference in Busy Areas","summary":"Researchers studied how crowding in cellular networks causes signal interference and reduces performance. They explored existing techniques that manage frequency use and examined how machine learning, specifically Q-Learning, can automatically adjust settings to improve signal quality. Using computer simulations, they found that their Q-Learning approach significantly increased signal quality for all users, especially those in crowded hotspots. The findings are based on simulated models and show potential benefits of dynamic interference management in busy network scenarios."},{"id":"2609.31404v1","title":"Periodic homogenization in scalar Hencky plasticity","abstract":"This paper investigates the periodic homogenization of a two-phase elasto-plastic material in the framework of $Γ$-convergence. It establishes the link between the homogenized functional and that which was previously obtained in [12] in the context of two-scale convergence. Although there is no gap between the two formulations and that involving two-scale has the hallmark of a (two-scale) elasto-plastic problem, we are unable to positively ascertain the elasto-plastic character of the homogenized energy except in a one-dimensional setting.","published":"2026-09-25T15:27:35Z","updated":"2026-09-25T15:27:35Z","authors":["Gilles A. Francfort","Alessandro Giacomini"],"categories":["math.AP"],"headline":"Studying Material Behavior Using Mathematical Homogenization","summary":"Researchers explored how two-phase elastic and plastic materials behave when averaged over repeating patterns using advanced mathematical techniques. They connected their new findings with earlier work but could only confirm detailed material behavior in a simplified one-dimensional case. The study’s conclusions are based on theoretical analysis and do not fully establish the general behavior of these materials in higher dimensions."},{"id":"2609.31403v1","title":"Sorry Robot, Happy Human: Vision-Language Models Read Only One of Two Legible Typographic Layers","abstract":"Vision-language models (VLMs), despite their success in optical character recognition (OCR) tasks, are vulnerable to typographic attacks and have a fragile structure for images with multiple text layers. In this study, the DecoyBench dataset was created using the Decoy Font method. The dataset consists of 300 images, each containing text with sharp contour lines superimposed on another text with soft shading. Six recent closed-source models from three different model families were evaluated using this dataset under two different prompting conditions (naive and guided) and at two different resolutions ($512\\times512$ and $64\\times64$). A validation study showed that human participants could read both text layers with high accuracy. In contrast, the models, with most variants and both prompting methods, read the contour text with near-human accuracy at high resolution, but almost never fully extracted the shading text. At low resolution, the contour text could not be read by either the models or humans, while the shading text could be extracted with high accuracy. The findings indicate that the evaluated VLMs exhibit a consistent behavioral limitation when processing typographic structures containing multiple spatial frequency layers.","published":"2026-09-25T15:27:21Z","updated":"2026-09-25T15:27:21Z","authors":["Mert İncidelen","Yamen Kashkash","Asya Berker","Murat Aydoğan"],"categories":["cs.CL","cs.AI","cs.CV"],"headline":"AI Vision-Language Models Struggle with Overlapping Text Layers","summary":"Researchers studied how vision-language AI models process images containing two overlapping text layers: one with sharp lines and another with soft shading. They created a special dataset of 300 such images and tested six AI models, comparing their reading ability to humans. While humans could accurately read both layers, the models usually only recognized the sharp-lined text, especially at high image resolution, showing a consistent limitation in handling complex text layouts. These findings are based on the researchers' tests and highlight challenges for current AI text reading methods."},{"id":"2609.31402v1","title":"AFA-Net: A Differential Attention Approach for Auditory Attention Detection","abstract":"Auditory Attention Detection (AAD) utilizes electroencephalographic (EEG) signals to identify a target speaker in a multi-speaker environment. Despite considerable progress, existing deep learning architectures often lack explicit mechanisms for handling noisy EEG data. To address this limitation, we propose Auditory Focus Attention Networks (AFA-Net), a machine learning framework that replaces vanilla attention with a simple yet flexible differential attention mechanism to help focus on task-relevant neural activity. AFA-Net achieves an upward accuracy of 96.8% at the 2s decision window, while using substantially fewer parameters than most existing methods. To the best of our knowledge, AFA-Net is among the first frameworks to explicitly try to combat EEG noise to improve AAD.","published":"2026-09-25T15:25:45Z","updated":"2026-09-25T15:25:45Z","authors":["Philip H. Lee","Shreeram Suresh Chandra","Karan Thakkar","John H. L. Hansen"],"categories":["cs.SD","cs.LG","eess.SP"],"headline":"New Method Helps Detect Focused Listening from Brain Signals","summary":"Researchers developed AFA-Net, a machine learning approach that analyzes brain signals (EEG) to identify which speaker a person is focusing on in noisy situations. This method uses a special attention mechanism designed to reduce the impact of noise in the data, improving accuracy while using fewer computational resources. Their results suggest AFA-Net can better identify auditory attention within short time frames, though findings are based on algorithm performance and need further validation."},{"id":"2609.31401v1","title":"Decodable In-Context State and Model Output Across Training","abstract":"Prior work established that a probe can decode an in-context binding on model errors and that probe-guided steering can repair some of them. We follow probe accuracy, model output, and steering response across public pretraining and post-training checkpoints. Probe accuracy rises during Pythia pretraining, while probe-guided steering moves from negligible all-trial benefit to a larger benefit at two model sizes. Saved scores distinguish probe-correct errors with low and above-uniform model probability for the correct candidate. Oracle-target steering already repairs many early errors, but saved aggregates cannot separate target quality from intervention sensitivity. A held-out comparison of decoders trained on the final state or candidate logits finds no detected final-state advantage on late-checkpoint model errors. An information-theoretic counterexample explains why decodability on errors alone cannot establish discarded output information. The connection to downstream omissions remains open.","published":"2026-09-25T15:25:38Z","updated":"2026-09-25T15:25:38Z","authors":["Manas Venkata Sai Ravulapalli","Samrath Singh Chadha"],"categories":["cs.LG"],"headline":"Researchers Track Decoding and Error Fixes in AI Training","summary":"The researchers studied how well a method called a 'probe' can identify and help fix errors made by AI models during different training stages. They found the probe's accuracy improves as training progresses, and that guiding the model based on the probe helps more in larger models. However, some measures don’t clearly separate error types or explain all improvements. The findings come from examining training checkpoints and theoretical analysis, leaving some questions about how the results affect real-world tasks unanswered."},{"id":"2609.31400v1","title":"Counterdiabatic quantum optimization for efficient state preparation in the Schwinger model","abstract":"The Quantum Approximate Optimization Algorithm (QAOA), whose design is underpinned by the adiabatic theorem, is one of the leading variational quantum algorithms for preparing the ground states of gauge theories. A recently proposed variant, DC-QAOA, incorporates counterdiabatic driving to accelerate the adiabatic process, reducing both circuit depth and runtime. In this work, we present a systematic analysis, from both theoretical and computational perspectives, of DC-QAOA and related counterdiabatic variants for preparing the ground state of the Schwinger model, the paradigmatic (1+1)-dimensional lattice gauge theory. Benchmarking these methods against standard QAOA, we show that counterdiabatic driving improves ground-state preparation while lowering circuit depth requirements. Our results establish counterdiabatic protocols as a practical route towards efficient quantum simulation of gauge theories.","published":"2026-09-25T15:25:34Z","updated":"2026-09-25T15:25:34Z","authors":["Ethan Laval","Bipasha Chakraborty","Stefano Cipolla"],"categories":["hep-lat","quant-ph"],"headline":"New quantum method speeds up state preparation in physics models","summary":"Researchers studied a quantum algorithm called DC-QAOA, designed to prepare specific states faster by using a technique called counterdiabatic driving. They tested this approach on a physics model known as the Schwinger model and found it can prepare ground states more efficiently than standard methods, requiring less computational effort. Their analysis, based on theory and simulations, suggests this method could help improve quantum simulations of certain physical theories, though the results are mainly theoretical and computational."},{"id":"2609.31399v1","title":"Differential Attention Unlocks Complementary EEG and Speech Fusion for Emotion Recognition","abstract":"Multimodal emotion recognition (MER) increasingly pairs EEG with speech, treating internal neural signals and external vocal expression as informative views of affect. In practice, naive fusion underperforms the stronger single modality, because EEG artifacts inject noise that corrupts the shared representation. We introduce EmoSpeechBrain, a multimodal framework built on the insight that noise suppression is a precondition for effective fusion. Its EEG encoder uses differential attention, taking the difference between two attention maps to cancel shared noise and isolate discriminative neural activity. An attention-based gating adapter aligns both modalities in a shared space and weights each one's contribution to the prediction. On two datasets - PME4 and EAV, EmoSpeechBrain improves MER accuracy by up to 12.9% over other state-of-the-art (SOTA) EEG encoders, and surpasses unimodal speech and EEG baselines by up to 13.1% and 23.1%. These results show that once EEG noise is suppressed, fusion delivers gains that naive combination cannot.","published":"2026-09-25T15:25:24Z","updated":"2026-09-25T15:25:24Z","authors":["Philip H. Lee","Shreeram Suresh Chandra","John H. L. Hansen"],"categories":["cs.LG","eess.SP"],"headline":"New Method Combines Brainwaves and Speech to Better Detect Emotions","summary":"Researchers developed EmoSpeechBrain, a system that improves emotion recognition by smartly combining brain signals (EEG) and speech. They addressed the problem of noisy brain data by using a special attention method to reduce interference before linking both types of information. Tests on two data sets showed their approach outperformed existing methods, suggesting careful noise reduction is key to effectively using brain and voice signals together for recognizing emotions."},{"id":"2609.31398v1","title":"Short Paper: Prefix Count Limits Can Increase First-Hit Discovery in Card Reissuance","abstract":"When a card number is compromised, an attacker may search for active numbers sharing its prefix. An issuer might respond by reissuing cards from heavily populated prefixes into less populated ones. We show that this intuitive count control can backfire. For fixed search regions, exposure weights, and total activity, we derive exactly when reducing the maximum prefix count increases the chance that a budgeted search finds an active number. In matched synthetic simulations over 50,000 candidates, targeted replacement meets the count limit in every run but raises supplied-12-digit-prefix discovery relative to equal-volume random replacement in three of twelve settings. Thus a lower prefix count does not by itself certify lower enumeration risk.","published":"2026-09-25T15:25:06Z","updated":"2026-09-25T15:25:06Z","authors":["Wasif Faisal","Suprava Saha Dibya"],"categories":["cs.CR"],"headline":"Limiting card number prefixes can sometimes increase fraud risk","summary":"Researchers studied how changing the distribution of credit card number prefixes affects the chance of attackers finding active cards after a breach. They found that reducing the maximum number of cards sharing a prefix does not always lower the risk of discovery; in some cases, it can actually make it easier for searches to find valid cards. Their findings come from mathematical analysis and computer simulations, showing that simply limiting prefix counts isn't a guaranteed way to reduce fraud exposure."},{"id":"2609.31397v1","title":"Intent2Tc: Automated Intent-to-Traffic Control Translation with Language Models","abstract":"Automated and highly usable Quality-of-Service (QoS) enforcement requires translating high-level service intents into deployable traffic-management policies. Although intent-based networking (IBN) has simplified policy specification, bridging the gap between business-level intents and executable network configurations remains complex, error-prone, and difficult to automate. This paper presents Intent2Tc, a closed-loop language-model-driven framework that translates business-level traffic-shaping intents into declarative sub-intents and subsequently into validated, executable Linux traffic control (tc) configurations. The framework integrates an Active Queue Management (AQM)-based digital twin (DT) semantic model, automated metadata extraction, critique-driven refinement, and Retrieval-Augmented Generation (RAG)-based knowledge reuse to improve semantic consistency and configuration reliability. We evaluate multiple open-source large language models (LLMs) and small language models (SLMs), together with Claude Sonnet-4.6, on 100 Request for Comments (RFC) 9315-compliant traffic-shaping intents. Across both translation stages, Intent2Tc achieves high semantic fidelity, configuration accuracy, and deployment readiness, with Claude Sonnet-4.6 reaching 0.98 semantic similarity, 1.0 semantic unit coverage, and 0.045 normalized edit distance. Furthermore, RAG reduces token consumption and inference latency while enabling compact models such as Phi-4-mini to approach the performance of substantially larger models. Linux tc serves as the target configuration platform, demonstrating the practical applicability of the proposed framework.","published":"2026-09-25T15:25:05Z","updated":"2026-09-25T15:25:05Z","authors":["Andrea Masini","Sudipta Acharya","Paolo Bellavista","Luca Foschini","Burak Kantarci"],"categories":["cs.NI","cs.AI","cs.CL"],"headline":"New AI Tool Translates Network Intentions into Traffic Controls","summary":"Researchers developed Intent2Tc, a system that uses language models to convert high-level business goals about network traffic into detailed and accurate technical configurations for Linux traffic control. The system includes automated checks and learning methods to improve reliability, tested on 100 network traffic scenarios. Their findings show Intent2Tc can produce configurations with high accuracy and efficiency, though these results are based on simulations and testing rather than real-world deployment."},{"id":"2609.31396v1","title":"Augmented Reality Interfaces for Human-Robot Collaboration: Development of a ROS 2-Based Sensor Streaming Framework and Validation via SLAM Algorithms","abstract":"In recent years, Human-Robot Collaboration (HRC) has taken on a central role in Industry 4.0 and collaborative robotics, demanding communication channels that are increasingly bidirectional, intuitive, and efficient. In this context, Augmented Reality (AR) presents itself as a fundamental enabling technology, capable of both displaying information to the operator and gathering spatial data about the surrounding environment. This thesis presents the development of a sensor streaming framework that connects the Magic Leap 2 AR headset with the ROS 2 (Robot Operating System) ecosystem. Using the Unity development environment and the ROSTCP-Connector package, an on-board application for the headset was developed, capable of acquiring real-time data from the integrated sensors (pose tracking, cameras, and environmental sensors) and publishing it to dedicated ROS 2 topics. In order to test the accuracy, latency, and robustness of the generated data stream, the framework was validated using SLAM (Simultaneous Localization and Mapping) algorithms known in the literature. The experimental results demonstrate that the proposed architecture ensures stable data transmission, laying the groundwork for safe real-time interaction and shared spatial awareness, and opening up new perspectives for the control and supervision of robotic systems in complex HRC scenarios.","published":"2026-09-25T15:25:03Z","updated":"2026-09-25T15:25:03Z","authors":["Alessandro Rubert","Stefano Ghidoni","Matteo Terreran"],"categories":["cs.RO"],"headline":"New Augmented Reality System Supports Human-Robot Teamwork","summary":"Researchers developed a system that links an augmented reality headset with a robot control platform to share real-time sensor data. They tested this system using mapping algorithms to check its accuracy and speed. Their results show stable data transmission, suggesting it could help humans and robots work together more safely and effectively in complex environments."},{"id":"2609.31395v1","title":"ActKV: Efficient LLM Agents through Action-Guided KV Cache Management","abstract":"Agentic LLM inference accumulates long KV caches across iterative observation-reasoning-action loops, imposing substantial memory overhead and limiting serving throughput. Existing compression methods emphasize overall output quality, overlooking the asymmetric importance of actions in driving task progress. Our key idea is to establish a compression criterion that values KV entries by their contribution to action generation and prioritizes action quality. However, iterative execution, dynamic memory demands, and scattered action-critical entries pose challenges to eviction policies, budget allocation, and paged memory integration. To this end, we propose ActKV, the first KV cache compression framework tailored for agentic LLM inference. (i) Action-oriented KV cache eviction exploits stable action access patterns to retain entries critical to future actions, supporting reliable task progress under compression. (ii) Confidence-driven adaptive budget allocation uses LLM's intrinsic confidence to adapt the budget to evolving action-critical memory demands. (iii) Page-aware compression management standardizes compression into three primitives with customized kernels, realizing practical throughput gains. On long-trace tasks, ActKV retains an average of 98.53% of FullKV's accuracy with only 25.98% of its peak KV cache memory. It also achieves 3.97 times and 3.58 times FullKV's token and task throughput, delivering state-of-the-art performance.","published":"2026-09-25T15:24:30Z","updated":"2026-09-25T15:24:30Z","authors":["Zihan Wang","Cheng Tang","Lei Gong","Chao Wang","Wenqi Lou","Teng Wang","Xuehai Zhou"],"categories":["cs.OS","cs.AI"],"headline":"New Method Reduces Memory Use in AI Task Planning Systems","summary":"Researchers studied how AI systems that plan actions store memory during repeated thinking steps, which usually requires a lot of memory and slows performance. They developed ActKV, a new way to compress and manage this memory that focuses on keeping information important for deciding actions. Their method adapts memory use based on confidence and manages memory efficiently, achieving similar accuracy with much less memory and faster task completion in computer tests. These results come from simulations of memory handling and have not been tested in real-world applications."},{"id":"2609.31394v1","title":"InternW0-$Δ$: A World Action Model Bridging Predictive Dynamics and Actions with 20K+ Hours of Open Data","abstract":"World Action Models (WAMs) jointly model visual dynamics and action generation for generalist robot manipulation. A central challenge is to integrate priors from large-scale pretrained models---including visual dynamics, scene semantics, geometry, and motion---into a unified framework for robot action generation. We introduce InternW0-$Δ$, a unified WAM pretrained on a heterogeneous corpus that outperforms prior methods across simulation benchmarks and real-robot platforms. InternW0-$Δ$ combines pretrained visual dynamics, scene-level semantics, 4D geometric and motion priors, and action generation within a Mixture-of-Transformers (MoT) framework. A pretrained video expert and an action expert interact under semantic guidance from a frozen VLM, while a pretrained 4D foundation model injects geometric and motion priors through training-only distillation. We further introduce Causal Imprint, which learns future-relevant scene changes from training-only future supervision and provides predictive representations directly to the action expert without future-video rollout at inference. For large-scale joint training, we construct a heterogeneous corpus of robot demonstrations, UMI data, egocentric human demonstrations, and Ego2Robot data, curated and aligned under a common state-action representation. The resulting corpus contains over 20K hours of processed training data, to our knowledge the largest open-source corpus of its kind. We pretrain InternW0-$Δ$ on this corpus and demonstrate strong performance across simulation benchmarks and real-robot platforms. We will open source the training code, model weights, infrastructure, data-processing pipeline, and processed data where licenses permit. Project page: https://internrobotics.github.io/InternW0-Delta/","published":"2026-09-25T15:24:25Z","updated":"2026-09-25T15:24:25Z","authors":["Xingyu Miao","Zizun Li","Baole Fang","Kaiwen Song","Tenghui Wang","Hanxue Zhang","Yating Wang","Xudong Li","Yuping He","Xueyuan Wei","Chao Gao","Xijie Yang","Yingxiang Xu","Kerui Ren","Wenqi Guo","Jianjun Zhou","Xinzhe Wang","Weiguang Zhao","Ni Yang","Zetao Cai","Yufei Xue","Hengjie Li","Zeyu He","Yuanzhen Zhou","Rong Fu","Jianyang Zhang","Siwei Cui","Fuxian Huang","Yunsong Zhou","Xing Gao","Yifei Yao","Qiaojun Yu","Kailin Li","Ming Zhou","Mu Huang","Xinyue Li","Wenze Cui","Bingqi Jiang","Xueyue Zhu","Junting Dong","Haoyu Guo","Tao Lu","Mulin Yu","Bowen Zhou","Bin Zhao","Tianfan Xue","Weinan Zhang","Chunhua Shen"],"categories":["cs.RO","cs.CV"],"headline":"New Robot Model Learns Actions from 20,000+ Hours of Diverse Data","summary":"Researchers developed InternW0-Δ, a robot model that combines visual understanding and action planning by training on over 20,000 hours of various robot and human demonstration videos. Their approach integrates multiple learning components, including visual patterns, scene details, and motion, into a single framework that improves performance in simulations and real robots. While the results are promising, the research is based on pretraining and testing within controlled setups, and the exact practical benefits are still being explored."},{"id":"2609.31393v1","title":"Hamilton's gradient estimates and Liouville theorems for $u_{t}=Δu^{m}+au\\log u+bu$ on Riemannian manifolds","abstract":"In this paper, we apply Nash-Moser iteration and Saloff-Coste's Sobolev inequalities to derive gradient estimates for positive solutions to a class of nonlinear parabolic equations of the form \\begin{equation*} u_{t}=Δu^{m}+a(x, t)u\\log u+b(x, t)u, \\quad 1-\\frac{2}{n}<m<1 \\quad \\text{or}\\quad 1<m<1+\\frac{1}{1+\\sqrt{2n}}, \\end{equation*} on a complete Riemannian manifold $(M,g)$ of dimension $n$, where $a(x, t)$, $b(x, t)$ are $C^{1}$ functions. For $a=b=0$, this equation becomes the porous medium equation (PME) or the fast diffusion equation (FDE) when $m>1$ or $m<1$, respectively. As consequences, we obtain Liouville type theorems for the corresponding constant coefficient equations when \\begin{equation*} 1-\\frac{1}{\\sqrt{n-1}}<m<1\\quad \\text{and}\\quad 1<m<1+\\frac{1}{\\sqrt{n-1}}. \\end{equation*}","published":"2026-09-25T15:22:26Z","updated":"2026-09-25T15:22:26Z","authors":["Jun Sun","Jiaming Yang"],"categories":["math.AP","math.DG"],"headline":"New gradient estimates and theorems for nonlinear heat equations on curved spaces","summary":"Researchers studied a type of nonlinear heat equation on curved spaces called Riemannian manifolds, focusing on equations involving powers and logarithms of the unknown function. They used advanced mathematical tools like Nash-Moser iteration and Sobolev inequalities to estimate how solutions change in space and time. Their work extends known results about these equations, including when solutions must be trivial (Liouville theorems), but the findings are theoretical and apply under certain mathematical conditions."},{"id":"2609.31392v1","title":"PANEL: An Open-Source, Self-Hosted Web Platform for Human Evaluation of Generative Models","abstract":"Human judgement is the reference measure for evaluating generative models, yet the software used to collect it lags behing the methodology. Researchers adapt listening-test frameworks designed for perceptual protocols such as MUSHRA, rely on closed commercial survey platforms, or implement single-use web applications. Live arenas such as Chatbot Arena and Music Arena rank publicly deployed systems at scale, but do not support controlled comparisons of a laboratory's own models with its own participants. We present PANEL, an open-source, self-hosted platform for such studies. A study is authored in the browser and distributed as a single link, with audio, video, image, and text stimuli, seven question types, and screening and skip logic. The platform reports per-question summaries, across-condition significance tests, pairwise win rates and Bradley--Terry scores, and supports power analysis from pilot data. Consent versioning, self-service withdrawal, retention enforcement, and audit logging support GDPR-compliant operation. Each study exports as a machine-readable specification. PANEL is available at https://github.com/matteospanio/panel.","published":"2026-09-25T15:21:47Z","updated":"2026-09-25T15:21:47Z","authors":["Matteo Spanio","Andrea Poltronieri","Martín Rocamora"],"categories":["cs.HC","cs.SD"],"headline":"New Open-Source Tool Helps Researchers Evaluate AI Models with Human Feedback","summary":"Researchers developed PANEL, a free, self-hosted web platform that makes it easier to collect human judgments for testing AI generative models. PANEL lets researchers design studies with various media types and question formats, run controlled comparisons with their own participants, and analyze the results—all while supporting privacy rules. The platform aims to improve on older methods that rely on commercial tools or simpler apps, but these findings are based on the software's features rather than user studies or real-world testing."},{"id":"2609.31391v1","title":"Shaping of the finite-size chaos crossover in Floquet circuits by coherent-mismatch architecture","abstract":"We study how the spatial profile of a deterministic coherent gate perturbation affects finite-size quasienergy statistics and operator dynamics in U(1)-symmetric XXZ Floquet circuits. Using a shared homogeneous reference, we compare a single modified bond with an alternating perturbation applied across the whole chain. After separating symmetry sectors and checking several known integrable forms, we find that the alternating circuits move toward circular orthogonal ensemble level statistics and change operator-space entanglement and out-of-time-order correlations at smaller local amplitudes than the single defects. For intermediate numbers of modified bonds, the response is nonmonotonic and also depends on where the bonds are placed, so bond density alone is not enough. A perturbative analysis identifies the additive weight of many modified bonds, while coherent spatial interference reshapes the coupling between nearby eigenstates.","published":"2026-09-25T15:21:37Z","updated":"2026-09-25T15:21:37Z","authors":["Xiangjun Tan","Wenqi Liu","Hanjie Zhu","Wenkai Bai","Zhanning Wang"],"categories":["quant-ph"],"headline":"How spatial pattern of changes affects quantum circuit chaos","summary":"Researchers studied how applying specific small changes to a complex quantum system, called a Floquet circuit, influences its behavior, focusing on how chaos emerges in systems of limited size. They compared effects of a single change versus alternating changes across the system and found that alternating patterns caused indicators of chaos to appear more readily. Their analysis showed that not just the amount but the arrangement of changes matters, due to interference effects. These findings come from theoretical and numerical studies of idealized quantum models."},{"id":"2609.31390v1","title":"AlphaOpsBench: Benchmarking End-to-End Alpha Strategy Operationalization in Prediction Markets","abstract":"Large language models increasingly generate quantitative trading strategies, yet existing benchmarks assume standardized assets, numerical features, or directly compilable strategy representations---assumptions that prediction-market strategies violate, since a coarse idea may leave the traded outcome, causal information source, signal definition, threshold, sizing, order policy, exit, and settlement behavior unspecified. We introduce \\textsc{AlphaOpsBench}, which evaluates end-to-end operationalization from source-grounded economic hypotheses to auditable executable programs over 581 source-preserving strategy records and a lifecycle-scale Polymarket dataset with 1.28 million binary markets, 183.6 million cleaned executions, settlement evidence, and limit-order-book history, comparing Direct generation against a Staged design-then-code protocol. In a corrected independent-generation study over 36 controlled tasks and 24 preregistered real strategies, strict end-to-end validity remains rare: Direct and Staged obtain 35/180 and 20/180 canonical passes on the controlled cohort and no confirmed pass on the real cohort, and repeated generations vary substantially in model-owned economic choices. By contrast, 775,725 of 783,655 scheduled historical replays complete, showing that replayability is a far weaker property than source-faithful operationalization. Financial outcomes depend on the declared execution model and available historical evidence, and fee and liquidity experiments show that execution costs alter subsequent trading paths rather than acting only as ex-post deductions. \\textsc{AlphaOpsBench} thus separates strategy fidelity, behavioral validity, historical executability, and financial performance in an evidence-aware benchmark for LLM-based quantitative research in prediction markets.","published":"2026-09-25T15:20:59Z","updated":"2026-09-25T15:20:59Z","authors":["Huaiyu Jia","Mingxuan Zhao","Jincheng Gao","Zifan Peng","Wentao Zhang","Siguang Li","Shuo Sun"],"categories":["cs.CE"],"headline":"Testing AI-Generated Trading Strategies in Prediction Markets","summary":"The researchers study how large language models create quantitative trading strategies, especially for prediction markets where strategy details are often incomplete. They developed AlphaOpsBench, a benchmark that evaluates how well AI-generated strategies can be turned from economic ideas into executable programs using a large dataset of prediction markets. Their findings show that fully valid and faithful AI strategies are still rare, and real-world factors like execution costs affect trading outcomes. This work highlights challenges in creating reliable, AI-driven trading strategies for complex markets."},{"id":"2609.31389v1","title":"Unbalancing unit vectors","abstract":"We show that for every $n$ unit vectors $v_1, \\dots, v_n$ in the $d$-dimensional Euclidean space, there exist signs $\\varepsilon_1, \\dots, \\varepsilon_n \\in \\{\\pm 1\\}$ such that $\\lVert \\varepsilon_1 v_1 + \\dots + \\varepsilon_n v_n \\rVert \\ge \\sqrt{2n - d}$, and we characterize the equality cases.","published":"2026-09-25T15:19:40Z","updated":"2026-09-25T15:19:40Z","authors":["Zilin Jiang","Jeck Lim","Skand Parvatikar"],"categories":["math.MG","math.CO"],"headline":"New finding on combining unit vectors in multi-dimensional space","summary":"The researchers studied collections of unit vectors in spaces with multiple dimensions and how assigning positive or negative signs to them affects their combined length. They proved that for any set of such vectors, it's always possible to choose signs so that the total length is at least the square root of (2 times the number of vectors minus the space's dimension). They also identified when this minimum length is exactly reached. The results are theoretical and based on mathematical analysis."},{"id":"2609.31388v1","title":"Adaptive Switching Between Leader-Based and Leaderless BFT Protocols","abstract":"Byzantine fault-tolerant (BFT) protocols are known for providing operational consistency and resilience in distributed systems. However, evolving network conditions, often driven by the network's inherent dynamism or adversarial influence, make it suboptimal to rely on a static protocol at all times. Existing BFT protocol adaptation solutions switch only among leader-based protocols and coordinate each switch through a separate consensus round, leaving them ineffective at handling severe asynchrony or situations in which an adaptive adversary targets the network's leader. We propose BFTide, a protocol adaptation architecture that enables a BFT system to intelligently and swiftly switch to a suitable protocol as network conditions shift. BFTide integrates a novel protocol switching layer that embeds protocol transition logic into the ongoing BFT operation, enabling safe and low-overhead transitions between partially synchronous leader-based protocols and asynchronous leaderless protocols. It further incorporates an offline-trained reinforcement learning policy that allows nodes to propose protocols at runtime based on observed system metrics. Experimental results show that BFTide reduces transaction latency under adverse network conditions compared with static BFT protocols and the state-of-the-art BFT protocol adaptation scheme BFTBrain (NSDI'25), while maintaining comparable throughput. The switching layer adds a modest 10-21% overhead to median latency when idle and requires no separate consensus round per switch.","published":"2026-09-25T15:19:14Z","updated":"2026-09-25T15:19:14Z","authors":["Sudip Bhujel","Yue Li","Ning Zhang","Y. Thomas Hou","Wenjing Lou","Yang Xiao"],"categories":["cs.DC"],"headline":"New System Improves Network Protocol Switching in Distributed Systems","summary":"Researchers studied how to make distributed computer networks more reliable by adapting their behavior to changing network conditions. They developed BFTide, a system that smartly switches between different communication protocols during operation without needing extra coordination steps. Tests showed BFTide can reduce delays when networks face problems and performs better than previous switching methods, though it adds some delay during idle times. The findings come from experiments and demonstrate potential improvements but focus on specific technical setups."},{"id":"2609.31387v1","title":"On the order of a finite group with trivial centre and bounded conjugacy classes","abstract":"A group is called an $n$-$\\bfc$-group if each of its conjugacy classes is finite and contains at most $n$ elements. By a theorem of B.~H.~Neumann the derived subgroup of such a group is finite of $n$-bounded order, whereas the order of the group itself need not be bounded, as extraspecial $p$-groups show. We prove that an $n$-$\\bfc$-group with trivial centre is finite of order at most $n^{64(\\log n)^{5}}$. Passing to the quotient by the hypercentre, we deduce that no assumption on the structure of the group is needed at all: for an arbitrary finite $n$-$\\bfc$-group $G$ one has $|G:Z_\\infty(G)|<n^{64(\\log n)^{5}}$. The finiteness of the group is required here only for the correct definition of the hypercentre.","published":"2026-09-25T15:18:59Z","updated":"2026-09-25T15:18:59Z","authors":["Ilya Gorshkov"],"categories":["math.GR"],"headline":"Researchers find size limits for special finite groups with small conjugacy classes","summary":"The researchers studied a type of mathematical group where each group element's related elements (conjugacy classes) are limited in size. They proved that if such a group has no center (elements that commute with all others), its total size is limited by a formula depending on the maximum size of these classes. They extended this to general finite groups by considering a related subgroup called the hypercentre. Their results rely on theoretical proofs and focus on group order bounds without real-world applications."},{"id":"2609.31386v1","title":"Modeling and Generative-AI-Based Design of Load-Adaptive Gravity Balancing Mechanisms","abstract":"Load-adaptive gravity balancing mechanisms (LA-GBMs) can accommodate various loading conditions by passively changing their characteristics in response to payload variations. However, their design is difficult because both the desired mechanism motion and static equilibrium under variable payloads must be satisfied simultaneously. This study proposes a general design methodology for LA-GBMs that does not depend on specific mechanism architectures or mechanical elements. The necessary conditions for the potential fields of LA-GBMs are formulated, and two general forms are derived: an affine form representing the effect of payload mass and a factorized form representing state transitions associated with load adaptation and gravity balancing. These forms are then provided to generative AI as design requirements to generate candidate potential functions. The generated functions are analytically verified in terms of their conformity to the two general forms and the conditions required for valid LA-GBMs. Furthermore, the obtained potential functions are decomposed into individual terms, and an example of a method for constructing an LA-GBM by combining springs, counterweights, and function-generating linkage mechanisms is presented. By using potential functions as an intermediate representation, the proposed framework enables the generation of LA-GBM design candidates without prescribing a mechanism architecture in advance. Mechanical realizability and manufacturability of the generated potential fields remain important issues for future work.","published":"2026-09-25T15:18:54Z","updated":"2026-09-25T15:18:54Z","authors":["Ryotaro Kayawake","Kazuki Abe","Shota Miyake","Masahiro Watanabe","Kenjiro Tadakuma"],"categories":["cs.RO"],"headline":"AI-Aided Design Method for Adaptive Gravity-Balancing Mechanisms","summary":"Researchers developed a general method to design load-adaptive gravity balancing mechanisms that adjust passively to different weights. They defined mathematical conditions these mechanisms must meet and used generative AI to create candidate designs based on these criteria. Their approach allows design exploration without fixing the mechanism type first, but the study focuses on theoretical models, and practical manufacturing challenges remain unaddressed."},{"id":"2609.31385v1","title":"A Criterion for the Sphericity of a CR Manifold","abstract":"We formulate and prove a simple criterion for the sphericity of a CR manifold (that is, for its equivalence to its model surface). We discuss the differences between the general case, in which the model surface is weighted homogeneous of arbitrary degree, and the case of Levi-nondegenerate manifolds, whose model surface is quadratic.","published":"2026-09-25T15:17:30Z","updated":"2026-09-25T15:17:30Z","authors":["M. A. Stepanova"],"categories":["math.CV"],"headline":"New test identifies when complex shapes resemble standard models","summary":"The researchers developed a straightforward way to determine when a particular type of geometric object called a CR manifold can be considered equivalent to a standard, well-understood shape. They explain how this test differs depending on specific properties of the manifold, such as whether it is Levi-nondegenerate. Their work clarifies these differences by comparing general cases to more specific ones, although the results focus on theoretical criteria rather than practical applications."},{"id":"2609.31384v1","title":"Multivariate conformal uncertainty propagation in multitask atomistic simulation: Successes and pitfalls","abstract":"Machine learning has become the standard tool for the design of interatomic potentials which balance efficiency and accuracy, but uncertainty quantification remains an open problem. Multiscale simulations introduce an additional challenge: robust uncertainty quantification across scales. Even within one scale, computations are often multistage, producing a sequence of target quantities, each dependent on the previous, and each with some uncertainty. Conformal methods have emerged as a model agnostic framework for recalibrating surrogate predictions to produce sets which contain the truth at a user-specified rate. For multistage workflows, we require uncertainty calibration for multiple chemical properties and atomistic configurations, and we want to propagate uncertainty sets to downstream quantities of interest. Such propagation should capture the error cancellations which occur in many downstream targets in materials science; for instance, an approximate energy difference is often more accurate than individual energy predictions. We present the first exploration of multivariate conformal methods for chemical properties, including Bonferroni-corrected hyperrectangles, hyperellipsoidal sets based on the Mahalanobis distance, and custom loss functions within conformal risk control. Calibration is applied directly to predicted energies, atomic forces, and virial stresses, then propagated to elastic constants and vacancy formation energies employing a variety of commonly considered approximate protocols in materials modeling. We highlight the benefits of building correlation predictions into the conformal procedure, making it possible to build sets which capture near symmetries and error cancellation. We conclude with a discussion of the interplay of the employed approximate computational protocol and conformal guarantees.","published":"2026-09-25T15:17:25Z","updated":"2026-09-25T15:17:25Z","authors":["Katharine Fisher","Michael Herbst","James Kermode","Youssef Marzouk"],"categories":["physics.chem-ph","stat.ML"],"headline":"New method tests uncertainty in computer simulations of materials","summary":"Researchers studied how to better estimate uncertainty in computer models that predict multiple properties of materials at the atomic level. They explored several mathematical techniques to recalibrate predictions and capture how errors might cancel out in related calculations, such as energy differences. Their work shows potential benefits of including relationships between predictions but also highlights challenges and limitations tied to the approximations used in these complex simulations."},{"id":"2609.31383v1","title":"Guiding End-to-End Driving Models with Endpoint-Constrained Trajectory Optimization","abstract":"End-to-end driving policies are commonly trained through open-loop behavior cloning, yet they must ultimately operate in closed-loop when deployed on a vehicle, creating a fundamental mismatch between training and execution. Beyond the commonly studied effects of covariate shift and causal confusion, we identify a complementary factor for this open-loop/closed-loop gap: waypoint-based supervision and displacement metrics do not ensure that the intermediate trajectory is physically coherent or easy for the controller to track. We observe that these inconsistencies concentrate primarily at intermediate waypoints, while the predicted endpoint remains comparatively reliable. Based on this observation, we introduce Endpoint-Constrained Optimization (ECO), a lightweight postprocessing layer that anchors the trajectory to the vehicle's executed history, preserves the policy's predicted endpoint, and reshapes the intermediate waypoints to improve feasibility. ECO requires no map, privileged simulator state, or additional training, and can be inserted between a broad range of waypoint-emitting policies and their controllers. Across two closed-loop simulators, it improves the aggregate closed-loop score of all six evaluated generative and regression-based driving policies, and the gains tend to increase with how often the base plans violate motion limits. On HUGSIM, ECO improves VaVAM from 18.1 to 31.0 HD-Score (+71%), achieving 1st place on the HUGSIM Closed-Loop Driving Challenge. Similarly, on AlpaSim, ECO increases the scene scores of VaVAM and DiffusionDrive by 123% and 22%, respectively. These results show that for a broad collection of end-to-end driving models, repairing the intermediate geometry of predicted trajectories without changing the policy's predicted endpoint can substantially improve closed-loop performance.","published":"2026-09-25T15:17:21Z","updated":"2026-09-25T15:17:21Z","authors":["Brayden Zhang","Mahsa Golchoubian","Igor Gilitschenski","Boris Ivanovic","Kashyap Chitta"],"categories":["cs.RO","cs.AI","cs.CV","cs.LG"],"headline":"Improving Self-Driving Car Paths with Smarter Route Adjustments","summary":"Researchers studied how self-driving car systems predict routes. They found that while the final destination predicted is usually accurate, the paths taken between points can be unrealistic or hard for the car to follow. To fix this, they created a method called Endpoint-Constrained Optimization (ECO) that reshapes the intermediate paths while keeping the endpoint the same. Tests in simulators showed ECO improved driving performance across different systems without needing extra training or maps, suggesting better practical route planning for autonomous cars."},{"id":"2609.31382v1","title":"Highlight-Then-Summarize: Learning to Compress Evidence for Long-Context Understanding","abstract":"Long-context understanding requires large language models (LLMs) to reason over lengthy documents, conversations, and code, yet task-relevant evidence is often sparse and scattered amid substantial irrelevant and redundant content. We propose Highlight-Then-Summarize (H2S), a compress-then-reason paradigm that first identifies source-grounded, question-relevant evidence and then integrates it into a compact, question-conditioned summary before producing the final answer. To train this behavior, we construct H2S-Dataset, comprising 6,647 examples from 11 benchmark families with an average context length of 43.9K tokens, and introduce H2S-RL, which provides process-level rewards for evidence selection and summary construction in addition to final-answer correctness. We evaluate on H2S-Bench, a seven-task long-context suite. Under a shared 128K input and 4K output budget, H2S-14B achieves an average score of 32.60, outperforming Qwen3.8-27B by 10.17 points and obtaining the strongest overall result among the evaluated open-source models. H2S-14B also achieves the highest Evidence-Summary Quality score and retains 97.1% of its 16K-budget performance with only a 4K output budget. These results show that explicitly selecting and integrating evidence improves long-context reasoning while enabling more compact generation.","published":"2026-09-25T15:16:35Z","updated":"2026-09-25T15:16:35Z","authors":["Zhaoyuan Xia","Qinghongbing Xie","Yung Xiang Hue","Jianguang Jiang","Gaofeng Lu","Zhenyu Jiao","Xing Yuan","Dai Dai","Tong Mo","Long Zeng"],"categories":["cs.CL","cs.AI"],"headline":"New Method Helps AI Understand Long Texts More Efficiently","summary":"Researchers developed a method called Highlight-Then-Summarize (H2S) that improves how large language models handle long documents by first picking out important evidence related to a question, then summarizing it before answering. They created a dataset and training approach to teach this process, tested it across multiple tasks, and found that their model performs better than others while using less output space. The findings suggest that focusing on relevant details before answering can help AI reason better with lengthy information."},{"id":"2609.31381v1","title":"Completed Pairs Hide Capped Failures: A ReVerPi Case Study of Selective Context Projection","abstract":"Context projection replaces older tool observations with compact, addressable excerpts, reducing repeated input while potentially adding evidence-retrieval turns. We study this trade-off in ReVerPi, a Pi extension with archived observations and matched full/projected continuations. In an 86-run source-reading campaign with 641 model requests, the 15 completed pairs show identical success: 12/15 per arm. Twelve further boundary runs stop, with the runner suppressing the companion whenever the first arm fails to complete. Restoring all 27 boundary runs bounds projected-minus-full success between $-$9 and +1 tasks. One omitted, selector-chosen projected continuation successfully retrieves archive text yet exhausts twelve requests; its full counterpart answers in three. The eleven jointly correct pairs form a fully observed success stratum within this recorded frame: projection reduces aggregate logical tokens by 25%, while increasing the median pair's tokens by 29% and total suffix requests from 35 to 55. Separating fitting from evaluation changes the selector's apparent tie: outside its four fitting pairs, it incurs one extra failure and 8.6% more logical tokens over thirteen comparable runs. This methodological case study connects stopping rules, known bounded failures, unexecuted companions, and resource aggregation. Its findings concern the recorded campaign, rather than population noninferiority or superiority over unrestricted Pi. Evaluations should retain every intervention boundary, execute both allocated arms independently of the first arm's completion, and report completion alongside interaction and token expenditure.","published":"2026-09-25T15:16:08Z","updated":"2026-09-25T15:16:08Z","authors":["Guangzhe Zhang"],"categories":["cs.AI"],"headline":"Study Examines Trade-offs in a Tool for Handling Text Context","summary":"Researchers analyzed how a method called context projection, which compresses tool input by replacing old observations with excerpts, impacts success and resource use in a system named ReVerPi. They found that while projection reduced total processed tokens by 25%, it sometimes increased the effort needed per task and could delay answers. The study highlights the importance of careful testing protocols and notes these findings apply to their recorded tests rather than proving overall tool effectiveness or inferiority."},{"id":"2609.31380v1","title":"Optimization with Region-Reduced ReLU Neural Networks","abstract":"Optimization of mathematical models involving integer decisions and neural networks with ReLU activation (ReLU ANNs) is a challenging task. Nevertheless, such models are an enabling technology in many application domains. A prominent example is superstructure optimization in chemical engineering, where ReLU ANNs are frequently employed as surrogate models for complex nonlinear processes. We survey recent developments in this area. We argue that in addition to network size and training options of the ANNs, the ReLU activation geometry and the number of linear regions on the domain of interest have a strong impact on computational optimization performance. While standard model compression approaches such as structured pruning reduce network size, they do not explicitly address geometric considerations. Therefore, we propose a novel \\emph{region-reduced} model compression approach that combines the stabilization of unstable neurons and the merging of redundant neurons to reduce the number of linear regions while maintaining predictive accuracy through error compensation. We evaluate our method against standard compression approaches on multiple optimization use cases. First, the two-dimensional peaks function for which we can visualize the activation geometry. Second, on optimization over individual surrogate ReLU ANNs for three chemical processes, and third, on a hybrid superstructure optimization problem that involves the three ReLU ANNs, additional process submodels, and binary variables. The results for the superstructure problem demonstrate the large potential of region reduction with a decrease of 40\\% to 50\\% in computational time, yielding solutions closer than 1\\% to the reference at negligible effort of obtaining the compressed model.","published":"2026-09-25T15:15:57Z","updated":"2026-09-25T15:15:57Z","authors":["Christoph Plate","Caroline Ganzer","Mirko Hahn","Alexander Klimek","Heyuan Liu","Sebastian Sager","Kai Sundmacher","Hanna Wilhelm"],"categories":["math.OC"],"headline":"Improving Optimization Using Smarter Neural Network Simplification","summary":"Researchers studied ways to make mathematical optimization easier when using neural networks with ReLU activation functions, which are common in modeling complex processes like those in chemical engineering. They found that reducing not just the size but also the number of regions where the network behaves differently (linear regions) can speed up computations. Their new method combines removing unstable and redundant parts of the network while keeping accuracy. Tests showed this approach can cut computing time nearly in half with minimal loss in solution quality."},{"id":"2609.31379v1","title":"How Much Must a Private Mempool Hide? Exact Leakage Thresholds for Sandwich Attacks","abstract":"Private and encrypted mempools hide pending transactions to stop sandwich attacks and other forms of maximal extractable value (MEV), but what they hide is rarely everything: a transaction's pair, direction, and a coarse range for its size can still leak. How much leakage makes sandwiching pay? We answer exactly for a fee-free constant-product automated market maker, the pricing rule behind Uniswap v2. Traders observe an interval containing the victim's size and bid in a first-price auction for the right to sandwich it, and the winning front-run must keep the victim's trade executable at every size in the interval. The answer turns on the smallest size consistent with the leak. It alone determines the feasible front-runs, the largest feasible front-run is optimal for pointwise, expected, and worst-case profit alike, and the guaranteed profit has a closed form. When execution is costly, a privacy layer that wants to rule out sandwiches profitable at every consistent size may therefore reveal anything about the size except a lower bound above an explicit threshold; the upper end of the range is irrelevant. With two or more symmetric traders, every pure-strategy perfect Bayesian equilibrium of the auction hands the entire expected net rent to the auctioneer. If the direction is hidden too, no non-contingent first leg front-runs both possible directions, while post-trade arbitrage can survive even perfect pre-trade hiding.","published":"2026-09-25T15:15:24Z","updated":"2026-09-25T15:15:24Z","authors":["Tingyi Lin","Jiazhuo Li","Ruoran Lai"],"categories":["cs.GT","cs.CR","q-fin.TR"],"headline":"Researchers find exact privacy needed to block blockchain sandwich attacks","summary":"Sandwich attacks exploit leaked info about pending blockchain trades, like their size range, to make profits. Researchers studied a specific trading model used by Uniswap v2 and calculated exactly how much information a private mempool must hide to stop these attacks from being profitable. They found that the minimum possible trade size leak determines attack feasibility, and hiding everything except a certain lower bound is crucial. Their results are theoretical and based on this model's assumptions about auctions and trade execution costs."},{"id":"2609.31378v1","title":"ContraFM-S2O: Flow Matching-Based One-step SAR-to-Optical Image Translation Model with Contrastive Learning","abstract":"In recent years, diffusion models and GAN-based models have become the mainstream approaches for SAR-to-optical image translation, owing to their advantages, such as high-quality generation and stable training. However, they have shortcomings such as high inference latency and the generated optical images suffer from low detail fidelity, often resulting in blurred edges and loss of fine textures. Thus, we propose ContraFM-S2O, which is a flow matching-based model for SAR-to-optical image translation. Unlike conventional diffusion models, ContraFM-S2O learns to predict the velocity field in training and solves ODE instead of SDE during inference to improve the sampling efficiency. In addition, ContraFM-S2O replaces instantaneous velocity with average velocity along the interpolation path to realize one-step SAR-to-optical image translation and uses contrastive learning to improve the quality of the generated optical images. Experiments show our model achieves state-of-the-art on SAR2Opt and QXS datasets, outperforming baselines, and reduces inference latency via one-step generation.","published":"2026-09-25T15:15:11Z","updated":"2026-09-25T15:15:11Z","authors":["Mingqian Yu","Wei-kuan Chiang","Qiurui Wang","Peilin Zhao"],"categories":["cs.CV"],"headline":"New Model Quickly Converts Radar Images to Clear Optical Images","summary":"Researchers developed a new method called ContraFM-S2O to translate radar images into optical images more efficiently. Unlike previous popular methods that can be slow and produce blurry results, this model predicts image changes in one step using a smarter process. Tests by the researchers showed it creates clearer images faster than existing techniques, though these findings are based on specific datasets and computational experiments."},{"id":"2609.31377v1","title":"Controlling three-dimensional falling liquid films: from weighted-residual models to direct numerical simulation","abstract":"We consider the problem of controlling a three-dimensional falling liquid film using same-fluid blowing and suction through the substrate. Building upon recent results for the corresponding two-dimensional problem, we employ a reduced-dimensional weighted-residual model for the film height and downstream flux as the basis for controller design. Linear stability analysis is used to estimate the number of actuators required to stabilise the film, while a linear-quadratic regulator (LQR) framework is used to construct feedback controls for the linearised system. The resulting gain matrix is then combined with observations from direct numerical simulation (DNS) of the three-dimensional Navier-Stokes equations, allowing controls derived from the reduced-dimensional model to be applied directly to the full flow. Through systematic numerical tests, we demonstrate that these controls successfully stabilise the flat-film solution across a physically relevant range of Reynolds numbers and domain configurations. We further show that both the number and placement of actuators play a crucial role in control performance, with spanwise instabilities requiring additional actuator coverage in the cross-stream direction. Although the extension from two to three dimensions substantially increases the cost of controller construction, this cost is incurred entirely offline during the computation of the gain matrix, making the overall framework computationally tractable. Our results demonstrate that LQR feedback controls derived from a three-dimensional weighted-residual model retain predictive value when applied to fully resolved DNS, extending previous two-dimensional control methodologies and helping bridge the gap between theoretical control design and experimentally relevant falling-film flows.","published":"2026-09-25T15:15:06Z","updated":"2026-09-25T15:15:06Z","authors":["Oscar A. Holroyd","Susana N. Gomes","Radu Cimpeanu"],"categories":["physics.flu-dyn","math.OC"],"headline":"Controlling 3D Falling Liquid Films Using Mathematical Models","summary":"Researchers studied how to control three-dimensional falling liquid films by applying gentle blowing and suction on the surface below. They used simplified mathematical models to design feedback controls and tested these controls with detailed computer simulations of fluid flow. Their findings suggest that controls based on simpler models can effectively stabilize the film in realistic conditions, though the number and placement of control points are important. The work shows promise for linking theory with practical fluid control, but results are based on simulations, not experiments."},{"id":"2609.31376v1","title":"Towards Whole-Study Screening for Congenital Heart Disease in Fetal Ultrasound Using Multiple Instance Learning","abstract":"Congenital heart disease (CHD) is the most common birth defect, yet a large fraction of cases remain undetected on prenatal ultrasound, in part because current artificial-intelligence methods assume that the key diagnostic frames have already been isolated from a study, by a clinician or by a view classifier. We remove that assumption and address CHD screening directly at the level of the whole ultrasound study. We propose a two-stage framework that first learns transferable frame representations by self-supervised masked-autoencoder pre-training on unlabeled fetal ultrasound, then identifies cardiac frames with a disease-robust module and aggregates them with a transformer-based multiple instance learning (MIL) model that produces a case-level diagnosis from study-level labels alone. The model further returns its highest-scoring frames for clinician review, and a hierarchical head separates critical from non-critical CHD. On the internal test set of our multi-source development cohort (FUSE), the proposed cardiac-gated MIL model reaches an area under the curve (AUC) of 0.985 with a specificity of 0.990, outperforming the reproduced NATMED ensemble (AUC 0.861, specificity 0.600) and the FetalCLIP foundation model (AUC 0.867, specificity 0.710). On an independent external cohort, all models initially perform near chance, but label-free CORAL adaptation raises the proposed model from an AUC of 0.513 to 0.944, whereas whole-study and view-dependent baselines do not recover. These results indicate that whole-study MIL with disease-robust cardiac-frame identification is an accurate and deployable route to prenatal CHD screening.","published":"2026-09-25T15:14:48Z","updated":"2026-09-25T15:14:48Z","authors":["Mohamed Azzam","Ruobing Liu","Esther C. Ugwueke","Ziyang Xu","Shibiao Wan","Alex Foy","Abraham Zabih","Jason Christensen","Neil Hamill","Ling Li","Jieqiong Wang"],"categories":["eess.IV","cs.CV"],"headline":"AI Approach Screens Whole Ultrasound for Fetal Heart Defects","summary":"Researchers aimed to improve detection of congenital heart disease (CHD) in prenatal ultrasound by analyzing entire studies, not just selected key images. They developed a two-step AI method that first learns image features from unlabeled ultrasound, then detects heart-related frames and assesses disease risk using a special model. Their tests show high accuracy on internal data and improved results on new data after adjustment, suggesting this method could help screen for CHD more effectively. However, findings are based on model testing and need further clinical validation."},{"id":"2609.31375v1","title":"Polar discontinuity screening by charge disproportionation in ferroelectric-nickelate superlattices","abstract":"We use density functional theory (DFT) to demonstrate rearrangement of charge disproportionation as a new mechanism for screening polar discontinuities at heterointerfaces. We focus on SmNiO$_3$, a III-III perovskite characterized by an insulating ground state that originates from disproportionation of the electrons at the Ni sites. When interfaced with ferroelectric, II-IV perovskite BaTiO$_3$, this system can display a polar discontinuity whose magnitude is determined by the nature of the interface (NiO$_2$-BaO or SmO-TiO$_2$) and the orientation of the BaTiO$_3$ spontaneous polarization. In the orientation maximizing the interface charge, SmNiO$_3$ compensates by transferring 1 $e$/f.u. between the two interfaces and modifying the disproportionation sequence from Ni$^{4+}$/Ni$^{2+}$ to Ni$^{2+}$/Ni$^{2+}$ at the SmO-TiO$_2$ interface and to Ni$^{4+}$/Ni$^{4+}$ at the NiO$_2$-BaO interface. While this mechanism is potentially universal in all perovskites exhibiting charge disproportionation, our formal analysis of BiNiO$_3$/BaTiO$_3$ superlattices suggests that larger polar discontinuities may need more complex disproportionation patterns for screening.","published":"2026-09-25T15:13:41Z","updated":"2026-09-25T15:13:41Z","authors":["Edith Simmen","Jonathan Amara","Philippe Ghosez","Nicola A. Spaldin"],"categories":["cond-mat.mtrl-sci"],"headline":"Charge Redistribution Helps Fix Interface Problems in Special Materials","summary":"Researchers used computer simulations to study how electrons rearrange at the boundary between two materials, SmNiO3 and BaTiO3. These materials have different electrical properties and meet at an interface that creates an imbalance called polar discontinuity. The team found that SmNiO3 adjusts by shifting electrons between edges, changing its internal charge pattern to balance the interface. This helps explain how such materials might naturally fix electrical mismatches, though more complex cases might need different adjustments, according to the researchers' theoretical study."},{"id":"2609.31374v1","title":"RECAST: From Log Replay to Closed-Loop Driving Simulation with View-Complete Actors","abstract":"Closed-loop driving simulation requires rendered observations to remain reliable as the ego vehicle and surrounding actors move beyond their recorded trajectories, exposing views absent from the source log. Existing data-driven simulators reconstruct dynamic actors from sparse observations, which can result in rendering artifacts under these viewpoint changes. We introduce RECAST (REconstructing Controllable Actors for Simulation and Testing), a 3D Gaussian Splatting framework that generates a view-complete actor from a single segmented vehicle observation in a driving log and registers the generated actor in the reconstructed scene. RECAST supports planner-in-the-loop rendering under controlled ego-actor interactions. To adapt an image-to-3D prior to real vehicles, we further introduce RECAR, a dataset of approximately 20K real vehicles with 600K background-free RGBA images spanning diverse vehicle colors and types. We use two-stage adaptation to improve vehicle generation from real driving-log observations. At the actor level, RECAST reduces $\\mathrm{FD}_{\\mathrm{incep}}$ from 9.788 to 7.992 relative to unadapted TRELLIS. At the scene level, under actor motion beyond logged trajectories, RECAST reduces $\\mathrm{FD}_{\\mathrm{incep}}$ from 129.35 to 112.10 and increases $\\mathrm{CLIP}_{\\mathrm{margin}}$ ($\\times1000$) from 0.14 to 3.47 relative to Street Gaussians. We demonstrate planner-in-the-loop simulation with the image-conditioned planner GTRS-Dense. Compared with native Street Gaussians actors, RECAST increases the no-collision (NC) rate from 22.2% (12/54) to 63.0% (34/54) and the mean minimum predicted time-to-collision (TTC) from 0.798 s to 2.150 s. These experiments show that RECAST supports closed-loop planner evaluation under controlled ego-actor interactions beyond log replay. Visit our project page at https://zijunkr.github.io/RECAST/","published":"2026-09-25T15:12:05Z","updated":"2026-09-25T15:12:05Z","authors":["Zijun Zhao","Liewen Liao","Kang Shen","Songan Zhang","Ming Yang"],"categories":["cs.RO","cs.CV"],"headline":"New method improves driving simulation with realistic vehicle views","summary":"Researchers developed RECAST, a system that creates complete 3D vehicle models from single images in driving logs, allowing simulations to show realistic views even when vehicles move beyond recorded paths. They introduced a large vehicle image dataset called RECAR to help their method learn real vehicle appearances. Compared to previous methods, RECAST reduces visual errors and improves safety measures in simulations, supporting better testing of driving plans when vehicle interactions differ from recorded data. The results are based on simulations and image data adaptations, not real-world tests."},{"id":"2609.31373v1","title":"Hamilton-Type Gradient Estimates for Porous Medium and Fast Diffusion Equations on Riemannian Manifolds for all exponents","abstract":"We establish local Hamilton-type gradient estimates for positive $C^{2,1}$ solutions of $u_t=Δu^m$ on complete Riemannian manifolds whose Ricci curvature is bounded from below. For every fixed $m>1$ and every fixed $0<m<1$, there exist $β=β(m,n)>0$ and $C=C(m,n)>0$ such that a solution $0<u\\leqslant A$ in $B_{2R}(x_0)\\times(t_0-T,t_0]$ satisfies the following local gradient estimate \\begin{equation*} \\sup_{B_R(x_0)\\times(t_0-T/2,t_0]}|\\nabla u^β| \\leqslant C A^β\\left(\\frac1R+\\sqrt{k}+\\frac{A^{(1-m)/2}}{\\sqrt T}\\right), \\end{equation*} where $\\mathrm{Ric}_M\\geqslant-k$. The proof uses an intrinsic quantitative alternative to locate, around each prescribed positive point, a cylinder on which the solution has a controlled upper-to-lower ratio. A local gradient estimate on this cylinder is combined with a stopping argument. As a consequence, every uniformly bounded positive ancient solution on a connected complete manifold with nonnegative Ricci curvature is constant.","published":"2026-09-25T15:10:48Z","updated":"2026-09-25T15:10:48Z","authors":["Jun Sun","Jiaming Yang"],"categories":["math.DG","math.AP"],"headline":"Mathematicians find new gradient estimates for heat-like equations on curved spaces","summary":"Researchers studied how heat-like processes described by porous medium and fast diffusion equations behave on curved spaces called Riemannian manifolds. They derived new local gradient estimates—that is, bounds on how rapidly solutions can change—in terms of geometric properties of the space and parameters of the equation. Their results apply to all relevant powers in the equation and imply that certain long-lasting, bounded solutions must be constant under nonnegative curvature conditions. These findings are theoretical and rely on mathematical analysis rather than experiments."},{"id":"2609.31372v1","title":"Quantum algorithms for the exponentiation of Toeplitz matrices and applications in partial differential equations","abstract":"We present quantum algorithms to approximate the exponential of banded Toeplitz matrices. These matrices have central importance in many PDE related problems, but their quantum implementation is hindered by their possibly large norms. Constructing directly the exponential we can circumvent this limitation. By relating the lower/upper shift operators to circulant and skewcirculant generators, which are diagonalised by the QFT, we construct (i) an LCU-based block encoding of banded Toeplitz matrices, (ii) an efficient, controllably truncated Pauli-string decomposition of the circulant eigenphases with a closed-form error bound, and (iii) a specialized QFT-Trotter product formula. As an application we build a block encoding of the propagator of the discretised heat equation with periodic, Dirichlet and Neumann boundary conditions, using a frequency cutoff.","published":"2026-09-25T15:08:29Z","updated":"2026-09-25T15:08:29Z","authors":["Xabier Gutiérrez","Nicola Mariella","Javier González-Conde","Sergiy Zhuk","Mikel Sanz"],"categories":["quant-ph"],"headline":"Quantum methods to handle special matrices in solving equations","summary":"The researchers developed quantum algorithms to approximate the exponential of banded Toeplitz matrices, which are important in solving partial differential equations (PDEs). They addressed challenges with implementing these matrices on quantum computers due to their large values by constructing their exponential form directly. Their work includes new techniques involving quantum Fourier transform and block encodings, and they demonstrated an application to the heat equation under various boundary conditions. These results are theoretical and involve complex quantum procedures."},{"id":"2609.31371v1","title":"Towards Understanding LLM-Based Log Anomaly Detection: An Empirical Study of Performance, Efficiency, and Robustness","abstract":"Large language models (LLMs) have demonstrated promising performance in log anomaly detection, yet how their adaptation strategies, architectures, and deployment configurations affect detection effectiveness remains insufficiently understood. To investigate these factors, we conduct a systematic empirical analysis across three public log datasets, examining different adaptation strategies, model architectures, parameter scales, and quantization settings. Our results reveal substantial performance differences across adaptation strategies, while model scaling yields varying detection gains across datasets. We further observe that models with comparable detection accuracy can exhibit markedly different computational costs, and that low-bit quantization largely preserves detection performance in the evaluated configurations. Finally, we examine detection robustness under structural, semantic, and label noise at different perturbation levels. These findings provide empirical insights into the performance, efficiency, and robustness of LLM-based log anomaly detection, highlighting practical considerations beyond conventional accuracy-oriented evaluation.","published":"2026-09-25T15:08:16Z","updated":"2026-09-25T15:08:16Z","authors":["Bin Li","Dongdong Wang","Siyang Lu"],"categories":["cs.LG","cs.CR"],"headline":"How Big Language Models Detect Problems in Computer Logs","summary":"Researchers studied how large language models (LLMs) detect unusual patterns in computer log data. They tested different ways of adapting the models, their sizes, and technical settings on three datasets. The study found that adaptation methods significantly affect detection accuracy, model size impacts vary, and simpler models can use less computing power without much accuracy loss. They also checked how well models handle various types of errors. These insights help understand LLM performance beyond just accuracy, based on experiments rather than theory."},{"id":"2609.31370v1","title":"Adaptive Dissipative State Preparation through Reinforcement Learning","abstract":"Dissipative algorithms approach the problem of ground state preparation by mimicking the natural thermalization of a quantum system in contact with a large, low-temperature thermal environment. The environment can be efficiently simulated by a single ancilla qubit with a variable energy gap that is repeatedly coupled to the system qubits to generate a dissipative channel, and then reset after each interaction. Here we present an adaptive implementation of the dissipative algorithm, RL-Adapt, that uses single-shot reinforcement learning to optimize the selection of ancilla frequency and system-bath interaction operator to maximize energy dissipation without relying on a priori knowledge of the system spectrum. The adaptive implementation results in significantly reduced convergence times and can successfully find the ground state even for non-ideal operator pools that fail to converge using non-adaptive, uniform random operator and ancilla frequency selection.","published":"2026-09-25T15:07:42Z","updated":"2026-09-25T15:07:42Z","authors":["Nathan M. Myers","Chenxu Liu","Yulong Dong","Nicholas P. Bauman","Karol Kowalski"],"categories":["quant-ph"],"headline":"Using AI to Improve Quantum State Preparation Methods","summary":"Researchers explored a way to prepare quantum systems in their lowest energy state by mimicking natural thermal processes. They developed an adaptive method called RL-Adapt that uses reinforcement learning, a type of artificial intelligence, to select system settings more effectively without prior knowledge of the system details. Their results suggest this approach speeds up the process and can succeed where non-adaptive methods fail, though these findings are based on simulations and algorithms rather than experimental tests."},{"id":"2609.31369v1","title":"Convergence of a fully discrete approximation for the stochastic 2D Euler equations with Kraichnan transport noise","abstract":"We study a fully discrete approximation theory for the two-dimensional incompressible Euler equations on $\\mathbb{T}^{2}$, in vorticity form, driven by Kraichnan-type transport noise and with $\\mathbb{H}^{-1}(\\mathbb{T}^2)$-valued initial vorticity. The main new ingredient is a finite-dimensional coercivity estimate for the torus Kraichnan model, which captures at the discrete level the regularizing mechanism induced by the rough transport noise. Together with a new compactness strategy based on continuous equation-based interpolants and localized discrete estimates, this coercivity estimate gives tightness of the approximations and shows that subsequential limit is a weak martingale solution of the Euler equations.","published":"2026-09-25T15:04:53Z","updated":"2026-09-25T15:04:53Z","authors":["Abhishek Chaudhary","Ujjwal Koley","Andreas Prohl"],"categories":["math.PR","math.AP","math.NA"],"headline":"New method improves simulation of 2D fluid flow with random effects","summary":"The researchers studied a way to approximate solutions for two-dimensional fluid flow equations affected by random noise, called Kraichnan transport noise. They developed a new mathematical estimate and a strategy to handle these approximations more effectively. Their work shows that these approximations converge to a valid weak solution, but the results are theoretical and focus on the model's behavior rather than practical computations."},{"id":"2609.31368v1","title":"Equation discovery with Bayesian tree-adjoining grammars","abstract":"Tree-Adjoining Grammars (TAGs) have recently been introduced to Nonlinear System Identification (NLSI) as a means of encoding an entire model class as a finite set of grammatical rules, from which candidate models are assembled as trees. Existing TAG-based identifiers rely on evolutionary optimisation and return point estimates of the model structure. This paper instead proposes the TAG framework within a Bayesian setting. A generative prior is defined over tree structures and their parameters, and a Reversible-Jump MCMC sampler with structure-preserving tree moves is used to infer the joint posterior over model structure, parameters and predictions. Two training objectives are considered; that is, a one-step-ahead objective with conjugate parameter proposals, and a simulation-based objective handled by likelihood-free inference. The approach is validated on a simulated polynomial NARX system, the Silverbox benchmark, and wave-loading data from the Christchurch Bay Tower, where embedding Morison's equation as a fixed initial tree yields a grey-box model that outperforms the physics-driven baseline. The results demonstrate that Bayesian TAGs are well suited to quantifying uncertainty in equation discovery for dynamical systems and to fitting physics-informed models.","published":"2026-09-25T15:04:40Z","updated":"2026-09-25T15:04:40Z","authors":["Christopher A. Lindley","Nikolaos Dervilis","Keith Worden"],"categories":["stat.ML","cs.LG","eess.SY","stat.CO"],"headline":"Using Bayesian Grammars to Discover Mathematical Models","summary":"Researchers studied a method called Tree-Adjoining Grammars (TAGs) to find equations that describe complex systems. They combined TAGs with Bayesian statistics to better estimate both the model structure and parameters, allowing for measuring uncertainty. Tested on simulated and real data, including a physics-based example, their approach showed promise in improving model accuracy and uncertainty quantification, though results are based on specific cases and simulations."},{"id":"2609.31367v1","title":"A quantitative phase-field model for grain boundary trapping and diffusion of hydrogen","abstract":"Grain boundaries influence hydrogen transport in polycrystalline metals by acting as both trapping sites and interconnected diffusion pathways. Conventional interpretations of thermal desorption spectroscopy (TDS) and permeation experiments often treat traps as isolated defects, neglecting grain boundary connectivity and potentially misinterpreting experimental data. Here, we develop a quantitative phase-field model for hydrogen diffusion, grain boundary trapping, and grain boundary-assisted transport. The formulation preserves the physical grain boundary volume independently of the numerical interface thickness and accounts for anisotropic diffusion parallel and perpendicular to grain boundaries. Benchmark simulations verify quantitative behavior under interface upscaling. The model shows that grain boundary diffusion can significantly shift TDS peaks, making Kissinger-type analyses unreliable when grain boundary transport is significant. Effective diffusion coefficients are also shown to depend strongly on grain size, trapping free energy, temperature, and grain boundary mobility. These results highlight the coupled role of trapping thermodynamics, bulk diffusion, and grain-boundary transport.","published":"2026-09-25T15:03:24Z","updated":"2026-09-25T15:03:24Z","authors":["Vincent Feyen","Martin Diehl","Nele Moelans"],"categories":["cond-mat.mtrl-sci","cond-mat.mes-hall"],"headline":"New model shows how grain boundaries affect hydrogen movement in metals","summary":"The researchers developed a detailed computer model to study how hydrogen moves and gets trapped along grain boundaries in metals, which are the interfaces between crystals. Their model considers how these boundaries connect and allow hydrogen to move, improving on simpler approaches that treat traps as isolated spots. Simulations suggest that grain boundary transport can change experimental measurements widely, meaning common analysis methods might be unreliable when this effect matters. The study highlights the complex interaction between hydrogen trapping, diffusion, and boundary effects in metal materials."},{"id":"2609.31366v1","title":"Renormalized perturbation theory in an intense background electromagnetic field","abstract":"Quantum electrodynamics in strong background electromagnetic fields or strong-field QED (SFQED) has been investigated in great detail at the tree level. The study of SFQED at higher loops has not been carried out in a correspondingly systematic way, with the notable exception of strong background atomic fields. Here, we investigate the renormalization of SFQED by writing the standard unrenormalized SFQED Lagrangian density, which differs from the vacuum-QED Lagrangian density by the additional interaction term of the Dirac four-current density with the background four-vector potential, in terms of renormalized quantities and counterterms. Within this framework, we confirm the necessity of renormalizing the background field as an electric charge rather than as a photon field. We compare this approach with an alternative one which features a different renormalized Lagrangian density and we show that they are physically equivalent. As a byproduct, it will become clear that the renormalizability of SFQED directly derives from the renormalizability of vacuum QED. Also, we obtain that the renormalized SFQED Lagrangian density features a new counterterm as compared to the vacuum Lagrangian density. This counterterm is necessary to renormalize the electron self energy, which undergoes a new contribution as compared to vacuum QED, physically due to the electromagnetic field produced by the four-current density induced in the vacuum by the background electromagnetic field. The vacuum-induced electromagnetic field diverges and needs to be renormalized. We show how the renormalization of the vacuum-induced electromagnetic field is implemented explicitly at one loop and at two loops. The renormalization at all loops is shown to be more easily deduced at the level of the Lagrangian density. Finally, some subtleties concerning the renormalization procedure in the case of a free background field.","published":"2026-09-25T15:01:40Z","updated":"2026-09-25T15:01:40Z","authors":["Misha A. Lopez-Lopez","Giulio Audagnotto","Antonino Di Piazza"],"categories":["hep-ph"],"headline":"Researchers study advanced math of quantum fields in strong electromagnetic backgrounds","summary":"The researchers examined how to adjust (renormalize) the mathematical description of quantum electrodynamics (QED) when strong background electromagnetic fields are present, going beyond basic approximations. They showed that the usual methods of renormalization from standard QED still apply but require an additional correction to handle new infinite terms arising from vacuum effects induced by the background field. Their work clarifies how these complex adjustments work at multiple calculation levels, although it is theoretical and based on mathematical formulations rather than experiments."},{"id":"2609.31365v1","title":"Categorical spin-networks: state-sum invariants of 4-manifolds from higher-gauge theory","abstract":"In this work, we define $(3+1)$-dimensional topological quantum field theories (TQFTs) and their lattice realizations with input data given by a pivotal fusion 2-category; more precisely, our framework applies in wider generality to presemisimple locally fusion pivotal tensor $2$-categories, which are equipped with a so-called \\textit{shadow trace}. We introduce a decoration of triangulated manifolds by the input data, and develop a $20j$-symbol formalism for computing 4-simplex scattering amplitudes. We verify the topological invariance of 20$j$-symbol summations under the 4-dimensional Pachner moves. Furthermore, we also construct explicitly a 3-dimensional Levin-Wen-type lattice Hamiltonian model realizing the state-sum invariant. Our construction can be understood as a higher-dimensional analogue of the spin-networks construction for the Turaev-Viro invariant. Examples arising from $2$-groups are discussed from the perspective of higher gauge theory.","published":"2026-09-25T15:01:07Z","updated":"2026-09-25T15:01:07Z","authors":["Hank Chen","Zhian Jia","Ran Luo","Wei Cui"],"categories":["math-ph","cond-mat.str-el","hep-th","math.QA"],"headline":"New Method to Study 4D Shapes Using Advanced Mathematical Tools","summary":"The researchers developed a mathematical framework to study 4-dimensional spaces using complex algebraic structures called fusion 2-categories. They created a way to decorate shapes divided into simple parts and calculated special numbers to understand how these shapes behave, ensuring these calculations don't change under certain transformations. Their work includes building a related 3D model and explores examples connected to advanced gauge theories. The findings come from theoretical constructions and mathematical proofs rather than experiments."},{"id":"2609.31364v1","title":"OpenVAM: Open-World Visual Attention Modeling with VLMs","abstract":"Predicting human gaze is a core capability for applications ranging from web/UI design analysis to robotics and human-computer interaction. Yet, most visual attention modeling methods output only a dense saliency map, which is often insufficient for action: practitioners need to connect attention peaks to discrete elements in the scene (what) and understand the drivers of those peaks in context (why), while remaining robust to domain shift across natural images, commercial content, and UI/web layouts. We, therefore, introduce OpenVAM (Open-world Visual Attention Modeling with VLMs), a unified framework that jointly addresses universality and explainability across heterogeneous domains (natural scenes, commercial imagery, and UI/web layouts) and supervision modalities. OpenVAM adopts a decoupled-but-aligned design: a dedicated dense visual pathway provides stable, spatially precise localization, while an instruction-following vision--language semantic head generates grounded what/why explanations conditioned on the same image and data-type prompts. A three-stage training strategy preserves strong localization priors while progressively introducing language grounding and improving explanation alignment via parameter-efficient adaptation without perturbing the saliency branch. We further propose a scalable pipeline to generate multi-domain saliency-reason annotations for training and systematic evaluation. Experiments across diverse datasets show that OpenVAM improves robustness under domain shift while producing image-grounded explanations that make saliency predictions more interpretable.","published":"2026-09-25T14:59:14Z","updated":"2026-09-25T14:59:14Z","authors":["Kiana Hooshanfar","Amirhossein Kazerouni","Alireza Hosseini","Michael Brudno","Babak Taati"],"categories":["cs.CV"],"headline":"New Model Explains Where and Why People Look in Images","summary":"The researchers address the problem of predicting where people focus their gaze in images, which is useful for web design, robotics, and human-computer interaction. They developed OpenVAM, a system that not only predicts attention areas but also explains what objects attract attention and why, across different types of images. Their experiments show OpenVAM improves prediction stability when images change and provides explanations grounded in the image content, making the model's outputs easier to understand."},{"id":"2609.31363v1","title":"Brenier Meets Adversarial Training: Optimal Transport Geometry for Robust Learning","abstract":"Distributionally robust optimization (DRO) provides a principled framework for learning under distribution shift, but its practical use is hindered by the difficulty of evaluating worst-case risks for nonconvex loss functions. We study a penalized DRO formulation in which the adversary may choose any distribution but incurs a Wasserstein penalty for deviating from the empirical distribution. We show that the adversary's problem can be reformulated as an optimization problem over transport maps that push empirical samples to adversarial ones, and we prove that optimal maps are cyclically monotone. We also show that standard adversarial training---based on per-sample local optimization---violates cyclical monotonicity and wastes transport costs unless the adversary is severely restricted. We propose two remedies. First, we introduce multi-start particle ascent, which alternates parallel gradient ascent with reassignment to enforce cyclical monotonicity across samples. Second, we parameterize adversarial maps as gradients of input-convex neural networks, which guarantees cyclical monotonicity by construction. Experiments on robust regression, image classification, and robust control show that our methods consistently outperform standard adversarial training and state-of-the-art baselines, achieving improved robustness and better generalization under distribution shift.","published":"2026-09-25T14:57:04Z","updated":"2026-09-25T14:57:04Z","authors":["Alireza Abdollahpoorrostam","Ehsan Sharifian","Buse Şen","Marco Cuturi","Daniel Kuhn"],"categories":["cs.LG","math.OC","stat.ML"],"headline":"New Approach Improves Robust Machine Learning Against Data Changes","summary":"Researchers studied a method to make machine learning models more reliable when data changes unexpectedly. They focused on how to measure the worst-case impact of shifts in data and found that existing adversarial training methods may be inefficient. By reformulating the problem using ideas from geometry and optimizing with new techniques, they developed methods that better handle these challenges. Their tests on various tasks showed improved robustness and adaptation to data changes, though the work is based on theoretical and simulation results provided by the researchers."},{"id":"2609.31362v1","title":"The Molecular Circumstellar Environments of Red Supergiants in Galactic Open Clusters","abstract":"We aim to expand the sample of RSGs with direct measurements of circumstellar gas through millimeter observations of two of the largest coeval RSG populations in the Galaxy: the open clusters RSGC1 and RSGC2. We present an interferometric molecular line study of 28 RSGs in RSGC1 and RSGC2 with ALMA. The primary targeted emission is CO $J=2-1$ and continuum, with additional thermal and maser SiO lines observed for RSGC2 targets. We measure terminal expansion velocities, systemic velocities, emission-region sizes, and provide updated dust models and luminosities using the new long-wavelength constraints. Finally, we perform 1D radiative transfer modeling of CO rotational-line emission to estimate mass-loss rates. We detect $^{12}$CO toward six RSGs in RSGC2 and five in RSGC1. Except for DFK 52, the spatial distribution of all $^{12}$CO emission is relatively compact or unresolved. The obtained mass-loss rates range from $\\log(\\dot{M}/M_\\odot\\,\\mathrm{yr}^{-1})=-5.5$ to $-3$, but are highly uncertain due to a lack of stringent constraints on the envelope sizes and temperatures. The lower-limit $\\dot{M}$-values we obtain are high compared to standard empirical prescriptions and favor dust models assuming a radiatively driven wind. DFK 49 is a clear outlier: its revised luminosity is the lowest in both clusters, yet it harbors a strong wind with $\\dot{M}>10^{-5}\\,M_\\odot\\,\\mathrm{yr}^{-1}$. The CO emission regions are much smaller than expected from nearby RSGs and theoretical models, and suggest a unique phase structure for the CSM around RSGs in cluster environments. This work demonstrates the mutual utility of molecular line and dust observations for constraining RSG winds, while highlighting the need for future spatially resolved, multi-tracer observations of RSGC targets to establish the physical structure of their circumstellar environments and improve mass-loss determinations.","published":"2026-09-25T14:56:56Z","updated":"2026-09-25T14:56:56Z","authors":["Mark A. Siebert","Elvire De Beck","Guillermo Quintana-Lacaci","Theo Khouri","Maryam Saberi","Matthias Maercker","Wouter H. T. Vlemmings"],"categories":["astro-ph.SR","astro-ph.GA"],"headline":"Studying Gas Around Red Supergiant Stars in Star Clusters","summary":"Researchers used millimeter-wave observations to study gas around 28 red supergiant stars in two large star clusters. They measured gas speeds, sizes, and mass-loss rates by analyzing carbon monoxide and other molecules, finding that these stars lose mass at rates higher than typical estimates. Their findings suggest the gas regions are smaller and have unique structures compared to isolated stars. However, the mass-loss estimates are uncertain, and the study highlights the need for more detailed observations to better understand these star environments."},{"id":"2609.31361v1","title":"Resolving the Dual-Timescale Forecasting Dilemma in High-Frequency Cyber-Physical Telemetry via Physics-Grounded Dynamic Asymptotic Decomposition","abstract":"High-frequency industrial cyber-physical systems (CPS) stream continuous sensor telemetry at sub-10-second intervals. Multi-horizon forecasting over hours is required for predictive maintenance and operational control. However, deploying machine learning models over extended horizons (H = 2,160 steps at dt = 5 s) encounters the Dual-Timescale Forecasting Dilemma: the trade-off between short-horizon kinetic momentum and long-horizon diurnal thermodynamic equilibrium. Autoregressive rollouts suffer from compounding error drift, while direct multi-output projectors exhibit variance explosion and noise extrapolation. In this work, we prove that autoregressive mutual information decays exponentially toward zero over extended lead times, and that the optimal minimum-variance estimator converges asymptotically to the periodic diurnal baseline (Lemma 1 and Theorem 1). Guided by these bounds, we propose the Dynamic Asymptotic Decomposition (DAD) framework. The architecture couples a continuous sigmoidal authority transition schedule \\(α(h)\\) with monotonic L2 regularization scaling \\(λ(h)\\), transferring prediction authority from kinetic autoregression to diurnal equilibrium while enforcing physical non-negativity. Evaluated across a 30-day industrial telemetry stream (N = 518,400 steps) over 16 walk-forward test windows, DAD achieves an overall Mean Absolute Error (MAE) of 0.1764, representing a 78.3% error reduction over Deep LSTM (0.8143), 60.2% over PatchTST (0.4431), and 20.8% over DLinear (0.2227). Computational profiling demonstrates edge execution in 0.72 ms with 0.203 MFLOPs and 72k parameters on a single CPU core (over 3400x faster than Transformer baselines), validated by an automated dual-stage CI/CD quality assurance harness and physical fail-safe architecture.","published":"2026-09-25T14:56:52Z","updated":"2026-09-25T14:56:52Z","authors":["Avik Kumar Das"],"categories":["eess.SP"],"headline":"New Method Improves Long-Term Forecasting of Industrial Sensor Data","summary":"Researchers studied the challenge of predicting high-frequency sensor data over long periods in industrial systems, where short-term changes and daily patterns both matter. They identified limits in current machine learning techniques and developed a new approach called Dynamic Asymptotic Decomposition (DAD) that blends short-term and long-term information effectively. Tested on real industrial data, their method reduced prediction errors significantly compared to existing models, and it runs efficiently on simple computer hardware. Their findings are based on mathematical proofs and extensive testing on a month's worth of sensor streams."},{"id":"2609.31360v1","title":"Programs-of-Layers in LLMs through the Lens of Cortical Areas","abstract":"Inference in LLMs is conventionally a fixed-depth, fixed-order forward pass through every layer, regardless of how difficult the input is. The human brain does not work this way: using the thalamus as a central hub, it routes information flexibly to all regions of the cortex according to demand. Li et al. (2026) recently showed, with a system they call program-of-layers (PoLar), that transformers can be given an analogous flexibility if their layers are treated as a library of functions rather than a fixed sequence. Performance improves over the standard forward pass when each input is dynamically routed through an adaptive sequence of skipped or repeated contiguous layer blocks. We reconstructed PoLar's diagnostic MCTS in more detail than the original paper and applied it across 5 models. We reproduced several of PoLar's findings: skipping outperformed the standard pass, repeating outperformed skipping, and combining both outperformed either alone. Shorter programs sufficed for easier questions, while harder questions required more layer repeats. However, we failed to replicate the main claim regarding their learned router for single-shot inference: its top-ranked prediction consistently collapsed back to the standard pass, even though its top-k predicted programs, taken together, did show a real accuracy gain. Beyond reproduction, we find that a small number of generic programs are enough to solve most of the questions. We also provide a much deeper analysis of these programs' structure and robustness: for example, we found that programs that correct errors are highly brittle: undoing even a single edit inside a program typically breaks the correction. Connecting this to the brain's routing mechanisms, PoLar mirrors principles of thalamo-cortical coordination between cortical-area-like transformer layers. We publicly release the code at https://datexis.github.io/RE-PoLar/","published":"2026-09-25T14:55:50Z","updated":"2026-09-25T14:55:50Z","authors":["Justus Westerhoff","Stephan Olbrich","Hatem Oraby","Matthew Evan Larkum","Felix Alexander Gers"],"categories":["cs.AI"],"headline":"New method adapts AI language layers inspired by the brain","summary":"Researchers studied a technique called program-of-layers (PoLar), which lets AI language models flexibly skip or repeat processing steps based on input difficulty, mimicking how the brain routes information. They confirmed PoLar improves performance over standard fixed processes and found most questions can be solved with a few common programs. However, they could not replicate a key claim that a specialized router can reliably pick the best processing path in one try. The work offers detailed insights but remains primarily tested through simulations."},{"id":"2609.31359v1","title":"EEG-based Word Association Paradigm for Adult ADHD Screening: An Exploratory Pilot Study","abstract":"With the prevalence of Attention Deficit Hyperactivity Disorder (ADHD) over the past decades, healthcare systems across the globe face critical diagnostic challenges due to long diagnostic waiting times and a reliance on subjective behavioural assessments that cannot distinguish ADHD from comorbid psychiatric disorders, especially for adult patients. This exploratory study investigates whether EEG-based word association paradigms show promise as a complementary approach to screening for ADHD in adults. Using a mixed-method approach, the study examines neurological and cognitive differences between neurotypical individuals (NT), clinically diagnosed ADHD participants (ND), and self-reported ADHD cases awaiting formal diagnosis (SR) across three word association tasks. While established EEG biomarkers (ERP N400, Theta/Beta Ratio, and Alpha Suppression) show no significant group differences, semantic distance analysis reveals a statistically significant main effect ($p=0.003$), with the SR group showing the most divergent associations. These preliminary findings suggest that word association paradigms may capture cognitive differences not detected by standard EEG metrics and encourage further large-scale investigation as a potential complement to existing adult ADHD screening tools.","published":"2026-09-25T14:55:39Z","updated":"2026-09-25T14:55:39Z","authors":["Caroline Peng","Tony Russell-Rose"],"categories":["cs.HC"],"headline":"Exploring Brain Tests to Help Screen Adult ADHD","summary":"Researchers looked at using brain activity during word association tasks to screen adults for ADHD, a condition that is hard to diagnose with current methods. They compared brain signals and word connections among adults with diagnosed ADHD, those awaiting diagnosis, and neurotypical individuals. While common brain signal measures did not differ, the study found differences in how people associated words, especially in those awaiting diagnosis, suggesting this approach might reveal unique cognitive patterns. These are early results needing further study."},{"id":"2609.31358v1","title":"A Safety-Bounded SDC-to-MCP Gateway for Medical AI Agents","abstract":"The Model Context Protocol (MCP) provides a common interface through which AI applications discover and use external resources and tools. It allows language-model agents to ground their reasoning in current system state and interact with heterogeneous services. In medical environments, however, exposing device state and action affordances requires deterministic constraints on possible effects. We present an IEEE 11073 Service-Oriented Device Connectivity (SDC)-to-MCP gateway that exposes metrics, alarms, context references, and semantic metadata as read-only resources, while representing selected action affordances as policy-validated dry-run tools. The term safety-bounded denotes a narrow no-execution property: agent-facing requests dispatch no SDC device operation. A Python prototype supports simulated fault and lifecycle experiments, a software-reference protocol path spanning independent Java and Python implementations, deterministic baselines, representation ablations, and multi-model agent evaluation. The results show semantically explicit resource exposure, visible rejection of invalid or outdated state, and preservation of the no-execution boundary across resource, proposal, and authorization paths. Explicit semantic metadata improved conformity to required metric identifiers in structured alarm outputs relative to a generic representation, while retained structured-output failures reveal a distinction between plausible narrative answers and task-compliant machine-readable results.","published":"2026-09-25T14:54:21Z","updated":"2026-09-25T14:54:21Z","authors":["Bennet Gerlach","Stefan Fischer"],"categories":["cs.DC","cs.AI"],"headline":"New Gateway Enhances Safety for Medical AI Device Use","summary":"Researchers developed a gateway that connects medical device standards to AI systems, allowing AI agents to access device information without risking unintended actions. Their Python prototype tested this approach in simulations, ensuring that AI requests do not trigger device operations, thus maintaining safety. The study shows that clear data labeling helps AI agents interpret medical alarms better, but some output challenges remain between generating understandable text and meeting strict task requirements."},{"id":"2609.31357v1","title":"Transformer-based Monte Carlo Localization in Construction Meshes","abstract":"To be able to perform inspection or digitization tasks, mobile robots on construction sites must be able to localize themselves reliably with respect to a global reference frame that is shared with a building map. Similar room layouts and low-texture surfaces pose a challenge for existing LiDAR- and vision-based localization methods. We approach this problem with a LiDAR-based global relocalization system that estimates the robot's pose relative to a building mesh and combines a PointNet++ encoder with a place recognition decoder, whose outputs serve as a learned observation model within a Monte Carlo Localization (MCL) framework. The pipeline is trained exclusively on synthetic LiDAR scans obtained by simulating the robot's sensors inside the building mesh. Our approach is robust in ambiguous environments due to an uncertainty-aware decoder that scales positional likelihoods and a resampling strategy that injects model hypotheses into the particle set, enabling recovery from potential particle depletion. Evaluations on real-world datasets show that our method outperforms both diffusion-based and ScanContext++ baselines while maintaining fast inference (18 ms per call), demonstrating the practicality of synthetic-data training for mesh-referenced global localization in construction robotics.","published":"2026-09-25T14:54:15Z","updated":"2026-09-25T14:54:15Z","authors":["Linus Kramer","William Talbot","Olga Vysotska","Marco Hutter"],"categories":["cs.RO"],"headline":"New Robot Localization Method Helps in Complex Construction Sites","summary":"Researchers developed a method to help robots find their exact location on construction sites, where similar rooms and plain surfaces make it hard to use traditional sensors. They combined a special sensor data processor with a recognition system inside a particle-filtering framework to estimate the robot's position relative to a building model. Trained only on simulated data, their approach showed better accuracy and speed than previous methods when tested on real data, though the study focuses on this specific simulated training and testing setup."},{"id":"2609.31356v1","title":"Open Vocabulary Domain Unlearning","abstract":"Vision-Language Models (VLMs) exhibit remarkable zero-shot generalization, yet they often encode unwanted or hazardous stylistic domains such as idealized textbook diagrams in medical AI or cartoon vehicles in autonomous driving. Approximate Domain Unlearning (ADU) aims to selectively erase a model's recognition of a target visual domain while preserving accuracy on the remaining domains. However, existing ADU methods operate under a flawed closed-vocabulary assumption: they evaluate unlearning solely on the specific object classes seen during the unlearning fine-tuning phase. Consequently, these methods do not unlearn the domain itself; they merely overfit to seen class-domain pairs, leaving the domain easily recognizable for unseen classes and providing a false sense of removal. We argue that true domain erasure must be class-agnostic. To address this, we formalize Open-Vocabulary Domain Unlearning (OVDU), a rigorous protocol that mandates domain forgetting must transfer to held-out classes. To solve the OVDU challenge, we propose a surgical parameter-editing framework. First, a Fisher Information mask isolates domain-sensitive weights, mathematically protecting foundational zero-shot generalization. Second, our Targeted Manifold Scattering (TMS) objective uses preference-based mining to locally scatter the forget domain's stylistic geometry. Evaluated across PACS, OfficeHome, and DomainNet, our method vastly improves open-vocabulary generalization over existing baselines. Crucially, it delivers exceptional sample efficiency, outperforming peak 8-shot baseline results with only 4 shots.","published":"2026-09-25T14:52:36Z","updated":"2026-09-25T14:52:36Z","authors":["Sumanth Udupa","Mehrtash Harandi","Yadan Luo","Mahsa Baktashmotlagh"],"categories":["cs.CV","cs.LG"],"headline":"New Method Helps AI Forget Specific Visual Styles Broadly","summary":"Researchers studied how vision-language AI models sometimes retain unwanted visual styles, like cartoonish images, which can be problematic. They identified a limitation in current methods that try to erase such styles but only work for specific known objects, not new ones. To fix this, the researchers proposed a new approach called Open-Vocabulary Domain Unlearning that aims to remove these styles regardless of object type. Their technique uses mathematical tools to selectively alter the model, showing better performance and efficiency in tests on multiple datasets. These results are based on evaluations with held-out classes to ensure true domain forgetting."},{"id":"2609.31355v1","title":"On the Brouwer-type Conjecture for Signless Laplacian Eigenvalues of Graphs","abstract":"Motivated by Brouwer's conjecture, Ashraf, Omidi and Tayfeh-Rezaie proposed the following Brouwer-type conjecture that for every graph $G$ on $n$ vertices with $m$ edges, the sum $S_k^+(G)$ of its $k$ largest signless Laplacian eigenvalues satisfies $S_k^+(G)\\le m+\\binom{k+1}{2}$ for $k=1, \\ldots, n$. In this paper, we prove that the above conjecture holds. Moreover, the equality holds if and only if $k=1$ and $G$ is either star $K_{1,a}$ or triangle $K_3$ with adding some isolated vertices. For split graphs, properties of block signless Laplacian matrices based on clique and independent set are adapted. While for non-split graphs, some spectral graph substructure are used to control the sum of signless Laplacian eigenvalues.","published":"2026-09-25T14:52:17Z","updated":"2026-09-25T14:52:17Z","authors":["Ruisong Yuan","Xiao-Dong Zhang"],"categories":["math.CO"],"headline":"Researchers Prove a Mathematical Conjecture About Graph Properties","summary":"The researchers studied a conjecture related to graphs, which are mathematical structures made of points and lines. The conjecture concerns a special measure called signless Laplacian eigenvalues and their sums. They proved that for any graph, this sum is limited by a specific formula involving the number of lines and a combination of numbers. They also identified the exact cases when this limit is reached. Their results are based on theoretical mathematical proofs dealing with different types of graphs."},{"id":"2609.31354v1","title":"Mutable Transcripts: Mitigating Context Pollution through Editable Conversation State","abstract":"Contemporary large language model (LLM) chat systems treat conversation history as an immutable sequence of turns that defines the model's working context. However, user intent in real interactions is not static: it evolves through correction, refinement, and shifting constraints. This mismatch between dynamic intent and static transcripts can result in context pollution, where outdated or irrelevant information persists and continues to influence subsequent responses. We introduce mutable transcripts, a new interaction paradigm that enables users to revise prior turns through natural language edit requests, allowing the conversation history itself to be updated rather than appended. This reframes the transcript from a passive record into an editable representation of conversational state. We present a working prototype that integrates transcript-level revision into a standard chat interface and evaluate its feasibility through a controlled user study (n=17) and an illustrative transcript analysis of representative interaction scenarios. Participants significantly preferred mutable transcripts over standard chat across measures of clarity, confidence, and ease of use, with reduced intent to restart conversations. Transcript analysis of representative user study conversations shows that mutable transcripts can reduce conversation length and eliminate obsolete retained context. These findings provide initial evidence that user-driven revision of conversational history can improve interaction quality and help maintain a more current representation of user intent. The source code and prototype can be accessed at https://github.com/QxLabIreland/ReChat","published":"2026-09-25T14:51:49Z","updated":"2026-09-25T14:51:49Z","authors":["Dan Barry","Andrew Hines"],"categories":["cs.AI"],"headline":"Editable Chat History Reduces Confusion in AI Conversations","summary":"Researchers studied a new system where users can edit previous parts of a chat with AI language models instead of only adding new messages. This approach aims to fix a problem called 'context pollution,' where outdated information in a chat misguides AI responses. Through a prototype and user study, participants found the editable chat history clearer and easier to use, helping keep the conversation more accurate and focused. The study offers preliminary evidence that allowing users to revise chat history can improve AI interactions, though results are from a small controlled study."},{"id":"2609.31353v1","title":"Semilinear Neumann boundary value problems for measure-valued maps","abstract":"We study the Wasserstein lift of nonlinear Poisson systems with inhomogeneous Neumann boundary conditions. We establish the existence of minimizers using new estimates involving the $p$-moments of measure-valued traces. We also derive the optimality conditions, which turn out to be matrix-valued generalizations of pressureless Euler equations. While deterministic minimizers exist under suitable convexity assumptions and for one-dimensional source domains, we show that genuinely measure-valued minimizers arise in higher dimensions. In particular, we construct examples in which the lifted and classical variational problems have different minimum values, thereby precluding the solutions to the lifted problem from being deterministic, despite the absence of any a priori mechanism enforcing measure-valued behavior. To the best of our knowledge, this phenomenon is new in the nonlinear elliptic theory.","published":"2026-09-25T14:51:22Z","updated":"2026-09-25T14:51:22Z","authors":["Aleksei Kroshnin","Hugo Lavenant","Dmitry Vorotnikov"],"categories":["math.AP","math.DG","math.FA","math.MG","math.PR"],"headline":"New Insights into Complex Boundary Problems in Mathematics","summary":"Researchers studied a type of mathematical problem involving nonlinear Poisson systems with specific edge conditions called Neumann boundaries. They proved that solutions minimizing a related quantity exist and developed conditions these solutions must meet, linking them to advanced fluid dynamics equations. Importantly, in higher-dimensional cases, they found solutions naturally involve measures rather than fixed values, revealing new behaviors previously unseen in this field. Their results rely on theoretical analysis and examples, without experimental validation."},{"id":"2609.31352v1","title":"Bellman's Forest Problem and Computability","abstract":"The goal of this paper is to refine methods in computable analysis and to employ them in the study of solutions of an optimization problem posed by Bellman. This problem asks how to find optimal (shortest) paths which do not fit into given plane figures. We show that each instance of Bellman's problem has an arbitrarily small uniformly computable perturbation for which the minimal length of path is computable, and thus the class of optimal paths is a $Π_1^0$ class, and there are uniformly computable paths which escape with arbitrarily small length greater than the minimum possible.","published":"2026-09-25T14:51:09Z","updated":"2026-09-25T14:51:09Z","authors":["Jacob Canel"],"categories":["math.LO","math.MG"],"headline":"Understanding Bellman's Problem with Computable Analysis","summary":"Researchers studied a problem posed by Bellman about finding the shortest paths that avoid certain shapes on a plane. They used advanced mathematical methods to show that small, computable changes to the problem make it possible to calculate the shortest path length. Their findings indicate that these optimal paths belong to a special class that can be described in a computable way, and there are paths that escape these shapes with lengths just slightly longer than the shortest possible."},{"id":"2609.31351v1","title":"Progressive Memory Transformer: Memory-Aware Attention for Time-Series","abstract":"Time-series carry structure simultaneously at multiple scales (fine-grained variation, mid-range motifs, and global properties) and downstream tasks operate at correspondingly different scales. Most existing self-supervised learning approaches supervise representations globally via instance-level contrastive losses and limited temporal neighborhood supervision, but do not explicitly exploit the structural hierarchy. We propose a learning framework that explicitly enforces a structural hierarchy across three scales independently: a local objective for token continuity, a mid-range objective for window-level motifs, and a global objective for sequence-level agreement. Realizing this framework requires the backbone to expose a representation at each scale; we introduce \\textbf{Progressive Memory Transformer} (PMT), which augments a transformer with writable, window-aligned memory that exposes the mid-range scale alongside the token and sequence-level representations conventional transformers already provide. Across seven UCR/UEA/UCI classification benchmarks, a cue-retention probe, and forecasting benchmarks, PMT learns representations that probe well at the global, mid-range, and local scales---strong low-label classification (1--5\\% labels), competitive forecasting performance across multiple horizons, and quantitative and qualitative evidence that memory states capture mid-range motifs.","published":"2026-09-25T14:49:42Z","updated":"2026-09-25T14:49:42Z","authors":["Tord Sture Stangeland","Andreas Köhler","Steffen Mæland","Adín Ramíres Rivera"],"categories":["cs.LG"],"headline":"New AI method improves time-series analysis at multiple scales","summary":"Researchers developed a method to better analyze time-based data by capturing patterns at small, medium, and large scales. They created a new type of transformer model that includes memory aligned to windows of data, allowing it to represent mid-range patterns in addition to short and long ones. Tests show this approach helps in tasks like classification with few labels and forecasting, with evidence that the model effectively learns important medium-scale structures. Results are based on benchmarks and probes described in the study."},{"id":"2609.31350v1","title":"Sharp Data-Processing Region for Quantum Conditional Rényi Entropies","abstract":"We study the three-parameter family of quantum conditional Rényi entropies $H_{α,z}^λ$ introduced in \\textit{Commun. Math. Phys.} 407, 180 (2026). We prove that the previously established data-processing region is sharp. We then characterize the club-sandwiched family through a classical coarse-graining principle: within the full data-processing region, universal monotonicity under deterministic coarse-graining of the classical conditioned register holds if and only if $z = α$. Finally, we establish rigidity of entropic duality by showing that any universal duality relation for finite-dimensional tripartite pure states must satisfy the known parameter relations. Thus, the data-processing region, the club-sandwiched subfamily, and the duality transformation are each uniquely characterized by their respective universal properties.","published":"2026-09-25T14:49:18Z","updated":"2026-09-25T14:49:18Z","authors":["Milad M. Goodarzi","Yucheng Bao","Marco Tomamichel"],"categories":["quant-ph"],"headline":"Researchers Define Limits of Quantum Entropy Measures","summary":"The researchers studied a complex family of quantum entropy measures that help understand information processing in quantum systems. They confirmed precise limits on how these measures behave when quantum information is processed and showed that certain special cases follow unique patterns related to simplifying the system. Their results rely on mathematical proofs and characterize key properties of these entropy measures, but do not explore practical applications or experimental tests."},{"id":"2609.31349v1","title":"DyMD: Preserving Interaction Dynamics through Distribution Matching Distillation in Few-Step Video World Models","abstract":"Large video diffusion models offer expressive priors for embodied prediction and learning, yet their many-step sampling remains costly for interactive downstream use. Distribution Matching Distillation (DMD) enables few-step video generation, but can suppress robot--object motion while preserving visual quality. Examining DMD's teacher and fake-score signals, we find that weak re-noising keeps the teacher posterior concentrated near motion-deficient rollouts, limiting motion-restoring guidance. Meanwhile, stronger-motion rollouts tend to incur larger fake-score fitting errors, which can hinder the generator's learning of interaction dynamics. We propose DyMD, a DMD framework that adapts both teacher supervision and critic fitting to the evolving student. Temporal affinity--conditioned re-noise sampling adapts the timestep distribution to each rollout's current interaction fidelity by mixing the base schedule with a teacher prior motivated by local posterior variation, thereby balancing motion recovery and appearance refinement. To better track stronger-motion rollouts, dynamics-guided fake-score tracking uses a noise-conditioned predictor to estimate noise-relative fitting difficulty from latent temporal dynamics, then upweights predicted-hard rollouts in the critic loss. Using DyMD, we distill a 14B teacher into a four-step 1.3B student with no auxiliary modules at inference. On embodied-video benchmarks, the student improves R-Bench task adherence by $9.6$ percentage points and PAI-Bench-G Domain score by $5.1$ points over Base DMD while maintaining comparable visual quality. As a backbone for downstream action planning, our student achieves 34% mean success across two WorldArena tasks, compared with 16% for Base DMD.","published":"2026-09-25T14:47:17Z","updated":"2026-09-25T14:47:17Z","authors":["Haojun Xu","Jie Huang","Xin Lu","Mingchen Zhong","Zihao Fan","Linjiang Huang","Si Liu"],"categories":["cs.CV","cs.AI"],"headline":"New Method Improves Video Models for Robot Interaction Prediction","summary":"Researchers studied how to make video models faster and better at predicting robot-object interactions. They developed DyMD, which adjusts how the model learns from examples to balance motion accuracy and visual quality. Their approach improved task performance and interaction understanding in simulations, but the results come from computer experiments rather than real-world testing."},{"id":"2609.31348v1","title":"Circular orbits and particle collisions in the space-time of boson stars and their frozen states","abstract":"In this paper, we discuss the dynamics of test particles in the space-time of spherically symmetric, charged boson stars and their frozen states. These latter objects consist of a de Sitter core and a black hole exterior with a shell of finite thickness interpolating between the two. We discuss charged and uncharged massive particles, find the ISCOs and determine whether collisions of massive particles close to the core of (frozen) boson stars can lead to large center of mass energies. We also discuss massless test particles and find that photons rings as well as light points are possible in the space-time of frozen boson stars. We compare our results with those in the corresponding Reissner-Nordstroem space-time and point out differences.","published":"2026-09-25T14:45:57Z","updated":"2026-09-25T14:45:57Z","authors":["Maria Chivers","Betti Hartmann","Katherine Horton","Yves Brihaye"],"categories":["gr-qc","hep-th"],"headline":"Particle Movements and Collisions Near Boson Stars Explored","summary":"Researchers studied how particles move and collide near special theoretical objects called boson stars and their 'frozen states,' which have a dense core and a black hole outer layer. They looked at both charged and uncharged particles, identifying stable orbits and whether collisions could produce high energies near the core. The study also examined light behavior near these stars, comparing findings to known black hole models. These results are based on theoretical analysis of particle dynamics in these unusual space-time environments."},{"id":"2609.31347v1","title":"Benchmark cases for gradient-based optimization in linear elastic solid mechanics using lattice Boltzmann methods","abstract":"Gradient-based optimization (GBO) of problems constrained by partial differential equations is central to parameter identification and structural design in solid mechanics. Lattice Boltzmann method (LBM) schemes for linear elastic solids have only recently been derived. Combined with automatically generated discrete adjoint collision kernels, they make GBO possible with a purely LBM-based primal and adjoint solver. We propose, to our knowledge, the first optimization benchmarks for solid LBM: three inverse problems, each with a manufactured solution of the forward problem and a known reference control. These are the identification of two scalar amplitudes of a body force field on a periodic domain, the distributed control of the same field with up to $12800$ degrees of freedom, and the identification of Young's modulus on an elliptic plate with Dirichlet and Neumann boundaries. The primal scheme converges with second order on the periodic domain and with first order on the elliptic plate. The recovered controls follow these rates, with experimental orders of convergence of $1.92$ and $1.06$, respectively. Gradients obtained via forward-mode automatic differentiation (f-AD) and the discrete adjoint LBM (ALBM) require the same number of optimization steps and yield control errors agreeing to at least eleven significant digits. The ALBM gradient is further validated node-wise against a finite difference quotient, with deviations decreasing under mesh refinement. The benchmarks are implemented in the open-source LBM framework OpenLB and provide a verification basis for optimization in LBM-based solid mechanics and its extension to fluid-structure interaction.","published":"2026-09-25T14:45:35Z","updated":"2026-09-25T14:45:35Z","authors":["Johannes L. Grafen","Florian Kaiser","Stephan Simonis","Shota Ito","Mathias J. Krause"],"categories":["math.OC"],"headline":"New benchmarks improve solid mechanics optimization with lattice Boltzmann method","summary":"Researchers created benchmark tests to help improve gradient-based optimization techniques used in solid mechanics, focusing on a method called the lattice Boltzmann method (LBM). They developed three example problems with known solutions to verify how well LBM can identify forces and material properties in elastic solids. Their results show good agreement between different gradient calculation methods and provide a reference for future work, though findings are currently based on simulations and method validations."},{"id":"2609.31346v1","title":"Uniform convergence of diffusion synthetic acceleration for heterogeneous slab transport","abstract":"The diffusion synthetic accelerated (DSA) source iteration is a standard solver for the radiative transfer equation. Using Fourier analysis, a convergence rate \\(ρ_\\infty(c)\\le0.2247\\,c\\) with \\(c\\) the maximum ratio of scattering to total cross section has been established for an infinite homogeneous medium. For slab geometry with inflow boundary conditions and arbitrary bounded cross sections we prove that the spectral radius of the DSA iteration is at most \\(ρ_\\infty(c)\\). We show that this convergence rate carries over to a variational discretization of the DSA iteration on every conforming tensor-product Galerkin space whose angular factor contains the constants. The analysis rests on an exact min--max characterization of the spectral radius. From it we derive a checkable sufficient condition for a given rate. We verify this condition using a suitable energy-stable projection, a weighted angular average whose weight is chosen so that the condition holds with the rate \\(ρ_\\infty(c)\\). Since the projection maps discrete spaces into discrete spaces, the argument applies verbatim to the discrete iteration. As a consequence, the condition number of the preconditioned system is uniformly bounded by \\(1+0.29\\,c\\), which implies rapid convergence of the preconditioned conjugate gradients method.","published":"2026-09-25T14:45:26Z","updated":"2026-09-25T14:45:26Z","authors":["Matthias Schlottbom"],"categories":["math.NA"],"headline":"Improving speed of calculations in radiation transport problems","summary":"Researchers studied a mathematical method called Diffusion Synthetic Acceleration (DSA) used to solve equations about how radiation moves through materials. They showed that, even in complex situations where material properties change, this method converges quickly at a rate previously known only for simpler cases. Their work uses theoretical analysis to establish conditions ensuring efficient computation, which also applies to practical numerical implementations. These findings suggest the method remains fast and reliable in more realistic scenarios, based on their mathematical proofs and simulations."},{"id":"2609.31345v1","title":"On the asymptotic shape of quantile surfaces","abstract":"This article is concerned with the asymptotic shape of quantile surfaces, defined as the set of quantiles at a given level $α$ generated by a controlled one-dimensional distribution. Specifically, when the distribution arises as a linear combination of log-normal random variables and the control is a vector of positive coefficients, we prove that quantile surfaces are globally concave in the left tail ($α\\to0$) and globally convex in the right tail ($α\\to1$). Moreover, these surfaces exhibit asymptotic separation of scale and shape.","published":"2026-09-25T14:44:45Z","updated":"2026-09-25T14:44:45Z","authors":["Florian Gach","Simon Hochgerner"],"categories":["math.ST","math.PR","q-fin.MF"],"headline":"Researchers study the shape of quantile surfaces in special distributions","summary":"The researchers examined the shape of quantile surfaces, which are sets of values representing specific statistical cutoffs in data, for certain complex distributions made from sums of log-normal variables. They showed that as these cutoff levels approach the extreme low end, the surfaces curve inward (concave), and as they approach the extreme high end, they curve outward (convex). Their findings come from mathematical analysis of these distributions and describe how the surfaces change scale and shape near extremes."},{"id":"2609.31344v1","title":"The memory equation of a characteristic function in surface diffusion: a sum rule","abstract":"The decay of a density wave in an interacting adlayer obeys an exact memory equation of Mori's form. Being a Fourier component of a conserved density, its correlation function is a characteristic function, and the kernel is built from a second one, that of a single jump. The local rate obeys a sum rule whose momentum dependence is one minus the characteristic function of a single jump, divided by the static structure factor. Detailed balance makes the decay a superposition of relaxation modes of non-negative weight whose moments are static equilibrium averages, so that the truncations of the associated continued fraction form a convergent hierarchy of closed forms, each of them a bound on the decay itself. The single rate a fit returns is shown to be the mean of the instantaneous rate over a time window, hence never above the local rate. At long wavelength, the two coincide, the relaxation spectrum separating there, one collective mode taking almost the whole weight of the density wave and the single exponential usually fitted to spin-echo data being the decay itself. What an experiment reaches at every wavevector is also the area under the decay, a correlation time. Together with the static structure factor and the single-jump geometry, it forms one combination which the theory identifies with the effective hop rate divided by the product of the local rate and the correlation time. That product never falls below unity and tends to unity at long wavelength whenever the rate at which an adparticle leaves a site is independent of the site it leaves for. In the hydrodynamic limit, the sum rule becomes a relation of Darken type in which the rate multiplied by the thermodynamic factor is the effective hop rate and not the mobility of a labelled adparticle.","published":"2026-09-25T14:44:42Z","updated":"2026-09-25T14:44:42Z","authors":["S. Miret-Artés"],"categories":["cond-mat.stat-mech"],"headline":"Understanding how particles move on surfaces using math rules","summary":"The researchers studied how waves in particle density on a surface fade away, using a precise mathematical equation called a memory equation. They showed that this fading relates closely to the patterns of individual particle jumps and certain equilibrium properties of the system. Their theory connects measurable quantities from experiments to effective rates at which particles hop on the surface but is based on mathematical analysis rather than direct experimental results."},{"id":"2609.31343v1","title":"The mean curvature blows up at nondegenerate neck pinch singularities of Lagrangian mean curvature flow","abstract":"We prove that at any nondegenerate neck pinch singularity of Lagrangian mean curvature flow of surfaces the mean curvature becomes unbounded, with control on the blow-up rate.","published":"2026-09-25T14:44:31Z","updated":"2026-09-25T14:44:31Z","authors":["Jason D. Lotay","Goncalo Oliveira"],"categories":["math.DG"],"headline":"Mean Curvature Becomes Infinite at Certain Surface Singularities","summary":"Researchers studied how surfaces evolve under a process called Lagrangian mean curvature flow. They proved that at specific points known as nondegenerate neck pinch singularities, the average curvature of the surface grows without bound, becoming infinite. Their work also provides details about how quickly this blow-up happens. The findings are theoretical and focus on understanding the behavior of these singularities in surface evolution."},{"id":"2609.31342v1","title":"Stale-Document Poisoning: When Outdated Retrieval Overrides Correct Model Answers","abstract":"Retrieval-augmented generation (RAG) is often used to address outdated knowledge by providing external evidence. But retrieval helps only when that evidence is still valid. We identify a temporal alignment failure, stale-document poisoning, in which outdated evidence makes a model wrong despite answering correctly without retrieval. We construct a benchmark of 317 verified knowledge reversals across medicine, law, software, and platform policy, grounded in dated official sources. Across 12 models, recent medical reversals are harder than long-established ones. More importantly, outdated retrieval flips 30% of Llama and 37% of Qwen answers even without instructions to trust the document; explicit follow instructions raise these rates to 66% and 75%. Across four open models and four domains, poisoning ranges from 17-91%, while matched up-to-date evidence is followed in 97-100% of trials. To isolate temporal applicability, we keep the historical evidence unchanged across 50 reversals and vary only the evaluation date. A clear pattern emerges: dates alone produce only modest adaptation, but when models are explicitly told when the old evidence stops applying, the larger models switch to the appropriate answer almost perfectly. Causal interventions confirm that this validity information directly shapes the final decision. The same internal components also support broader comparison tasks, suggesting that temporal applicability can recruit a general reasoning mechanism used for other comparisons. Finally, a fixed recency-aware hybrid re-ranker reduces poisoning by 4.6-10.0 points when dates are accurate, with gains that depend on reliable temporal metadata. Reliable RAG therefore requires selective trust: models must determine not only what retrieved evidence says, but whether it still applies.","published":"2026-09-25T14:43:43Z","updated":"2026-09-25T14:43:43Z","authors":["Md Shamim Ahmed","Lukas Galke Poech","Richard Röttger"],"categories":["cs.CL"],"headline":"Outdated Information Can Mislead AI Models Despite Correct Answers","summary":"Researchers studied how AI models using external documents for answers can be misled by outdated information, a problem they call 'stale-document poisoning.' They tested 12 models across topics like medicine and law, finding that outdated evidence often causes models to change correct answers incorrectly. The researchers found that giving models information about when evidence expires helps them choose the right answers. Their work shows that AI needs to judge not just the evidence itself but also whether it is still valid."},{"id":"2609.31341v1","title":"The Right Information Extraction Pipeline Depends on the Document: Accuracy-Energy Trade-offs for Small, Local Models","abstract":"Whether an information extraction pipeline should process page images or parsed text depends on the document, and the answer flips across the layout spectrum. We study this trade-off under a constraint that rules out (closed) cloud services: privacy-sensitive documents processed on-premise by small ($\\le 8\\mathrm{B}$ parameter) text-only and vision--language models, evaluated on both accuracy and energy over a design space spanning input representation, model family, and inference configuration. Benchmarking on the near-plain-text Kleister-NDA contracts and the layout-rich VRDU forms, we find that batching is the dominant energy lever, cutting energy per page by 38-85% at no cost in accuracy, while FP8 quantization saves 27-32% when requests are served one at a time but less than 1mWh per page (9-19%) once batching is applied. Preprocessing dominates what remains: neural OCR costs $17\\times$ more energy per page than classical OCR and never reaches the Pareto frontier. Which representation wins flips with the type of document: vision--language models on layout-rich documents and small text-only models with a cheap parser on near-plain text, where they are both more accurate and cheaper than any vision--language configuration. Our work yields concrete guidelines for energy-efficient, privacy-compliant local information extraction.","published":"2026-09-25T14:43:18Z","updated":"2026-09-25T14:43:18Z","authors":["Christoph Walser","Mauricio Fadel Argerich","Jonathan Fürst"],"categories":["cs.AI","cs.CL"],"headline":"Choosing the Best Document Processing Depends on Document Type","summary":"Researchers studied how to extract information from documents using small on-site models to protect privacy, focusing on accuracy and energy use. They found that grouping tasks together saves the most energy without losing accuracy, and simpler document types work better with text-only models, while complex layouts benefit from models that analyze both text and images. Their findings highlight trade-offs in energy use and model choice depending on the document's layout, offering guidance for efficient, privacy-conscious processing."},{"id":"2609.31340v1","title":"Natural Saturation Of The Sterile Neutrino Dark Matter Resonant Production By a High Lepton Flavor Asymmetry In Primordial Plasma","abstract":"Sterile neutrinos remain a well-motivated dark-matter candidate whose cosmological abundance can be enhanced by resonant active--sterile conversion in the presence of a lepton asymmetry in primordial plasma. In the standard picture, a larger initial asymmetry increases the matter potential and can therefore promote resonant production. However, the same increase also shifts the resonance to later stages of the cosmological evolution when the plasma temperature is lower. We show that this delayed production epoch can overlap with the onset of active neutrino oscillations, which redistribute the flavor asymmetries and may substantially reduce, or in some cases nearly erase, the lepton asymmetry needed for a successful resonant sterile-neutrino production. This interplay provides a natural saturation mechanism for the final sterile-neutrino abundance: beyond a certain range of initial asymmetries, increasing the primordial asymmetry no longer leads to a proportional increase in the produced dark-matter density. We identify this effect as an additional constraint on resonant sterile-neutrino production at large lepton asymmetry and discuss its dependence on the flavor structure of the initial asymmetry and on the background cosmological evolution. Our numerical results reveal two order of magnitude range in sterile-active neutrino mixing parameter $\\sin^2(2θ)$ presently consistent with the resonant mechanism of the dark matter sterile neutrino production. It must be investigated by the next generation X-ray telescopes to fully explore this mechanism and corresponding minimal models suggesting sterile neutrinos as viable dark matter. We also consider pion condensation at large flavor lepton asymmetries, which may lead to a first-order QCD phase transition and associated gravitational-wave production in the early Universe.","published":"2026-09-25T14:43:14Z","updated":"2026-09-25T14:43:14Z","authors":["Dmitry Gorbunov","Dmitry Kalashnikov"],"categories":["hep-ph","astro-ph.CO"],"headline":"How early universe conditions affect sterile neutrino dark matter","summary":"Researchers studied how variations in lepton asymmetry—a difference in types of neutrinos—in the early universe influence the production of sterile neutrinos as dark matter candidates. They found that while higher asymmetry initially helps produce more sterile neutrinos, interactions and oscillations among neutrinos later reduce this effect, preventing unlimited increase in production. Their simulations suggest limits on the conditions that support this mechanism and highlight areas for future observational tests. The findings are based on theoretical models and numerical analysis, without experimental confirmation yet."},{"id":"2609.31339v1","title":"ChronoFuseGS: Multi-Temporal Gaussian Fusion with Per-Splat Persistence and Change Visualization","abstract":"Reconstructing environments where parts of the scene change between captured image sets poses a challenge for 3D scene reconstruction. We present ChronoFuseGS, a multi-temporal Gaussian Splatting approach that addresses this issue by taking multiple separately trained Gaussian Splatting models, each representing a distinct timestep and partially overlapping in geographic coverage, and merging them into a single combined model. By allowing Gaussians from one timestep to contribute to the reconstruction at other timesteps, our approach leverages data across all captured timesteps to refine persistent parts of the scene. The model supports incremental extension, allowing new timesteps to be added while preserving the existing merged reconstruction. It encodes, for each Gaussian primitive, at which timesteps it contributes to the reconstruction. To support visual exploration of the reconstructed scene, we present a change-aware visualization approach that highlights the parts of the scene that have changed across a user-defined time selection, while preserving the color of persistent parts. Since the persistence encoding operates at the Gaussian primitive level, changes are visualized at sub-object granularity rather than being limited to object-level changes. We evaluate our approach on a real-world outdoor dataset of a flood management area, captured over 7 months across eight recording days and covering seasonal vegetation changes, snow cover, and flooding events, which we make publicly available. Our results demonstrate that the combined model consistently outperforms individually trained single-timestep models in novel-view synthesis quality, recovers structural details absent in the individual reconstructions, and reliably highlights changes in fine details and sub-parts of objects and natural structures.","published":"2026-09-25T14:43:04Z","updated":"2026-09-25T14:43:04Z","authors":["Tobias Batik","Diana Marin","Peter Kán","Hannes Kaufmann"],"categories":["cs.GR","cs.CV"],"headline":"New Method Fuses Multiple 3D Scene Snapshots Over Time","summary":"Researchers developed ChronoFuseGS, a way to combine several 3D models of a scene taken at different times into one unified model. This method helps capture changes and persistent parts of environments, like seasonal or flood-related changes in a natural area. Their approach visualizes detailed changes over time and improves the quality of the 3D reconstruction compared to single-time models. These findings are based on testing with outdoor data collected over seven months."},{"id":"2609.31338v1","title":"Modeling dependent degradation data considering inherent causal relationships for reliability analysis and remaining useful life prediction","abstract":"Accurate modeling of dependent degradation processes is essential for credible reliability assessment and remaining useful life (RUL) prediction in complex systems. Existing dependent degradation models usually describe dependence using correlation-based methods with symmetric characteristics. However, they ignore inherent causal directionality between degradation paths, which may lead to biased reliability evaluation and RUL predictions when physical causality exists. To address this issue, this paper proposes a causality-driven framework for modeling dependent degradation data. Firstly, univariate degradation models with multi-source uncertainties are established for each performance indicator based on the Wiener process. Then, the stable Peter-Clark algorithm is employed to uncover inherent causal relationships between degradation processes, and an uncertainty-aware neural network is employed to quantify causal effects considering uncertainties. Next, univariate degradation predictions and causal predictions are integrated within a Bayesian framework to construct a causally dependent degradation model, and the corresponding loss function is derived for model training. Finally, system reliability and RUL predictions are derived via Monte Carlo simulation. The proposed methodology is validated on the C-MAPSS dataset. Results show that considering inherent causal directionality between degradation processes helps eliminate physically unrealistic degradation behaviors, yielding more accurate degradation and RUL predictions than independent and correlation-based dependent degradation models.","published":"2026-09-25T14:43:03Z","updated":"2026-09-25T14:43:03Z","authors":["Shi-Shun Chen","Xiao-Yang Li"],"categories":["stat.AP"],"headline":"New method models how parts wear out by their true cause-and-effect links","summary":"The researchers address the problem that current models predicting how parts degrade over time overlook the direction of cause-and-effect between parts, which can lead to less accurate reliability and remaining life predictions. They developed a new approach that identifies causal relationships between different degradation processes and combines this with uncertainty-aware modeling to better predict system wear and remaining useful life. Their tests on a benchmark dataset showed their method produced more realistic and accurate predictions than models ignoring causal directions or only using correlations."},{"id":"2609.31337v1","title":"Representation-Guided Generation and Integration of Executable Programs for Robot Manipulation","abstract":"Building a robotic manipulation system requires connecting perception, planning, and control through carefully designed representations and interfaces. VLM code generation offers a way to automate this construction, but independently generated components may operate on incompatible geometric and task-level information. We present Representation-guided Integration of VLM-generated Executable Task programs (RIVET), a framework for generating complete manipulation systems around a shared object-centric representation. The representation combines per-object 6D poses, which preserve the metric information required for action grounding, with a relation graph that exposes the task-level structure required for planning. Guided by this representation, a VLM generates cooperating perception, rendering, relation-inference, and planning programs, each combining task-specific computation with available packages where useful. The resulting programs are authored once for a manipulation domain and reused on unseen start and goal configurations without code regeneration. We evaluate RIVET on cube stacking, tangram rearrangement, and three-dimensional assembly in simulation and on a physical robot, where we achieve 83% overall success rate in the real world by reusing offline-generated systems. Our results demonstrate that representation-guided program generation can adapt a common manipulation framework to tasks with different geometric, relational, and sequential requirements.","published":"2026-09-25T14:41:24Z","updated":"2026-09-25T14:41:24Z","authors":["Ruixiao Yang","Mingxin Yu","Chuchu Fan"],"categories":["cs.RO"],"headline":"New method helps robots learn tasks from shared object info","summary":"Researchers developed RIVET, a system that creates robot programs by using a shared way to describe objects and their relationships. This approach helps different robot tasks, like stacking cubes or assembling parts, use the same basic program without redoing the code each time. The team tested RIVET in both simulations and on a real robot and found it worked successfully in 83% of physical trials, showing potential for flexible robot manipulation systems."},{"id":"2609.31336v1","title":"Greedy queens and the golden ratio","abstract":"Place a queen in each successive column of an infinite $\\mathbb{N}\\times\\mathbb{N}$ chessboard, always choosing the lowest row such that no two queens may attack one another. We prove that the row $q_n$ occupied by the queen in the $n$th column satisfies $q_n=nφ+O(1)$ or $q_n=n/φ+O(1)$, where $φ=(1+\\sqrt5)/2$ is the golden ratio.","published":"2026-09-25T14:41:09Z","updated":"2026-09-25T14:41:09Z","authors":["Boon Suan Ho"],"categories":["math.CO"],"headline":"Researchers Study Queen Placements and the Golden Ratio","summary":"Researchers looked at placing queens one by one on an infinite chessboard, always putting each new queen in the lowest row that doesn't threaten the others. They found that the row positions of the queens closely relate to the golden ratio, a special number about 1.618. Their findings show a mathematical pattern in the queen placements but are theoretical and based on this specific placement rule."},{"id":"2609.31335v1","title":"The Wiener norm of lifts of dense Sidon sets","abstract":"We describe, with proof, an infinite class of examples of finite sets of integers $A$ for which $\\|\\widehat{1_A}\\|_1\\geq (c_\\mathcal{R}-o_{|A| \\rightarrow \\infty}(1))\\sqrt{|A|}$ where $c_\\mathcal{R}=0.953\\dots$.","published":"2026-09-25T14:40:57Z","updated":"2026-09-25T14:40:57Z","authors":["Tom Sanders"],"categories":["math.CA"],"headline":"New examples of special integer sets with high Wiener norms","summary":"The researchers studied certain finite sets of integers known as Sidon sets and how their Fourier transforms behave. They provided a mathematical proof describing an infinite class of such sets where a specific norm (called the Wiener norm) of their indicator functions is at least approximately 0.953 times the square root of the size of the set. Their findings are theoretical and based on mathematical analysis rather than experiments."},{"id":"2609.31334v1","title":"Formalizing Carleson's Theorem in Lean","abstract":"We present the formalization of Carleson's theorem in the proof assistant Lean. This paper describes the mathematical content, organization of the project, the blueprint, and the main design decisions behind the formalization. It is the result of a large collaborative effort, written and developed in public.","published":"2026-09-25T14:39:36Z","updated":"2026-09-25T14:39:36Z","authors":["Lars Becker","María Inés de Frutos-Fernández","Leo Diedering","Floris van Doorn","Sébastien Gouëzel","Evgenia Karunus","Edward van de Meent","Pietro Monticone","Jasper Mulder-Sohn","Jim Portegies","Joris Roos","Michael Rothgang","James Sundstrom","Jeremy Tan"],"categories":["math.CA","cs.LO"],"headline":"Researchers formalize Carleson's theorem using Lean software","summary":"Researchers have formalized Carleson's theorem—a significant mathematical result—using a proof assistant tool called Lean. Their work details the mathematical content, project setup, and key design choices involved in this collaborative public effort. The paper mainly focuses on describing the formalization process rather than new mathematical findings."},{"id":"2609.31333v1","title":"Probing the formation of femtosecond laser-induced periodic surface structures on silica films by ultrafast small-angle X-ray scattering","abstract":"Ultrashort X-ray pulses, as provided by X-ray Free Electron Lasers (XFELs), offer unique opportunities to probe the formation of laser-induced nanostructures with sizes even below the optical diffraction limit, with sub-micometer and sub-picosecond spatial and temporal resolution. Using ultrafast small-angle X-ray scattering (SAXS), we probe the formation of nanoscatterers and nanocracks in reciprocal space as they evolve into different types of laser-induced periodic surface structures (LIPSS) on silica films. Our focus is on the dynamics and evolution of nanoscatterers and high spatial frequency LIPSS (HSFL), with spatial periods below 100 nm. Our experiments include an in-situ analysis of the pulse-by-pulse evolution of HSFL at a fixed sample location, as well as fs time-resolved delay and fluence scans using an ultrafast pump-probe scheme. We find that nanoscatterers and periodic nanocracks form at delay times of hundreds of picoseconds and subsequently evolve into HSFL. At low laser fluences, inter-pulse incubation effects become important during multi-pulse irradiation. Simultaneously, inter-pulse feedback leads to an ordering and improve the regularity of the HSFL. The experiments are complemented by theoretical analyses, including numerical Finite-Difference Time-Domain (FDTD) simulations. These combined analyses allow us to propose a five-stage model of LIPSS formation on silica films upon irradiation with fs-laser pulses.","published":"2026-09-25T14:38:35Z","updated":"2026-09-25T14:38:35Z","authors":["Dominik Kaczmarek","Jean-Luc Déziel","Thies Johannes Albert","Matthias Weise","Jerzy Antonowicz","Robert Carley","Devesh Chopra","Loïc Le Guyader","Laurent Mercadier","Giuseppe Mercurio","Roman Minikayev","Mianzhen Mo","Andreas Scherz","Ryszard Sobierajski","Yanwen Sun","Martin Teichmann","Peter Zalden","Klaus Sokolowski-Tinten","Jörn Bonse"],"categories":["physics.optics"],"headline":"Studying ultrafast laser patterns on glass-like films with X-rays","summary":"Researchers used extremely short X-ray pulses to observe how tiny surface patterns form on silica films when hit by femtosecond laser pulses. They tracked the growth of nanoscale cracks and structures over very short times and multiple laser shots, finding that these features develop and organize over hundreds of picoseconds. The team combined experiments and computer simulations to propose a detailed five-step model explaining how these periodic surface patterns emerge under laser exposure."},{"id":"2609.31332v1","title":"Factorial Multivariate Bayesian Causal Forests: heterogeneous main and interaction effects of multiple treatments on correlated outcomes","abstract":"Many studies expose units to several binary treatments at once and record several correlated outcomes, yet analysts usually estimate one treatment's average effect on one outcome at a time -- discarding how treatments interact and how their effects vary across units. We introduce the factorial multivariate Bayesian causal forest, a Bayesian nonparametric model that decomposes the factorial response surface over the treatment lattice into a prognostic sum-of-trees plus one sum-of-trees per main or interaction effect, each with correlated multivariate leaf parameters, estimated jointly with coherent uncertainty. A single indicator-weighted kernel samples every component -- the prognostic term is the special case whose indicator is unity -- so the two-treatment model, the single-treatment multivariate causal forest, and a general order-r truncation are one and the same sampler. We give the ANOVA/Möbius identification of each estimand, an efficient Rcpp engine, and an interpretability layer using value-suppressing uncertainty maps. In simulations the average-effect estimators are unbiased with nominal coverage, consistent, and robust under misspecification, failing only under unmeasured confounding, which we flag. We illustrate the method on clinical, agricultural and economic data and on a deeper application to the NHANES survey. Software is provided as an R package.","published":"2026-09-25T14:38:17Z","updated":"2026-09-25T14:38:17Z","authors":["Danilo A. Sarti"],"categories":["stat.ME","stat.AP","stat.CO"],"headline":"New method analyzes multiple treatments and outcomes together","summary":"Researchers developed a statistical model called factorial multivariate Bayesian causal forest to study how multiple treatments affect several related outcomes at once. Unlike typical studies that look at one treatment and outcome separately, this method captures interactions between treatments and variation across individuals. The approach estimates effects jointly with uncertainty and is tested in simulations showing reliable results except when important factors are unmeasured. The researchers also applied it to real-world data and provide software for use."},{"id":"2609.31331v1","title":"The number of touching pairs of congruent sphere packings in Euclidean 3-space","abstract":"A packing of $n$ congruent balls in $\\mathbb{R}^3$ is a family of interior-disjoint Euclidean balls all having the same radius. The contact number of a packing is the number of touching pairs of balls. In this paper we investigate the problem of determining the maximum contact number, $c(n)$, of a packing of $n$ congruent balls in $\\mathbb{R}^3$. We first show that all packings of $n$ congruent balls that have a contact number of $c(n)$ are minimally rigid. Furthermore, we show that $c(n)=3n-6$ for $n=6,7,8,$ and $9$. These two results resolve a conjecture of K. Bezdek and Khan. During the proof of the latter result, we also enumerate the contact structures of all packings of $n$ congruent balls with contact number $c(n)$ for $n=6,7,$ and $8$. Additionally, we provide a lower bound construction which shows $c(n)> 6n-6\\sqrt[3]{2}n^\\frac{2}{3}$ when $n=16k^3-33k^2+24k-6$ where $k\\in \\mathbb{N}$. We also look at the restricted problem where each ball is centered on the face-centered cubic lattice $A_3$. In this case let $c_{A}(n)$ denote the maximum contact number. We show that $c_{A}(n)\\leq 6n-\\frac{6}{\\sqrt[6]{2}}n^\\frac{2}{3}$ for all $n$, and determine the asymptotics of $c_{A}(n)$ to be $c_{A}(n)=6n-(1+o(1))6\\sqrt[3]{2}n^\\frac{2}{3}$.","published":"2026-09-25T14:37:43Z","updated":"2026-09-25T14:37:43Z","authors":["Cameron Strachan"],"categories":["math.CO","math.MG"],"headline":"Researchers study how many touching pairs form in packed equal spheres","summary":"The researchers explored the maximum number of touching pairs (contact number) in arrangements of equal-sized balls packed in three-dimensional space. They proved that for 6 to 9 balls, the maximum touching pairs follow a specific formula and confirmed related conjectures. Their work also includes characterizing these packings and providing bounds for larger numbers of balls, especially when the balls are arranged on a specific lattice pattern. The findings are mostly theoretical and based on mathematical proofs and constructions."},{"id":"2609.31330v1","title":"Experiments on $3$-isogeny Selmer groups of elliptic curves with a $3$-torsion point","abstract":"Let $E_{A,B}\\colon y^2 +Axy + By = x^3$ be an elliptic curve over $\\mathbb{Q}$. The $3$-torsion point $(0,0)$ induces a $3$-isogeny $φ\\colon E_{A,B}\\to E_{A,B}'$. Assuming $E_{A,B}$ has good reduction at $3$, we construct an explicit $(m+t)\\times m$ matrix $M_{A,B}'$ over $\\mathbb{F}_3$, whose kernel encodes the dual isogeny Selmer group $\\operatorname{Sel}_{\\widehatφ}(E_{A,B}')$ modulo the image of the torsion point $(0,0)$. Here, $m = ω(B) - 1$, and $t$ encodes the \\emph{global Selmer ratio} $3^{t-2}$. We compute $M_{A,B}'$ in various regimes for billions of elliptic curves $E_{A,B}$. Based on our data, and motivated by prevalence of random linear algebraic models throughout number theory, we conjecture that, for fixed $m$ and $t$, the matrices $M_{A,B}'$ become uniformly distributed, and we formulate a corresponding conjecture for the distribution of $\\operatorname{Sel}_φ(E_{A,B})$. Our model predicts that, for fixed $t$, the average size of $\\operatorname{Sel}_φ(E_{A,B})$ is $1 + 3^t$.","published":"2026-09-25T14:37:31Z","updated":"2026-09-25T14:37:31Z","authors":["Ariel Weiss","Dongchen Zou"],"categories":["math.NT"],"headline":"Studying patterns in special groups of elliptic curves with a 3-point","summary":"The researchers focused on certain mathematical objects called elliptic curves that have a special 3-torsion point. They built explicit matrices to represent related algebraic structures called Selmer groups, which help understand properties of these curves. By analyzing billions of these curves, they observed patterns suggesting these matrices behave randomly in a precise way and proposed a conjecture predicting the average size of these Selmer groups. Their findings are based on extensive computations and theoretical reasoning."},{"id":"2609.31329v1","title":"Bridging Body and Brain: Gene-Driven Morphology--Control Co-Design","abstract":"Morphology--control co-design jointly optimizes an agent's body structure and control policy as an integrated embodied system. However, existing methods typically model morphology design and control with separate networks coupled only indirectly through a shared task objective, limiting explicit high-level coordination. Inspired by natural genes that coordinate biological development, we introduce \\textbf{Morphogene}, a compact latent blueprint that bridges an agent's body and brain. Through AdaConcat, Morphogene jointly conditions morphology and control generation at the limb level, allowing its variations to induce coordinated changes in both components. Building on this representation, we propose \\textbf{GeCode}, which formulates co-design as exploration in the compact Morphogene space. Each Morphogene anchors a local design region in which nearby body--brain designs are explored, while performance-guided updates move these anchors toward promising regions for more efficient exploration of the broader design space. This process combines local refinement with global exploration while preserving body--brain compatibility. Extensive experiments across diverse 2D and 3D co-design tasks demonstrate that GeCode consistently outperforms existing state-of-the-art methods, achieving substantially faster convergence and higher final performance.","published":"2026-09-25T14:36:18Z","updated":"2026-09-25T14:36:18Z","authors":["Fu Feng","Ruixiao Shi","Yucheng Xie","Jing Wang","Xin Geng"],"categories":["cs.LG"],"headline":"New method links body and brain design for smarter robots","summary":"Researchers developed Morphogene, a compact gene-like code that connects a robot's body structure and control system. Using this, their GeCode approach explores designs more efficiently by refining both parts together, inspired by biological coordination. Tests on 2D and 3D tasks showed their method converges faster and performs better than previous techniques, though findings are based on computer simulations and not real-world trials."},{"id":"2609.31328v1","title":"An ETH-Tight, Constructive FPT Algorithm for the Cone and Polytope Intersection Problem","abstract":"In a landmark paper, Goemans and Rothvoss (2020) established an XP algorithm running in time $\\text{enc}(P)^{2^{O(d)}} \\cdot \\text{enc}(Q)^{O(1)}$ for the \\emph{Cone and Polytope Intersection} problem: finding a vector $y \\in \\text{int.cone}(P \\cap \\mathbb{Z}^d) \\cap Q$ together with a sparse certificate $λ\\in \\mathbb{Z}_{\\ge 0}^{P \\cap \\mathbb{Z}^d}$ supported on at most $2^{2d+1}$ generators, where $P \\subseteq \\mathbb{R}^d$ is a bounded rational polyhedron and $Q \\subseteq \\mathbb{R}^d$ is an arbitrary rational polyhedron. For high-multiplicity bin packing, this gives a running time of ${|I|}^{2^{O(d)}}$, where $|I|$ denotes the encoding length of the input. Recently, Koana and Kumabe (2026) proved that the decision variant of this problem is fixed-parameter tractable (FPT) parameterized by the number of item types $d$ with running time $2^{d^{O(d)}} \\cdot {|I|}^{O(1)} = 2^{2^{O(d \\log d)}} \\cdot {|I|}^{O(1)}$. In this work, we generalize the framework of Koana and Kumabe from standard bin packing to the full Cone and Polytope Intersection Problem of Goemans and Rothvoss, directly encompassing high-multiplicity bin packing, point-in-cone, and scheduling. Secondly, by combining Carathéodory-type integer cone bounds (Eisenbrand and Shmonin, 2006) with active support enumeration, we reduce the running time to: $$2^{2^{O(d)}} \\cdot (\\text{enc}(P) + \\text{enc}(Q))^{O(1)}$$. Under the Exponential Time Hypothesis (ETH), the double-exponential lower bound of Kowalik, Lassota, Majewski, Pilipczuk, and Sokołowski (2024) for point-in-cone and Jansen, Ohnesorge, and Pirotton (2026) for high-multiplicity bin packing implies that this parameter dependence is asymptotically optimal. Finally, we provide an explicit decompression algorithm that extracts a solution with sparse support $|\\text{supp}(λ)| \\le 2^{2d+1}$ in single-exponential FPT time.","published":"2026-09-25T14:34:46Z","updated":"2026-09-25T14:34:46Z","authors":["Klaus Jansen","Felix Ohnesorge"],"categories":["cs.DS"],"headline":"New Efficient Algorithm Solves Complex Geometry and Packing Problems","summary":"Researchers improved methods for finding special points where two geometric shapes intersect, a problem important in packing and scheduling. They expanded previous work to create an algorithm that runs faster—specifically in a time that grows double-exponentially with a key parameter but is optimal under current theoretical limits. Their approach also produces a compact certificate confirming the solution, and the findings mainly hold for these high-dimensional, theoretical problems rather than practical real-world cases."},{"id":"2609.31327v1","title":"Eulerian and Bipartite Partial Duals of Hypermaps","abstract":"We study hyperedge partial duals of finite hypermaps in a purely combinatorial framework, without assuming orientability. A hypermap is represented by three fixed-point-free involutions $(τ_0,τ_1,τ_2)$ on its flag set. We first give an explicit construction of the medial map from this model: $02$-orbits become the medial vertex discs, while $τ_1$-transpositions become the medial bands; a local orientation system and its twist data then provide a signed rotation description of the medial map. We next prove that the state circles associated with a chosen set of hyperedges are in natural bijection with the vertex orbits of the corresponding partial dual, yielding a crossing-total characterization of all Eulerian hyperedge partial duals. For bipartiteness, the twist data lead to a modified medial map in which inserted bars record the obstruction to a global orientation. We prove that a partial dual is bipartite if and only if its dualized hyperedge set is exactly the set of $c$-type hyperedges identified by an all-crossing orientation of this modified medial map. When the hypermap is orientable, these constructions specialize to the known orientable-hypermap results; when every hyperedge has valence two, they specialize to the ribbon-graph results.","published":"2026-09-25T14:31:13Z","updated":"2026-09-25T14:31:13Z","authors":["Yufan Han","Metrose Metsidik"],"categories":["math.CO"],"headline":"Researchers analyze complex shapes called hypermaps using combinatorics","summary":"This research explores special transformations of hypermaps, which are abstract mathematical structures made of interconnected elements, without assuming these shapes can be oriented in a usual way. The researchers created a method to build related maps called medial maps and studied properties like when these transformed shapes have certain patterns (Eulerian) or can be split into two distinct groups (bipartite). Their findings extend previous knowledge from simpler or orientable cases, but the work is theoretical, relying on abstract mathematical models rather than real-world experiments."},{"id":"2609.31326v1","title":"CG-HAF: An Interpretable Global-Local Lesion-Burden Fusion Framework for Ordinal Acne Severity Grading in Agentic Skincare Support","abstract":"Ordinal acne severity grading requires distinguishing visually similar neighboring grades while jointly weighing holistic facial appearance and localized lesion burden - evidence that most existing approaches collapse into a single opaque representation. We introduce CG-HAF, a global-local fusion framework that instead keeps this evidence explicit: averaged holistic severity probabilities from independently trained classifiers are combined with structured lesion-burden descriptors from an object detector (lesion count, detection confidence, lesion area) into a compact representation, from which a lightweight, interpretable classifier produces the final grade. On a widely used benchmark, this fusion yields a clear, statistically supported improvement over global-evidence-only baselines, with the largest gains on the most severe cases. Testing on an independent dataset with a different grading standard shows that strong within-dataset performance does not transfer automatically, and a follow-up diagnostic attributes much of this gap to mismatched grading criteria rather than detection failure alone. These findings support interpretable global-local fusion as an effective strategy for ordinal acne grading while highlighting criterion alignment as key to cross-dataset portability, with a further illustration of how the resulting severity signal can support transparent, non-diagnostic decision-making in skincare applications.","published":"2026-09-25T14:30:16Z","updated":"2026-09-25T14:30:16Z","authors":["Muhammad Muhtasim Shahriar","Md. Naimur Asif Borno","Saad Aloteibi","Mohammad Ali Moni"],"categories":["cs.CV","cs.AI"],"headline":"New Method Improves Acne Severity Grading by Combining Global and Local Data","summary":"Researchers developed CG-HAF, a method that grades acne severity by combining overall facial appearance and detailed lesion information into an interpretable system. Tested on a common benchmark, their approach improved grading accuracy, especially for severe acne. However, when applied to a different dataset with different grading rules, performance dropped, suggesting that differences in grading criteria, not just detection issues, affect results. The study highlights the potential of combining global and local features for clearer acne assessment but notes challenges in applying models across different grading systems."},{"id":"2609.31325v1","title":"More Sensors Only One Field: Rethinking Continual Spatio-Temporal Forecasting","abstract":"Continual spatio-temporal forecasting supports traffic management and environmental monitoring under evolving dynamics and expanding sensor networks. However, conventional graph-based continual learning methods tie forecasting representations to the current sensor layout, so sensor expansion can alter the representation of learned spatial relationships. Our key insight is that sensor expansion changes the evidence available about a process without necessarily changing the dynamics to be learned. We propose STFO (Spatio-Temporal Field Operator), which parameterizes forecasting knowledge as a shared field-evolution operator and handles changing sensor layouts through observation and query interfaces. Normalized coordinate-based aggregation lifts irregular sensor histories onto a fixed latent grid, enabling reuse of learned spatial maps across observation sets without sensor-specific parameters. To accommodate process drift, a spectral descriptor summarizes variation across spatial scales and conditions Fourier propagation and attention to adapt operator responses to the current spatial regime. Coordinate-based decoding queries the evolved field at sensor locations and combines spatial corrections with local-history predictions. Experiments on PEMS-Stream, CA-Stream, and AIR-Stream demonstrate state-of-the-art average forecasting performance. STFO-Large reduces average MAE over DOL by 8.4% on PEMS-Stream and 4.7% on CA-Stream. Our code is available at https://github.com/Xielewei/Spatio-Temporal-Field-Operator.","published":"2026-09-25T14:29:22Z","updated":"2026-09-25T14:29:22Z","authors":["Lewei Xie","Haoyu Zhang","Jiajun Zhou","Yulong Chen","Guanxing Chen","Yu-An Huang","Hau-San Wong","Yifan Zhang","Zhi-An Huang"],"categories":["cs.LG"],"headline":"New Method Improves Forecasting with Expanding Sensor Networks","summary":"Researchers studied how to make forecasts when the number and arrangement of sensors change over time, which usually disrupts learning methods tied to fixed sensor layouts. They developed a technique called STFO that treats forecast knowledge consistently, regardless of changing sensors, by mapping data onto a stable grid and adapting to shifts in conditions. Tests showed STFO improved prediction accuracy in traffic and environmental data. Results come from experiments and simulations, not real-world deployment, and highlight a new way to handle changing sensor data."},{"id":"2609.31324v1","title":"A statistical hardness classification of ultraluminous X-ray sources using long-term Swift/XRT monitoring","abstract":"We characterize long-term hardness behavior in 29 ultraluminous X-ray sources (ULXs) using over 3500 archival Swift/XRT observations. We classify observations as soft-band dominated, hard-band dominated, or uncertain from the hardness ratio (HR; 1.5--10/0.3--1.5 keV count rates) and its confidence bounds. We derive $f_{\\rm soft}$ and $f_{\\rm hard}$ and define the Crossing Index (CI) as the fraction of consecutive confidently classified observations on opposite sides of $HR=1$. We analyze these descriptors with rule-based classification and unsupervised hierarchical clustering. The rule-based analysis identifies eight soft-band-dominated, five hard-band-dominated, and 16 mixed-band systems; clustering yields a similar three-group structure. Known or candidate accretor types tend to occupy band-dominated groups, whereas unknown accretors are more common in the mixed-band group. All four sources with reported long-term X-ray modulation discussed in an orbital or super-orbital context are classified as mixed-band. The three-cluster solution is preferred in 25 of 29 leave-one-out trials, while bootstrap tests find the band-dominated groups more stable than the mixed-band group. As the classification uses observed count rates, the groups represent Swift/XRT signatures rather than intrinsic spectral states. Response-folding tests show that modeled HR depends on absorption column density and photon index. Bayesian reanalysis of a representative low-count source reduces CI from 56.3\\% to 38.5\\% while preserving its mixed-band classification. Random downsampling produces no systematic shift in CI; halving the observations changes median CI by at most $\\sim10\\%$ across four sources. Larger samples with independent absorption constraints can test links among these observational classes, accretor type, long-term periodic modulation, and other X-ray properties.","published":"2026-09-25T14:28:03Z","updated":"2026-09-25T14:28:03Z","authors":["Sinan Allak","Wei Yu","Menglei Zhou","Honghui Liu","Andrea Santangelo"],"categories":["astro-ph.HE"],"headline":"Classifying ultraluminous X-ray sources by long-term X-ray color patterns","summary":"Researchers analyzed over 3500 Swift/XRT observations of 29 ultraluminous X-ray sources to study their long-term X-ray color, or hardness. They grouped sources as soft-band dominated, hard-band dominated, or mixed based on how the X-ray emission changes over time. These groups relate to the type of material falling onto compact objects, but the classification reflects observational data rather than intrinsic properties. The study supports searching for links between these groups and the sources’ nature, though the findings rely on observed signals and statistical analysis."},{"id":"2609.31323v1","title":"See to Reach, Feel to Grasp: Learning A Blind Grasp Reflex for Anthropomorphic Robotic Hands","abstract":"In this work we study if a robotic hand using proprioception alone can grasp diverse objects with no visual observation. We present a modular dexterous grasping architecture that separates global arm motion from local contact control. An independently controlled arm guides the hand toward the object, while a reinforcement learning policy grasps and stabilizes it using only hand proprioceptive feedback. We call this \\textit{a blind grasp reflex}: grasping without images, object poses, or geometric observations. A learned stable-grasp score determines when the object is securely held, allowing the arm to begin post-grasp manipulation. This separation makes grasping a reusable hand-level skill that can be combined with independently designed arm controllers for various manipulation tasks. Experiments in simulation and on hardware demonstrate robust blind grasping across diverse objects and seamless composition with a range of arm controllers. Moreover, despite never observing contact geometry, the learned grasp score closely aligns with an independent physics-based measure of grasp stability. The resulting approach follows a simple principle: see to reach, feel to grasp. Project page: https://blindgraspreflex.github.io.","published":"2026-09-25T14:27:54Z","updated":"2026-09-25T14:27:54Z","authors":["Alexander Alexiev","Tzu-Yuan Lin","Sang Min Kim","Ho Jae Lee","Yonghyeon Lee","Sangbae Kim"],"categories":["cs.RO"],"headline":"Robotic Hand Learns to Grasp Objects Without Visual Help","summary":"Researchers explored whether a robotic hand could grasp various objects using only its sense of position and movement, without any visual input. They developed a system where a robot arm moves the hand near an object, and a learning-based method uses internal feedback to grasp and hold it securely. Tests in simulations and real robots show the hand can reliably grasp objects blindly, with a learned score that matches physics-based measures of grasp quality. This method separates arm movement from hand control, allowing flexible use in different tasks."},{"id":"2609.31322v1","title":"On convergence of formal Dulac series, satisfying an algebraic $q$-difference equation","abstract":"An algebraic $q$-difference equation for $|q|>1$ is considered. A sufficient condition is proposed for the convergence of a solution to such an equation in the form of a Dulac series within a sector centered at the origin. An example illustrating the application of this sufficient condition is provided.","published":"2026-09-25T14:27:42Z","updated":"2026-09-25T14:27:42Z","authors":["Nikita V. Gaianov","Anastasia V. Parusnikova"],"categories":["math.CA"],"headline":"When certain series solutions to equations truly converge","summary":"The researchers study a specific type of equation called an algebraic q-difference equation, focusing on cases where the parameter q is greater than 1 in absolute value. They provide a condition that ensures a special kind of series solution, known as a Dulac series, actually converges near zero within a certain region. Their work includes an example showing how this condition can be used. The findings are theoretical and concern the mathematical properties of these series solutions."},{"id":"2609.31321v1","title":"Suitability studies of exotic muon decay at the High Intensity heavy-ion Accelerator Facility","abstract":"We present a detailed feasibility study of exotic muon decay searches using the muon beam at the High Intensity heavy-ion Accelerator Facility (HIAF). For the decay channel mu to eX0, where X0 denotes a new boson featuring LFV couplings, the produced electron carries very low energy when the mass of X0 approaches the muon mass, making detection challenging. Nevertheless, the high-energy beam at HIAF can boost these electrons up to a measurable energy range. The detector system proposed here comprises front-end scintillators for particle identification (PID) and a main detector consisting of an RPC stack to track decay products and a scintillator for energy measurement. Leveraging the kinematic properties of low momenta and small emission angles of signal electrons, we employ optimized momentum and angular acceptance cuts on the RPC detector to efficiently suppress background events. With this strategy, we obtain an upper limit of 1e-5 on the branching ratio of the exotic muon decay at 95% confidence level, achieving state-of-the-art sensitivity.","published":"2026-09-25T14:27:11Z","updated":"2026-09-25T14:27:11Z","authors":["Lingzhi Dong","Jinning Li","Leyun Gao","Cheng-en Liu","Yu Xu","Xuecheng Zhang","Liangwen Chen","Qite Li","Chen Zhou","Qiang Li","Zhiyu Sun"],"categories":["hep-ex","hep-ph","physics.ins-det"],"headline":"Studying rare muon decay to detect new particles at HIAF","summary":"Researchers investigated the possibility of detecting a rare muon decay involving a new particle using the muon beam at the High Intensity heavy-ion Accelerator Facility (HIAF). They proposed a detector setup designed to identify very low-energy electrons boosted by the high-energy beam, and used techniques to reduce background noise. Their study, based on simulations, suggests they could set an upper limit on how often this exotic decay happens, demonstrating promising sensitivity but not an actual detection."},{"id":"2609.31320v1","title":"A new operator splitting method for coupled PDE problems, with an application to poromechanics","abstract":"The present work proposes a new operator splitting technique. The method leverages ideas used in PDE-constrained domain decomposition, adapting them to the objective of decoupling equations defined on the same geometric domain. The methodology is applied here to the resolution of coupled flow and deformation in the linear quasi-static Biot problem. An overview of its main characteristics is provided, also in comparison with the well established fixed-stress split for poroelasticity.","published":"2026-09-25T14:26:25Z","updated":"2026-09-25T14:26:25Z","authors":["Francesca Marcon","Stefano Scialò"],"categories":["math.NA"],"headline":"New Method to Solve Coupled Physical Equations Separately","summary":"Researchers developed a new technique to break down complex equations that describe interacting physical processes into simpler parts that can be solved independently. They focused on equations about fluid flow and material deformation in porous materials. The study compares this new method with an existing one, highlighting its main features. The findings are theoretical and based on mathematical modeling, providing an alternative approach to solving these coupled problems."},{"id":"2609.31319v1","title":"Criticality enabled long-range order in a U(1)-symmetric spin-1 Heisenberg chain with biquadratic interactions","abstract":"In a recent paper Nahum argued on the basis of a renormalization-group analysis that, contrary to standard lore, ground states of 1D spin chains with short-range interactions can spontaneously break U(1) ``easy-plane'' spin rotation symmetry. True long-range order of $(S^x,S^y)$ arises at the phase transition between two quasi-long-range-ordered phases. Here we present detailed numerical results for an anisotropic spin-1 chain model. We find a magnetization exponent $β\\approx0.29$, consistent with the $ε$-expansion prediction $β\\approx0.28$, a divergence of the Luttinger parameter on approaching the transition from both sides with an exponent that we relate to Nahum's transition, and finite-size spectra consistent with a dynamical critical exponent $z\\approx2$. Our results support the critical behavior predicted by Nahum's field theory.","published":"2026-09-25T14:26:20Z","updated":"2026-09-25T14:26:20Z","authors":["Natalia Chepiga","Fabian H. L. Essler"],"categories":["cond-mat.str-el","quant-ph"],"headline":"Scientists study unusual magnetic order in 1D spin chains","summary":"Researchers investigated a special type of magnetic system where tiny quantum spins are arranged in a line. Contrary to previous beliefs, they found evidence supporting a theory that these systems can develop definite long-range magnetic order under certain conditions. Using detailed computer simulations, they observed critical behavior matching recent theoretical predictions, though their findings are specific to this model and primarily support the proposed mathematical theory."},{"id":"2609.31318v1","title":"AgentXploit: Autonomous Repository-to-Runtime Red-Teaming for AI Agents","abstract":"AI agents combine language models with external data and tools that can modify files, call APIs, or execute code. Security failures can arise when adversarial content changes an agent's tool use or when the surrounding software contains vulnerabilities such as path traversal or command injection. We study authorized white-box pre-deployment auditing, where the auditor has access to the target repository and a controlled runtime, but successful attacks must still act through the task-defined attacker interface and be confirmed by an external verifier. We present AgentXploit, a two-role auditing system that separates repository-level attack-path discovery from runtime exploitation. The Analyzer Agent traces attacker-controlled inputs to sensitive operations and records code-supported candidate attack paths; the Exploiter Agent turns these paths into concrete attacks and revises them using runtime feedback. We also introduce AgentXploit-Bench, containing 72 reproducible vulnerabilities across 12 open-source AI-agent systems and frameworks. Across three runs, AgentXploit reaches 59.3% end-to-end success, compared with 38.4% for Codex. Under a token-budget-matched comparison, Codex reaches 46.3%. On AgentDojo, where injection points are provided, the Exploiter Agent reaches 79.2% attack success versus 52.7% for AgentVigil. These results highlight repository discovery and runtime exploitation as distinct challenges in end-to-end agent security auditing.","published":"2026-09-25T14:26:04Z","updated":"2026-09-25T14:26:04Z","authors":["Weida Liang","Shi Qiu","Zhun Wang","Simon Sure","Xiaoyuan Liu","Tianneng Shi","Zhaorun Chen","Wenbo Guo","Dawn Song"],"categories":["cs.CR","cs.AI"],"headline":"New System Tests AI Agent Security by Mimicking Attacks","summary":"Researchers developed AgentXploit, a method to test how AI agents can be exploited by finding and using weaknesses in their code and behavior before they are deployed. It separates the process into discovering potential attack paths in the code and then trying to exploit them during operation. They created a benchmark with real vulnerabilities to compare AgentXploit's effectiveness against other tools, showing it identifies security issues more successfully. Their work highlights that both discovering and exploiting weaknesses are important challenges in securing AI agents."},{"id":"2609.31317v1","title":"Highly elastic collisions of ultracold asymmetric top molecules in a static electric field","abstract":"We propose that the application of static electric fields to reactive CaNH$_2$ molecules can result in large ratios of elastic to lossy collisions at ultracold temperatures. In the first rotationally excited manifold, the asymmetric top structure of CaNH$_2$ results in a $\\approx 226$ MHz splitting of the opposite parity $K$-doublet states. This modest splitting allows electric fields below a kilovolt per centimeter to polarize the molecules and induce long-range dipole-dipole interactions. By increasing the electric field, we find that dipole-dipole attraction can become sufficiently large to induce weakly bound octatomic field-linked states of [CaNH$_2$]$_2$. In the absence of these field-linked states, we identify select molecular states where dipolar interactions allow elastic collisions to dominate over lossy collisions by up to three orders of magnitude, while maintaining loss rate coefficients at the $10^{-14}$ cm$^3$s$^{-1}$ level. These results establish that evaporative cooling is a viable route to quantum degenerate gases of laser coolable asymmetric top molecules. Access to such low entropy ensembles of asymmetric top molecules holds promise to push the frontiers of quantum simulation, state-resolved chemistry, and searches for beyond the Standard Model physics.","published":"2026-09-25T14:24:05Z","updated":"2026-09-25T14:24:05Z","authors":["Reuben R. W. Wang"],"categories":["physics.atom-ph","cond-mat.quant-gas","quant-ph"],"headline":"Electric fields may enable gentle collisions in cold molecules","summary":"Researchers studied how applying static electric fields to certain complex molecules (CaNH2) at very low temperatures can increase the number of gentle (elastic) collisions compared to damaging (lossy) ones. They found that moderate electric fields can polarize the molecules and induce interactions that encourage elastic collisions, sometimes creating weakly bound molecular pairs. These findings suggest that a cooling technique called evaporative cooling could work for these molecules, potentially aiding future studies in quantum science. The conclusions are based on theoretical modeling and simulations."},{"id":"2609.31316v1","title":"Excitation exchange between identical atoms having arbitrary angular momentum","abstract":"Equations of motion are derived for excitation exchange between two atoms having arbitrary angular momentum.","published":"2026-09-25T14:23:45Z","updated":"2026-09-25T14:23:45Z","authors":["P. R. Berman","Zeyuan Wang"],"categories":["quant-ph"],"headline":"Study of Energy Transfer Between Atoms with Different Spins","summary":"Researchers derived mathematical equations describing how energy is exchanged between two atoms that have any amount of angular momentum, which is a type of atomic spin. The work focuses on the theoretical motion equations governing this process. The abstract reports the development of these general equations but does not provide experimental results or applications."},{"id":"2609.31315v1","title":"LUCID: Learning Under Confounding for Inference and Discovery in Time Series","abstract":"Unobserved common causes are pervasive in real-world time series and can induce spurious associations that causal discovery methods mistake for direct edges. We propose LUCID (Learning Under Confounding for Inference and Discovery, a regime-adaptive deconfounding layer that first estimates the confounding regime from data using a Marčenko--Pastur spectral router, then applies a deconfounding strategy matched to that regime. When the spectrum indicates pervasive factor confounding, LUCID attenuates factor-dominated variation and recovers contemporaneous (lag-$0$) structure from the resulting innovations, with edge selection calibrated against a data-driven edge-free null. Rather than being tied to a particular discovery algorithm, it can wrap existing discovery engines; we demonstrate consistent improvements across three such methods. On a diverse synthetic out-of-distribution benchmark spanning changes in confounder strength and sparsity, loading density, lag structure, volatility dynamics, edge heterogeneity, persistence, intermittency, and tail behavior, LUCID achieves the best family-weighted directed, lag-resolved graph $F_1$ ($0.60$), improving over the strongest baseline by $0.19$ absolute ($\\approx\\!46\\%$ relative). Its advantage widens relative to looser lag-collapsed scoring, and remains robust under intermittent and heavy-tailed confounding. Code reproducing the method, the benchmark generators, and every reported experiment is available at https://github.com/bloomberg/causal-ts.","published":"2026-09-25T14:21:21Z","updated":"2026-09-25T14:21:21Z","authors":["Mohammad Fesanghary"],"categories":["cs.LG","stat.ML"],"headline":"New Method Helps Untangle Hidden Causes in Time Series Data","summary":"Researchers developed LUCID, a technique to better detect genuine connections in time series data that may be distorted by unseen common influences. LUCID first identifies the type of hidden confounding using a spectral method, then applies an appropriate correction before discovering relationships. Tested on synthetic data with varied complexities, LUCID improved accuracy in identifying direct connections compared to existing methods. The findings are based on simulations and currently limited to scenario benchmarks, as detailed in the researchers' reported experiments."},{"id":"2609.31314v1","title":"CytoSPM: Open-Vocabulary Cytopathology Detection with Structured Prompt Bank","abstract":"Cytopathology detection requires open-vocabulary recognition because cellular categories are fine-grained, long-tailed, and continuously evolving across different organ systems. However, existing cytology detectors are mostly single-domain and closed-set, and there is still no unified benchmark for evaluating open-vocabulary cytopathology detection. We present PentaCyto, a multi-domain benchmark covering cervical, urinary, respiratory, serous fluid, and thyroid cytology, with 24 base categories and 9 held-out novel categories. Each category is associated with structured cytomorphology prompts that describe diagnostic morphological attributes and provide clinically grounded textual knowledge. We further propose CytoSPM, an efficient detector based on a decoupled two-stage design. It first extracts reusable class-agnostic visual representations, and then performs class-aware structural prompt matching with class names and cytomorphology prompts. On PentaCyto, CytoSPM outperforms existing methods in novel-category detection and open-vocabulary detection while maintaining efficient inference.","published":"2026-09-25T14:21:21Z","updated":"2026-09-25T14:21:21Z","authors":["Wenjie Li","Zishan Xu","Jinyang Huang","Zhengxin Nie","Shichao Kan","Yixiong Liang"],"categories":["cs.CV"],"headline":"New AI tool improves cell disease detection across multiple organs","summary":"Detecting diseases in cells is challenging because the types of cells vary widely and keep changing. The researchers created PentaCyto, a new testing set covering cells from five body areas with many known and new categories. They also developed CytoSPM, a two-step AI system that first finds general cell features and then matches detailed cell descriptions. Their tests show CytoSPM works better than current methods at recognizing new cell types, though this is based on their specific benchmark data."},{"id":"2609.31313v1","title":"Towards VLA-Dreamer: Refining VLA Behavior Using World Models","abstract":"Vision-Language-Action models (VLAs), while showing strong potential for robot control, require massive amounts of high-quality imitation learning data. Moreover, the absence of an explicit world model casts further doubt on their control capabilities. In this concept paper, we propose a novel architecture that addresses sample efficiency in VLAs by training a predictive world model on the embedding space of the VLA's vision encoder. We hypothesize that these embeddings are action-relevant and usable for future prediction. To this end, we propose using the suggested architecture to investigate how well these embeddings predict the future based on actions, as the inability to do so would mark a key limitation of VLA architectures: the lack of a non-lossy implicit world model to simulate real-world dynamics. The proposed architecture differs from the standard world model dynamics as the loss comes from the embedding space rather than the pixel space, similar to joint embedding predictive architectures. Furthermore, the trained world model can be utilized for short-term planning tasks by sampling VLA actions given goal images. We intend to examine the richness of vision embeddings in VLAs and reduce their high data requirements through a world model that can also generate plans during inference.","published":"2026-09-25T14:21:20Z","updated":"2026-09-25T14:21:20Z","authors":["Parsa Mastouri Kashani","Jan-Gerrit Habekost","Stefan Wermter"],"categories":["cs.RO","cs.AI"],"headline":"Improving Robot Control by Predicting Future from Visual Data","summary":"Researchers address challenges in robot control models that combine vision, language, and action but need large amounts of data and lack clear world understanding. They propose a new system that predicts future states using a compressed visual representation, aiming to improve learning efficiency and planning. This paper presents a concept and hypothesis about the usefulness of these visual embeddings for prediction, but actual performance and benefits remain to be tested in future work."},{"id":"2609.31312v1","title":"Causal Query Compression for Lindblad Dynamics: Optimal Queries and Nearly Linear Local Simulation","abstract":"We construct a causal query compiler for time-dependent Lindblad dynamics and establish query and gate bounds under distinct input models. With coherent block encodings of the Hamiltonian and individual jump operators, a Lipschitz generator can be simulated to diamond-norm error $\\varepsilon$ using $O(1+τ+\\log(1/\\varepsilon)/\\log(e+\\log(1/\\varepsilon)/τ))$ controlled queries, where $τ=T(α_H+\\sum_μα_μ^2)$. This bound is independent of the time-grid size and is worst-case optimal for $τ\\ge1$. Weak active entry and clean return yield a factorial history tail; combining two reuse lengths removes the auxiliary input. A three-call polar correction supplies the Lindblad cells. The compiler also characterizes the minimum oracle action of bounded adaptive Markovian protocols by quantum query complexity and the general adversary bound, up to constant factors. For finite-range dynamics on $n$ lattice sites, efficient coherent evaluation of the local matrices and fixed local parameters give $O(n(T+1)\\operatorname{polylog}X)$ elementary gates, where $X=\\max\\{e,n(T+1)(1+K_t+B)/\\varepsilon\\}$, for piecewise Hölder generators with fixed exponent, local variation bound $K_t$, and at most $B$ breakpoints per term. The jumps need not commute. This gate bound includes evaluation, arithmetic, selection, routing, and environment storage. Its proof combines a spatial decomposition on a common bath, compilation on occupied inputs, and routing on compressed records. Reusing bath storage between short segments gives depth $(T+1)\\operatorname{polylog}X$ and space $n\\operatorname{polylog}X$.","published":"2026-09-25T14:20:11Z","updated":"2026-09-25T14:20:11Z","authors":["Jacob Kitchen"],"categories":["quant-ph"],"headline":"Optimizing Quantum Simulation Queries for Complex Time-Dependent Systems","summary":"The researchers developed a method to efficiently simulate certain quantum systems that change over time, specifically those described by Lindblad dynamics. They created a compiler that optimizes the number of queries and computational steps needed to approximate the system's behavior accurately, independent of specific time discretization. Their approach also applies to large systems with local interactions, improving resource use while maintaining error bounds. These results are theoretical and based on mathematical and complexity analyses rather than experimental demonstrations."},{"id":"2609.31311v1","title":"Residual Fisher Information in Measurement-Error Uncertainty","abstract":"Measurement error in the Watanabe-Sagawa-Ueda formulation is the excess estimation error of a classical record over the quantum statistical limit. We show that, at the quantum-instrument level, this excess has an exact decomposition. The record-only error matrix equals a residual-aware error matrix plus a positive Schur-complement term fixed by Fisher information left in the unread post-measurement system. Thus measurement error separates into intrinsic instrument loss and record-restriction loss","published":"2026-09-25T14:19:57Z","updated":"2026-09-25T14:19:57Z","authors":["Masayuki Ohzeki","Taiga Suzuki"],"categories":["quant-ph"],"headline":"New Breakdown of Measurement Error in Quantum Systems","summary":"Researchers studied how errors occur when measuring quantum systems, focusing on extra mistakes beyond fundamental quantum limits. They found that this excess error can be exactly divided into two parts: one due to the measurement device itself and another due to limitations of the recorded data. Their findings help clarify where measurement errors come from by analyzing information left unused in the system after measurement. The results are theoretical and relate to quantum measurement frameworks."},{"id":"2609.31310v1","title":"Revisiting Certified Defense with Differential Privacy on Vision Transformers","abstract":"Certified defenses that incorporate differential privacy have proven effective on Convolutional Neural Networks (CNNs), furnishing rigorous robustness guarantees against norm-bounded adversaries. However, the certified robustness behavior of Pixel Differential Privacy (PixelDP) remains largely unexplored with the self-attention architecture now dominating the deep-learning landscape. Given that the Transformer has a profound impact on our daily applications from the digital world to the physical world, it is crucial to study certified robustness through differential-privacy-style stability. To fill this research gap, we revisit this construction in Vision Transformers and identify a failure mode that is largely hidden in the convolutional setting. When noise is injected after the patch embedding, the Laplace mechanism with the inherited grouped $\\ell_1$ sensitivity bound collapses to chance-level accuracy across noise scales, whereas the Gaussian mechanism remains trainable. This contrast isolates the source of failure: not the injected noise itself, but the geometry of the sensitivity constraint. We show that the attenuation induced by the inherited $Δ_{1,1}$ projection increases with layer width and kernel size according to a random-matrix scale $C/(\\sqrt{M}+\\sqrt{N})$. Replacing the $\\ell_1$-type constraint with a spectral-norm constraint eliminates the collapse across datasets and architectures, but creates a fundamental obstacle: the repaired models no longer satisfy the sensitivity condition required by the standard Laplace certificate. We resolve this mismatch by deriving a dimension-free $(\\varepsilon,\\ δ)$-privacy guarantee for the Laplace mechanism under $\\ell_2$ sensitivity through concentration of the privacy loss.","published":"2026-09-25T14:17:57Z","updated":"2026-09-25T14:17:57Z","authors":["Jun Yan","Weiquan Huang","Qixian Zhang","Yan Bai","Shutai Zhang"],"categories":["cs.CR","stat.AP"],"headline":"Improving Certified Defense for Vision Transformers Using Differential Privacy","summary":"Researchers studied how to protect vision transformers—a type of AI model—from certain attacks by adding privacy-preserving noise. They found that previously used methods worked poorly when applied to transformers, causing accuracy to drop to chance levels. By changing the mathematical constraints used to measure sensitivity, they improved stability, but this broke existing privacy guarantees. To fix this, they developed a new privacy guarantee that works with their improved method. Their findings are based on theoretical analysis and simulations, highlighting challenges in adapting privacy techniques to new AI architectures."},{"id":"2609.31309v1","title":"Fate of Bouncing Singularities at Finite $G_N$","abstract":"The gravity approximation to a thermal two-point function can develop a complex-time singularity associated with a trajectory that reaches the black-hole singularity. We show that the corresponding singularity is absent at finite $G_N$. This is done by extending Conformal Field Theory analyticity properties to the second sheet (and to the universal cover). We also put two different upper bounds on the amplitude of the correlator at the would-be singularity, by certain finite partial sums, and by a certain partition function on a manifold with nontrivial topology, where two thermal geometries are glued. We discuss the possible implications for the black hole singularity.","published":"2026-09-25T14:17:01Z","updated":"2026-09-25T14:17:01Z","authors":["Zohar Komargodski"],"categories":["hep-th"],"headline":"Researchers Study Black Hole Singularity Behavior at Finite Gravity Strength","summary":"The researchers investigate a specific type of mathematical singularity that appears in models of black holes when using an approximation of gravity. They find that this singularity disappears when the strength of gravity is considered finite, by extending certain mathematical properties of quantum field theories. Their work provides new upper limits on related quantities and discusses potential implications for understanding black hole singularities. Their findings are theoretical and based on advanced mathematical frameworks."},{"id":"2609.31308v1","title":"SQuIGG$\\vec{L}$E: Ubiquitous Cool Gas Outflows with Large Mass-Loading Factors in $z\\sim0.7$ Post-starburst Galaxies","abstract":"While recent James Webb Space Telescope studies have highlighted cool gas outflows as an important mechanism for rapid quenching at $z\\gtrsim2$, such space-based spectroscopic samples remain limited in size. In this work, we search for and characterize cool gas outflows traced by Fe II $λλ$2586,2600 and Mg II $λλ$2796,2803 absorption in post-starburst galaxies (PSBs) at $0.5<z<0.9$, which are accessible to ground-based spectroscopy. Utilizing the SQuIGG$\\vec{L}$E survey containing over 1300 PSBs, we stack spectra in bins of stellar mass, (specific) star formation rate, burst mass, burst mass fraction, and time since quenching. We observe blueshifted line centroids indicative of outflows in all stacks except the one containing galaxies that quenched more than 260 Myr ago. Outflow velocities decline with increasing time since quenching, while showing little correlation with other galaxy properties. The inferred outflows have large mass loading factors of 1 -- 120, maximum velocities above the interstellar medium escape velocities of their host galaxies, and energy and momentum fluxes exceeding those expected from the current star formation. Our results suggest that cool gas outflows are ubiquitous in intermediate-redshift PSBs and are likely relic outflows launched by previous starburst or AGN activity. These outflows have the potential to quench star formation by transporting gas into the galactic halos, consistent with outflows playing an important role in rapid quenching from intermediate to high redshift.","published":"2026-09-25T14:15:02Z","updated":"2026-09-25T14:15:02Z","authors":["Yuanze Luo","Justin S. Spilker","David J. Setton","Timothy Heckman","Rachel Bezanson","Vincenzo R. D'Onofrio","Alex Geiger","Andy D. Goulding","Jenny E. Greene","Kate Rowlands","Katherine A. Suess","Elizabeth Taylor","Margaret E. Verrico","Andrew Weldon","Vivienne Wild","Pengpei Zhu"],"categories":["astro-ph.GA"],"headline":"Widespread Gas Outflows Found in Post-Starburst Galaxies","summary":"Researchers studied cool gas flowing out of galaxies that recently stopped forming stars, using data from over 1300 post-starburst galaxies observed with ground-based telescopes. They found gas outflows are common shortly after star formation ends but weaken with time. These outflows are strong and fast enough to escape their galaxies, suggesting they may help stop star formation by moving gas away. The findings come from analyzing stacked spectra and support the idea that past events like starbursts or black hole activity drive these gas flows."},{"id":"2609.31307v1","title":"SQuIGG$\\vec{L}$E: Spatially resolved NIRSpec Pa$α$ reveals that the most CO-luminous, $z\\sim0.7$ post-starburst galaxies exhibit falling---but still active---star formation behind optically thick dust","abstract":"One of the fundamental uncertainties in the study of galaxies at the interface between star formation and quiescence--post-starburst galaxies--is their instantaneous star formation rate. While these galaxies appear quiescent in the rest UV-optical, many of the youngest systems are highly CO luminous, implying dense molecular gas, and potentially star formation. However, their instantaneous star formation rates have proved elusive, as measurements with rest-optical emission lines or UV-to-IR SED fitting yield order-of-magnitude uncertainty. Here, we present JWST/NIRSpec G395M/170LP integral field spectroscopy of the three most CO-luminous massive galaxies in the SQuIGG$\\vec{L}$E survey, spatially resolving Pa$α$/Br$γ$/Br$β$ to map the dust-corrected star formation rate. We find that these galaxies host star formation rates of $27-80$ $M_\\odot \\ \\mathrm{yr^{-1}}$, though $\\sim30-40\\%$ of the emission may arise from non-star-forming ionization (AGN or shocks), implied by high [FeII]/Pa$α$ in central spaxels. We also find effective $A_{\\mathrm{Pa}α}\\sim1$ and individual $\\sim700$ pc spaxels with $A_{\\mathrm{Paα}}\\sim2$, implying significant dust obscuration of nebular regions that contribute little to the rest-optical lines. These galaxies are not consistent with total quiescence; the measured star formation rates place these galaxies on or above the main sequence. However, their star formation rates \\textit{are} significantly lower ($\\sim0.5-1$ dex) than the peak star formation rates inferred from spectrophotometric fitting. We propose that the CO luminous post-starburst phase represents the last gasp of star formation in galaxies on the rapid quenching pathway, and that dust-obscured star formation, outflows, and (speculatively) an uncertain CO-to-H2 conversion reconcile their gas exhaustion with observed timescales of fading.","published":"2026-09-25T14:15:02Z","updated":"2026-09-25T14:15:02Z","authors":["David J. Setton","Jenny E. Greene","Rachel Bezanson","Katherine A. Suess","Justin S. Spilker","Vincenzo R. D'Onofrio","Elia Cenci","Robert Feldmann","Alex J. Geiger","Andy D. Goulding","Mariska Kriek","Anika Kumar","Yuanze Luo","Desika Narayanan","Margaret E. Verrico","Pengpei Zhu"],"categories":["astro-ph.GA"],"headline":"Dusty post-starburst galaxies still form stars at moderate rates","summary":"Researchers studied three post-starburst galaxies that are bright in carbon monoxide (CO), which suggests they have dense gas and could form stars. Using JWST's NIRSpec, they measured star formation rates of about 27 to 80 solar masses per year, finding these galaxies are not fully inactive but have lower star formation than their past peaks. The study also revealed significant dust obscures some star-forming regions, complicating earlier measurements. The findings suggest these galaxies are in a rapid transition phase toward shutting down star formation."},{"id":"2609.31306v1","title":"Benchmarking Attention for Tabular Foundation Models","abstract":"Tabular in-context learners such as TabPFN, Mitra, or ConTextTab rely on alternating row and column attention over 2D sequences of latent embeddings. These attention patterns differ markedly from the one-dimensional case in language models: row attention involves longer sequences while column attention operates on much shorter ones, and the strided memory layout of tabular data makes producing contiguous tensors costly. Moreover, the hidden dimensions used in current models are small compared to recent language models. Yet efficient attention has been studied mostly for one-dimensional sequences, leaving the two-dimensional tabular setting unexplored. To this end, we create a reproducible benchmarking setup and study the unique characteristics of tabular attention across several backends -- Torch SDPA (efficient and cuDNN), FlashAttention-2/3/4, and the inference-only backends vLLM and SageAttention -- measuring forward and backward throughput across realistic tabular shapes on three GPU generations (A100, H100, B200). We find that the optimal backend choice differs between column and row attention and varies across hardware as well as model specifics: While the FlashAttention implementations tailored for each GPU generation perform overall best, they are at times outperformed by CuDNN in the case of column attention at longer sequences with cross-over points depending on the head dimension. Among inference-only backends, SageAttention performs well for row attention and large sequences beyond 16\\,k rows. Our reproducible benchmark lays the foundation for future improvements to table-native attention. The self-contained benchmarking and evaluation code is openly available at: https://github.com/SAP-samples/tabular-attention-benchmark","published":"2026-09-25T14:14:48Z","updated":"2026-09-25T14:14:48Z","authors":["Maximilian Schambach","Clemens Biehl","Sam Thelin"],"categories":["cs.LG"],"headline":"Researchers Benchmark Attention Methods for Table Data Models","summary":"The researchers studied how different attention techniques work for models analyzing table-like data, which is arranged in rows and columns. They tested several software backends on different graphics hardware to see which handles row and column attention best, noting performance varies based on method, hardware, and data layout. Their findings highlight that no single approach is best in all cases, and they provide an open benchmark tool to help improve attention methods specifically designed for tabular data. The results come from controlled experiments and benchmarking software performance."},{"id":"2609.31305v1","title":"A solution to the paradox of youth in the Galactic centre","abstract":"The supermassive black hole at the Galactic centre, Sgr A*, is surrounded by an S cluster of young stars. These stars are predominantly B-type stars, but their origin remains uncertain. Unlike massive stars in the Galactic field, which exhibit a binary fraction of about $70\\%$, those in the S cluster show a significantly lower fraction. It is key to understand this discrepancy for constraining their formation. We show that the binary deficit arises naturally from dynamical interactions between stellar binaries and Sgr A* in an in situ formation scenario. Using K-amplitude data, we derived an observational binary fraction of $f_{\\mathrm{obs}} = 0.43 \\pm 0.09$ for the S-cluster stars. Our dynamical modelling predicts that interactions with Sgr A* cause $18\\%$ of the binaries to merge, $20\\%$ to be disrupted, and $62\\%$ to survive, yielding a theoretical binary fraction of $f_{\\mathrm{theo}} = 0.38 \\pm 0.10$. The predicted binary fraction with distance from Sgr A* is also consistent with independent observational constraints. These results show that the decline in binarity towards the Galactic centre is driven by stronger tidal forces, and they support an in situ origin of the S cluster in the same event that produced the Fermi bubbles.","published":"2026-09-25T14:14:46Z","updated":"2026-09-25T14:14:46Z","authors":["Rodrigo P. Silva","Alexandre C. M. Correia","Tjarda C. N. Boekholt","Paulo J. V. Garcia"],"categories":["astro-ph.GA","astro-ph.SR"],"headline":"Black Hole Forces Explain Fewer Binary Stars Near Galactic Center","summary":"Researchers studied young stars near the Milky Way’s central black hole, Sgr A*, which have fewer pairs of stars orbiting each other (binaries) compared to stars farther away. By analyzing star movements, they found that the black hole’s strong gravitational pull causes many binary stars to merge or break apart. Their model predicts a binary star fraction close to what is observed, supporting the idea that these stars formed right there near the black hole despite the intense forces."},{"id":"2609.31304v1","title":"Topology and Dirichlet spectrum of free boundary minimal and CMC surfaces","abstract":"We find bounds for the genus and for the number of boundary components of a surface that is either index one as a free boundary minimal surface or stable as a free boundary constant mean curvature surface in a Riemannian three-manifold with convex boundary and non-negative Ricci curvature, provided the first Dirichlet eigenvalue of its Jacobi operator is non-negative. As an application, we establish strong topological restrictions on such surfaces in compact strictly convex domains of three-dimensional normal homogeneous spaces.","published":"2026-09-25T14:14:34Z","updated":"2026-09-25T14:14:34Z","authors":["Alcides de Carvalho","Roney Santos"],"categories":["math.DG","math.AP"],"headline":"Researchers bound shapes of special surfaces in curved spaces","summary":"The researchers studied certain surfaces that meet the boundary of a curved three-dimensional space freely and have special curvature properties. They found limits on how complex these surfaces can be in terms of holes and boundary loops, assuming specific mathematical conditions hold. Their work applies to surfaces in spaces with smoothly curved, convex boundaries and reveals strong restrictions on the possible shapes these surfaces can have, based on mathematical analysis rather than physical experiments."},{"id":"2609.31303v1","title":"Geometric Moment Contraction for Stochastic Nesterov Acceleration","abstract":"We study geometric moment contraction (GMC) of the constant-parameter stochastic Nesterov recursion \\[ Y_k=Θ_k+β(Θ_k-Θ_{k-1}),\\qquad Θ_{k+1}=Y_k-γG(Y_k,X_{k+1}). \\] Under mean strong monotonicity and stochastic $L^p$ Lipschitz continuity, an explicit Perron comparison proves synchronous $L^p$ contraction when $βγL_p<(1-β)(1-q_{γ,p})$. This direct criterion includes infinite-variance gradients for $1<p<2$, but its small-step regime requires $β<μ/(μ+L_p)$. A complementary power-Lyapunov argument establishes a positive, generally much smaller, step-size interval for every fixed $β<1$ and every $p>1$, using only a finite $p$th gradient moment. At $p=2$, a simpler explicit certificate gives \\[ 0<γ<\\frac{2μ(1-β)^2}{L_2^2(1-β+2β^2)}. \\] Its quadratic high-momentum scaling is a limitation of the chosen metric, not a sharp stability boundary. We quantify this loss, provide a general mean-only quadratic $S$-procedure, and exploit endpoint Lyapunov inequalities under stronger samplewise sector information. Verified endpoint certificates can be orders of magnitude less conservative than the explicit metric.","published":"2026-09-25T14:14:14Z","updated":"2026-09-25T14:14:14Z","authors":["Wei Biao Wu"],"categories":["stat.ML","cs.LG"],"headline":"Researchers analyze stability conditions for stochastic Nesterov method","summary":"The researchers examine how a variant of the Nesterov accelerated method, used for optimization with randomness, behaves under certain mathematical conditions that ensure contraction or stability. They provide explicit formulas and criteria showing when the method reliably shrinks errors, considering cases with infinite-variance gradients and different step sizes. Their work includes new theoretical tools and comparisons of different stability conditions, but the results focus on mathematical criteria rather than practical outcomes or specific applications."},{"id":"2609.31302v1","title":"KCensus: Synthesizing Latency-Optimal Consensus Fast Paths (Extended Version)","abstract":"Strongly consistent geo-replication often relies on fast paths to reduce latency in the common case of no failures or contention. Existing fast-path schemes, however, are ad hoc and restrictive: each corresponds to a point in a broad design space shaped by network topology, workload, and latency objective, so no single scheme works best across settings. This paper looks at fast-path schemes from a new perspective, as mechanisms that spread knowledge about proposals. With this view, we identify a fundamental condition on the spread of knowledge for a fast-path scheme to work. We then introduce KCensus, a framework that turns this condition into an optimization problem, synthesizing new fast-path schemes that are optimal for a given setting. We use KCensus to build a geo-replicated key-value store and evaluate it across AWS regions. Our system outperforms competing protocols, with up to 16% lower average latency.","published":"2026-09-25T14:11:57Z","updated":"2026-09-25T14:11:57Z","authors":["Clément Burgelin","Antoine Murat","Gal Sela","Marcos K. Aguilera","Rachid Guerraoui"],"categories":["cs.DC"],"headline":"New Method Improves Speed of Data Agreement Across Networks","summary":"Researchers studied how computer systems agree on data changes quickly without errors, focusing on reducing wait times when conditions are smooth. They analyzed existing methods and found a key factor called 'knowledge spread' that affects performance. Using this insight, they created KCensus, a tool that designs the fastest agreement methods tailored to specific network setups. Testing in cloud environments showed KCensus can reduce waiting times by up to 16%, though the results depend on the tested conditions."},{"id":"2609.31301v1","title":"Beyond Approved Actions: Runtime Validation of Persistent Outcomes in Agent Workflows","abstract":"Large language model agents increasingly act on software systems, no longer merely generating text but also changing databases and online services. However, an approved database update may succeed yet leave an unapproved notification because execution can produce persistent effects beyond the requested change. Current safeguards can approve an action or record its aftermath, but without checking the persistent result before continuation, an unapproved outcome can be accepted as success and propagated to later steps. We present EffectMatch, a runtime that collects persistent changes within a controlled execution boundary and compares them with what the application approved for the current state and execution. The comparison governs commit and dependent execution. In comparative evaluation on 206 public business tasks, EffectMatch preserved all clean executions and prevented all tested incorrect commits. Six 20-run ablations exposed the failure caused by each removed mechanism, while 80 task-topology cases preserved truthful handoffs and blocked invalid continuation. Together, these results show that EffectMatch blocks the silent acceptance and downstream propagation of persistent outcomes inconsistent with application approval.","published":"2026-09-25T14:11:14Z","updated":"2026-09-25T14:11:14Z","authors":["Haoran Zhang","Hengtong Zhang","Zhiyu Liang","Yu Yan","Decheng Zuo","Hongzhi Wang"],"categories":["cs.SE","cs.AI"],"headline":"New Method Checks Software Changes to Prevent Unapproved Effects","summary":"Researchers studied how software systems controlled by AI can unintentionally cause changes beyond what was approved, such as unwanted notifications. They developed EffectMatch, a system that monitors and compares actual changes during execution with what was approved before allowing processes to continue. Tested on 206 tasks, EffectMatch blocked incorrect changes while allowing correct ones. The study shows how checking persistent outcomes at runtime helps avoid unapproved results being accepted or passed along. Results come from controlled experiments and evaluations."},{"id":"2609.31300v1","title":"New distance bounds for one-generator quasi-cyclic codes with applications to Hermitian LCD codes and entanglement-assisted quantum codes","abstract":"We introduce lower bounds on the minimum distance of one-generator quasi-cyclic codes of arbitrary index, obtained by puncturing and shortening the code at the level of its blocks. For balanced codes, these bounds form a nondecreasing chain whose first term is the Jensen bound. We apply these bounds to conduct a computer search and obtain many Hermitian LCD codes with good parameters. Among them there are six codes of length at most 30 which improve the corresponding lower bounds of Araya and Harada. These codes further give maximal-entanglement entanglement-assisted quantum codes, which we compare with the parameters currently recorded in the literature.","published":"2026-09-25T14:09:44Z","updated":"2026-09-25T14:09:44Z","authors":["Kanat Abdukhalikov","Duy Ho","Rasha M. Shat"],"categories":["cs.IT"],"headline":"New methods improve certain error-correcting code designs","summary":"The researchers developed new ways to estimate the minimum distance—a measure of error resistance—of a class of structured codes called one-generator quasi-cyclic codes. They used these estimates to find several Hermitian LCD codes with better parameters than previously known for code lengths up to 30. These improved codes can be used to create certain quantum codes with maximal entanglement, and the researchers compared their results to existing literature. Their findings are based on mathematical methods and computer searches."},{"id":"2609.31299v1","title":"A Mathematical Theory of Near-Field Super-Resolution","abstract":"Finite-aperture near-field sensing leads to a super-resolution geometry fundamentally different from the translation-invariant Fourier setting. In the Fresnel regime, wavefront curvature introduces a range-dependent quadratic aperture phase, so distinguishability is governed by incomplete quadratic exponential sums of the form \\( \\sum_{n=0}^{N_r-1} a_n e^{i(ω_1 n+ω_2 n^2)} \\), rather than by angular separation alone. We develop a deterministic recovery theory for sparse measures with ranges on a finite grid and continuous angles, and introduce a support-uniform quadratic-phase aperture criterion replacing classical minimum separation. Under this criterion, total-variation minimization exactly recovers every sparse measure in the admissible support class, uniformly over all nonzero complex amplitudes. The proof develops nonasymptotic support-uniform bounds for finite quadratic sums and combines them with a gauged Hermite dual certificate controlling interpolation, local curvature, and off-support leakage. We further construct a finite-harmonic Bessel-Vandermonde lift with explicit truncation error. In the far-field limit, the quadratic phase disappears and the theory reduces to Fourier-type angular super-resolution.","published":"2026-09-25T14:09:01Z","updated":"2026-09-25T14:09:01Z","authors":["Sajad Daei","Gabor Fodor","Mikael Skoglund"],"categories":["cs.IT","math.NT"],"headline":"New Math Theory Improves Understanding of Near-Field Imaging Limits","summary":"Researchers studied near-field sensing, where waves curve differently than in traditional far-field cases, making resolution depend on more complex factors than simple angles. They developed a mathematical framework to exactly recover sparse signals under this new setting using a criteria based on wave phase, rather than just distance between points. Their results are theoretical and provide recovery guarantees for signals arranged on grids with continuous angles, explaining how classical far-field results are a special case of their theory."},{"id":"2609.31298v1","title":"UniAR: A Unified Framework for Autism Recognition Enhanced by Multi-View Prompt Learning","abstract":"Autism Spectrum Disorder (ASD) is a complex neurodevelopmental disorder for which early and accurate diagnosis is critical to improving long-term developmental outcomes. However, existing ASD recognition methods are often constrained by the scarcity of diagnostic text data, forcing them to rely mainly on visual analysis and limiting their ability to model clinically meaningful semantic reasoning. To address this challenge, we propose UniAR, a unified framework enhanced by multi-granularity prompt learning for robust ASD recognition under heterogeneous data variations. Specifically, UniAR leverages a large multimodal model to generate hierarchical diagnostic descriptions at the word, phrase, and sentence levels, compensating for the lack of paired clinical reports. To align the generated semantics with visual evidence, we further design a Mixture-of-Experts-based Multi-Scale Alignment Module, which dynamically matches vector-quantized visual prototypes with semantic representations at corresponding granularities. Extensive experiments on four benchmarks covering brain MRI and facial expression scenarios show that UniAR consistently outperforms existing state-of-the-art methods, achieving average accuracies of 75.9\\% on MRI benchmarks and 91.6\\% on facial benchmarks, while improving average Accuracy on MRI benchmarks by 1.5 percentage points and average Accuracy on facial benchmarks by 1.2 percentage points over baselines. These results demonstrate that UniAR offers a robust and interpretable framework for ASD screening under semantic scarcity.","published":"2026-09-25T14:07:52Z","updated":"2026-09-25T14:07:52Z","authors":["Lei Xin","Zeheng Wang","Jiayin Zhu","Shihong Huang","Fanhu Zeng","Changjiang Jiang","Dengbo He","Yutao Yue","Zhenglun Kong"],"categories":["cs.CV","cs.AI"],"headline":"New AI Method Improves Autism Recognition from Brain and Facial Data","summary":"Autism Spectrum Disorder (ASD) diagnosis is important but existing methods often lack enough clinical text data and focus mainly on images. Researchers developed UniAR, a system that combines visual data with generated diagnostic descriptions at multiple levels to better interpret signs of ASD. They tested UniAR on brain MRI and facial expression data, finding it improved accuracy over current methods. The findings come from computational experiments and suggest UniAR could provide more reliable and understandable ASD screening."},{"id":"2609.31297v1","title":"Who Belongs Together? Topical and Social Structure in Bluesky Starter Packs","abstract":"Bluesky starter packs are human curated collections of accounts and feeds aimed at helping users discover new communities, especially during onboarding. Prior research has examined their effects on platform growth and account visibility. However, whether these packs capture meaningful topical and social structure remains unanswered. We study 2718 active starter packs containing approximately 144K accounts, combining pack metadata, follow graphs, user post histories and inferred demographic attributes to answer this question. We assign starter packs to 17 topical categories, and examine semantic coherence, curator--member resemblance, shared audiences, and follow reciprocity. We found that pack members are closely aligned with both their fellow pack members and their pack's description compared to accounts included in other packs of the same topic, indicating topical specificity beyond the broader topical categories. We also identify semantic resemblance between curators and their pack members, while demographic resemblance varies by attribute and topic. Network analyses show higher follow reciprocity among members of the same pack and between members and curators than between a pack member and either another member from the same broader topic or a nonmember. Members also share substantial portions of their follower networks with fellow pack members. Despite this coherence, a typical member's follower follows only a small fraction of other members in the pack, indicating limited audience coverage. These results suggest that starter packs capture coherent topical and social groups while leaving opportunities for further account discovery. More broadly, results show human-curation can be useful to capture relevant communities and connections, and platforms could integrate human-curation into recommendation systems as a complement to algorithmic systems to expand opportunities for discovery.","published":"2026-09-25T14:07:19Z","updated":"2026-09-25T14:07:19Z","authors":["Sima Adleyba","Onur Varol"],"categories":["cs.SI","cs.CY"],"headline":"Study Finds Bluesky Starter Packs Reflect Real Social and Topic Groups","summary":"Researchers analyzed over 2,700 Bluesky starter packs containing about 144,000 accounts to see if these human-curated collections represent meaningful social and topical groups. They found that pack members closely align with each other and with their pack's theme, showing both topical and social coherence. However, audience overlap is limited, meaning follower networks only partially cover the pack members. The study suggests curated starter packs capture relevant communities but leave room for discovering more accounts. These findings come from analyzing metadata, user posts, demographics, and network connections."},{"id":"2609.31296v1","title":"Improving predictions for the $pp \\to t\\bar{t}W^+$ process at the LHC with the MINLO method","abstract":"We compare the standard NLO and $\\texttt{MiNLO}$ approaches for the full off-shell $pp \\to t\\bar{t}W^+ + X$ process at the LHC, at both the integrated and differential cross-section levels. Contrary to NLO, in the $\\texttt{MiNLO}$ approach, which is now also available within the HELAC-NLO framework, renormalization and factorization scales are dynamically extracted and Sudakov form factors are incorporated. Results are particularly presented for the $pp \\to t\\bar{t}W^+j$ process at NLO accuracy in perturbative QCD using two different scale choices. Merged predictions up to two jets to improve the overall modeling of the full off-shell $pp \\to t\\bar{t}W^+ + X$ process are also discussed.","published":"2026-09-25T14:07:18Z","updated":"2026-09-25T14:07:18Z","authors":["Nikolaos Dimitrakopoulos"],"categories":["hep-ph"],"headline":"New method improves predictions for top quark and W boson production at LHC","summary":"Researchers compared standard and improved calculation methods for predicting how often certain particles (top quarks and W bosons) are produced together in proton collisions at the Large Hadron Collider. They tested a new approach called MiNLO that adjusts calculation scales more dynamically and includes additional effects, leading to better modeling of the process, especially when extra jets (streams of particles) are involved. The study supports that MiNLO offers improved predictions but is based on theoretical simulations rather than experimental data."},{"id":"2609.31295v1","title":"On harmonic centers of graphs","abstract":"The harmonic centrality of a vertex $v$ in a graph $G = (V,E)$ is the sum of reciprocals of distances of vertices of $G$ from $v$. The vertices of $G$ which have the maximum (minimum) harmonic centrality form the harmonic center (or periphery, resp.) of $G$. We study harmonic centers of graphs and their localization in graph blocks, presenting sufficient conditions for graphs (in terms of diameter or number of edges) to have those centers contained in a single block; in addition, we show that each connected graph is the harmonic center as well as harmonic periphery of some graphs.","published":"2026-09-25T14:05:56Z","updated":"2026-09-25T14:05:56Z","authors":["Tomáš Madaras","Matúš Paralič"],"categories":["math.CO"],"headline":"Researchers Explore Central Points in Network Structures","summary":"This research examines harmonic centrality, which measures how close a network point is to all others by adding up reciprocal distances. The researchers study where the most central points, called harmonic centers, appear within parts of the network and identify conditions that keep these centers within one section. They also show that any connected network can be the most or least central part of some other network. Their findings are theoretical and focus on properties of graph structures."},{"id":"2609.31294v1","title":"On Shnirelman's inequality in 2D: Irreversible behavior and Euler-Lagrange variational formulation","abstract":"In this paper we prove a version of Shnirelman's inequality in the two-dimensional square $[0,1]^2$. More precisely, given a target fluid configuration $f$ (a volume-preserving diffeomorphism) we construct a divergence-free velocity field $v\\in L^q_tL^p_x$ whose flow connects $f$ to the identity and whose $L^q_{t} L^p_x$-norm can effectively be bounded by the $L^p$-difference of $f$ and $id$. However, higher regularity of the vector field might be affected. The main difference with the higher dimensional case is that, because of the topological obstructions of dimension $ν=2$, we have to allow that different fluid trajectories 'intersect' in space. We also observe that this choice leads to the emergence of irreversible behaviors in the Eulerian dynamics.","published":"2026-09-25T14:05:00Z","updated":"2026-09-25T14:05:00Z","authors":["Stefan Schiffer","Martina Zizza"],"categories":["math.AP","math.OC"],"headline":"Researchers Explore Fluid Flow Inequality and Irreversible Behavior in 2D","summary":"The researchers studied a mathematical inequality related to fluid movement in a two-dimensional square area. They constructed a fluid velocity field connecting a target fluid state to the original state, showing that certain mathematical bounds hold. Their work reveals that in two dimensions, fluid paths can intersect, which differs from higher dimensions and leads to irreversible behavior in the fluid motion. These findings are theoretical and based on mathematical proofs rather than experiments."},{"id":"2609.31293v1","title":"Why Alzheimer's Speech Screening Fails to Generalize: Bridging the Deployment Gap via Cross-Corpus Evidence Anchoring","abstract":"Speech-based screening is a promising, non-invasive approach for detecting Alzheimer's disease and related cognitive risks. However, models trained on a single domain often generalize poorly to unseen languages, tasks, or recording protocols. This paper investigates this deployment gap using a leave-one-corpus-out evaluation across four distinct datasets. Among 70 interpretable speech and language features, 59 exhibit direction conflicts between healthy control and cognitive risk groups across corpora, with pause, silence, and speech rate showing high protocol sensitivity. Furthermore, while the XLM-R text baseline achieves strong average performance, its Area Under the ROC Curve (AUC) drops to 0.520 on the weakest held-out domain. A standard GroupDRO baseline reaches a 0.766 mean speaker AUC and a 0.504 worst-domain AUC under the same protocol. To address this, we propose a fusion method that integrates XLM-R text baseline scores with evidence anchors selected during training. Balanced fusion achieves a 0.785 mean speaker AUC, while anchor-heavy fusion raises the worst-case speaker AUC to 0.615. This work highlights the need to audit feature transferability and report worst-case domain robustness in cognitive speech screening.","published":"2026-09-25T14:02:52Z","updated":"2026-09-25T14:02:52Z","authors":["Zijian Lu","Sizhe Liu","Yin Zhang","Jixuan Deng","Xinrong Lin","Xinchen Yuan","Chicheng Jin","Yiping Zuo","Yuanchao Li"],"categories":["eess.AS","cs.CL","cs.SD"],"headline":"Improving Alzheimer's Speech Tests Across Different Settings","summary":"Researchers studied why speech-based tests for Alzheimer's often fail when used in new languages or settings. They analyzed multiple datasets and found many speech features behave inconsistently across tests, causing poor performance when models are applied elsewhere. To improve reliability, the team developed a new method combining text analysis with stable 'evidence anchors' from training data, which enhanced detection accuracy in varied conditions. Their findings emphasize checking how well speech features work across different groups and reporting results in the least favorable cases."},{"id":"2609.31292v1","title":"Dynamic Sampling for Telemetry in Microservices: A Reinforcement Learning and Entropy-Based Approach","abstract":"Microservices architectures are increasingly deployed in cloud-based distributed environments, making application development and maintenance more dynamic, but also increasing the complexity of troubleshooting and observability. Distributed tracing tools are therefore essential for request analysis and debugging, despite introducing additional overhead that can be amplified by excessive and inefficient data collection. This article proposes RADAR (Reinforcement Learning Agent for Dynamic And Relevant trace sampling), an agent that combines reinforcement learning with a data entropy assessment to achieve more efficient capture of traces relevant to system monitoring, based on the OpenTelemetry standard. RADAR tests different sampling rules to discover which combination is most efficient. A test environment simulating a minimalist online store with several microservices distributed across a Kubernetes cluster served as the basis for the experiments, which evaluated the agent's convergence and the system's performance in terms of resource consumption and collected data quality. Results showed that RADAR reduced network bandwidth consumption by 97.4% and CPU usage by 99.0% compared to full data collection, also outperforming a fixed-rate sampling baseline. Beyond these resource savings, the approach preserved observability of critical scenarios, retaining approximately 85.6% of rare trace patterns and increasing the average entropy of the stored information by approximately 25%, validating the feasibility of using entropy to orchestrate telemetry autonomously and efficiently.","published":"2026-09-25T14:02:46Z","updated":"2026-09-25T14:02:46Z","authors":["Renan Martins Alves","Jéferson Campos Nobre","Juliano Araujo Wickboldt"],"categories":["cs.NI"],"headline":"Smart Trace Sampling Reduces Microservices Monitoring Costs","summary":"Researchers developed RADAR, a system using reinforcement learning and data entropy to improve how monitoring data is collected from complex cloud-based microservices. By testing different sampling strategies in a simulated online store environment, they showed RADAR drastically cuts network and CPU use compared to collecting all data or simple fixed-rate sampling. Importantly, it still captures most unusual events and maintains useful information for troubleshooting, supporting the idea that entropy can guide efficient data collection. This study is based on simulation results and focuses on system performance and data quality."},{"id":"2609.31291v1","title":"Softmax Reparameterization for Output-Head Quantization","abstract":"Large vocabularies make output heads a substantial inference cost in small language models. We propose softmax reparameterization, a post-training method that selects a functionally equivalent output head before quantization. The method subtracts a scalar multiple of the vocabulary-row mean from every output row and selects the coefficient by validation KL separately for RTN, activation-weighted MSE, and full-Hessian GPTQ. This one-dimensional search includes the original head and fixed mean-centering, preserves the full-precision softmax distribution, and leaves the trained decoder unchanged; a rank-one correction handles nonlinear logit paths such as soft-capping. Across seven heads, W4 gains concentrate where baseline quantization substantially distorts predictions: on Phi-4-mini, AW-MSE KL falls from 0.936 to 0.256. The gains survive stronger GPTQ calibration and remain complementary to exact per-channel scaling and affine quantization. Across four heads and three W4 quantizers, frozen WikiText-selected coefficients also transfer to C4 and OpenWebMath, outperforming mean-centering in all 18 comparisons where the frozen coefficient differs from $1$ and matching it in the remaining six. At W2, used as a compression stress test, benefits broaden across nearly the full model--quantizer matrix. Matched residual analysis shows that improved fidelity can accompany greater logit reconstruction error while reducing the residual's Fisher-weighted cost. For shift-compatible heads, reparameterization adds no inference operation and preserves packed W4 execution: with the decoder held in BF16, quantizing the Phi output head reduces batch-one generation latency by 10.8% relative to the BF16-head baseline.","published":"2026-09-25T14:01:25Z","updated":"2026-09-25T14:01:25Z","authors":["Asim Kadav","Christian Flores","Chirag Arora","Varun Kotte","Hongbo Zheng","Lan Yan","Priya Shanmugasundaram","Tracy Holloway King"],"categories":["cs.LG","cs.AI"],"headline":"Improving small language model output efficiency using softmax tweaks","summary":"The researchers address the high computation cost of output layers in small language models with large vocabularies. They propose a method that adjusts the output layer before compressing it, aiming to keep prediction quality while reducing size. Their tests show improved prediction accuracy and reduced processing time in simulated scenarios, though results are based on specific model settings and compression tests."},{"id":"2609.31290v1","title":"An Optimal Structure for All-Pairs Nearest Mincuts and Sensitivity Oracles for Edge Insertions","abstract":"Given an undirected weighted graph $G=(V,E)$ on $n$ vertices, the classical Gomory-Hu tree of $G$ is a structure that encodes an arbitrary minimum $s,t$-cut for every $s,t\\in V$ using just $O(n)$ space. In this work, we ask whether the same compactness is achievable for the natural and structured family of all-pairs \\textit{nearest minimum cuts}. The nearest minimum $(s,t)$-cut is the unique inclusion-wise minimal one among all minimum $s,t$-cuts containing $s$. This family has proven useful in a wide range of applications, including fault-tolerant reachability, minimum cut sensitivity oracles, cactus representations, and fast Gomory-Hu tree constructions. Despite its fundamental role, no subquadratic space representation is known for them to date. The $O(n)$ space representations are known only in single-source settings, where given a source $s$, one can report the nearest minimum $(t,s)$-cut for any $t\\in V\\setminus \\{s\\}$. We close this gap by presenting the first optimal space representation of all-pairs nearest minimum cuts, providing a natural analogue of the Gomory-Hu tree. Our main result is an $O(n)$ space structure that encodes the nearest minimum cut between every pair of vertices. Furthermore, given any pair $s,t\\in V$, it can report the nearest minimum $s,t$-cut in $O(n)$ time. Both bounds match those of the Gomory-Hu tree and are worst-case optimal. As an application, we design an all-pairs minimum cut sensitivity oracle for edge insertion: a data structure that occupies $O(n)$ space and, given any edge $e$, can determine for all pairs $s,t\\in V$ whether the minimum $s,t$-cut value increases upon insertion of $e$ in $O(n^2)$ total time. Existing insertion sensitivity oracles were either limited to the single-source setting or used $O(n^2)$ space for all-pairs [Baswana, Gupta, and Knollmann, Algorithmica'22; Baswana and Pandey, SODA'22].","published":"2026-09-25T14:00:23Z","updated":"2026-09-25T14:00:23Z","authors":["Koustav Bhanja","Yotam Kenneth-Mordoch","Asaf Petruschka"],"categories":["cs.DS"],"headline":"Compact structure encodes nearest minimum cuts for all vertex pairs","summary":"Researchers developed a new data structure that uses only linear space to represent the smallest cuts separating every pair of vertices in a weighted graph. This builds on a classical method known as the Gomory-Hu tree but focuses on nearest minimum cuts, which are important in applications like network reliability. Their structure can report these cuts efficiently and matches known optimal limits. They also created a tool to quickly check how adding an edge affects all pairs' minimum cuts, improving on previous methods that required more space or handled fewer cases."},{"id":"2609.31289v1","title":"The word problem in verbal products of groups","abstract":"We prove that nilpotent and metabelian products of groups with decidable word problem have decidable word problem. In contrast, we exhibit families of groups with decidable word problem for which their solvable product have undecidable word problem. In doing so, we exhibit hyperbolic groups with undecidable membership problem in its derived subgroups.","published":"2026-09-25T14:00:02Z","updated":"2026-09-25T14:00:02Z","authors":["Javier Brude","Román Sasyk"],"categories":["math.GR"],"headline":"Researchers study word problem in special group combinations","summary":"The researchers investigated the 'word problem,' which asks whether certain expressions in a mathematical group can be simplified to a basic form. They proved that for two specific types of combined groups, called nilpotent and metabelian products, if the original groups had a solvable word problem, the combined groups also did. However, they found that for another type called solvable products, this does not always hold. Their work highlights limitations in solving these problems and includes examples involving complicated groups known as hyperbolic groups."},{"id":"2609.31288v1","title":"When the Model Retires: An Empirical Study of LLM Migration in Open-Source Applications","abstract":"Applications built on commercial large language model (LLM) APIs depend on model versions that providers retire on their own schedule, with notice periods ranging from one year to two weeks. We ask what actually happens to applications when a model is retired. We mine GitHub for commits that migrate away from officially deprecated models and endpoints of OpenAI, Anthropic, and Google, matching each commit to the provider's published announcement and shutdown dates. From 22,555 commits in 17,703 non-fork repositories (2024-2026), 5,139 are matched to an official event; two independent coders validated a stratified sample of 300 (kappa = 0.89-0.95), and we reweight all estimates by their labels. We find that an estimated 82% (95% CI 79-84) of migrations away from retired models were committed after the shutdown date - after the application had started failing - regardless of repository popularity, prior retirement experience, or the presence of a provider-abstraction layer. The share tracks the provider's notice policy: 89% for Anthropic's 60-114-day notices versus 13% for OpenAI's one-year Assistants API notice, and each e-fold increase in notice length reduces the odds of post-shutdown migration by about three quarters. Model identifiers are hard-coded in 94% of migrating applications, migration effort scales from a median of 6 added lines for prompt-only applications to nearly 700 for fine-tuned ones, and only 8% of migrations switch provider. We release the dataset and pipeline and discuss implications for deprecation policy, dependency-risk assessment of LLM products, and tooling.","published":"2026-09-25T13:57:50Z","updated":"2026-09-25T13:57:50Z","authors":["Hyungjin Lukas Kim"],"categories":["cs.SE"],"headline":"Most Apps Update AI Models Only After They Stop Working","summary":"Researchers studied how open-source applications respond when their AI language models are retired by providers like OpenAI and Google. By analyzing over 22,000 software updates, they found that about 82% of migrations to new models happen after the old models have already stopped working, regardless of app popularity or preparation. Longer provider notice periods reduced this delay. The study highlights challenges developers face updating hard-coded AI models and suggests implications for better planning and tool support."},{"id":"2609.31287v1","title":"H2 emission in the Orion Bar, NGC7023 and the Horsehead Nebula observed with the James Webb Space Telescope","abstract":"Molecular hydrogen (H2), the most abundant molecule in the interstellar medium, is a handy tool to study Photodissociation Regions. The JWST, with its high spatial resolution, sensitivity, and wavelength coverage, provides unique access to the detection and spatial morphology of the H2 rotational and rovibrational lines. Our goal is to analyze H2 line emission detected with JWST in several PDRs (the Orion Bar, NGC7023, and the Horsehead Nebula) to constrain the physical structure of dissociation front, and compare the impact of the ultraviolet field and the gas density across different regions. The analysis of H2 spatial morphology reveals similar filamentary structures across all regions. Spatial shifts between H2 lines (0.5\") are observed and linked to differences in excitation mechanisms (thermalized vs FUV-pumped lines) and the temperature gradient. Despite differences in incident irradiation conditions, analysis of H2 rotational excitation yields high gas temperatures that are similar across the three PDRs. This is expected for a certain regime of excited PDRs (such as Orion and NGC7023), where the G0 at the DFs and the density are similar. However, it is surprising for less excited PDRs like the Horsehead Nebula. OPRs in the rotational and rovibrational levels differ across the three PDRs. We also used H2 rovibrational lines and the 0-0 S(3) line to estimate the visual extinction across the field of view, yielding markedly different results that constrain both the geometry and the grain composition. Due to the complex 3D geometry, H2 emission is observed throughout the FOV and is dominated by the thin UV illuminated layer, making H2 an unreliable tracer of gas temperature in deeper and coolers molecular layers of the PDR. Therefore, H2 emission is not sufficient to trace the possible temperature and density gradient across PDRs and must therefore be associated with other tracers.","published":"2026-09-25T13:57:20Z","updated":"2026-09-25T13:57:20Z","authors":["M. Zannese","A. Sidhu","A. Guerras","M. Pound","M. Wolfire","A. G. G. M. Tielens","E. Peeters","E. Habart","A. Abergel","M. Baes","O. Berné","C. Boersma","E. Bron","J. Cami","R. Chown","E. Dartois","P. Dell'Ova","J. R. Goicoechea","K. D. Gordon","P. Guillard","O. Kannavou","K. Misselt","N. Monnier","A. Noriega-Crespo","T. Onaka","D. Van de Putte","L. Verstraete","A. N. Witt"],"categories":["astro-ph.GA"],"headline":"Researchers Study Hydrogen Emission in Space with JWST","summary":"Molecular hydrogen is common in space and helps study regions affected by ultraviolet light. Using the James Webb Space Telescope, researchers examined hydrogen emissions in three space regions and found similar structures and temperatures despite different conditions. They discovered that while hydrogen emissions reveal some details, they alone cannot fully measure temperature and density changes in these regions, suggesting additional tools are needed to understand them better."},{"id":"2609.31286v1","title":"G2MAF: Test-Time Gradient Guidance for Multi-Agent Flow Policies","abstract":"Offline multi-agent reinforcement learning (MARL) learns cooperative policies from fixed datasets without further environment interaction and a learned policy is frozen at deployment. Such a frozen policy typically proposes a single joint action and executes it directly at deployment time. However, this one-shot deployment often commits to a suboptimal proposal, even when better nearby alternatives remain consistent with the behavior data. To address this issue, we propose Gradient Guided Multi Agent Flow (G2MAF), a refinement framework for optimizing joint policies at test-time. G2MAF applies one globally normalized, projected critic gradient to guide and coordinate all agents' corrections while keeping the action both feasible and close to the frozen policy proposal. Across 24 MPE and SMAC settings, its canonical variant improves 20 frozen settings, with mean relative gains of 9.2% on MPE and 8.9% on SMAC, with model inference latency increased by about 6% only.","published":"2026-09-25T13:57:02Z","updated":"2026-09-25T13:57:02Z","authors":["Guowei Zou","Haitao Wang","Guoxin Wang","Zhiquan Chen","Beiwen Zhang","Guojie Wang","Hejun Wu"],"categories":["cs.AI"],"headline":"Improving Multi-Agent Decisions After Training with Gradient Guidance","summary":"Researchers studied a way to enhance decisions made by multiple cooperating agents trained using fixed data, without further learning. They introduced G2MAF, a method that refines the agents' joint actions during deployment by using gradient guidance to subtly adjust actions for better coordination. In tests across 24 scenarios, their approach improved outcomes compared to fixed trained policies, though the results come from controlled simulations and include a small increase in decision-making time."},{"id":"2609.31285v1","title":"MoTop: Motion-Topological Model For Micro AU Detection","abstract":"Facial micro-expressions are spontaneous, brief, and subtle facial movements that reveal suppressed emotions in high-stakes environments. In contrast to classic expression analysis, detecting action unit (AU) yields a finer representation of facial movements, serving as a preliminary step before defining expression classes and other downstream tasks. Therefore, it represents a crucial upstream task in facial analysis, and improving an AU detection module increases the precision of facial analysis. Despite that, detecting AU is challenging because of the constrictive nature of the AU activation regions, leading to confusion among different AUs known as AU ambiguity. To model the fine-scale changes, we propose \\textbf{MoTop}, a motion-topological model that is augmented with a learnable motion context, yielding regional soft guidance for facial activity, followed by facial landmarks that capture the fine-scale topological changes of micro AUs. To increase the micro facial landmark representations, we amplify the encoded facial landmark transitions via linear extrapolation, thereby increasing the spatial proximity of landmarks and enhancing the low-intensity landmark dynamics. In addition, we design anatomical facial clusters that enhance the hierarchical representation, facilitating multi-scale modelling of facial geometry and improving micro-topological representations. With these contributions, we have achieved state-of-the-art performance on the CD6ME protocol for the micro AU detection task.","published":"2026-09-25T13:56:09Z","updated":"2026-09-25T13:56:09Z","authors":["Huai-Qian Khor","Mengting Wei","Yante Li","Chu Kiong Loo","Guoying Zhao"],"categories":["cs.CV"],"headline":"New Model Improves Detection of Subtle Facial Movements","summary":"Researchers studied tiny facial movements called micro-expressions, which reveal hidden emotions but are hard to detect accurately. They developed a model named MoTop that tracks small facial muscle movements and their relationships using motion cues and facial landmark points, enhancing subtle detail detection. Their approach, tested on a specialized dataset, shows improved accuracy in identifying these fine facial activity patterns, though findings are based on this specific test setup and focus on improving preliminary facial analysis steps."},{"id":"2609.31284v1","title":"Rate-Preserving Shrinking-Support Gaussian Process Prediction","abstract":"Gaussian process prediction is a central tool in spatial statistics, but standard implementations require dense matrix operations that become prohibitive for large datasets. We propose a scale-adjusted compactly supported working correlation for Gaussian process prediction, using the generalized Wendland family with a support radius $φ_n$ that is allowed to decrease with the sample size $n$. Under fixed-domain asymptotics with quasi-uniform designs, a fixed support radius does not yield asymptotic sparsity, whereas a shrinking support radius can make the covariance matrix sparse. We show that $φ_n$ and the regularization parameter can be jointly chosen so that the resulting predictor preserves the optimal integrated mean squared prediction error rate while reducing the number of nonzero covariance entries and the associated sparse matrix-vector multiplication cost. We also establish an analogous rate-preserving sparsification result for kernel ridge regression. Simulations and an ERA5 temperature application show that the proposed method achieves competitive prediction accuracy while retaining the computational efficiency provided by sparse linear algebra.","published":"2026-09-25T13:54:39Z","updated":"2026-09-25T13:54:39Z","authors":["Xiaopeng Xiang","Wenlin Dai","Marc G. Genton","Wenjia Wang"],"categories":["stat.ME"],"headline":"Faster Gaussian Process Predictions for Large Spatial Data","summary":"Researchers address the problem that standard Gaussian process prediction methods become too slow for large datasets because of complex calculations. They developed a new approach that adjusts how data points influence each other, allowing the method to use less computational effort while keeping prediction accuracy. Their findings, supported by mathematical analysis and simulations, suggest this technique can predict spatial data efficiently without losing accuracy, although results are based on theoretical models and specific data examples."},{"id":"2609.31283v1","title":"MATISSE-py: a high-level Python interface for efficient, automated VLTI/MATISSE data reduction","abstract":"The Multi AperTure mid-Infrared SpectroScopic Experiment (MATISSE) at the Very Large Telescope Interferometer (VLTI) delivers rich mid-infrared interferometric datasets that are reduced through a robust, multi-stage ESO pipeline. In its native form, however, the reduction relies on command-line calls to esorex, hand-edited set-of-files (SOF) and configuration files, and manual iterations that are difficult to automate, reproduce, or integrate into modern Python workflows. We present MATISSE-py, a modular Python interface that wraps the official ESO esorex recipes behind a single, user-friendly matisse command. It automates discovery, reduction, calibration, and inspection of the data, and ships as a tested, open-source package built on industry-standard practices. Beyond convenience, MATISSE-py introduces science-grade improvements developed over seven years of MATISSE operations: visibility- and photometry-factor (V-factor and P-factor) corrections derived from the GRAVITY fringe tracker to mitigate residual visibility losses, per-baseline data filtering, beam-commuting-device (BCD) \"magic-number\" corrections, and an absolute spectrophotometric flux calibration using state-of-the-art reference spectra. On faint targets observed under degraded conditions, the V-factor correction recovers up to a few percent of lost visibility and reduces the statistical uncertainties by nearly 20%, while yielding calibrator diameters that are consistent across BCD modes. MATISSE-py thus consolidates the consortium's expertise into a maintainable package and shortens the reduction-to-science loop for the community.","published":"2026-09-25T13:54:21Z","updated":"2026-09-25T13:54:21Z","authors":["A. Soulain","A. Matter","P. Berio","F. Millour","B. Lopez","R. Petrov","S. Lagarde"],"categories":["astro-ph.IM","astro-ph.GA","astro-ph.SR"],"headline":"New Python Tool Simplifies Mid-Infrared Telescope Data Processing","summary":"Researchers developed MATISSE-py, a Python interface that automates and improves data reduction for the MATISSE instrument at the Very Large Telescope Interferometer. This tool replaces complex manual steps with a single command, adding corrections to enhance data quality especially for faint targets under poor conditions. The findings show that using MATISSE-py can recover some lost data and reduce uncertainties, helping astronomers analyze observations more efficiently. The reported benefits are based on the researchers' experience over seven years of MATISSE operations."},{"id":"2609.31282v1","title":"Resource-Optimized and Energy-Aware Agentic AI Framework Anchored on Blockchain for Secure Software Supply Chains","abstract":"This paper proposes a blockchain-backed agentic security framework designed to safeguard the complete software development lifecycle (SDLC) while also securing the agentic AI components responsible for monitoring it. The framework coordinates a set of specialised security agents, covering source integrity, dependency and SBOM analysis, CI configura tion auditing, artifact verification, and runtime policy evaluation, each supported by a large language model (LLM) that interprets artefacts, reasons over tool outputs, and produces structured security reports. To ensure agent trustworthiness, every agent generates a cryptographically signed attestation that is recorded in a permissioned blockchain via smart contracts, including an agent registry, an immutable attestation log, and an enforceable release-policy module. Communication among agents and with blockchain nodes is secured using a consortium-operated certificate authority, ensuring authenticated and tamper-resistant interactions. A detailed use-case and sequence flow demonstrate how a source code security agent performs analysis, anchors its attestation on-chain, and triggers a verifiable allow/block deployment decision. The proposed framework of fers decentralised integrity transparent provenance, uninterrupted security assurance and a generalisable architecture to incorporate the agentic AI into the modern software supply chain security.","published":"2026-09-25T13:53:50Z","updated":"2026-09-25T13:53:50Z","authors":["Toqeer Ali Syed","Asadullah Abdullah Khan"],"categories":["cs.CR","cs.AI"],"headline":"New Blockchain-Based AI System Aims to Secure Software Development","summary":"Researchers propose a security framework that uses AI agents and blockchain technology to protect the entire software development process. Each AI agent checks different security aspects and creates verified reports stored on a secure blockchain, helping ensure trustworthy and tamper-proof monitoring. The framework offers transparent, continuous security but is described through a detailed example rather than real-world testing."},{"id":"2609.31281v1","title":"MA-WAM: Multi-Agent World-Action Model for Test-Time Planning","abstract":"Multi-agent cooperative tasks require different agents to execute a joint action simultaneously, and each agent's action affects both the observations and responses of the other agents. Hence, a world model is needed to predict the team return resulting from the joint actions of all agents. A naive extension directly applies a single-agent world model to each agent's action when predicting the team return step by step. However, such an extension fails to capture the dependencies among the simultaneous actions of multiple agents. We propose Multi-Agent World-Action Model (MA-WAM), a test-time planning framework that enables a frozen multi-agent flow policy to evaluate futures of candidate joint actions. To our knowledge, MA-WAM is the first test-time world-model planner for multi-agent flow policies. MA-WAM predicts the consequences of each joint action according to cross-agent dependencies and enables efficient candidate scoring. Across 30 offline multi-agent reinforcement learning (MARL) settings on MAMuJoCo, SMAC, and MPE, MA-WAM achieves mean relative gains of 22.0% over direct execution and 25.6% over uniform action selection. Under the standard evaluation protocol on an A100 GPU, MA-WAM adds 12.1 ms, accounting for 2.5% of the measured generation-and-scoring time.","published":"2026-09-25T13:52:39Z","updated":"2026-09-25T13:52:39Z","authors":["Guowei Zou","Haitao Wang","Guoxin Wang","Beiwen Zhang","Zhiquan Chen","Guojie Wang","Hejun Wu"],"categories":["cs.AI"],"headline":"New Model Helps Multiple Agents Plan Better Together","summary":"Researchers developed a model called MA-WAM to help teams of agents predict the outcomes of their combined actions. Unlike simpler methods, MA-WAM considers how agents' simultaneous actions affect each other to better estimate future results. Tested in simulated cooperative tasks, the model improved team performance by about 22% compared to direct action use. These results come from computer simulations and show potential for more coordinated decision-making in multi-agent systems."},{"id":"2609.31280v1","title":"Quantum environment afterglow from broadband excitation spectroscopy in superconducting qubits","abstract":"Understanding the qubit environment is central to fault-tolerant superconducting quantum computation. Characterization typically relies on relaxation, leaving excitation largely underexplored. Here, we access this information with time-resolved broadband excitation spectroscopy. The resulting qubit excitation spectrum reveals a highly structured landscape, interspersed with cold regions. Combined with postselection, this technique enables full reconstruction of the noise power spectral density (PSD) and separates quantum from classical noise. With feed-forward, it exposes long-lived two-level-systems (TLSs), whose relaxation times span tens of microseconds to milliseconds - uncovering an intrinsic link between the qubit-TLS coupling and TLS relaxation. The data suggest that the long-lived TLSs are intrinsic to the qubit environment, and can cause excitation that lasts for many qubit operation cycles. Consequently, the environment retains a memory of prior dynamics, and properties like gate fidelity become non-Markovian and protocol-dependent. The presented approach enables identifying and bypassing the hidden roadblocks formed by long-lived TLSs in fault-tolerant quantum computation.","published":"2026-09-25T13:52:16Z","updated":"2026-09-25T13:52:16Z","authors":["J. R. Guimarães","Y. Liu","Y. Gao","Y. Haddad","A. Galicia","D. A. Volkov","J. T. Schmieder","H. Bhardwaj","M. Neis","J. Cereijo","M. Jerger","P. A. Bushev","R. Barends"],"categories":["quant-ph","cond-mat.mes-hall"],"headline":"Researchers Reveal Hidden Noise in Quantum Computer Components","summary":"The researchers studied the environment around superconducting qubits, crucial for quantum computing, using a new spectroscopy method that explores qubit excitation rather than just relaxation. This method allowed them to map detailed noise patterns and distinguish quantum noise from classical noise. Their findings suggest that certain long-lived defects within the qubit environment can retain memory of past activity, potentially affecting the performance of quantum operations in ways that depend on the previous states and protocols used."},{"id":"2609.31279v1","title":"Generalized Harmonic Measures: Synchronized Approximation and Positive Harmonic Representation","abstract":"We study synchronized boundary and kernel approximation for compatible harmonic measure systems. Joint limits of inner-boundary measures and kernels give a representation on a boundary retaining both position and kernel information. Finite decomposition gives concentration on minimal kernels, and kernel concentration gives uniqueness of the function projection. Boundary concentration along one common subsequence realizes this representation on a prescribed compact refinement. A gluing map then gives a unique representing measure on the minimal boundary. A neighborhood nonvanishing condition gives synchronized selection along the full sequence. Classical harmonic functions, a disk with a split boundary point, and finite weighted metric graphs illustrate the distinction between these conclusions.","published":"2026-09-25T13:52:13Z","updated":"2026-09-25T13:52:13Z","authors":["Duchao Liu","Yunjie Wang"],"categories":["math.FA"],"headline":"New Approach to Representing Harmonic Measures on Boundaries","summary":"Researchers explored how certain mathematical measures related to harmonic functions can be approximated and represented together on boundaries. They studied how combining information about positions and specific kernel functions leads to unique ways to describe these measures. Their work involves complex conditions and examples but is theoretical, focusing on understanding the structure of these representations rather than practical applications."},{"id":"2609.31278v1","title":"Hexagon decompositions and the Weil-Petersson metric","abstract":"We give an explicit quasi-isometry from the flip graph of triangulations $\\mathscr{F}_{g,n}$ to the $ε$-thick part of Teichmüller space equipped with the Teichmüller metric, by mapping each ideal triangulation to the hyperbolic surface whose shearing coordinates along that triangulation are all equal to zero. Extending this construction, we obtain a quasi-isometry $Q$ from the hexagon graph $\\mathscr{H}_{g,n}$ to the augmented Teichmüller space $\\overline{\\operatorname{Teich}}_{g,n}$ equipped with the Weil-Petersson metric, and we estimate its width, that is, the Hausdorff distance between $Q(\\mathscr{H}_{g,n})$ and $\\overline{\\operatorname{Teich}}_{g,n}$. By bounding the Weil-Petersson distance along grafting rays and the Teichmüller distance along shearing deformations, we show that the width is at most $\\sqrt{g+n}\\log(g+n)$. We also provide an explicit projection from any point in $\\overline{\\operatorname{Teich}}_{g,n}$ to the thick part of boundary strata, maintaining simultaneous control over the Weil-Petersson distance and the shearing coordinates.","published":"2026-09-25T13:51:36Z","updated":"2026-09-25T13:51:36Z","authors":["Marie Abadie"],"categories":["math.GT"],"headline":"New connections found between surface shapes and geometry graphs","summary":"The researchers studied mathematical spaces that describe shapes of surfaces with holes. They created a precise mapping between graphs built from certain surface decompositions and these shape spaces equipped with special distance measures. Their work shows that this mapping closely matches the complex geometry, bounding how far off it can be. They also developed a way to project any shape back to simpler boundary parts while controlling distances. These results are theoretical and focus on abstract properties of these geometric spaces."},{"id":"2609.31277v1","title":"High-throughput computational discovery of CuI-based ternary $p$-type transparent conductors","abstract":"P-type transparent conducting materials (TCMs) remain scarce, limiting progress in transparent electronics and tandem photovoltaics. Here, we perform a high-throughput computational search across CuI-based ternary X--Cu--I compounds using the minima hopping method combined with density functional theory. Filtering for thermodynamic stability, optical transparency, and light hole effective mass, we identify 58 candidate p-type TCMs, of which 49 are previously unreported. Twenty are p-type degenerate semiconductors with intrinsic hole carriers, including Cu$_2$FI$_2$ (calculated gap 3.75 eV, hole effective mass $0.262\\,m_0$). We further use this dataset to revisit the chemical modulation of the valence band (CMVB) design rules, originally formulated for Cu-based delafossite oxides. Four generalizations emerge: linear Cu coordination is not required, as tetrahedral coordination (CN = 4) is compatible with p-type TCM character; the mechanism extends to halide-based systems with I $5p$ anion states; the Cu $d$ state energy center $ε_d$ is a quantitative descriptor for valence band dispersion; and Cu undercoordination in monovalent-anion hosts drives intrinsic p-type carrier generation through stoichiometric charge balance. These results establish a generalized computational framework for the design of p-type halide TCMs.","published":"2026-09-25T13:49:52Z","updated":"2026-09-25T13:49:52Z","authors":["Michael Seifert","Miguel A. L. Marques","Silvana Botti"],"categories":["cond-mat.mtrl-sci"],"headline":"Researchers Find New Copper-Iodine Materials for Transparent Electronics","summary":"Scientists used computer simulations to search for copper-iodine based compounds that can conduct electricity while remaining transparent. They found 58 promising materials, 49 of which were previously unknown, that could work as p-type transparent conductors needed for electronics and solar cells. Their study also expanded understanding of how the arrangement of atoms affects electrical behavior, but these results are based on theoretical calculations rather than experiments."},{"id":"2609.31276v1","title":"Explicity Defined Norms on $JT_*$ Spaces","abstract":"We present a Banach space $X_α$, equipped with an explicit norm, whose properties resemble those of the predual $JT_*$ of the James Tree space. In particular, $X_α$ admits a shrinking dyadic basis with the property that the sums of its elements over any segment are uniformly bounded. The space $X_α$ is complementably $\\ell_2$-saturated, and its dual $X_α^*$ does not contain an isomorphic copy of $\\ell_1$. Finally, the bidual $X_α^{**}$ is also complementably $\\ell_2$-saturated, and the quotient $X_α^{**}/X_α$ is isometric to $c_0(2^{\\mathbb{N}})$.","published":"2026-09-25T13:49:51Z","updated":"2026-09-25T13:49:51Z","authors":["Spiros A. Argyros","Pavlos Motakis"],"categories":["math.FA"],"headline":"Researchers Define a New Mathematical Space Similar to James Tree Space","summary":"Researchers describe a new mathematical structure called Xα with an explicit way to measure size (norm). This space shares important properties with a known space called the predual of the James Tree space, such as having a special basis and certain boundedness traits. Their findings come from theoretical definitions and do not include practical experiments or applications."},{"id":"2609.31275v1","title":"Semi-Automatic Quantification of Bayesian Networks for Software Decision Support: Comparing WSA and RNM in a Software R&D Organization","abstract":"Expert-driven Bayesian networks can support recurring software decisions when historical data are limited, but quantifying their conditional probability tables (CPTs) requires many probability judgments. Semi-automatic methods reduce direct elicitation, yet practitioners have little comparative evidence from operational decision processes. We report an embedded case study comparing the Weighted Sum Algorithm (WSA) and Ranked Nodes Method (RNM) as complete elicitation-and-quantification pipelines in two contexts within a single software research and development (R&D) organization: feature selection for Internet of Things projects and user interface design selection. Within each context, the pipelines shared the graph, root priors, and decision evidence. We evaluated them through 15 expert-defined model-walkthrough scenarios and retrospective reconstructions of alternatives recorded in decision meetings. WSA matched 7/7 and 6/8 walkthrough expectations, whereas RNM matched 4/7 and 4/8. Both pipelines placed the selected alternatives near the top of the retrospective rankings. Their complete distributions nevertheless differed: the median total variation distance was 0.304 across four distinct feature patterns and 0.340 across seven distinct design patterns, rising to 0.890 for one design pattern. These results concern two ordinal value-estimation models in one organization. They show that similar shortlist behavior does not imply equivalent representations of uncertainty. For comparable software decision-support models, method selection and validation should consider node semantics, the judgments experts can provide, calibration requirements, complete output distributions, and the intended downstream use of the probabilities.","published":"2026-09-25T13:49:10Z","updated":"2026-09-25T13:49:10Z","authors":["Mirko Perkusich","João Nunes","Emilia Mendes","Emanuel Dantas","Ademar Sousa","Danyllo Albuquerque","Kyller C. Gorgônio","Angelo Perkusich"],"categories":["cs.SE"],"headline":"Comparing Two Methods to Support Software Decisions with Bayesian Networks","summary":"Researchers studied two ways to help software experts make decisions using Bayesian networks, which combine expert opinions when historical data is limited. They compared the Weighted Sum Algorithm (WSA) and Ranked Nodes Method (RNM) in real software project settings within one company, assessing how well each matched expert expectations and past decisions. The study found that while both methods helped identify good options, they produced different uncertainty estimates, suggesting careful method choice and validation are important. Results apply only to these methods in one organization."},{"id":"2609.31274v1","title":"Metastable homotopy theory via Tate coalgebras","abstract":"We develop foundations for metastable homotopy theory using Heuts' 2-truncated Tate coalgebras.","published":"2026-09-25T13:49:01Z","updated":"2026-09-25T13:49:01Z","authors":["Marco Nervo"],"categories":["math.AT"],"headline":"Foundations for a New Approach to Metastable Homotopy Theory","summary":"The researchers propose a new foundational framework for studying metastable homotopy theory, a complex area of mathematics dealing with shapes and spaces, using a tool called 2-truncated Tate coalgebras developed by Heuts. Their work focuses on building the theoretical groundwork rather than practical applications. The abstract is brief and primarily indicates the development of these foundational ideas without detailed results or examples."},{"id":"2609.31273v1","title":"Revisiting three-dimensional GLSMs and K-theoretic I-functions","abstract":"We revisit the correspondence between three-dimensional gauged linear sigma models and K-theoretic Gromov-Witten theory. We show that the K-theoretic Hori-Vafa formula for the grassmannian arises from the three-dimensional gauge theory upon including an appropriate anomaly-cancellation factor. Furthermore, we derive K-theoretic I-functions for isotropic grassmannians by incorporating symmetric and anti-symmetric tensor representation fields together with suitable boundary conditions.","published":"2026-09-25T13:48:22Z","updated":"2026-09-25T13:48:22Z","authors":["Taro Kimura","Yinda Li","Xiaohan Yan"],"categories":["hep-th","math.AG"],"headline":"New insights into 3D gauge theories and mathematical functions","summary":"The researchers studied the relationship between three-dimensional gauge theories and a branch of mathematics called K-theoretic Gromov-Witten theory. They showed that a known mathematical formula comes from these gauge theories when including a specific correction factor related to anomalies. They also developed new mathematical functions for special geometric spaces by adding certain fields and boundary conditions to the theory. Their findings are theoretical and based on mathematical derivations within this framework."},{"id":"2609.31272v1","title":"Cognitive Skills in the Age of AI: Computing Students and Experts Perceptions","abstract":"AI is becoming increasingly integrated into daily workflows, especially in computing. We are gradually shifting towards an AI-rich future, an impending yet unknown one. One important emerging concern is whether we are accordingly preparing our future computing workforce. Further, we need to know what the important cognitive skills are to remain relevant in the computing workforce and if there are changes in cognitive skill importance. To investigate this direction, we conducted a mixed-methods study, collecting perceptions from computing students and computing experts regarding the importance of cognitive skills in the past, present, and future. We report that the perceived importance of most cognitive skills will decrease in the future, with an AI-rich environment, but critical thinking skills remain important. Further, we report reasons collected through interviews on why the importance of cognitive skills will change and how future computing students can prepare for it.","published":"2026-09-25T13:48:02Z","updated":"2026-09-25T13:48:02Z","authors":["Neha Rani","Vu Minh Anh Le","Austin M. Spangler","Erta Cenko"],"categories":["cs.HC","cs.AI"],"headline":"How AI Might Change Important Thinking Skills in Computing Jobs","summary":"Researchers explored how AI may affect which thinking skills matter most in computing work. They asked students and experts about the importance of different cognitive skills in the past, present, and future. Their findings suggest most skills are expected to become less important in an AI-rich future, except for critical thinking, which remains key. The study also shares reasons why these changes may happen and advice for future students to prepare, based on interviews."},{"id":"2609.31271v1","title":"Learning Aerosol Coagulation Dynamics in Latent Space with a Scale-Covariant Neural ODE","abstract":"Particle-resolved aerosol models preserve the joint size-composition structure that governs cloud activation, optical properties, and freezing, but their computational cost limits their use in large-scale atmospheric models. We extend AeroMELD from a compact representation of aerosol populations to a prognostic model of coagulation. The resulting AeroMELD-Coag advances nine learned coordinates for population shape and one for total particle number through a scale-covariant neural ordinary differential equation that incorporates the known concentration dependence of coagulation. Across 2,000 held-out trajectories spanning 48 h, median symmetric errors are 3.5% for latent shape and 0.52% for total number. This ten-coordinate state also retains the evolving size-resolved composition and associated cloud-condensation-nuclei (CCN) activation, optical properties, and frozen fraction. Compared with the evaluated 20-bin sectional model with 320 prognostic coordinates, AeroMELD-Coag lowers pooled mean CCN activation error from 1.98% to 1.38% and frozen-fraction error from 1.72% to 0.56%. In a matched integration benchmark, GPU integration time per trajectory, amortized over a batch, is more than three orders of magnitude lower than for both single-core CPU references. These coagulation results provide a proof of concept for advancing aerosol microphysics efficiently in a compact learned state while retaining information about aerosol mixing state. They establish a foundation for future large-scale atmospheric models to represent aerosol microphysics with greater mixing-state detail within practical computational budgets.","published":"2026-09-25T13:47:20Z","updated":"2026-09-25T13:47:20Z","authors":["Wenhan Tang","Ruqi Yang","Jeffrey H. Curtis","Ehsan Saleh","Lekha Patel","Peter A. Bosler","Nicole Riemer","Matthew West"],"categories":["physics.ao-ph"],"headline":"New AI Model Efficiently Predicts Air Particle Clumping","summary":"Researchers developed a compact neural network model called AeroMELD-Coag to simulate how tiny airborne particles stick together over time, a process important for clouds and climate. This model uses only ten variables to closely track particle changes and performs much faster than traditional detailed simulations while maintaining accuracy. The study shows the model's potential to improve large atmospheric simulations by efficiently capturing complex particle behaviors, though findings are based on simulated test data rather than real-world measurements."},{"id":"2609.31270v1","title":"Multiscale computational study of the dielectric response of semi-crystalline polyethylene with chemical defects","abstract":"Polymers are widely used as functional insulating materials, but predicting their dielectric behavior is challenging because multiple mechanisms, acting across different spatial and temporal scales, contribute to their response. In polyethylene (PE), one of the most common polymers, electronic, atomic, and mesoscale dynamics must all be considered. In this work, semicrystalline models of polyethylene were developed and investigated using a multiscale approach. Quantum simulations were employed to describe electronic and vibrational contributions to the dielectric response, while classical molecular dynamics simulations captured the behavior of polymer chains at room temperature and frequencies down to the 10-100 MHz range. The study focuses on the impact of radio-oxidation defects on the dielectric properties of polyethylene. Quantum calculations reveal how electronic and vibrational contributions depend on the local atomic environment surrounding these defects. Comparison with molecular dynamics results highlights the influence of temperature on the static dielectric constant. Structural analyses further assess the effect of each defect type on PE crystallinity. Analysis of dipolar correlation functions shows that different defects interact with one another and with the polymer matrix in distinct ways, affecting permittivity and dielectric loss peaks. For all defect types, defect-defect interactions provide the largest contribution to dielectric response. While most oxidized groups exhibit positive coupling with the PE matrix, alcohol defects display a negative cross-correlation, partially offsetting their impact on the dielectric properties. Our results also show that ketone groups produce the largest dielectric loss between the defects studied.","published":"2026-09-25T13:47:11Z","updated":"2026-09-25T13:47:11Z","authors":["Roshal Perepadan Shaju","Guido Roma","Xavier Colin"],"categories":["cond-mat.mtrl-sci","cond-mat.soft","physics.app-ph","physics.comp-ph"],"headline":"How Defects Affect Electrical Behavior of Polyethylene Plastic","summary":"Researchers studied how tiny chemical defects change the way polyethylene, a common plastic, responds to electrical fields. Using detailed computer simulations at different scales, they looked at electronic effects and the movement of atoms and molecules. They found that different defects interact in unique ways, influencing the material’s ability to store and lose electrical energy, with some defects having stronger effects than others. This study helps explain the complex behavior of polyethylene but is based on computer models rather than experiments."},{"id":"2609.31269v1","title":"An abelian model for the Goodwillie tower of the circle","abstract":"We study the Goodwillie tower of the identity evaluated at the circle. Our main result is the p-local equivalence $ΩP_{p^k}I(S^1) \\simeq \\mathbb{Z} \\times Ω^{\\infty+2k}τ_{>0}\\mathbb{Sp}^{p^k}$ obtained from the work of Behrens and Kuhn on the Whitehead conjecture. In particular, this identifies the Goodwillie approximations of the circle as infinite loop spaces, a phenomenon which does not hold in general. The equivalence also reduces the computation of their homotopy groups to a stable problem, making them accessible to tools such as the Adams spectral sequence and to computer calculations. We place these computations in a more general framework by developing a Goodwillie-calculus analogue of the Gray sequence, which relates the homotopy groups of successive stages of the tower.","published":"2026-09-25T13:44:23Z","updated":"2026-09-25T13:44:23Z","authors":["Marco Nervo"],"categories":["math.AT"],"headline":"New insights into mathematical structure of the circle using Goodwillie tower","summary":"The researchers studied a complex theoretical tool called the Goodwillie tower, focusing on its application to the circle shape in mathematics. They found a special equivalence that lets them understand the approximations of the circle as infinite loop spaces, which is unusual and not generally true for other shapes. Their work also simplifies calculations related to the circle's structure and connects these computations through a new sequence that relates different stages in the tower. The results are theoretical and based on advanced mathematical frameworks."},{"id":"2609.31268v1","title":"Vibrational Mpemba relaxation in a linear damped elastic system","abstract":"We show that a linear damped elastic system can display a Mpemba-like reversal of vibrational relaxation without any nonlinear constitutive response or amplitude-dependent damping. The mechanism is purely modal. For a one-dimensional Kelvin--Voigt elastic medium, the decay rate of the $n$-th vibrational mode scales as $γ_n\\propto n^2$. We formulate relaxation in terms of a positive modal envelope energy, thereby excluding crossings caused merely by oscillation phase. A one-parameter family of initial states is then constructed such that increasing the initial elastic excitation simultaneously reduces the projection onto the slow fundamental mode. In the invariant two-mode subspace, the relaxation-order crossing is obtained analytically and, for this family, occurs at a universal dimensionless time independent of the selected pair of initial states. In the limiting case where the slowest mode is absent, the dominant relaxation rate changes discontinuously, yielding a direct vibrational analogue of the strong Mpemba effect. We further show that the same relaxation-order reversal occurs in the ordinary mechanical vibrational energy, which decreases monotonically despite retaining the oscillatory dynamics of the underlying modes.","published":"2026-09-25T13:43:11Z","updated":"2026-09-25T13:43:11Z","authors":["Kiwamu Yoshii","Satoshi Takasa"],"categories":["cond-mat.soft"],"headline":"Study finds unusual vibration relaxation in simple elastic systems","summary":"Researchers investigated how vibrations calm down in a basic damped elastic system without complex behaviors or varying friction. They discovered that changing initial vibration levels can cause faster relaxation in initially more excited states, a reversal known as a Mpemba-like effect. Their work uses mathematical models to explain this phenomenon in specific vibration modes, but it is based on theoretical analysis and simulations rather than experiments."},{"id":"2609.31267v1","title":"Electron lattice potentials for ultracold atoms using circular Rydberg orbitals","abstract":"Recent advances in experiments using individually trapped atoms have enabled precise control over circular Rydberg electrons with exceptionally long lifetimes. We show that these giant and stable electron orbits can form toroidal lattice potentials for ultracold atoms with a period set by the electron's de Broglie wavelength. Unlike conventional static optical lattices, this electron lattice is formed via the electron-atom interaction, which mixes Rydberg circular states with opposite azimuthal phase winding into a standing electronic matter wave. The lattice phase is thereby intrinsically coupled to the atom position. For a pair of atoms, this results in ballistic tunneling motion in the two-atom spatial correlations along the ring, while the single particle dynamics is essentially free rotation. We simulate the dynamics for an experimentally realistic setting that exploits optical tweezers for individual atom control. Our results open a route toward incorporating long range atom-atom interactions mediated by a single electron and, ultimately, realizing small Bose and Fermi gases confined in these microscopic electronic atom traps.","published":"2026-09-25T13:42:22Z","updated":"2026-09-25T13:42:22Z","authors":["Aileen A. T. Durst","Einius Pultinevicius","Homar Rivera-Rodríguez","Tilman Pfau","Matthew T. Eiles","Florian Meinert"],"categories":["cond-mat.quant-gas","physics.atom-ph","quant-ph"],"headline":"Using electron orbits to trap ultracold atoms in tiny rings","summary":"Researchers explored how very large, stable electron orbits called circular Rydberg states can create ring-shaped traps for ultracold atoms. These traps form through interactions between the electron and atoms, producing a standing wave pattern that depends on atom position. Their simulations suggest pairs of atoms could tunnel around the ring, while individual atoms move freely. This work, based on theoretical modeling of realistic experiments, points to new ways of controlling atoms with long-range interactions mediated by a single electron."},{"id":"2609.31266v1","title":"Locality and filter design for dissipative ground-state preparation","abstract":"Ground-state preparation is a fundamental challenge in quantum many-body physics, chemistry and materials science. Dissipative state-preparation techniques engineer a Lindbladian coupling to a bath that drives the system to its ground state, needing no initial overlap. Their cost is dominated by the Hamiltonian simulations inside the filtered jump operators, since each jump is a weighted sum of Heisenberg-evolved copies, or branches, of a coupling operator. We reduce this cost by following three simple observations to their logical end. First, on a fixed time grid, linear programming optimises the branch weights, trading the heating response against the circuit cost. Second, Lieb--Robinson bounds let each branch evolve on a patch whose radius grows with its duration. Third, if the dynamics converges quickly and truncation changes the jumps only slightly, the stationary state stays near the ground state even on patches smaller than the light cone. We validate our approach numerically. On a grid of $81$ evolution times, the optimised filter reduces the worst-case heating response, or leakage, more than $116$-fold at equal cost. Across fifteen molecules and six chains, the cost falls by 27 to 56\\% at matched design-grid leakage; at matched cost, the molecular leakage falls by a factor of $2.9$ to $2\\times10^{3}$. Density-matrix simulations show high ground-state fidelity for both filters, and on Ising and Hubbard chains the cheaper filter keeps a ground-state population of at least $0.89$ on patches of three or four sites. The same patches fail on a Heisenberg chain whose truncation error falls more slowly with patch size and whose Lindbladian gap is 12 times smaller than the Ising chain's, pointing to a role for the gap in how far patches can be truncated. At fixed radius the per-branch cost becomes independent of system size, reducing the cost of dissipative state preparation.","published":"2026-09-25T13:41:56Z","updated":"2026-09-25T13:41:56Z","authors":["Samuel J. Elman"],"categories":["quant-ph"],"headline":"Improving quantum ground-state preparation with efficient local filters","summary":"The researchers address the challenge of preparing a quantum system's lowest energy state using dissipative methods, which typically involve costly simulations. They propose an approach that optimizes filter design using a combination of mathematical optimization and limiting the simulation to smaller local patches guided by physical bounds. Their numerical tests show significant cost reductions and improved accuracy for several quantum models, although limitations appear for certain systems with smaller energy gaps. These results suggest a way to reduce computational effort in quantum state preparation while maintaining high fidelity, supported mainly by simulations and theoretical arguments."},{"id":"2609.31265v1","title":"Steady--State Current Signatures of Strong Light--Matter Coupling in Single--Molecule Junctions","abstract":"Strong light-matter coupling offers exciting opportunities for controlling molecular properties, yet resolving single-molecule behavior from collective effects remains a challenge. Recent experiments in scanning tunneling microscope break junctions (STM-BJs) have demonstrated strong light-matter coupling at the single-molecule level. Because STM-BJs provide direct access to molecular-junction currents, we investigate whether strong coupling leaves a measurable fingerprint in transport, potentially probing polaritonic states beyond optical spectroscopy. Using a semiclassical mapping approach for nonequilibrium quantum transport, we find that steady-state current indeed carries a direct signature of single-molecule strong coupling. We test the robustness of this signature against electrode connectivity, nuclear motion, spatially structured electromagnetic modes, and solvent effects. While environmental factors and nuclear motion reshape the signature, the current is not a passive response but is actively influenced by coupled molecular dynamics. Our work establishes current as a tool to interpret and control strong light-matter interactions in nanoscale junctions.","published":"2026-09-25T13:41:53Z","updated":"2026-09-25T13:41:53Z","authors":["Kritanjan Polley","Norah M. Hoffmann"],"categories":["quant-ph","cond-mat.mes-hall"],"headline":"Detecting Light-Molecule Interaction in Single-Molecule Electronics","summary":"Researchers explored how strong interactions between light and single molecules affect electrical current in tiny junctions. Using theoretical methods, they found that steady electrical current shows clear signs of this strong light-matter interaction, even when considering environmental and molecular motion effects. Their findings suggest electrical current measurements can help study and control light-driven molecular behavior at the nanoscale, though this work is based on theoretical simulations rather than experiments."},{"id":"2609.31264v1","title":"Identifying Scientists on X","abstract":"With the growing importance of science-related discourse on the Web and the erosion of the classical knowledge order, it is important to identify different user groups, such as scientists, automatically. This work proposes an approach for identifying scientists and non- scientists on X/Twitter based on their user biographies and tweets. We show that we are able to classify accounts as scientists and non- scientists on two different datasets, reaching an F1 score of up to 0.88 using Random Forests with linguistic features and up to 0.96 using a contrastively fine-tuned DeBERTa model in an ensemble setup. Furthermore, we provide two datasets with X users labeled as scientists or non scientists and their respective tweets and user biographies.","published":"2026-09-25T13:39:48Z","updated":"2026-09-25T13:39:48Z","authors":["Philipp Meier","Katarina Boland","Laura Kallmeyer","Stefan Dietze"],"categories":["cs.CL"],"headline":"Researchers Develop Method to Identify Scientists on Twitter","summary":"Researchers tackled the challenge of distinguishing scientists from non-scientists on the social media platform X (formerly Twitter). They used information from users' biographies and tweets to train computer models that classify accounts, achieving high accuracy scores. Their study provides two labeled datasets of users and contributes new techniques for automatically identifying scientists online based on written content. The findings come from data analysis and machine learning models applied to user text."},{"id":"2609.31263v1","title":"Taming the Option Factor Zoo: A High-Dimensional Analysis","abstract":"Option-implied factors are largely, but not entirely, spanned by the equity factor zoo. We construct 137 option-implied characteristics for optionable U.S. stocks from 2004 to 2023, screen them to 20 representative long-short factors, and test each against 160 equity factors with the double-selection LASSO of Feng, Giglio, and Xiu (2020). In-sample, the equity factors account for about 91% of the option factors' variance, and adding the option factors does not significantly raise the maximum Sharpe ratio. Individually, however, 8 of the 20 option factors, mainly measures of jump tails, kurtosis, and implied-volatility convexity, have non-zero SDF loadings after controlling for the equity zoo, and 4 remain significant after a Bonferroni correction. Options thus add a few pricing-relevant dimensions rather than a new factor zoo.","published":"2026-09-25T13:39:09Z","updated":"2026-09-25T13:39:09Z","authors":["Alexander Walter","Lukas Zimmer","Maxim Ulrich"],"categories":["q-fin.CP"],"headline":"Researchers analyze option factors compared to known stock factors","summary":"The researchers studied 137 characteristics derived from stock options on U.S. stocks between 2004 and 2023 and narrowed them down to 20 key factors. They compared these option-based factors to 160 known stock market factors using advanced statistical methods. They found that most option factors overlap with existing stock factors, explaining about 91% of their variations, but a few option factors still provide some unique information. Their findings suggest options add only a small number of new insights beyond established stock factors."},{"id":"2609.31262v1","title":"Deduplication-while-Training: A Resilient Paradigm for Privacy-Preserving Cross-Client Deduplication in Federated Learning","abstract":"Cross-client duplicate data in large language model training corpora degrades the efficiency of federated learning (FL) while exacerbating model memorization and privacy risks. Privacy-preserving cross-client deduplication effectively mitigates this issue by eliminating duplicate training data. However, existing schemes all follow a \"Deduplication-before-Training\" paradigm. This serially coupled paradigm incurs high fault-tolerance costs and lacks support for dynamic client joining. To this end, we propose an unexplored paradigm called \"Deduplication-while-Training (DwT)\", which enables concurrent deduplication and training. DwT transforms cross-client deduplication from a one-time, globally synchronous preprocessing operation into a continuous online service with state management, concurrent claiming, and failure recovery. By enabling state synchronization and task takeover, it minimizes the impact of client disconnections on the overall training progress while supporting the dynamic joining of clients. We design DwT-FL, a privacy-preserving deduplication system, to support DwT. By designing a concurrent state-claim mechanism and a hot-cold dual-queue scheduling strategy, DwT-FL enables the parallel execution of secure deduplication and model training, while effectively handling client disconnections and dynamic joins. Experimental evaluations demonstrate that, compared to the state-of-the-art scheme, DwT-FL significantly reduces the time overhead of failure recovery and dynamic joining by up to 93.04% and 94.18%, respectively. This provides an efficient and elastic concurrent deduplication scheme for dynamic and unstable FL environments.","published":"2026-09-25T13:39:07Z","updated":"2026-09-25T13:39:07Z","authors":["Rongxi Wang","Guanxiong Ha","Chunfu Jia","Yongsheng Lin","Minfen Gao","Hanmiaomiao Wang"],"categories":["cs.CR","cs.DC"],"headline":"New Method Speeds Privacy-Safe Data Deduplication in Federated Learning","summary":"Researchers address the problem of duplicate data slowing down and raising privacy risks in federated learning, where many users train a shared model without sharing raw data. They propose a new method called 'Deduplication-while-Training' (DwT) that removes duplicate data continuously during training instead of before it starts. Their system, DwT-FL, supports users joining or leaving dynamically and reduces delays caused by failures or changes, based on experimental tests. These findings reflect improvements in efficiency and resilience under certain conditions simulated by the researchers."},{"id":"2609.31261v1","title":"MoSAR: Mixture of Semantic Attention Regimes for Learning Adaptive and Approximable Attention Geometries","abstract":"The quadratic complexity of dense self-attention remains a central bottleneck for long-context language modeling. Many efficient alternatives address this cost by deciding in advance where attention should be sparse or local. We argue that attention approximation should instead be approached as a geometric problem, with the relevant interaction geometry learned from data: natural-language dependencies are input-dependent and difficult to prescribe in advance, so the model should learn where positional relevance can decay and where broader interactions must be preserved. We introduce Mixture of Semantic Attention Regimes (MoSAR), which learns such an adaptive, controlled-decay geometry over query--key interactions. Input-conditioned query and key routers, applied after positional encoding, select mixtures over short, medium, and global regimes, inducing a continuous distance-dependent attention field rather than a fixed sparsity pattern. This geometry is learned during training and can subsequently be discretized through top-1 routing. In controlled pre-training experiments with matched 500M-parameter models, MoSAR learns a substantially lower-reach attention geometry without degrading language-modeling quality, improving perplexity over dense RoPE at the training context length. Under length extrapolation, MoSAR achieves the best perplexity among all evaluated variants, including strong baselines such as ALiBi. Moreover, the learned geometry remains stable under deterministic top-1 discretization, suggesting that it is not only adaptive, but also amenable to low-cost approximation at inference time.","published":"2026-09-25T13:38:28Z","updated":"2026-09-25T13:38:28Z","authors":["Michele Paolicelli","Alessandro Petruzzelli","Alessandro Franceso Maria Martina","Cataldo Musto","Giovanni Semeraro"],"categories":["cs.CL","cs.AI"],"headline":"New Method Learns Smarter Attention Patterns for Language Models","summary":"Researchers address the problem of slow processing in language models caused by complex attention calculations. They developed MoSAR, a system that learns to focus attention adaptively across words based on their relationships, rather than using fixed patterns. Tests show MoSAR can handle long text efficiently without losing accuracy, and its learned focus patterns can be simplified later to speed up predictions, though these findings are based on controlled experiments with specific model sizes."},{"id":"2609.31260v1","title":"Agentic Limit Order Books: Phase Transitions and Market Impact","abstract":"We investigate the systemic macroscopic dynamics emerging from Limit Order Books (LOBs) populated exclusively by autonomous reinforcement-learning agentic traders. By formalizing agent interactions within a microscopic order-matching engine, we examine two fundamental quantitative phenomena: equilibrium phase transitions in order flow regime shifts, and the structural dynamics of market impact. We show that agentic LOBs exhibit distinct phase boundaries separating orderly price discovery from hyper-volatile cascade states, governed by critical thresholds in the number of agents and observable market depth. Furthermore, we demonstrate that market impact under agentic liquidity provision deviates from classical square-root dynamics, exhibiting distinct dissipative, balanced, and non-dissipative regimes under non-linear feedback loops.","published":"2026-09-25T13:38:26Z","updated":"2026-09-25T13:38:26Z","authors":["Jan Rosenzweig"],"categories":["q-fin.TR","cs.AI","q-fin.CP"],"headline":"Study Explores How Automated Traders Affect Market Stability and Prices","summary":"Researchers studied a simplified market where all trades come from computer agents that learn and adapt. They found these agent-driven markets can switch between stable price setting and sudden volatile crashes depending on agent numbers and market conditions. The study also reveals that the way trades influence prices differs from traditional models, showing complex patterns due to feedback effects. These findings come from simulations of theoretical agent behavior, not real markets."},{"id":"2609.31259v1","title":"Sensitivity-driven Personalization of a Glucoregulatory Model for Digital Twin Therapeutics in Type 1 Diabetes","abstract":"Digital twins are increasingly used in diabetes research, but reproducing individual glucose dynamics requires accurate identification of glucoregulatory model parameters. Traditional sensitivity analysis can identify influential parameters, yet a ranking based on limited conditions may miss parameters that matter during specific disturbances or for particular individuals. We therefore examine both the magnitude and timing of parameter influence across dynamic input-output conditions and assess whether a common ranking holds across participants. We analyze the Hovorka glucoregulatory model using data from 192 participants receiving automated insulin delivery therapy in the Type 1 Diabetes and Exercise Initiative dataset. We extend Sobol sensitivity analysis to time series and rank parameter influence under four conditions: full-day profiles, isolated meal disturbances, insulin bolus injections, and postprandial responses. We combine the condition-specific results into a global ranking and use it to select parameters for participant-specific identification. Compared with population parameters, identification restricted to the sensitivity-derived subset reduces the root mean square error of 60-minute glucose predictions by 60%, to approximately 31 mg/dL. These findings suggest that a global ranking can capture parameter influence across individuals and dynamic conditions. By narrowing the parameters requiring identification, this approach reduces computational cost and could accelerate the development of personalized diabetes digital twins.","published":"2026-09-25T13:38:15Z","updated":"2026-09-25T13:38:15Z","authors":["Clara Escorihuela-Altaba","Vihangkumar V. Naik","Eleonora Manzoni","Jose Garcia-Tirado"],"categories":["eess.SY"],"headline":"Improving Personalized Diabetes Models by Focusing on Key Factors","summary":"Researchers studied how to better personalize digital models that predict blood sugar in people with Type 1 diabetes. They analyzed data from 192 individuals to find which model parameters most affect blood sugar under various daily conditions. Their method improved short-term glucose predictions and could reduce the complexity and time needed to tailor treatment models for individuals. The findings are based on analyzing existing data and simulations, highlighting which factors matter most for accurate personalized diabetes modeling."},{"id":"2609.31258v1","title":"Multiphysics Simulation Framework for Continuous Ion Transport and Energetics in Atmospheric Pressure Interfaces","abstract":"Optimizing atmospheric pressure interfaces requires balancing efficient ion transport with unwanted collision activation. The complex environment at the inlet region, including rapid pressure drops, RF/DC electric fields, ion-gas collisions, and space charge effects, makes simulation of this system challenging. In this work, we introduce MuCITE (Multiphysics simulation for Continuous Ion Transport and Energetics), which combines continuous ion injection, precomputed electric and gas flow fields, and particle-in-cell space charge updates. Furthermore, by employing improved impulsive collision theory, it can not only predict transport but also analyze the loss position of ions and the evolution of their kinetic energy, internal energy, and dissociation probability during transport. Transport predictions with ion currents and activation trends of in-source fragmentation were compared with experimental measurements using a custom-built source and a Q-Exactive HF source. The results show that the simulation predictions agree well with the experimental results and can serve as a reference before experiments. MuCITE allows users to derive ion transport, output beam characteristics, and ion energy through the configurable interface geometry and operating conditions. It supports simulation-guided design and tunable parameters to achieve predicted efficient ion transport or in source activation.","published":"2026-09-25T13:35:53Z","updated":"2026-09-25T13:35:53Z","authors":["Yihui Yan","Jozef Lengyel","Frederic Rosu","Valerie Gabelica"],"categories":["physics.chem-ph","q-bio.BM"],"headline":"New Simulation Tool Models Ion Movement in Atmospheric Pressure Devices","summary":"Researchers developed MuCITE, a simulation framework to better understand how ions move and gain energy in complex environments at atmospheric pressure interfaces. This tool considers multiple factors like electric fields and gas flow to predict ion transport and energy changes, which are important for optimizing instrument performance. By comparing simulation results with experiments, the researchers found good agreement, suggesting MuCITE can help guide design and operation before doing physical tests. The findings are based on simulation and experimental comparison, not yet real-world application outcomes."},{"id":"2609.31257v1","title":"Scalable Variable Selection under Predictor Dependence with Adaptive Virtual Dummies","abstract":"Reliable high-dimensional variable selection requires scalable error-controlling methods. The Terminating-Random Experiments (T-Rex) selector estimates the false discovery rate (FDR) by aggregating early-terminated forward-selection paths in which predictors compete with synthetic dummies. We address two remaining challenges: i) predictor dependence can bias dummy-predictor competition; ii) computation is wasted on recomputing terms shared across experiments. Building on memory-efficient virtual dummies, which sequentially sample projections from their exact conditional law, we estimate the conditional Gaussian law of inactive predictors and map draws onto the remaining sphere radius. A shared lazy Gram cache computes the response product and each requested Gram column once across experiments. Simulations across three covariance structures show that uniform spherical dummies can exceed the target FDR, whereas the proposed method empirically controls FDR. Caching yields more than a sixfold speedup, and the method remains feasible with 100000 predictors, where competing FDR-controlling methods become computationally impractical.","published":"2026-09-25T13:34:34Z","updated":"2026-09-25T13:34:34Z","authors":["Taulant Koka","Michael Muma"],"categories":["eess.SP","stat.ME"],"headline":"New Method Improves Variable Selection by Handling Predictor Dependence","summary":"Researchers developed an improved approach to select important variables in large datasets while controlling errors. They addressed challenges from related predictors biasing selection and reduced extra computation by using efficient 'virtual dummies' and caching techniques. Simulations showed their method better controls error rates and speeds up computation, remaining practical with very large numbers of predictors, unlike some existing methods. However, these findings are based on simulations, so real-world performance needs further study."},{"id":"2609.31256v1","title":"The planted tensor problem over finite fields: algorithms and cryptography","abstract":"Inspired by the planted clique problem for random graphs, we introduce the planted totally-isotropic space problem for random tensors as follows. Let $U\\cong \\mathbb{F}_q^n$ and $W\\cong \\mathbb{F}_q^m$ be finite-dimensional vector spaces over a finite field $\\mathbb{F}_q$. Given $d\\in \\mathbb{N}$, choose a random \\(d\\)-dimensional subspace \\(V\\leq U\\), and construct a random alternating bilinear map $φ:U\\times U\\to W$ subject to the constraint \\(φ(V,V)=0\\). Such a $V$ is known as a totally-isotropic space of $φ$, and the goal is to recover $V$. Building on the recent probabilistic analysis of random tensors (Pham--Qiao--Wigderson--Wigderson, \\emph{in progress}), we initiate the study of the algorithmic hardness of this problem. Setting $m=\\lceil n/\\log n\\rceil$, we show that this problem admits an average-case polynomial-time algorithm for $d\\geq n/2$, by leveraging recent advances on the non-commutative rank problem. We also show that this problem admits a $q^{O(n\\log n)}$-time algorithm. We carry out algorithmic experiments using polynomial-system solving. From these results, we conjecture that the planted totally-isotropic space problem for $d=\\lceil n/C\\rceil$ with some constant $C\\geq 3$ is exponentially hard. Based on this evidence of computational hardness, we explore cryptographic applications of the planted totally-isotropic space problem and related planted tensor problems. We present private simultaneous messages and secret sharing protocols based on planted tensor problems, following the protocols based on planted subgraphs in (Abram--Beimel--Ishai--Kushilevitz--Narayanan, \\emph{TCC}'23). At the same security level, the public information size of protocols based on planted subgraphs is (moderately) exponential in that of protocols based on planted tensors, while the communication costs of these protocols are polynomially related.","published":"2026-09-25T13:32:45Z","updated":"2026-09-25T13:32:45Z","authors":["Yuxuan Liu","Youming Qiao","Gang Tang","Chuanqi Zhang"],"categories":["cs.DS","cs.CR"],"headline":"New Problem in Mathematics May Help Develop Cryptography","summary":"Researchers introduced the \"planted totally-isotropic space problem,\" which involves finding a special subspace hidden inside a random mathematical structure called a tensor over finite fields. They showed that for some cases, the problem can be solved efficiently, while in others it likely requires an exponentially long time. Based on these findings, the researchers suggest that this problem could be useful for creating new cryptographic protocols, potentially offering advantages over existing methods. Their conclusions rely on theoretical analysis and computational experiments."},{"id":"2609.31255v1","title":"PIA: A Personal Intelligence Agent Turning Health Conversations into Records and Records into Understanding","abstract":"General-purpose agent memory summarizes conversations: it extracts salient snippets, embeds them, and retrieves the top-k into the prompt. A health agent cannot run on summaries: a dose becomes a sentence, \"since last week\" is resolved at the model's discretion, and a three-month glucose trend cannot be answered by text similarity. We present PIA, a personal intelligence agent deployed alongside a consumer health agent. PIA receives the agent's natural-language requests, decides for itself whether and how to write or read, and turns conversations into typed clinical records and records into a synthesized understanding of the user. Its memory harness consists of four controls -- extraction, memory, retrieval, and understanding -- each a domain-agnostic mechanism with a pluggable health module: schema, medical alias dictionary, knowledge graph, and temporal rules. We show how the same query receives a different answer as the memory injected into the response context deepens from one-dimensional recall, to a two-dimensional health snapshot, to a three-dimensional trajectory with causality, and report lessons from operation: self-reported health data are missing not at random, question phrasing governs the quality of synthesized understanding, and nearly a third of candidate causal links are structural noise that rules alone remove.","published":"2026-09-25T13:32:20Z","updated":"2026-09-25T13:32:20Z","authors":["Jeonghun Yoon","Dongchan Kim","Hongyeon Yu","Young-Bum Kim","Jaegul Choo"],"categories":["cs.CL"],"headline":"PIA Helps Turn Health Talks into Detailed Records and Insights","summary":"Researchers developed PIA, an intelligent system designed to improve how health conversations are recorded and understood. Unlike simple summaries, PIA converts talks into detailed clinical records and builds a synthesized understanding using various tools like medical dictionaries and timelines. Their work shows that deeper memory use provides richer answers, but they also note challenges like missing self-reported data and the need for careful question phrasing to ensure accurate understanding."},{"id":"2609.31254v1","title":"A General Non-Markovian Electronic Friction and Langevin Dynamics Framework for Nonequilibrium Systems","abstract":"We present a general framework for simulating the nonequilibrium vibronic dynamics of molecules interacting with metal surfaces, which utilizes non-Markovian electronic friction and the corresponding generalized Langevin equation. The method employs a Markovian embedding scheme to sample coordinate-dependent quantum colored noise, with the underlying memory kernels represented systematically by exponential decompositions obtained directly from the electronic forces. In doing so, the approach extends the widely used electronic friction and Langevin dynamics formalism beyond the conventional Markovian approximation while retaining the coordinate dependence of the electronic back-action. We demonstrate the method for nonequilibrium charge transport through a vibrationally coupled molecular junction, where comparisons with numerically exact hierarchical equations of motion simulations show that the non-Markovian formulation is both more accurate and more robust than its Markovian counterpart. In particular, we find that non-Markovian electronic forces play a decisive role in the nonequilibrium vibrational dynamics, stabilizing regimes in which the conventional Markovian approximation fails.","published":"2026-09-25T13:27:57Z","updated":"2026-09-25T13:27:57Z","authors":["S. L. Rudge","R. J. Preston","D. S. Kosov","M. Thoss"],"categories":["physics.chem-ph","cond-mat.mes-hall"],"headline":"New Method Improves Simulation of Molecule-Metal Interactions","summary":"Researchers developed a new way to simulate how molecules interact with metal surfaces during electrical charge flow. Their approach uses advanced techniques to better capture complex effects over time, going beyond older simpler models. Tests showed that their method more accurately reflects real behavior, especially in situations where traditional approximations do not work well, according to the researchers' simulations and comparisons."},{"id":"2609.31253v1","title":"Enriching Sequential Recommendation with Graph Laplacian Positional Embeddings","abstract":"Sequential recommenders typically rely on learnable positional embeddings to encode the order of user interactions. In this work, we ask whether this ordinal signal can be replaced by a structural one derived from the item space. We propose to use Laplacian positional embeddings in SASRec: we build an item co-occurrence graph from training interactions, compute eigenvectors of its symmetric normalized Laplacian, and use them as frozen graph-derived positional embeddings. The backbone architecture and training objective remain unchanged. Experiments on four public sequential-recommendation benchmarks show that this simple replacement improves SASRec performance on most ranking metrics and remains competitive with strong positional and temporal encoding baselines. These findings indicate that item-item graph structure can be an effective substitute for standard ordinal positional embeddings in sequential recommendation.","published":"2026-09-25T13:27:23Z","updated":"2026-09-25T13:27:23Z","authors":["Ekaterina Trushkova","Artur Gimranov","Anton Lysenko"],"categories":["cs.IR"],"headline":"Improving Recommendations Using Graph-Based Position Embeddings","summary":"Researchers explored whether the usual way of showing order in recommendation sequences could be replaced with a structure-based signal from how items relate. They proposed using special embeddings derived from a graph representing item co-occurrences, without changing the recommendation model's core. Tests on four datasets showed this method improved performance on many ranking measures and competed well with existing approaches. These results suggest that item relationships in a graph can effectively capture order in recommendations. The study is based on experiments with existing models."},{"id":"2609.31252v1","title":"Peregrino: A Full-Hardware Accelerator for the Complete Falcon Post-Quantum Digital Signature Scheme on Resource-Constrained Edge Devices","abstract":"The arrival of quantum computers threatens the security guarantees of classical cryptography, since quantum algorithms can break schemes that remain secure against conventional attacks. The National Institute of Standards and Technology (NIST) has therefore standardized a set of post-quantum cryptographic algorithms, among them Falcon, a lattice-based digital signature scheme with the most compact signature and public-key sizes of the standardized candidates. Falcon's reliance on floating-point arithmetic makes it hard to implement in hardware, and prior work offers only partial accelerators for specific operations such as signature generation or verification, or a single full implementation generated through high-level synthesis (HLS). This work presents Peregrino, the first hardware accelerator of the complete Falcon digital signature scheme designed from scratch in HDL, targeting resource-constrained edge devices without a native floating-point unit through an emulated floating-point datapath. Implemented on a single Artix 7 XC7A200T FPGA, Peregrino uses 85261 LUTs, 41382 FFs, 44 BRAMs, and 142 DSPs for the Falcon-1024 variant. Against the only prior full implementation, the HLS-based FalconTakesOff, it uses 1,9$\\times$ fewer LUTs, 3,4$\\times$ fewer FFs, 2,7$\\times$ fewer BRAMs, and 9,9$\\times$ fewer DSPs, fitting the entire scheme on one FPGA where the HLS design requires at least two. Operating as a peripheral of an on-chip MicroBlaze soft-core, the accelerator additionally reduces key-pair generation, signature generation, and signature verification clock cycles by 92\\%, 96\\%, and 85\\% over the emulated floating-point reference software.","published":"2026-09-25T13:27:03Z","updated":"2026-09-25T13:27:03Z","authors":["Antonio Carreño","Jaime Señor","Jorge Portilla"],"categories":["cs.CR","cs.AR"],"headline":"New Hardware Speeds Up Quantum-Resistant Digital Signatures on Small Devices","summary":"Researchers developed Peregrino, a specialized hardware accelerator that runs the Falcon post-quantum digital signature scheme on small, resource-limited devices. Falcon is a security method designed to resist quantum computer attacks but is hard to run efficiently because it uses complex calculations. Peregrino implements the entire Falcon scheme using less hardware and fits it on a single chip, improving speed significantly compared to previous designs that only partially accelerated it or required multiple chips. The findings show Peregrino works well on simulations and hardware but are limited to a specific FPGA device."},{"id":"2609.31251v1","title":"Structure-dependent failure modes of neural priors in acoustic full-waveform inversion","abstract":"Frequency continuation does not improve every neural representation, and better velocity recovery need not imply better prediction of unseen waveforms. We compare a velocity grid with total-variation regularization, an adapted sinusoidal coordinate network (SIREN), and a frozen image generator with a full-grid residual in two-dimensional acoustic full-waveform inversion. All methods use the same frozen SWEEP discretization, acquisition, observations, background, and budget of 1,500 shot--model gradient evaluations. Three exposed synthetic targets represent inclined layering, curved layering, and a high-contrast curved interface drawn from a fault-labelled family. With 2-to-10 Hz continuation, the generative-residual model attains velocity RMSEs of 218.7, 108.0, and 341.9 m/s, compared with 232.5, 100.3, and 412.9 m/s for Grid+TV. Thus the latter wins on curved layering, while the former improves model RMSE by 5.9% and 17.2% on the other targets. Grid+TV nevertheless predicts held-out 10 Hz shots better on all three. SIREN is worse than the common background in every cell; continuation worsens two targets but improves one. A five-weight TV sensitivity study preserves the CurveFault model-error ranking. An equal-budget variational extension fails both recovery and interval coverage. These results establish conditional rankings, not universal superiority or an isolated spectral-bias mechanism. Full-grid residuals also prevent attributing gains uniquely to the generator. We advocate joint reporting of model recovery, unseen-shot prediction, optimization cost, and negative outcomes when auditing neural FWI.","published":"2026-09-25T13:26:55Z","updated":"2026-09-25T13:26:55Z","authors":["Ziye Yu","Xin Liu","Yuqi Cai"],"categories":["physics.geo-ph"],"headline":"Evaluating neural methods for acoustic waveform inversion accuracy","summary":"The researchers studied different computational approaches to improve acoustic full-waveform inversion, a method to map underground structures. They compared traditional grid methods with neural networks on synthetic data representing geological layers. Results showed no single method excelled in all aspects: some had better accuracy in velocity but worse prediction on new data. The authors highlight the importance of reporting multiple performance measures and caution against assuming one method is always best based on limited results."},{"id":"2609.31250v1","title":"Deterministic Regime Switching and Feasibility Inversion in Dynamic Tensor Rematerialization","abstract":"We report fine-grained, deterministic instability in Dynamic Tensor Rematerialization (DTR), an online eviction policy for memory-constrained DNN training, measured on the reference DTR simulator (simrd) using public execution traces. On an LSTM trace, memory budgets differing by 0.10% of unconstrained peak memory select fast and slow execution regimes whose overheads differ by as much as 7.3x; the slow regime is driven by broadly repeated re-eviction of the same storages (evictions per storage rise from 1.33 to 8.27 while the set of distinct evicted storages is essentially unchanged: 5,233 vs 5,236, with the two sets overlapping at Jaccard 0.999). On a ResNet-32 trace, a fine budget sweep reveals a deterministic feasibility inversion: the run is feasible at ratio 0.101, infeasible (OOM) across 0.102-0.106, and feasible again from 0.107. We trace the immediate cause of the OOM to a fully pinned recursive rematerialization frontier that exceeds the budget after every evictable tensor has been evicted. Ablations using the DTR authors' own variants implicate the joint size-staleness scoring term in the observed LSTM instability. We argue these are at least two distinct budget-sensitive pathologies rather than one mechanism, and we separate what is demonstrated from what remains hypothesised. All results concern the reference simulator; reproduction in a production runtime is future work. Code, instrumentation, and raw results accompany this preprint.","published":"2026-09-25T13:26:36Z","updated":"2026-09-25T13:26:36Z","authors":["Mahesh Reddy Pagadala"],"categories":["cs.LG","cs.DC"],"headline":"Researchers find memory issues in AI training simulator","summary":"Researchers studied Dynamic Tensor Rematerialization (DTR), a method to manage memory during neural network training, using a simulator and public data. They found that tiny changes in memory limits caused big, predictable shifts in performance and occasional out-of-memory errors. Two different problems affected how memory was reused and evicted. These findings are limited to simulations; real-world testing is yet to be done. The researchers provided their code and data to support further study."},{"id":"2609.31249v1","title":"The derivative function of the zeta-star correspondence","abstract":"We study the right derivative of the zeta-star correspondence between binary expansions and multiple zeta-star values of infinite depth. At a finite-index point, this derivative is a normalized difference of two finite-depth multiple zeta-star values. We use this identity to express its local behavior in terms of lower-truncated star sums and their boundary coefficients. The homogeneous contribution at summation variable three contracts with ratio $2/3$; the full coefficient also receives a positive forcing term. A weighted sum of the remaining coefficients contracts by at most $1/2$ at each binary step. On the open parameter interval, we prove that the derivative is right-continuous everywhere and continuous precisely at the non-dyadic points. Its left jump at a dyadic point is given by an explicit positive multiple series, with divergence exactly at $1/2-2^{-M}$, $M\\ge2$. A refinement of the index gives an infinite upper right Dini derivative at every interior point. We also prove that the Hausdorff, packing, and modified upper box dimensions of the graph are all $\\log_2(8/3)$. The Hausdorff lower bound uses the dimension of the Bernoulli convolution with parameter $2/3$, together with a separate measure-transfer argument. Finally, the pointwise Hölder exponent at a non-dyadic point is $\\log_2(3/2)$ divided by its dyadic approximation exponent, and the corresponding Hausdorff spectrum is linear.","published":"2026-09-25T13:26:02Z","updated":"2026-09-25T13:26:02Z","authors":["Jiangtao Li"],"categories":["math.NT","math.CA"],"headline":"Researchers analyze a complex math function's behavior and dimensions","summary":"The researchers examine the derivative of a mathematical function linking binary numbers and special infinite sums called multiple zeta-star values. They describe how this derivative behaves at specific points and prove properties about its continuity and jumps. Using advanced mathematical tools, they determine detailed geometric properties of the function's graph, including its fractal dimensions. Their findings are theoretical and rely on complex formula manipulations and dimension theory, without direct real-world applications discussed."},{"id":"2609.31248v1","title":"Gauss What You Need: Compact Gaussian Splatting Across Scene Scales","abstract":"3D Gaussian Splatting reconstructs a scene as a collection of Gaussian primitives from a set of posed photographs called the capture. The number of primitives used to represent the scene affects reconstruction quality, storage, and rendering cost. How to select this number automatically across capture scales remains unresolved: configurations effective on standard benchmarks can leave larger captures with too few Gaussians to reconstruct fine details. We observe that the surface to represent, given by the capture's extent and resolution, is known before training, whereas its content complexity becomes apparent during training, through the reconstruction quality on the training views. We introduce TangoGS, which combines capture-derived model sizing with training-based adaptation: the capture determines the scale of the model, and training feedback determines its final size within that scale. Before training, TangoGS derives a learning allowance for model growth from the capture's total pixels after discounting views that re-observe the same scene points. During training, reconstruction quality guides how many Gaussians to add and remove. On 13 standard benchmark scenes, TangoGS matches the mean PSNR of the best-performing evaluated baseline, LeGS, with $48\\%$ fewer Gaussians. On eight large captures, the same configuration automatically scales to larger models when necessary, achieving the highest mean PSNR among evaluated methods: $0.54$ dB above the runner-up with $2.3\\times$ as many Gaussians. Together, capture-derived learning allowances and training-quality guided density control enable a state-of-the-art quality--size compromise across scene scales without retuning.","published":"2026-09-25T13:25:59Z","updated":"2026-09-25T13:25:59Z","authors":["Afif Boudaoud","Jiayi Liu","Alexandru Calotoiu","Torsten Hoefler"],"categories":["cs.CV"],"headline":"New Method Adjusts 3D Scene Details Automatically for Better Quality","summary":"Researchers studied how to reconstruct 3D scenes using small glowing shapes called Gaussians from multiple photos. They developed TangoGS, a technique that chooses the right number of these shapes based on the photo set's size and how well the reconstruction looks during training. Their results show TangoGS can keep high image quality while using fewer shapes on standard tests and handle larger scenes automatically, improving quality without extra manual tweaking. These findings come from tests on benchmark scenes and large data sets."},{"id":"2609.31247v1","title":"Geometric Inconsistency Localization in Multi-View Image Sets","abstract":"Novel view synthesis (NVS) models can produce realistic new views of the same scene from different viewpoints. However, these generated views are not always geometrically consistent with one another. Multi-view (MV) consistency has shown promise as a tool for evaluating these NVS models. Its potential for multimedia forensics, however, remains largely unexplored, particularly for localizing geometric inconsistencies across wide-baseline image pairs. To enable research in this direction, we introduce DeformView, a wide-baseline MV dataset with pixel-level annotations of geometric inconsistencies. Using DeformView, we evaluate state-of-the-art MV consistency-scoring methods and show that approaches developed for NVS evaluation transfer poorly to the forensic task of geometric inconsistency localization. To address this limitation, we propose DEFECt3R, a lightweight learning-based classifier that uses cross-view feature relationships to localize geometric inconsistencies at the pixel level. By learning from explicit supervision, including hard negatives from geometrically consistent yet deformed views, DEFECt3R improves localization performance and substantially reduces false positives compared to existing consistency-scoring methods. Ablation experiments further show that both feature representations and correspondence quality contribute to localization performance. Overall, our findings demonstrate that MV geometric consistency is a promising yet underexplored signal for multimedia forensics and establish a benchmark and baseline for geometric inconsistency localization in wide-baseline MV image pairs. Code and dataset are available at https://github.com/IDLabMedia/DeformView-DEFECt3R","published":"2026-09-25T13:25:56Z","updated":"2026-09-25T13:25:56Z","authors":["Xander Staelens","Albéric Loos","Bert Ramlot","Hannes Mareen","Peter Lambert","Glenn Van Wallendael"],"categories":["cs.CV","cs.AI","cs.CR","cs.MM"],"headline":"New tool finds geometric errors in images from different viewpoints","summary":"The researchers studied how to detect when images of the same scene from different viewpoints don't match up correctly, which can happen in generated views of a scene. They created a new dataset named DeformView with detailed labels showing where such geometric inconsistencies occur. Using this dataset, they found existing methods don't work well for precisely locating errors, so they developed a new approach called DEFECt3R that better identifies these mismatches at the pixel level. Their work highlights challenges and opportunities in using multi-view consistency to check image authenticity."},{"id":"2609.31246v1","title":"Provenance of HAVING Queries in Semirings with Monus","abstract":"The semiring framework and its extensions form the basis of a rich collection of theoretical results and implementations for provenance tracking of database queries. Many real-world queries use aggregation and conditions on the aggregate values. Support for such queries has been proposed by introducing semimodule elements as aggregate values and formal comparisons between aggregate values as tuple annotations, which takes the approach outside the standard semiring framework. In this work, we show how to introduce a semantics for the provenance of such queries in arbitrary commutative semirings with monus (or m-semirings), without the need for additional operators. This semantics is shown to agree with the standard provenance of the aggregation-free self-join rewriting of HAVING COUNT(*) queries in semirings that are absorptive and where times distributes over monus. We derive algorithms for this semantics and implement them within the ProvSQL system, with viable performance on a real-world dataset for probabilistic query evaluation.","published":"2026-09-25T13:23:05Z","updated":"2026-09-25T13:23:05Z","authors":["Aryak Sen","Pratik Karmakar","Silviu Maniu","Angelo Saadeh","Pierre Senellart"],"categories":["cs.DB"],"headline":"Improving Tracking for Certain Database Queries Using Semirings","summary":"Researchers studied how to track the origins of database results for queries that include conditions on aggregated values, like counts. They developed a method using mathematical structures called semirings with monus, avoiding more complex extensions. Their approach matches known methods for simpler queries and allows new algorithms, which they tested in a system called ProvSQL with real-world data. Their findings are theoretical but supported by implementation and performance tests."},{"id":"2609.31245v1","title":"RupeeBias: Auditing Demographic Bias in Indian Economic Guidance from Large Language Models","abstract":"Individuals turn to large language models (LLMs) for guidance across a wide range of economic tasks, from comparing loan options and planning savings to deciding what raise to ask for or how much to charge for their services. LLMs are known to reproduce social biases, and biased economic guidance may influence what users believe they are worth, what they ask for, and what they ultimately accept. This risk is especially salient in India, where economic outcomes are shaped by demographic categories such as caste and urban-rural location. Existing LLM bias benchmarks, however, are largely designed around Western demographic categories and therefore miss key axes of economic disparity in the Indian context. We introduce RupeeBias, a benchmark for auditing demographic bias in LLM-generated economic guidance across Indian economic settings. RupeeBias consists of 39,150 prompts spanning four use cases: salary estimation, salary increment estimation, counter-offer recommendation, and service pricing recommendation. The benchmark follows a single-attribute counterfactual design, holding the description of the user's qualifications, experience, or service offering fixed while varying one demographic identifier at a time. RupeeBias covers 87 India-specific demographic identifiers across six axes: caste, religion, regional identity, gender, disability, and urban-rural location, with all prompts constructed in both English and Hinglish. We evaluate nine LLMs on RupeeBias and find systematic demographic disparities across all six axes. For otherwise identical prompts that differ only in demographic identifier, LLM-generated economic outputs differ by 20.2% on average. We publicly release RupeeBias to support future research on demographic bias in LLM-generated economic guidance across India-specific demographic and economic contexts.","published":"2026-09-25T13:22:24Z","updated":"2026-09-25T13:22:24Z","authors":["Pavithra P M Nair","Bhavik Talaviya","Shourya Bhushan","Rahul Pankajakshan","Seema Guruvadoo","Avinash Agarwal","Gilad Gressel","Krishnashree Achuthan"],"categories":["cs.CL"],"headline":"Study Reveals Demographic Bias in AI Economic Advice in India","summary":"Researchers created RupeeBias, a tool to check if large language models (LLMs) give different economic advice based on Indian demographics like caste, religion, or location. They tested nine AI models with over 39,000 prompts and found these models gave significantly different advice—on average 20% different—just by changing demographic details. The study highlights that AI tools can reproduce social biases, especially in India, and RupeeBias can help future work reduce such bias."},{"id":"2609.31244v1","title":"N-Body Dynamics as a Mechanism for Mean Motion Resonance Formation in the Post-Gas Era","abstract":"Mean motion resonances (MMRs) are traditionally thought to form during the gas-rich phase of protoplanetary disks via disk-driven migration. Whether MMRs can form solely through collisional damping after disk dispersal has not been systematically investigated. To study MMR formation in the post-gas-disk era, we perform a large suite of N-body simulations, each containing 10,000 planetesimals around an M-dwarf star and evolving over 10 Myr. We compare two groups of simulations that differ only in their initial mass distributions. Each group consists of 10 runs with statistically identical initial planetesimal orbital distributions, generated using different random seeds. One group adopts a uniform planetesimal mass distribution, whereas the other features a bimodal distribution with massive embryos embedded in a swarm of smaller planetesimals. At 1 Myr, 40% of systems (8/20) exhibit near-resonant period ratios, but this fraction drops to 25% (5/20) by 10 Myr as higher-order commensurabilities dissolve. Resonant-angle tracking shows that all persistent systems have circulating resonant angles at 10 Myr, although one system undergoes a transient libration phase between about 2 and 7 Myr before returning to circulation. These results demonstrate that collisional merging can drive MMR formation even without a gas disk, although in this scenario near-resonant configurations are strongly favored over true resonances. Furthermore, this mechanism produces similar MMR fractions in both groups, suggesting that MMR formation is insensitive to the initial mass distribution, although the latter may modulate the formation pathway to a limited extent.","published":"2026-09-25T13:21:35Z","updated":"2026-09-25T13:21:35Z","authors":["Kai Sun","Sheng Jin","Dong-Hong Wu","Meijuan Wang"],"categories":["astro-ph.EP"],"headline":"Planetesimal Collisions Can Create Planetary Resonances Without Gas","summary":"Researchers used computer simulations of 10,000 small bodies orbiting a star to study how planets can settle into synchronized orbits, known as mean motion resonances (MMRs), after the surrounding gas in their system has disappeared. They found that collisions alone can produce near-resonant orbital patterns, but exact resonances are rare and tend to break down over time. This suggests that while gas drives resonance formation traditionally, collisional processes can also shape planetary orbits even without gas, regardless of initial object mass distributions."},{"id":"2609.31243v1","title":"Generalized Quantifiers: Scope Dominance and Branching","abstract":"Let $P$ and $Q$ be proper upward monotone unary generalized quantifiers on a fixed domain $D$. We study the validity, over all structures $M$ with domain $D$, of \\[ M \\models Px\\,Qy\\,R(x,y) \\rightarrow Qy\\,Px\\,R(x,y), \\] called scope dominance. The iteration of two quantifiers lies between their branching product, which is generated by rectangles, and the dual of the branching product of their duals. We give criteria for equality with these bounds, obtain a new proof of the countable characterization of scope dominance, and characterize the case in which the inner quantifier is \"at least $κ$ many.\" Under the Generalized Continuum Hypothesis, Goldberg's theorem shows that, for ultrafilters, scope dominance is equivalent to equality between iteration and the branching product, whereas an elementary counterexample on a domain of size $\\aleph_2$ shows that this equivalence fails for arbitrary upward monotone quantifiers. For the case where both quantifiers are filters, the question becomes set-theoretic: the existence of a counterexample is equiconsistent with the existence of a measurable cardinal.","published":"2026-09-25T13:21:23Z","updated":"2026-09-25T13:21:23Z","authors":["Fredrik Engström"],"categories":["math.LO"],"headline":"Researchers Study How Two Counting Rules Interact in Logic","summary":"The researchers investigate how two specific counting methods, called generalized quantifiers, behave when combined in logical statements over a fixed set. They analyze when the order of these quantifiers can be switched (scope dominance) and how their combination relates to known mathematical structures called branching products. Their results include new proofs and conditions involving large infinity concepts, showing that some properties depend on advanced set theory assumptions. The findings are theoretical and based on logical and mathematical analysis rather than experiments."},{"id":"2609.31242v1","title":"Partial regularity for local minimizers of functionals with double-phase Orlicz growth","abstract":"In this paper, we consider the local minimizers to a class of non-autonomous functional with double-phase Orlicz growth: \\begin{equation*} u \\in W^{1,1}(Ω; \\mathbb{R}^N) \\mapsto \\int_Ω\\Big( {\\varphi_1}\\big(|Du|_{\\mathbb{A}^u}\\big) + a(x)\\,{\\varphi_2} \\big(|Du|_{\\mathbb{A}^u}\\big) \\Big)\\, dx, \\end{equation*} where $|Du|_{\\mathbb{A}^u} := \\big\\langle \\mathbb{A}(x,u)\\,Du, Du \\big\\rangle^{\\frac{1}{2}}$ with $\\mathbb{A}(x,u) = \\big\\{A^{αβ}_{ij}(x, u)\\big\\}_{i,j = 1\\cdots N}^{α,β= 1\\cdots n}$ as the uniformly elliptic bounded symmetric tensor field, and $\\varphi_k(\\cdot)$ for $k=1,2$ are two different $N$-functions. We prove the partial regularity of their local minimizers if the continuity modulus of $a(\\cdot)$ and $\\varphi_k(\\cdot)$ are fit for the gap conditions, and $\\mathbb{A}(x,u)$ meets an optimal regularity in $(x,u)$. This is an extension from the single Orlicz growth functional to the double-phase one based on an essential improvement of several important inequalities.","published":"2026-09-25T13:20:52Z","updated":"2026-09-25T13:20:52Z","authors":["Wenrui Chang","Shenzhou Zheng"],"categories":["math.AP"],"headline":"Mathematicians study smoothness of complex energy-minimizing functions","summary":"Researchers examine certain mathematical expressions that describe how energy depends on changes in functions, focusing on a complex case called double-phase growth. They prove that local minimizers—functions that make these expressions as small as possible nearby—are partially smooth under specific technical conditions about the involved functions and coefficients. Their results extend known smoothness properties from simpler cases to this more complicated setting, based on improved mathematical inequalities. The work is theoretical and concerns abstract mathematical functionals."},{"id":"2609.31241v1","title":"Frequency- and Power-Dependent Optical Response of a Cesium Rydberg Microwave Receiver","abstract":"We show experimentally obtained frequency- and power-dependent optical response of a room-temperature cesium vapor cell used as a Rydberg microwave receiver based on electromagnetically induced transparency. An applied microwave field couples the ${70S_{1/2}}$ Rydberg state to the nearby ${70P_{1/2}}$, ${70P_{3/2}}$, ${69P_{1/2}}$, and ${69P_{3/2}}$ states, modifying the probe-transmission spectrum. Probe transmission is recorded over a 800 MHz range of coupling laser detunings. These EIT spectra are measured for RF frequencies from 9 to 13.5 GHz and over a 40 dB range of RF power. We analyze the RF induced splittings in the recorded EIT spectra. We observe both limiting cases, where the EIT peak separation is either dominated by the RF detuning or by the RF power. At higher powers we are also able to observe complex features due to coupling of multiple nearby Rydberg states. We further analyze the probe-transmission signal at zero coupling-laser detuning. Sigmoidal fits to the power-response curves are used to determine the midpoint power and the 10-90% transition interval as functions of RF frequency.","published":"2026-09-25T13:19:08Z","updated":"2026-09-25T13:19:08Z","authors":["Jure Pirman","Ema Stopar","Katja Gosar","Erik Zupanič","Peter Jeglič"],"categories":["quant-ph","physics.app-ph","physics.atom-ph"],"headline":"Researchers study how cesium vapor detects microwaves differently by frequency and power","summary":"The researchers experimentally examined how a room-temperature cesium vapor cell reacts to microwaves of varying frequency and strength using a technique called electromagnetically induced transparency. They measured changes in light transmission influenced by microwave fields coupling specific high-energy atomic states across a range of frequencies (9-13.5 GHz) and powers. Their findings reveal how microwave power and frequency affect atomic state interactions and probe light behavior, with more complex effects observed at higher powers. The study is based on observed spectra and fits and is limited to these experimental conditions."},{"id":"2609.31240v1","title":"First-principles calculations of the thermodynamic factor of multicomponent alloys","abstract":"Computing non-dilute diffusion coefficients from Onsager transport coefficients of a multicomponent alloy requires the thermodynamic factor, which measures the curvature of the free energy with respect to composition. As diffusion in metals is mediated by vacancies, the curvature that matters is not that of the vacancy-free alloy, which is straightforward to compute, but that of an alloy carrying a dilute concentration of vacancies. The resulting matrix is nearly singular, and composition fluctuations from which it is obtained converge slowly in Monte Carlo simulations. Here we develop two routes that avoid sampling vacancy composition fluctuations. The first constructs the thermodynamic factor from two quantities, the free-energy curvature of the vacancy-free alloy and the vacancy concentration. The second truncates the semi-grand canonical partition function at a single vacancy and recovers the full thermodynamic factor of the vacancy-containing alloy. The same separation shows that the largest eigenvalue of the diffusion matrix is the vacancy tracer diffusion coefficient for any number of components and any degree of non-ideality, while the remaining eigenvalues scale with the vacancy concentration. Across binary through quinary alloys of the Hf--Mo--Nb--Ti--Zr system, both routes reproduce eigenvalues and eigenvectors of diffusion matrices obtained from fully converged Monte Carlo simulations. The single-vacancy expansion reaches this accuracy with roughly one tenth of the sampling effort. A scaling relation extends both routes to local vacancy chemical potentials away from equilibrium, without additional simulations. These results allow thermodynamic descriptions from atomistic models or CALPHAD assessments to be used directly in mesoscale simulations of mass transport.","published":"2026-09-25T13:19:02Z","updated":"2026-09-25T13:19:02Z","authors":["Damien K. J. Lee","Shashank Saxena","Anton Van der Ven","Anirudh Raju Natarajan"],"categories":["cond-mat.mtrl-sci"],"headline":"New methods improve calculations of metal alloy diffusion factors","summary":"Researchers studied how atoms move in complex metal mixtures, focusing on a factor called the thermodynamic factor that relates to the material's energy landscape. They developed two new calculation approaches that avoid difficult simulations of tiny defects called vacancies, making predictions more efficient and accurate. Their methods match detailed computer simulations in several metal alloy systems and offer ways to use theoretical models to better understand how elements spread within metals."},{"id":"2609.31239v1","title":"Chiral Transfer and Entanglement Generation of Even-Parity Bell States with Engineered Two-Photon Loss","abstract":"We investigate chiral transfer and dissipative generation of even-parity Bell states in a two-qubit system with coherent two-photon driving and engineered two-photon loss. We find that adiabatic encirclement of a second-order exceptional point induces direction-dependent transfer between the Bell states $|Φ^+\\rangle$ and $|Φ^-\\rangle$, governed by a time-integrated low-loss branch-selection mechanism. We develop a hybrid-Liouvillian description with a control parameter $q$ that interpolates between conditional no-jump dynamics and unconditional Lindblad evolution, and use it to assess how exceptional-point-induced chirality survives in the presence of quantum jumps. When the same parameter loop is initialized in the separable state $|00\\rangle$, it directly generates strong even-parity entanglement. Together, these results extend dissipative Bell-state control beyond the single-excitation manifold. They further demonstrate that exceptional-point-based protocols can unify chiral state transfer and entanglement generation within a single engineered two-photon platform.","published":"2026-09-25T13:18:47Z","updated":"2026-09-25T13:18:47Z","authors":["Lin Xiao","Jian Li","Mu Zhou","Bin Wei","Qing-Xu Li","Jia-Ji Zhu"],"categories":["quant-ph"],"headline":"Researchers Explore Directional Control of Quantum Entanglement States","summary":"The researchers studied how to control two-qubit quantum states called even-parity Bell states using special two-photon driving and loss processes. They found that moving around a particular parameter point (an exceptional point) causes directional transfer between these states, and this process can create strong entanglement starting from a simple initial state. Their work uses a mathematical framework to understand how these directional effects persist despite quantum noise. These findings are theoretical and highlight how engineered two-photon systems might unify control over state transfer and entanglement."},{"id":"2609.31238v1","title":"Wave-Propagation Geometry Emerges from Scalar Travel-Time Relations","abstract":"Scalar travel-time relations can identify propagation geometry without velocity inputs, ray labels or equation supervision during learning. Physical interpretation is applied after learning through the high-frequency isotropic eikonal relation (Aki and Richards, 2002). Here we show that local value differences constrain derivatives through sampling geometry and smoothness. Controlled neural fields recover propagation directions with $5.34\\pm0.07^{\\circ}$ median angular error, and local P- and S-wave velocities in a prospectively locked, test-sealed realization. Classical representations also recover geometry. Scalar-matched perturbations separate value accuracy from derivative reliability; regularized inversion can recover structure when direct gradients fail. A field trained on 2.90 million Chinese arrivals retains continental crust--mantle boundary structure: correlation with a withheld reference is $ρ=0.910$ raw and 0.592 after quadratic detrending (spatial-shift $p=0.020$). Direct catalogue values yield $ρ=0.909$, and thickness contrasts are compressed. Manual-only California arrivals support broad velocity organization with limited lateral fidelity. Synthetic quantum fields containing phase information support probability-flow and trajectory readouts without supervising those quantities. These results establish conditional identification across representations and physical settings: informative scalar relations can constrain geometry, while known physical relations supply its interpretation. They do not imply that accurate scalar fitting alone guarantees reliable physical derivatives.","published":"2026-09-25T13:18:47Z","updated":"2026-09-25T13:18:47Z","authors":["Ziye Yu"],"categories":["physics.geo-ph"],"headline":"Researchers Use Travel-Time Data to Map Wave Directions and Velocities","summary":"The researchers studied how timing measurements of waves can reveal their travel paths and speeds without needing detailed velocity models or labels. They used neural networks and other methods to recover wave directions with small errors and estimated velocity structures in the Earth's crust and mantle. Their work shows that scalar timing information can indicate wave geometry, but accurate timing alone does not always ensure reliable physical details about wave behavior. The findings mainly come from controlled data and theoretical models."},{"id":"2609.31237v1","title":"Conformal correlator systems","abstract":"In critical limits of statistical models, there exist non-local correlators that are conformally covariant without belonging to a conformal field theory. To describe such correlators, we define conformal correlator systems, whose axioms are weaker than those of CFT. As a result, $N$-point correlators are determined by $4$-point correlators, instead of $3$-point correlators in CFT. In two dimensions, conformal spins may take arbitrary complex values, leading to correlators with nontrivial monodromies. We show that sphere $4$-point correlators and torus $1$-point correlators obey nontrivial monodromy constraints. We construct correlators with abelian monodromies in free bosonic theories and in critical loop models.","published":"2026-09-25T13:16:04Z","updated":"2026-09-25T13:16:04Z","authors":["Sylvain Ribault"],"categories":["hep-th"],"headline":"New framework for describing complex patterns in statistical models","summary":"Researchers study special mathematical objects called conformal correlators found in statistical models at critical points, which behave differently from those in traditional conformal field theory (CFT). They introduce a new, less strict set of rules called conformal correlator systems to describe these patterns, showing that knowing certain 4-point correlators can determine others. Their work, mainly theoretical and using simulations, explores how these systems behave in two dimensions and constructs examples in specific model types, highlighting novel mathematical constraints and behaviors."},{"id":"2609.31236v1","title":"Catch-Up and Come Home: Economic Convergence and Return Migration","abstract":"The return migration of skilled workers provides origin countries many important benefits. The likelihood of return migration is thought to depend on relative economic opportunities, but quantitative evidence of its relative importance remains limited. Here we measure return migration rates using the public profiles of software developers on GitHub, geocoded across ten waves between 2012 and 2026, in a panel containing more than 270,000 international movers. Within five years of departure, 8.8% of movers have returned. Origin-country rates range from under 1\\% to 16\\%. Hazard models with corridor fixed effects show that migrants whose home countries experience greater catch-up growth return at significantly higher rates. Immutable features of country pairs such as cultural or geographic distance do not explain the estimate. We forecast future return rates using IMF growth projections, predicting, for example, that India's return rate will rise 12\\% by 2035 and China's 7\\%. The difference is driven mostly by India's faster projected growth at home. As developing economies close income gaps with the destinations their emigrants have settled in, they recover more of the workers they lost.","published":"2026-09-25T13:15:46Z","updated":"2026-09-25T13:15:46Z","authors":["Kinga Varga","Bence Kollányi","Johannes Wachs"],"categories":["econ.GN","cs.SI"],"headline":"Skilled Workers Return More as Home Economies Improve","summary":"Researchers studied the return of skilled workers, using data from over 270,000 software developers on GitHub from 2012 to 2026. They found that about 8.8% return to their home countries within five years, with higher return rates linked to faster economic growth at home. The researchers predict that countries like India will see increased return migration due to predicted economic growth. Results are based on data analysis and forecasting, not on experimental or field studies, and focus on the relationship between economic growth and return migration."},{"id":"2609.31235v1","title":"Purin: A Biology-inspired Mechanism for Artificial Neural Networks","abstract":"Artificial neural networks (ANNs) usually represent neural transmission with fixed trainable weights during a training batch, which omits short-term changes in synaptic efficacy. In addition, the discrete time-step simulation requires additional temporal processing that many conventional ANN architectures do not use. To overcome these challenges, we propose Purin, a biology-inspired and ANN-compatible mechanism, that introduces synaptic efficacy modulation into conventional convolutional neural networks. Purin uses a time-interval-based abstraction for neural activities, which allows Purin to introduce short- and long-term synaptic efficacy changes without using discrete time-steps. Purin introduces a bounded factor to represent temporary synaptic efficacy changes, together with two weight matrices that represent input-side and output-side efficacy. The weight matrices are updated by backpropagation and interpreted as the long-term synaptic efficacy changes. Experimental results show that after removing the confounding factors in the AlexNet, VGG11, and GoogLeNet architectures, Purin improves the classification accuracies in all three models across the evaluated datasets.","published":"2026-09-25T13:15:22Z","updated":"2026-09-25T13:15:22Z","authors":["Zishu Liu","Chunbo Luo","Christos Grecos"],"categories":["cs.AI","cs.NE"],"headline":"New Method Adds Short-Term Changes to Improve Neural Networks","summary":"Researchers developed Purin, a new mechanism inspired by biology to improve artificial neural networks by including short-term changes in connections, called synaptic efficacy. Purin captures these changes using a time-interval approach rather than fixed steps, combining temporary and long-term adjustments in connection strengths. Testing showed that Purin improved accuracy in popular image recognition models, but the findings are based on specific datasets and simulations presented by the researchers."},{"id":"2609.31234v1","title":"WeaveAgent: A Two-Stage Tool-Routing Agent for Ultra-High-Resolution Remote Sensing Imagery","abstract":"Problem. Ultra-high-resolution (UHR) remote sensing with vague user intents has two bottlenecks: visual tokens are expensive, and tool calling must be format-reliable (pretrained models emit zero tool calls zero-shot). Method. WeaveAgent, a two-stage tool-routing agent, decouples routing from visual perception. Stage A is routing-first: emission is trained, not elicited. Stage B executes conditionally: intrinsic queries enter visual answering (full-scene thumbnail; a WeaveEarth-style evidence board as an optional fixed-budget, approx. 5k-token compression interface); extrinsic queries execute tool call on original full-resolution imagery, answering from tool observations in a second, observation-masked round. Training: alignment SFT, then GRPO under reward R_WA2. Results. Alignment SFT lifts extrinsic routing from 0% to 80.75% (323/400); GRPO suppresses 9 intrinsic mis-emissions while tool selection is unchanged. The trained 2B system does not beat the zero-shot 8B baseline overall (0.263 vs. 0.250), a diagnostic contribution. Oracle attribution separates two repair ingredients: loading the observation into context lifts extrinsic answer accuracy from 0.025 to 0.425 under marker-free cross-mode returns, and the two-turn SFT stage adds a further +9.3 points to 0.518 at a small routing cost. A +/- image ablation shows emission suppression is visually grounded, and a query-register matrix shows LLM-rewritten queries cost trained checkpoints 2-11 points. Scope. All training and evaluation use the 5,000 / 3,273 / 1,000-record VagueUHR corpus (600 intrinsic + 400 tool-requiring; the base seeds synthesis and is not used for optimization). Single-pass evidence construction runs at 7.31 s per image on an RTX 4090. Code, data, and evaluation protocols will be released.","published":"2026-09-25T13:13:11Z","updated":"2026-09-25T13:13:11Z","authors":["Zhongyu Pang"],"categories":["cs.CV"],"headline":"New Two-Step Method Helps Analyze Ultra-High-Resolution Satellite Images","summary":"Researchers addressed challenges in interpreting very detailed satellite images when user questions are unclear. They developed WeaveAgent, a two-stage system that first decides which analysis tools to use before processing image details, improving tool selection accuracy significantly. Testing showed the method improved some aspects but did not outperform larger existing models overall. The study focused on simulations with a specific dataset, and results highlight components important for future improvements in analyzing complex satellite imagery."},{"id":"2609.31233v1","title":"Failure of the Pólya-Szegő inequality for the fractional gradient","abstract":"We prove that the Pólya-Szegő inequality for the $L^p$ norm of the Riesz fractional $s$-gradient fails for every $p\\in[1,2)$ and $s\\in(0,1)$, thus providing a negative answer to a question posed by Nguyen and Squassina in 2017.","published":"2026-09-25T13:12:33Z","updated":"2026-09-25T13:12:33Z","authors":["Giorgio Stefani"],"categories":["math.FA"],"headline":"Researchers Disprove a Mathematical Inequality for Fractional Gradients","summary":"The researchers studied a specific mathematical inequality called the Pólya-Szegő inequality, which relates to a concept known as the fractional gradient. They proved that this inequality does not hold true for certain values of parameters p and s, specifically for p between 1 and 2, and s between 0 and 1. Their work answers a question raised by other mathematicians in 2017, showing that the inequality fails under these conditions."},{"id":"2609.31232v1","title":"Cybflight: An Embedded Rust Autopilot for Aerial Robotics Research","abstract":"Bringing an aerial robotics method from simulation to flight should not require rebuilding a mature autopilot or adding a companion computer. Cybflight is an open-source embedded Rust research autopilot whose typed, replaceable interfaces connect hardware access, perception, state estimation, trajectory planning, and control. This modular development and compile-time optimization workflow is demonstrated with replaceable Rust implementations of model predictive contour-tracking control (MPCTC) and incremental nonlinear dynamic inversion (INDI) running on one STM32H743 without a companion computer. Using this configuration, the vehicle reaches 12.38 m/s during indoor flight, while an outdoor flight using global navigation satellite system (GNSS) position updates reaches 31.4 m/s. These flights show that running demanding estimation and nonlinear control entirely on a flight-controller microcontroller need not come at the expense of a modular autopilot structure.","published":"2026-09-25T13:10:12Z","updated":"2026-09-25T13:10:12Z","authors":["Yifan Lin","Chao Qin","H S Helson Go","Hugh H. -T. Liu"],"categories":["cs.RO"],"headline":"New Rust-based autopilot enables modular drone flight control","summary":"Researchers developed Cybflight, an open-source autopilot written in Rust that runs entirely on a flight controller microchip without needing extra computers. This system connects various flight functions like sensing and control through replaceable software parts, allowing efficient and flexible drone operation. The researchers demonstrated Cybflight controlling fast indoor and outdoor flights, showing that advanced drone control can run on embedded hardware without losing modularity."},{"id":"2609.31231v1","title":"On the shape of the numerical range of some tridiagonal matrices","abstract":"In this paper, we study the elliptical shape of the numerical range of tridiagonal matrices of orders $3,4$ and $5$ with zero entries on the main diagonal, a positive constant value $a$ along the super-diagonal, and sub-diagonal entries are \\begin{align*} \\begin{cases} r, sr^2, r^3, sr^4,\\ldots, r^{n-1} & \\mbox{ if $n$ is even, and } r, sr^2, r^3, sr^4,\\ldots, sr^{n-1} & \\mbox{ if $n$ is odd}, \\end{cases} \\end{align*} where $n$ denotes the order of the matrix, $s>0$ and $r$ is a real number. We further investigate the flat portions on the boundary of the numerical range of these $n\\times n$ matrices in the special case $r=-1$. Some results given in \\cite{chien2011numerical} are obtained as a particular case of our results.","published":"2026-09-25T13:09:39Z","updated":"2026-09-25T13:09:39Z","authors":["Arobinda Ghosh","Riddhick Birbonshi","Sarita Ojha","Kallol Paul"],"categories":["math.FA"],"headline":"Researchers study shapes of special tridiagonal matrices' numerical ranges","summary":"The researchers investigate the elliptical shapes formed by the numerical ranges (a type of value set) of certain tridiagonal matrices with zero main diagonals and specific patterns on their sub- and super-diagonals. They focus on matrices of sizes 3 to 5 and examine special cases when one parameter equals -1, exploring flat parts on the boundary of these shapes. Their work includes and extends some previous findings, but it is limited to theoretical analysis of these matrix properties."},{"id":"2609.31230v1","title":"Excitation and Identifiability in the 2D Navier-Stokes equations","abstract":"Convergence in parameter estimation classically requires ``persistency of excitation,\" which is a non-degeneracy condition on a trajectory-dependent signal. This paper develops, to the best of our knowledge, the first such excitation theory for a nonlinear partial differential equation. In the context of identifying the unknown viscosity from spectral observations in the two-dimensional Navier--Stokes equations for incompressible fluids, our excitation condition is computable, verifiable a priori, and sharp with respect to scaling. Our approach employs a data assimilation methodology to account for an unknown initial state and select candidate viscosities by minimizing the loss between the low-mode observations of the fluid velocity and low-mode projection of a nudging-based filter that assimilates these observations. The main novelty of our framework is to distinguish a new functional, $\\mathsf{W}$, representing the work done by the filter's associated sensitivity variable on the enstrophy, around which our entire analysis is centered. We show that $\\mathsf{W}$ is explicitly comparable to the observability Gramian of the associated Gauss--Newton iteration, and subsequently establish the following dichotomy at every critical point of the observational loss: either $\\mathsf{W}$ exceeds a certain threshold, in which case the candidate viscosity obeys an explicit error estimate that depends inversely on $\\mathsf{W}$ and the observational density $N$, but directly on the error between initial conditions, or else $\\mathsf{W}$ is below the threshold and the observations are quantitatively insensitive to parameter updates in a way that is detectable to the user. Notably, our approach is energy-based, and therefore expected to be adaptable to many other nonlinear dissipative systems.","published":"2026-09-25T13:08:04Z","updated":"2026-09-25T13:08:04Z","authors":["Vincent R. Martinez","Sarah Strikwerda","Xiang Wan"],"categories":["math.OC","math.AP"],"headline":"New method improves fluid viscosity identification in 2D Navier-Stokes equations","summary":"Researchers developed the first excitation theory for a nonlinear equation to better estimate fluid viscosity in two-dimensional incompressible flows. They introduced a computable condition and used data assimilation to handle unknown initial states, focusing on minimizing errors between observed and modeled fluid velocities. Their analysis centers on a new functional measuring work done in the system, which helps identify when parameter estimates are reliable or when observations lack sensitivity. These findings are theoretical and rely on mathematical properties and simulations."},{"id":"2609.31229v1","title":"From Image Morphology to Multiphysics Earth Posteriors","abstract":"Structural learning and physical-property inference can draw on different sources of knowledge. We learn morphology from 33,140 natural, satellite, and texture images, without geological training models, then infer Earth properties from physical observations. Identical frozen weights map 128 coordinates into each $256\\times256$ compressional-velocity, shear-velocity, or density field. Fixed adapters assign physical meaning; observation-specific likelihoods determine new posteriors without retraining. Across 256 unseen Earth models, image coordinates reduce projection error fourfold relative to a dimension-matched cosine basis. Matched inversions favor image coordinates for the tested curved-velocity targets and a smooth radial basis for the tested faulted targets. A simplified 2.5-dimensional application with 915 measured observations along one 120.9-km southern-California profile constrains a velocity--density contrast. Grouped arrival-time and gravity errors fall by 43% and 44% relative to the pretrained image prior, although the matched smooth radial basis predicts better. Direct layered-Rayleigh updating reduces blind phase-velocity error by 28% at five sites, whereas P-and-gravity-only sharing worsens it, distinguishing reusable morphology from beneficial cross-property coupling. These results establish a practical separation: images supply structural alternatives, and heterogeneous physical observations determine their experiment-specific posterior probabilities.","published":"2026-09-25T13:07:42Z","updated":"2026-09-25T13:07:42Z","authors":["Ziye Yu","Xin Liu","Yuqi Cai"],"categories":["physics.geo-ph","astro-ph.EP"],"headline":"Researchers link image patterns to Earth's physical properties without retraining","summary":"The researchers used over 33,000 natural and satellite images to understand shapes (morphology) without relying on geological models, then inferred Earth's physical properties like velocities and density from real observations. They developed a method mapping coordinates to velocity and density fields, improving prediction accuracy compared to traditional approaches. Testing on Earth models and real California data showed this image-based approach helps separate structural possibilities from physical measurements, though some other methods still performed better in certain cases. Their findings come from simulations and measurements along a profile, indicating a way to connect structural images to Earth properties."},{"id":"2609.31228v1","title":"Joule-Profiler: Profiling the Energy Consumption of Build Automation Tools Made Easy","abstract":"Build pipelines are integral to modern software development, yet their energy footprint remains largely invisible to practitioners. Existing CI energy tools either rely on model-based estimation (due to hardware access restrictions in cloud runners) or report only total pipeline energy without decomposing it into meaningful phases. Joule-Profiler is an open-source command-line tool for Linux that measures hardware energy consumption via Intel RAPL (CPU), NVML (NVIDIA GPU), and attributes it to user-defined program phases by monitoring standard output. In this tool paper, we demonstrate Joule-Profiler in the context of Maven build pipelines using Google Gson as a case study. By applying a token pattern-matching Maven plugin invocations, we analyze builds across the 5 most recent Gson releases into per-phase energy profiles, comparing cold builds (empty local repository) and warm builds (cached dependencies).","published":"2026-09-25T13:07:41Z","updated":"2026-09-25T13:07:41Z","authors":["Jérémy Woirhaye","François Gibier","Romain Rouvoy"],"categories":["cs.SE"],"headline":"New Tool Measures Energy Use in Software Build Processes","summary":"Researchers developed Joule-Profiler, an open-source tool that tracks hardware energy use during software build steps on Linux systems. Unlike previous tools, it breaks down energy consumption by specific build phases using existing hardware sensors. They tested it on Maven builds for a software library, comparing fresh and cached builds to produce detailed energy profiles. The findings come from this case study and show how Joule-Profiler can help understand where energy is used in build pipelines, though results are specific to the tested scenarios."},{"id":"2609.31227v1","title":"Charged Letelier--Alencar Black Strings with a Quintessence and the Cosmolgical Constant","abstract":"This study derives charged black string solutions in the presence of a Letelier--Alencar cloud of strings and quintessence within Einstein--Maxwell--$Λ$ theory. We study the curvature scalars in order to investigate the asymptotical behavior of these solutions. We calculate the relevant conserved and thermodynamic quantities, confirming their compliance with the first law of thermodynamics. System stability is rigorously analyzed through heat capacity and Gibbs free energy, identifying regimes of local and global equilibrium. Additionally, by interpreting the cosmological constant as thermodynamic pressure within an extended phase space, we establish the validity of the extended first law and derive the corresponding Smarr relation for these solutions. We also investigate the null geodesic structure, deriving the equation for the photon cylinder radius, which generalizes the known result for charged black strings by incorporating the effects of the Letelier--Alencar cloud of strings and quintessence. The numerical analysis of this equation determines the existence and location of photon cylinders as functions of the model parameters.","published":"2026-09-25T13:06:50Z","updated":"2026-09-25T13:06:50Z","authors":["B. Eslam Panah","F. M. da Silva","L. G. Barbosa"],"categories":["gr-qc"],"headline":"Researchers Explore New Charged Black String Models with Dark Energy Effects","summary":"The researchers derived new theoretical solutions describing charged black strings influenced by a special cloud of strings and a form of dark energy called quintessence. They analyzed properties like stability, thermodynamics, and light paths near these objects, confirming they obey expected physical laws in this theoretical setup. Their findings are based on mathematical modeling and numerical analysis, expanding understanding of black string behavior under these conditions but remain theoretical without experimental confirmation."},{"id":"2609.31226v1","title":"ATLASGAL: The HCN-IR Star Formation Relation for the Milky Way","abstract":"We examine the relation between star formation and dense gas, traced by HCN(1-0), using a sample of 343 Galactic clumps spanning the full range of evolutionary stages representative of high-mass and cluster formation, including quiescent, protostellar, young stellar object (YSO), and HII regions. Earlier Galactic studies, based primarily on luminous HII regions, reported a linear correlation between infrared (IR) and HCN luminosities, consistent with extragalactic surveys, and interpreted it as a universal star formation law. Our larger evolutionary sample yields a significantly steeper slope of $1.80\\pm0.08$, inconsistent with extragalactic trends. However, we find that each of the star-forming evolutionary stages individually has a shallower slope ($\\sim$1.2) than that of the full sample, with quiescent clumps showing no correlation. We estimate the Milky Way's total IR and HCN luminosities from dense clumps by integrating the bolometric luminosities and clump masses of Hi-GAL sources. Both fall significantly below extragalactic values, demonstrating that clumps alone cannot account for the IR-HCN relation. Only $\\sim$5-25 per cent of Galactic HCN luminosity arises from dense clumps, implying that most extragalactic HCN traces gas that is not involved in star formation. Additionally, dense clumps contribute only $\\sim$2-3 per cent of the Milky Way's IR luminosity, which is dominated instead by extended HII regions and PDRs tracing longer star formation timescales. As a result, the observed linear relation in nearby galaxies reflects a nearly constant star formation efficiency averaged over large areas and long timescales, rather than a direct link between SFR and dense gas mass.","published":"2026-09-25T13:06:30Z","updated":"2026-09-25T13:06:30Z","authors":["I. I. Grozdanova","J. S. Urquhart","W. -J. Kim","F. Wyrowski","T. Csengeri","A. Karska","A. Giannetti","D. Colombo","D. Eden","M. A. Thompson","T. J. T. Moore"],"categories":["astro-ph.GA","astro-ph.SR"],"headline":"Star Formation and Dense Gas in the Milky Way Show Complex Links","summary":"Researchers studied 343 regions in our galaxy where stars form, focusing on how dense gas relates to star formation. They found that the connection between gas traced by HCN and infrared light, a star formation indicator, is more complex than previously thought and differs from patterns seen in other galaxies. Their findings suggest that only a small part of the galaxy's dense gas and infrared light come from active star-forming clumps, and that broader galactic processes affect the observed relations over large areas and long times."},{"id":"2609.31225v1","title":"Imp-ACT: Adaptive Impedance Control and Action Chunking with Transformers to Learn Contact-Rich Manipulation from Demonstrations","abstract":"Contact-rich manipulation requires robots to balance accurate motion tracking with compliant interaction, yet most visual-action policies leave compliance fixed at the controller level. We present Imp-ACT, a methodologically grounded and practical approach to incorporating direction-dependent Cartesian stiffness modulation directly into demonstration collection, without manual stiffness selection or offline target reconstruction. During teleoperation, a self-tuning impedance controller adapts stiffness along the instantaneous direction of motion while maintaining compliance in orthogonal directions. The adapted stiffness is applied and recorded alongside visual observations and motion commands, capturing motion and compliance under the same dynamics. We implement this pipeline using Action Chunking with Transformer (ACT) to predict end-effector pose, gripper action, and motion-direction stiffness from visual, proprioceptive, and wrench observations. The performance of Imp-ACT is evaluated on wiping and plug insertion using both success rate and quantitative measures of contact behavior. Compared with fixed low- and high-stiffness baselines, Imp-ACT achieves comparable or higher success while maintaining low interaction forces. In wiping, it reduces contact-force vibration by approximately $29\\times$ relative to the compliant baseline and $180\\times$ relative to the stiff baseline. In plug insertion, it reduces forces orthogonal to the insertion direction by $43\\%$ relative to the better fixed-stiffness baseline. These results highlight the benefit of maintaining sufficient stiffness along the direction needed for task execution while preserving compliance in other directions to limit contact forces and accommodate environmental constraints.","published":"2026-09-25T13:04:37Z","updated":"2026-09-25T13:04:37Z","authors":["Luca Zanetti","Doganay Sirintuna","Idil Ozdamar","Pietro Balatti","Heng Zhang","Arash Ajoudani"],"categories":["cs.RO"],"headline":"New Robot Method Balances Firmness and Flexibility in Touch Tasks","summary":"Researchers developed Imp-ACT, a robot control method that adjusts how stiff or flexible a robot's movements are based on the direction of motion during tasks involving contact, like wiping or plug insertion. By tuning this 'stiffness' during demonstration, without manual setup, the robot learns to apply the right amount of force, improving success and reducing excess contact forces. Tests show Imp-ACT lowers force vibrations and unwanted pressure compared to fixed stiffness controls. These findings come from controlled experiments and highlight the benefit of adaptive stiffness in robotic manipulation."},{"id":"2609.31224v1","title":"Acoustic-to-Text KV Compression for Full-Duplex Speech Models","abstract":"Full-duplex speech language models continuously accumulate acoustic key-value (KV) states, making long-running interactions memory-intensive. During listening, the model can finish processing an audio unit before the next arrives; we term the remaining interval listening-time slack. We propose acoustic-to-text KV compression, which introduces a transcription side channel to convert incoming speech into compact textual memory within this interval. When the cache exceeds a target budget during inference, older acoustic states are evicted while transcripts and recent acoustic context remain. We train the side channel with LoRA using cross-entropy on transcription segments. To preserve listening and speaking behavior, we apply knowledge distillation to the original model's token-level output distributions at native prediction positions. On ten-minute LongSpeech sessions, our MiniCPM-o 4.5 implementation reduces peak streaming KV-cache size by 64.6% compared with the same model without eviction. The proposed method also improves transcription, temporal question answering, and summarization over the baseline. Full-Duplex-Bench evaluations further show comparable pause-handling, turn-taking, and interruption performance.","published":"2026-09-25T13:04:09Z","updated":"2026-09-25T13:04:09Z","authors":["Yejin Lee","Seungbeom Kim","Yongha Lee","Kyuhong Shim"],"categories":["cs.SD","cs.AI","eess.AS"],"headline":"Reducing Memory Use in Real-Time Speech Models with Text Compression","summary":"Researchers addressed the problem of memory buildup in continuous speech models that keep acoustic data over time. They developed a method that converts spoken audio into compact text summaries during brief waiting periods, allowing older audio data to be removed while retaining recent context and transcripts. Their experiments with ten-minute speech sessions show this approach reduces memory use by over 60% and improves transcription and understanding tasks, while maintaining natural conversation behaviors like pauses and turn-taking, according to their tests."},{"id":"2609.31223v1","title":"Quantum interaction can superactivate cheating under parallel repetition","abstract":"We study interactive multiprover games and many-round protocols in which the communication between the verifier and the provers is quantum. Parallel repetition is known to suppress soundness error of classical two-prover games arbitrarily close to zero, as shown by Raz (STOC '95), even with quantum strategies as shown by Yuen (ICALP '16), and Bavarian, Vidick and Yuen (STOC '17). Yet, we show that quantum communication can have the opposite, unexpected effect. Specifically, we exhibit a one-round quantum game and two-round $\\mathsf{QMIP}$ protocols whose local (unentangled) value is strictly less than one for a single instance, yet equals one under $n$-fold parallel repetition for every $n \\geq 2$. The key is that parallel repetition not only imposes additional winning conditions but also supplies additional quantum resources as entanglement in the exchanged states can be exploited jointly across copies. Our multiround examples build on superactivation of zero-error capacities of quantum channels. To establish this connection, we introduce quantum games and interactive protocols which capture the one-shot zero-error classical and quantum capacities, both with and without entanglement assistance. Our constructions use quantum-state verification and a teleportation-based reduction from two rounds to one. In contrast, when shared entanglement is allowed, we show that parallel repetition cannot increase the entangled value when at most $3$ messages are exchanged, consistent with classical results of Bellare, Impagliazzo and Naor (FOCS '97), and that of single-prover systems by Kitaev and Watrous (STOC '00). Combining this monotonicity with our quantum game-channel correspondence, we show that entanglement-assisted zero-error classical and quantum capacities cannot be superactivated.","published":"2026-09-25T13:03:16Z","updated":"2026-09-25T13:03:16Z","authors":["Archishna Bhattacharyya","Laura Mančinska","Yuming Zhao"],"categories":["quant-ph","cs.CC"],"headline":"Quantum communication can unexpectedly help cheating in repeated games","summary":"Researchers studied games where a verifier communicates with multiple provers using quantum messages. Classical results show repeating a game reduces cheating chances, but the researchers found cases where repeating quantum communication games actually allows perfect cheating across repetitions, even if a single game has less than perfect cheating. This happens because repeated quantum interactions create new combined resources. Their results highlight surprising differences between quantum and classical communication in multi-round protocols, though entanglement-assisted cases behave more like classical ones. These findings are based on theoretical constructions and mathematical analysis."},{"id":"2609.31222v1","title":"Budgeted Quotient-Residual Guidance for Frozen Pocket-Conditioned Molecular Diffusion","abstract":"Pocket-conditioned molecular diffusion updates ambient atom coordinates, but many lead-optimization objectives are expressed on quotient features such as distances, contacts, and anchored substructures. We introduce budgeted quotient-residual guidance (QRG), an inference-time correction that makes these quotient objectives active without retraining the molecular generator. QRG lifts quotient covectors to metric-horizontal ambient directions and delivers them through a trust budget set by the frozen sampler's own step norm: quotient geometry chooses the direction, while sampler motion bounds the scale. We derive the horizontal lift, closed-form sampler-budget update, KL/kinetic interpretation around a frozen reverse step, equivariance conditions, and a product-budget split for budget-capped section and residual controls. Controlled quotient tasks confirm that sampler-relative delivery activates signals that raw local quotient gradients leave dormant. On frozen TargetDiff backbones, official seed-0 CBGBench ligand-generation/editing sweeps show practical quality-runtime gains: Local-QRG improves validity from 0.815 to 0.864 on fragment growing, 0.664 to 0.707 on scaffold hopping, and 0.681 to 0.712 on linker design, while PredNext-QRG improves fragment/scaffold and remains near-neutral on linker. Novelty remains 1.000 and diversity is preserved in the matched multi-seed molecular slice, giving task-dependent improvements without sampler retraining or backbone modification. Overall, QRG provides a lightweight route to quotient-aware inference for frozen molecular samplers with explicit runtime accounting.","published":"2026-09-25T13:02:11Z","updated":"2026-09-25T13:02:11Z","authors":["Xinyu Wang","Jinbo Bi","Minghu Song"],"categories":["cs.LG"],"headline":"New method improves molecular design without retraining models","summary":"The researchers address challenges in molecular design where optimization goals focus on relationships like distances or contacts between atoms, rather than direct atom positions. They introduce a method called budgeted quotient-residual guidance (QRG) that adjusts the design process during inference to better account for these relational goals, without retraining the molecular generation model. Their tests show QRG improves the quality of generated molecules in tasks like fragment growing and scaffold hopping, while maintaining diversity and novelty, using only small runtime adjustments."},{"id":"2609.31221v1","title":"Distinct Feedback-Strength Requirements for Quantum-State Ensemble Preparation under Channel-Equivalent Monitoring","abstract":"Different measurements of the same environment can preserve the average quantum channel yet change the feedback strength needed to prepare a state ensemble. We establish this distinction for a multiqubit register under complete Pauli monitoring, targeting the uniform ensemble of real-amplitude pure states. With unit detection efficiency, public records, and bounded Hamiltonian feedback, fixed transverse noise requires a minimum peak strength proportional to $ε^{-1}$ at distributional error $ε$. Tangent noise instead admits a construction requiring only $O[\\sqrt{\\log(1/ε)}]$. At fixed preparation time, we determine the optimum over all allowed feedback policies and derive a geometric lower bound beyond radial symmetry. The separation identifies the measurement realization as part of the control-resource specification for monitored ensemble preparation.","published":"2026-09-25T13:01:52Z","updated":"2026-09-25T13:01:52Z","authors":["Haitao Huang","Qinglin Zhao","Xiaoyu Li"],"categories":["quant-ph"],"headline":"How Measurement Methods Affect Quantum State Preparation Effort","summary":"Researchers studied how different ways of measuring a quantum system's environment influence the effort needed to prepare specific quantum states. They found that while overall quantum channel effects stay the same, the required feedback to control the system varies greatly depending on the measurement method used. Their findings, based on theoretical analysis of multi-qubit systems and feedback controls, highlight that the choice of measurement technique impacts the resources needed for quantum state preparation."},{"id":"2609.31220v1","title":"A unitary invariant for $C^*$-algebras generated by isometries with a twisted commutation relation","abstract":"We investigate the classification of $C^*$-algebras generated by pairs of isometries $(V_1, V_2)$ satisfying $V_1V_2=qV_2V_1$, where $q=e^{2πiθ}$ with $θ$ irrational, up to unitary equivalence. Using a functional model that realizes pure $q$-commuting pairs on vector-valued Hardy spaces in terms of a uniquely associated triple $(\\mathcal{F}, P, U)$ consisting of Hilbert space $\\mathcal{F}$, an orthogonal projection $P$ and a unitary operator $U$, we analyze the internal structure of $C^*(V_1,V_2)$ via an extension of the $C^*$-algebra generated by its minimal $q$-commuting unitary extension by the defect ideal $\\mathcal{I}_{\\mathcal{D}} =\\langle [V_1^*, V_1],[V_2^*, V_2]\\rangle$. For pairs that are essentially doubly $q$-commuting, we show that $PUP|_{\\operatorname{ran} P}$ is semi-Fredholm and prove that $|\\operatorname{ind} PUP|_{\\operatorname{ran} P}|$ is invariant under unitary equivalence of the corresponding $C^*$-algebras, thereby providing an explicit criterion for distinguishing non-equivalent pairs. This work extends the seminal 1978 results of Berger, Coburn, and Lebow for commuting isometries, while addressing the key technical nuances of this noncommutative setting: most notably, navigating the failure of Gelfand theory due to noncommutativity and replacing the classical Toeplitz extension by an extension of the noncommutative torus by the ideal of compact operators.","published":"2026-09-25T13:01:17Z","updated":"2026-09-25T13:01:17Z","authors":["Sourav Ghosh","Haripada Sau"],"categories":["math.OA","math.FA"],"headline":"Classifying special operator pairs with a new invariant tool","summary":"Researchers study certain mathematical objects called pairs of isometries that almost commute but with a twist defined by an irrational angle. They develop a way to represent and analyze these pairs through associated mathematical structures, leading to a new invariant number that helps distinguish different cases. Their work extends older results from simpler, commuting situations to this more complex noncommuting setting, but the findings are theoretical and focus on understanding the internal structure of the related operator algebras."},{"id":"2609.31219v1","title":"Research with AI Agents: How Agentic Systems Are Changing Scientific Work","abstract":"Background. Agentic AI systems independently decompose tasks such as literature search, data analysis, and programming into subtasks, search the web, access databases, and execute code. This allows them to perform digital research tasks at high speed. Objectives. Under what conditions does the use of agentic systems produce reliable efficiency gains, and which tasks remain with researchers? Materials and methods. Summary of current studies on literature searches, data analysis, software development, and clinical decision support. Results. For digital activities, work shifts from execution to steering and review. Efficiency gains are greatest when expected behavior can be formalized in advance and tested automatically. In complex agentic systems, recorded sequences of reasoning steps and tool calls can quickly become too extensive for human review. Furthermore, explanations generated by the model do not reliably reflect how an output was produced. One possible step toward more reliable systems is the validation of individual components. The limited reviewability extends beyond research itself; the review of scientific articles and grant proposals is also reaching capacity limits. In pathology, curated and annotated data, researchers' own analytical skills, and institutional exchange of experience are becoming increasingly important. Conclusions. Researchers remain responsible for their results. They must determine what to delegate and how to review the results. The importance of a research question to patients, the field, and society cannot be fully assessed using formalized criteria and remains a matter of expert judgment. Agentic systems can free up time for this.","published":"2026-09-25T13:01:02Z","updated":"2026-09-25T13:01:02Z","authors":["Johannes Lotz","Markus Wenzel"],"categories":["cs.CY"],"headline":"AI Agents Shift Research Focus from Doing to Guiding Tasks","summary":"Researchers studied how AI systems that can independently handle parts of research tasks like searching literature or analyzing data change scientific work. They found that these AI tools speed up digital tasks when the work steps can be clearly defined and tested, but complex reasoning steps by the AI become hard for humans to review fully. The researchers emphasize that humans must still guide, check, and judge important aspects of research, as AI explanations may not fully reveal how results were produced."},{"id":"2609.31218v1","title":"POPT: Physics-Informed Deep Learning for Phase-Only Positioning in Distributed MIMO","abstract":"This article develops a physics-informed deep learning framework for single-snapshot 3D phase-only positioning in narrowband phase-coherent distributed multiple-input multiple-output (D-MIMO) networks under two-ray propagation. Unlike prior phase-coherent D-MIMO localization methods that largely assume line-of-sight (LoS)-only channels, we consider a LoS path and a specular ground reflection. Since the narrowband single-snapshot observation cannot resolve these components in delay, their coherent superposition induces structured perturbations in the carrier phase measurements. To address this, we develop a Gaussian process (GP)-based model that learns the quasi-periodic phase distortion caused by ground reflection from carrier phase measurements collected at a small number of training locations and uses it to generate high-quality synthetic samples. Building on these samples, we propose the Phase-Only Positioning Transformer (POPT), an encoder-only transformer that captures inter-antenna point (AP) phase relationships without solving the highly non-convex maximum-likelihood (ML) problem underlying model-based estimators. We also derive the fundamental position error bound (PEB) and develop maximum-likelihood estimation (MLE) for the considered multipath phase-only D-MIMO positioning problem. Numerical results show that, with only 50 GP training locations, the proposed method achieves near-PEB accuracy. The analytical MLE is highly sensitive to relative permittivity mismatch, whereas the proposed method mitigates this dependence by learning the phase perturbation directly from measurement data. Compared with MLE, the proposed approach also reduces floating-point operation (FLOP) complexity and inference time by about 1.7 and 3.6 orders of magnitude, respectively.","published":"2026-09-25T12:59:04Z","updated":"2026-09-25T12:59:04Z","authors":["Fatih Ayten","Ossi Kaltiokallio","Jukka Talvitie","Akshay Jain","Mehmet C. Ilter","Musa Furkan Keskin","Elena Simona Lohan","Henk Wymeersch","Mikko Valkama"],"categories":["eess.SP"],"headline":"New Deep Learning Method Improves 3D Positioning with Limited Signals","summary":"Researchers developed a deep learning approach called POPT to locate devices in 3D using limited phase-only signals in wireless networks that experience signal reflection from the ground. Unlike traditional methods assuming direct signal paths, their method learns complex signal distortions from real measurements, allowing it to create realistic synthetic data for training. Their results show that POPT can accurately estimate positions with less computational effort and less sensitivity to environmental changes compared to classical mathematical approaches. The findings are based on simulations and measured data from a limited number of locations."},{"id":"2609.31217v1","title":"Deep Reinforcement Learning for Misbehavior Detection Under Partially Observable V2X Data","abstract":"Misbehavior detection in vehicle-to-everything (V2X) systems is essential for ensuring the semantic correctness of exchanged messages and preventing the dissemination of falsified information. Existing data-centric misbehavior detection approaches largely rely on statistical validation or supervised machine learning models under the implicit assumption of fully observable V2X streams. In practice, however, vehicular environments are inherently partially observable due to hardware failures, intermittent connectivity, and environmental occlusions. Moreover, missingness itself can be strategically exploited by adversaries to evade detection. In this paper, we study misbehavior detection under incomplete V2X observations and propose a deep reinforcement learning (DRL)-based detection framework that learns adaptive policies with incomplete data. We further introduce an adversarial threat model in which attackers exploit or deliberately induce missingness to evade detection, including evasion via natural occlusions and adversarial feature suppression. Extensive experiments conducted on the VeReMi dataset under various missingness patterns demonstrate that DRL significantly outperforms a powerful XGBoost baseline under natural partial observability. However, results also reveal a critical vulnerability: DRL policies can be highly susceptible to evasion attacks that strategically exploit natural missingness. In contrast, DRL exhibits more gradual degradation under direct feature suppression compared to static tree-based models.","published":"2026-09-25T12:58:02Z","updated":"2026-09-25T12:58:02Z","authors":["Roshan Sedar","Charalampos Kalalas"],"categories":["cs.NI"],"headline":"Deep Learning Helps Detect Fake Vehicle Messages Despite Missing Data","summary":"Researchers studied how to spot false or misleading messages in vehicle communication systems, even when some data is missing due to hardware or connectivity issues. They developed a deep reinforcement learning method that adapts to incomplete information and tested it on a dataset simulating real-world conditions. Their results show this method works better than traditional approaches but can still be tricked by attackers who cleverly exploit missing data, revealing important challenges for future safety efforts."},{"id":"2609.31216v1","title":"Koebe uniformization, John domains, and transboundary modulus","abstract":"We prove Koebe's conjecture for John domains by showing that every planar John domain is conformally equivalent to a circle domain. Our proof is based on a geometric property of John domains, which we prove to be sufficient for uniformization by a circle domain. As a consequence, we obtain an alternative proof of Koebe's conjecture for inner uniform and Gromov hyperbolic domains, recovering results established in earlier work of the authors. Finally, we derive a necessary and sufficient condition for a domain to satisfy Koebe's conjecture by using Schramm's transboundary modulus and well-behaved exhaustions of circle domains constructed by the second author and Rajala.","published":"2026-09-25T12:56:46Z","updated":"2026-09-25T12:56:46Z","authors":["Christina Karafyllia","Dimitrios Ntalampekos"],"categories":["math.CV"],"headline":"Researchers prove Koebe's conjecture for specific plane shapes","summary":"The researchers addressed Koebe's conjecture for John domains, a class of shapes in the plane, by showing these shapes can be transformed into circle domains through conformal mapping. They used geometric features unique to John domains to support this proof. Their work also provides an alternative proof for other related domains and offers a precise condition for when Koebe's conjecture holds, based on advanced mathematical tools. These results are theoretical and rely on complex geometric and analytic methods."},{"id":"2609.31215v1","title":"DIAL: Position-Debiased LLM Judges with Adaptive Human Preference Calibration","abstract":"Large language models (LLMs) as a judge enable scalable evaluation, but their judgments can be sensitive to response order and, even after removing such position effects, can still diverge systematically from human preferences.We introduce DIAL, a unified framework that combines abundant LLM comparisons with limited human comparisons to separate judge-specific position effects, learn shared structure in position-debiased LLM preferences, and adaptively calibrate that structure toward the human preference target. Theoretically, we study three aspects of DIAL: (i) identification of latent LLM preferences, position effects, and human calibration; (ii) adaptive estimation that balances LLM anchoring against limited human evidence; and (iii) fixed-weight uncertainty quantification for the calibrated human preference. Empirically, we evaluate position debiasing and human alignment separately in controlled simulations and on three human-preference benchmarks, showing that DIAL remains robust to unbalanced response order, achieves strong human-aligned rankings with limited labels, and adapts toward human evidence when LLM information is imperfect. Our real-data study collects over 410K judgments from 21 LLM judges in both display orders, providing a resource for future studies of LLM-judge bias, heterogeneity, and human alignment.","published":"2026-09-25T12:56:36Z","updated":"2026-09-25T12:56:36Z","authors":["Zesheng Cai","Yingqi Fan","Sichang Chen","Jin-Hong Du"],"categories":["cs.AI","stat.AP","stat.ME","stat.ML"],"headline":"New Method Helps AI Judge Responses Like Humans Do","summary":"Researchers developed DIAL, a system to improve how AI language models judge responses by adjusting for biases caused by answer order and differences from human preferences. They combined many AI comparisons with fewer human judgments to better match AI rankings to what people prefer. Their tests show DIAL can correct biased rankings and align AI judgments closer to human opinions, even with limited human data. This work mainly focuses on methods and simulations rather than real-world applications."},{"id":"2609.31214v1","title":"Which Influence Are We Estimating? The Role of Counterfactual Specifications in Data Attribution","abstract":"Estimating the influence of training examples on model behavior is essential for data debugging, valuation, and attribution. Existing influence estimators often produce incompatible rankings, which are commonly ascribed to approximation error. We argue that a more fundamental source of disagreement is specification mismatch: influence depends on the behavior being attributed, the intervention applied to each training example, and the counterfactual training process that maps the intervention to a model response. These choices are especially important when the target behavior requires a tractable surrogate, such as query loss, a logit, or a margin. We formalize influence as a counterfactual estimand, distinguish specification mismatch across estimands from approximation error in estimating a fixed estimand, and organize representative estimators by their implied specifications. We further derive a local decomposition that exposes how behavior signals, training signals, and counterfactual parameter responses interact. Controlled experiments show that exact estimands under different specifications can induce different rankings, whereas approximation error grows as perturbations move farther from their linearization points. Experiments on noisy label detection and LLM attribution show that specification choices significantly affect attribution quality, especially for the choice of behavior surrogate. Behavior-aligned specifications can identify target-specific training examples obscured by default loss-based or similarity-based specifications. These results establish specification analysis as a necessary first step for interpreting and comparing data influence estimators.","published":"2026-09-25T12:56:12Z","updated":"2026-09-25T12:56:12Z","authors":["Zhe Li","Wei Zhao","Peixin Zhang","Jun Sun"],"categories":["cs.AI","cs.LG"],"headline":"Understanding How We Measure Data Influence on AI Models","summary":"Researchers studied how different methods estimate the influence of training data on AI model behavior. They found that disagreements often come from how the influence is defined and measured, not just errors in calculation. Their experiments show that choices in these definitions significantly affect results, suggesting careful consideration is needed when interpreting such influence estimates."},{"id":"2609.31213v1","title":"The Paper-Stack Trampoline: Acoustic Restitution in Layered Media","abstract":"A ball bouncing on a thicker cushion should bounce less. We show that this is not always so . Measuring the coefficient of restitution $e$ of a steel ball dropped on a stack of $N$ sheets of standard A4 printer paper, we find that $e$ first drops, then rises to a pronounced maximum, and only then decreases towards a minimum. The maximum occurs where the acoustic round trip through the stack matches the contact time of the impact. For a contact set by the stack itself, one recovers the classical bar-impact condition of a stack-to-ball mass ratio of order one. Bar impact, however, predicts only a weak recovery, insensitive to the surrounding gas. For paper stacks, the recovery is large and no maximum is resolved when the stack is evacuated: the air trapped between the sheets, not the paper alone, makes the stack a trampoline.","published":"2026-09-25T12:55:33Z","updated":"2026-09-25T12:55:33Z","authors":["Manou Liesker","Colton Kawamura","Maria Kieft","Joshua A. Dijksman","Antoine Deblais"],"categories":["cond-mat.soft"],"headline":"Paper stacks can make a ball bounce higher than expected","summary":"Researchers studied how a steel ball bounces on stacks of printer paper and found that the bounce height doesn't just decrease with thicker stacks as expected. Instead, the bounce first lowers, then rises to a peak before falling again. They discovered that trapped air between the paper sheets plays a key role, acting like a trampoline and boosting the bounce. These findings come from controlled experiments and measurements, highlighting the surprising effects of air cushioning in layered paper stacks."},{"id":"2609.31212v1","title":"An all-dimensional Lewy theorem for pluriharmonic mappings via Milnor monodromy","abstract":"Hengartner asked whether a pluriharmonic mapping between equidimensional complex Euclidean domains is locally one-to-one precisely at the points where its real Jacobian is nonzero. Naser had earlier claimed the conclusion in complex dimension two, and a recent theorem of Kalaj supplies a complete proof in that dimension. We prove the result in every complex dimension. The main ingredient is a singularity-theoretic characterization which is of independent interest: if $q\\colon(\\C^n,0)\\to(\\C,0)$ is a nonconstant holomorphic germ, then $dq(0)\\ne0$ if and only if one (equivalently, every) phase hypersurface \\( \\{\\Repart(e^{-iθ}q)=0\\} \\) is a rational homology $(2n-1)$-manifold at the origin. The difficult direction combines local conical structure, Alexander duality, the Milnor fibration, and A'Campo's vanishing theorem for the Lefschetz number of local monodromy. The argument requires neither an isolated critical point nor a reducedness hypothesis. It follows that the real critical set of any pluriharmonic self-dimensional mapping agrees exactly with its local noninjectivity set. As a quantitative application, combining our theorem with Kalaj's boundary regularity theorem shows that, in every complex dimension, a quasiconformal pluriharmonic homeomorphism of the unit ball onto a bounded $C^1$-Dini domain is bi-Lipschitz.","published":"2026-09-25T12:54:28Z","updated":"2026-09-25T12:54:28Z","authors":["Deguang Zhong","Zhi-Gang Wang"],"categories":["math.CV"],"headline":"New math theorem clarifies behavior of complex pluriharmonic maps","summary":"Researchers addressed whether pluriharmonic mappings between complex spaces are locally one-to-one exactly where their real Jacobian is nonzero, confirming this holds in all complex dimensions. They used advanced tools from singularity theory and topology, such as Milnor fibrations and monodromy, to prove this without restrictive conditions previously needed. Their work also links the set of points where these mappings fail to be injective with the real critical points, and shows certain smooth mappings of the unit ball are bi-Lipschitz. The results are theoretical and rely on complex mathematical structures."},{"id":"2609.31211v1","title":"CoralPlan: Observation Skill Selection and Execution for Underwater Robotic Inspection","abstract":"Underwater robotic inspection depends on acquiring views that reveal task-relevant structure. For a structurally complex coral colony, recognising the target is only the starting point: the robot must select and execute a viewing motion suited to the inspection task. We present CoralPlan, a vision-language system that selects an observation skill from a current camera image and task text supplied by an episode manifest. A shared motion interface executes orbit, patch, or survey as target-relative trajectories; the remaining plan fields provide operator guidance. Observation completion requires target keeping and primitive-specific coverage, while joint success also requires selection to match the recorded reference. We evaluate this interface in 144 simulated episodes and 36 matched simulation-hardware pairs. In a clear-water pool with external target-reference poses, hardware observation completion reaches 77.8% and joint success reaches 63.9%. The experiments identify both reference-mismatched completions and incomplete observations after a matching skill selection. These results connect observation-skill choice to measurable underwater execution outcomes and identify where task-directed acquisition succeeds or fails.","published":"2026-09-25T12:53:55Z","updated":"2026-09-25T12:53:55Z","authors":["Yuer Gao","Yu Zhao","Yi Cai"],"categories":["cs.RO"],"headline":"New system helps robots inspect coral underwater more effectively","summary":"Researchers developed CoralPlan, a system that helps underwater robots choose how to move their cameras to inspect complex coral structures based on images and task instructions. The system selects specific viewing motions and controls the robot to keep the target in view and cover required areas. Tested mainly in simulation and pool experiments, CoralPlan showed it could successfully complete observations about 78% of the time, revealing strengths and challenges in underwater robotic inspections."},{"id":"2609.31210v1","title":"Black silicon avalanche photodiode with dopant-free multiplication region achieves >98% EQE","abstract":"Conventional silicon avalanche photodiodes (Si APDs) rely on highly doped regions to enable impact ionization and achieve internal gain. However, inefficient charge collection in these regions, combined with front surface reflection, degrades the external quantum efficiency (EQE) of these devices. In this work, we mitigate both loss mechanisms by integrating black silicon (b-Si) surface nanostructuring with Al2O3 induced carrier collection in an inverted-mesa Si APD architecture. This design confines the electric field and defines the multiplication region without requiring ion implantation. The resulting devices display near-ideal responsivity (at M = 1) across the UV-visible spectrum, with EQE exceeding 100% at 200-400 nm (peak ~130%) and exhibiting 92-100% at 400-700 nm. Avalanche gains of M ~ 4 at 80 V, M ~ 10 at 100 V, and M ~ 23 near breakdown at ~110 V are obtained. Dark current remains in the picoampere range up to ~80 V but rises to the nanoampere range near breakdown, narrowing the practical bias window. Capacitance-limited rise times of ~30-570 ns are measured for device diameters of 1-5 mm. The results demonstrate that the developed architecture enables APDs capable of detecting every single photon over wide range of wavelengths.","published":"2026-09-25T12:52:12Z","updated":"2026-09-25T12:52:12Z","authors":["O. Leiviskä","O. Setälä","A. Haarahiltunen","J. Heinonen","V. Vähänissi","H. Savin"],"categories":["physics.app-ph"],"headline":"New silicon photodiode design improves light detection efficiency","summary":"Researchers developed a silicon avalanche photodiode (APD) that avoids traditional doping, using black silicon surface structures and a special coating to better collect electric charges. Their design reduces light reflection and improves charge collection, resulting in very high efficiency exceeding 98% across much of the visible light spectrum. Tests showed strong internal amplification of signals, but performance decreases near breakdown voltage, limiting operating range. These findings are based on reported measurements and simulations of the new device structure."},{"id":"2609.31209v1","title":"Bypassing no-go theorems on mixed classical-quantum systems: the (counter)example of hybrid van Hove theory","abstract":"Motivated by models of mixed quantum-classical hybrid theories, we consider the representation of classical observables as Hilbert-space operators based on van Hove's unitary representation of contact transformations. This leads to a reformulation of classical mechanics where the phase space density satisfies the Liouville equation and, more importantly, classical observables are represented by van Hove operators that satisfy a commutation algebra isomorphic to the classical Poisson algebra of functions in phase space. Thus, this Hilbert space approach captures all the essential features of the phase space formulation of classical mechanics in that it reproduces its algebraic structure -- something that is not achieved by other approaches such as the Koopman-von Neumann theory. As an important feature, the definition of appropriate physical states in this formalism imposes certain natural requirements on the phase of the classical wave function, which ensure correct expectation values for the observables. Furthermore, we discuss how to extend the approach to the hybrid van Hove theory of mixed classical-quantum systems and illustrate its application with the example of interacting classical and quantum oscillators. Finally, we focus on the question of how hybrid van Hove theory evades no-go theorems on mixed classical-quantum systems and point to the various assumptions of the no-go theorems that do not apply to our approach. This demonstrates that these no-go theorems are not universally applicable, contrary to their original claims.","published":"2026-09-25T12:49:58Z","updated":"2026-09-25T12:49:58Z","authors":["Marcel Reginatto","Andrés Darío Bermúdez Manjarres","Sebastian Ulbricht"],"categories":["quant-ph"],"headline":"New approach challenges limits in mixing classical and quantum models","summary":"The researchers studied ways to combine classical and quantum physics using a mathematical framework called van Hove theory. They showed how classical systems can be represented in a way that preserves important mathematical properties and then extended this to mixed classical-quantum systems. Their work also explains why certain previous 'no-go' theorems, which claimed such mixtures are impossible, do not apply to their approach, suggesting these theorems are not always valid."},{"id":"2609.31208v1","title":"Electron paramagnetic resonance of Dy${^3+}$-doped CaWO$_4$: spin Hamiltonian, crystal-field analysis, and linear electric field effect","abstract":"We report an electron paramagnetic resonance (EPR) study of Dy${^3+}$ in CaWO$_4$ single crystals spanning dopant concentrations from 80~ppb to 247~ppm. The $g$ factors of the ground Kramers doublet at the $S_4$ point-group site, $g_\\parallel = 7.22 \\pm 0.01$ and $g_\\perp = 5.45 \\pm 0.01$, and the hyperfine constants of \\ce{^{161}Dy} and $^{163}$Dy are determined with an order-of-magnitude improvement in precision over the only previous report. A crystal-field analysis constrained jointly by published optical levels and by the measured $g$ tensor identifies the ground doublet as the lower branch of an anticrossing between the near-degenerate $\\|{\\pm}11/2>$ and $\\|{\\mp}13/2>$ states, and explains why optical data alone left the $g$ values unconstrained. \\add{The same parameter set, tested on the excited multiplets of the $^6H$ term, reproduces the ${}^6H_{13/2}$ crystal-field levels better than the optical parameters themselves.} The angular dependence of the linewidth in the $ab$ plane reveals broadening by random internal electric fields through the linear electric field effect. Calibrating these fields with the residual Er$^{3+}$ present in the same crystals yields the first electric-field coupling parameters of Dy$^3+$, $(B_{31}^2+B_{36}^2)^{1/2} = (34\\pm5)\\times10^{-6}$~cm/V, three times the Er$^{3+}$ value and the largest reported for a rare-earth ion in this host.The residual linewidth and the concentration-dependent lineshape asymmetry follow the quadratic field scaling expected for a second-order Stark shift, which escapes the inversion-site cancellation constraining the linear effect. Both couplings trace back to the $\\sim 20$~cm$^{-1}$ gap to the first opposite-symmetry doublet, a consequence of the dense crystal-field structure of the $J=15/2$ manifold.","published":"2026-09-25T12:46:39Z","updated":"2026-09-25T12:46:39Z","authors":["Achuthan Manoj Kumar","Larissa Aboudem-Joumessi","Remy Dassonneville","Adrien Savoyant","Karolina Waszowska","Pengrui Jiao","Patrice Bertet","Philippe Goldner","Sylvain Bertaina"],"categories":["cond-mat.mtrl-sci"],"headline":"Studying magnetic properties of Dy3+ ions in CaWO4 crystals","summary":"Researchers examined how dysprosium ions (Dy3+) behave magnetically inside calcium tungstate crystals using a precise technique called electron paramagnetic resonance. They improved measurements of key magnetic properties and explained complex interactions within the ions using crystal-field theory. Additionally, they discovered how internal electric fields affect these ions, measuring the electric coupling for the first time and noting it is unusually strong compared to similar ions. Their findings are based on experimental data and theoretical analysis."},{"id":"2609.31207v1","title":"Enabling a Unified Cross-Domain Representation for Two-Finger Gripper Manipulation via Interaction-Centric Modeling","abstract":"Achieving robust cross-embodiment generalization in imitation learning demands overcoming a critical representation flaw that inextricably entangles task semantics with hardware-specific visual geometry. We propose an interaction-centric framework that leverages the shared structure of two-finger grippers via a parameterized universal gripper abstraction, yielding a canonical gripper-frame representation. Given language and RGB-D observations, a VLM infers the subtask and grounds an interaction triplet (gripper, held, target), while SAM~2.1 tracks masks to reduce VLM queries. We design concise hybrid features that combine target/collision artificial potential fields for global guidance with segmented gripper-frame point clouds for local geometry, and use a Flow-Matching Transformer to predict smooth 7-DoF action chunks. Experiments in simulation and real-world tasks demonstrate that ours is the first imitation learning approach to simultaneously achieve competitive benchmark scores and extreme cross-embodiment/cross-viewpoint zero-shot sim-to-real transfer to completely distinct, heterogeneous robot platforms.","published":"2026-09-25T12:45:36Z","updated":"2026-09-25T12:45:36Z","authors":["Guanlin Li","Shifeng Bao","Yihan Zhao","Haitao Shen","Haoyang Li","Chen Zhao","Tong Yang","Jie Tang","Jing Zhang"],"categories":["cs.RO","cs.CV"],"headline":"New Method Helps Robots Use Two-Finger Grippers Across Different Robots","summary":"Researchers addressed a problem where robot learning mixes up task details with specific hardware visuals, limiting flexibility. They created a new framework that focuses on the interaction itself, using a universal model for two-finger grippers and combining language, vision, and geometric features to guide actions. Their tests in simulations and real-world tasks showed this is the first approach enabling robots to learn by imitation and successfully transfer skills across very different robots without extra training."},{"id":"2609.31206v1","title":"Self-Supervised Representation Learning: From Spectral Foundation Models to Auroral Emission Spectra","abstract":"Auroral spectrographs such as the Auroral Spectrograph In Skibotn (ASIS) record hundreds of thousands of emission spectra, but only a few hundred can be labelled by an expert. To exploit the rest, we pretrain a 1D Vision Transformer with a masked autoencoder on 223,000 unlabelled spectra. Without labels, its representation recovers the emission-line intensity ratios that physicists use to diagnose the precipitating particles (R^2 0.91 vs. 0.77 for an untrained control) and, under one linear probe, classifies as well as 13 features designed by experts. Fine-tuned, the model outperforms the previous supervised auroral classifier on its own benchmark (macro-AP 88.5 vs. 77.8), reaches 0.870 mAP, and exceeds the same architecture trained from scratch by +0.159 with 10% of the labels; attribution shows that it uses both N2+ bands. Could an existing pretrained model replace it? Two astronomical spectral foundation models and a time-series model transfer according to their spectral window: SpectraFM, trained in the infrared, falls below the untrained control, whereas SpecFormer, trained in the optical, approaches in-domain pretraining without reaching it.","published":"2026-09-25T12:45:06Z","updated":"2026-09-25T12:45:06Z","authors":["Matthieu Le Lain","Gaël Cessateur","Sébastien Lefèvre"],"categories":["cs.LG","physics.space-ph"],"headline":"Teaching AI to Understand Auroral Light Spectra Without Labels","summary":"Researchers trained an AI model on over 220,000 unlabeled auroral light spectra to learn useful features without expert labels. Their model could identify key particle information and classify spectra as well as expert-designed methods, outperforming previous models when fine-tuned with limited labeled data. They also tested if existing astronomy models could be used but found their specialized training was more effective. Findings are based on AI analysis of auroral emission data, with limitations on transferability from other domains."},{"id":"2609.31205v1","title":"Complexity, approximation, and extension of proper $\\{a,b\\}$-edge-weightings","abstract":"For distinct integers $a$ and $b$, an $\\{a,b\\}$-edge-weighting assigns $a$ or $b$ to each edge and labels each vertex by the sum of its incident weights. Such a weighting is proper if adjacent vertices receive distinct labels. We prove that, for every fixed pair of distinct integers, deciding whether a proper weighting exists is NP-complete even for simple cubic planar graphs. On planar multigraphs with $m$ edges, we give an exact $2^{O(\\sqrt m)}$-time algorithm and, assuming the Exponential Time Hypothesis (ETH), exclude $2^{o(\\sqrt m)}$-time algorithms even for simple cubic planar graphs. As a consequence, locally irregular $2$-edge-coloring is NP-complete on simple cubic planar graphs, admits a deterministic $2^{O(\\sqrt n)}$-time algorithm on $n$-vertex graphs in this class, and admits no $2^{o(\\sqrt n)}$-time algorithm under ETH. For maximizing the number of edges joining vertices with distinct labels, we give a deterministic efficient polynomial-time approximation scheme (EPTAS) on planar multigraphs, a polynomial-time $1/2$-approximation on multigraphs, and APX-completeness even on simple cubic graphs. Extending a partial $\\{a,b\\}$-edge-weighting to a proper one is NP-complete for every fixed pair even on simple cubic planar bipartite graphs, while it is polynomial-time solvable on trees. The hardness persists even when the prescribed edges form disjoint paths of length $6$ and all edges of each path have the same prescribed weight.","published":"2026-09-25T12:44:15Z","updated":"2026-09-25T12:44:15Z","authors":["Péter Madarasi","Máté Simon"],"categories":["cs.DM","cs.CC","math.CO"],"headline":"Researchers study hard problems of assigning two weights to graph edges","summary":"The researchers explore the challenge of assigning one of two distinct numbers to edges in a network so that connected points have different total sums. They show deciding if such an assignment exists is computationally difficult (NP-complete), even for simple network types called cubic planar graphs. They develop algorithms that run in sub-exponential time and prove limits on faster methods. They also study approximate solutions and find some cases allow efficient approximations, while others remain hard. Extending partial assignments is also shown to be hard except on tree structures."},{"id":"2609.31204v1","title":"FlatClip: A Geometry-Aware Surface-Level Baseline for fMRI Representation Learning","abstract":"Recent fMRI foundation models differ substantially in the spatial scale at which they represent brain activity. ROI- and connectivity-based models are efficient but coarse, whereas voxel-level models preserve fine-grained spatial structure but require specialized 3D/4D architectures and costly fMRI-specific pretraining. We ask how effectively an image-pretrained encoder can reuse the spatial organization of cortical activity. Motivated by evidence that macroscale brain activity is strongly constrained by brain geometry, we introduce FlatClip, a frozen-encoder surface-level baseline that renders cortical activity as geometry-aware flatmap sequences and reuses a frozen SigLIP2 image encoder with only a lightweight downstream probe. Across resting-state benchmarks, FlatClip serves as a competitive middle-ground representation, outperforming ROI-level baselines on HCP and ADNI tasks while remaining weaker on PPMI and below the strongest voxel-level models overall. On visual-fMRI decoding, restricting the input to visual or NSD-provided task-active cortex improves performance, highlighting the value of task-relevant cortical coverage. Spatial perturbation controls reduce the predictive performance of flatmap features under both retrained and fixed readouts, and anatomy-linked arrangements consistently outperform vertex permutations across three colormaps. Together, these results position surface-level flatmap sequences as a practical middle-ground baseline between ROI and voxel models, and support the utility of anatomy-linked spatial organization for reusing image-pretrained features. Code is available at https://github.com/OneMore1/FlatClip.","published":"2026-09-25T12:42:40Z","updated":"2026-09-25T12:42:40Z","authors":["Mo Wang","Wenhao Ye","Zihan Ning","Jiayu Zuo","Junfeng Xia","Hongkai Wen","Quanying Liu"],"categories":["cs.CE","cs.CV","q-bio.NC"],"headline":"New Method Links Brain Geometry to fMRI Data Analysis","summary":"Researchers developed FlatClip, a way to analyze brain activity that respects the brain's shape by turning it into flat images for processing with existing image tools. This method strikes a balance between simple regional models and detailed voxel-level models, performing well in some tests but not others. Their work shows that considering brain anatomy helps reuse image analysis techniques for brain data, offering a practical middle approach for future fMRI studies."},{"id":"2609.31203v1","title":"Orbital photoinduced Faraday effect in a lattice of conducting disks","abstract":"We present a theoretical study of pump-induced Faraday and Kerr rotation in a two-dimensional lattice of conducting disks. We propose a mechanism of Faraday and Kerr responses in which a circularly polarized pump directly induces a high-frequency Hall component $σ_{xy}=-σ_{yx}$ of the electron conductivity tensor at the probe frequency. Microscopically, $σ_{xy}$ originates from the third-order nonlinear response of the electron gas rather than from the real magnetic field created by solenoidal charge currents in the inverse Faraday effect. We show that the Faraday and Kerr rotation angles are significantly enhanced when the pump and probe frequencies are tuned close to the plasmon resonance of the disks, reaching $\\sim 0.1^\\circ$ per 1~kW/cm$^2$ of incident pump intensity in the terahertz range. This mechanism can explain recently observed giant pump-induced Faraday rotation in graphene disk lattices.","published":"2026-09-25T12:40:29Z","updated":"2026-09-25T12:40:29Z","authors":["A. A. Gunyaga","M. V. Durnev"],"categories":["cond-mat.mes-hall"],"headline":"New theory explains light rotation in conducting disk lattices","summary":"Researchers studied how shining circularly polarized light (pump) on a lattice of tiny conducting disks changes the way another light beam (probe) passes through, causing Faraday and Kerr rotation effects. They show this rotation comes from an unusual electrical response inside the disks, not from an actual magnetic field. Their calculations suggest these rotations get stronger near certain resonance frequencies and may explain recent experimental observations in graphene disk lattices. These results are theoretical and focus on predicted optical behavior at specific frequencies."},{"id":"2609.31202v1","title":"Preserve-and-Compose Training for Composed Image Retrieval","abstract":"Composed image retrieval (CIR) aims to retrieve images that satisfy a user-specified modification while preserving relevant visual content from a reference image. Collecting target images for this purpose is costly, motivating zero-shot CIR methods that instead use target captions as supervision. However, target captions may omit source details that should be preserved. We therefore propose, Preserve-and-Compose Training, which complements target-caption supervision with visual evidence from the source image. PACT learns from image--text--text (ITT) triplets without target images or gallery updates, aligning composed queries with target captions while preserving source evidence through visual supervision. We further introduce Chord scoring, which combines target similarity with source-relative directional agreement in the frozen image space. Results across four ZS-CIR benchmarks show that combining target-caption supervision with source-image evidence leads to strong retrieval performance across datasets, backbone scales, and external galleries. The code is available on https://github.com/sehyunkwon/PACT.","published":"2026-09-25T12:37:06Z","updated":"2026-09-25T12:37:06Z","authors":["Sehyun Kwon"],"categories":["cs.CV"],"headline":"New Method Improves Image Search with User-Specified Changes","summary":"Researchers studied how to find images that match a requested change based on a starting image. They created a training approach called Preserve-and-Compose Training (PACT) that uses descriptions of the target image along with visual information from the original image, even when no target images are available. Their method improves retrieval accuracy in tests and combines text descriptions with image details to better match user modifications. These findings are based on experiments with existing image search datasets and testing scenarios."},{"id":"2609.31201v1","title":"Samples, Sources, Space: Decomposing Data Scale in Spatially Structured Representation Learning of Human Brain Microarchitecture","abstract":"Scaling studies typically represent training data by a single count of samples. For hierarchically and spatially structured data, however, the same number of samples can be drawn from few or many sources and distributed differently across the underlying domain. We therefore study data scaling as an allocation problem, separating unique sample count, source diversity, and spatial coverage. We study this decomposition in microscopic whole-brain histology, where a source is an individual brain, and a sample is an image patch at a specific spatial location. Across 93 controlled pretraining runs of a contrastive model that uses spatial proximity for supervision, we vary data allocation, compute, and model capacity over 11.6 million spatially anchored image patches from 21 human brains. Performance improves with more unique samples, broader spatial coverage, additional compute, and larger model capacity. At fixed sample count, distributing samples across one to 18 subjects produces no detectable improvement, even though representations generalize substantially better to subjects encountered during pretraining. Inter-subject variation therefore strongly affects generalization, but additional subjects provide no benefit when a fixed sample budget is distributed across more sources. These results establish sample count, source diversity, and spatial coverage as distinct axes of data scaling in spatially structured representation learning.","published":"2026-09-25T12:36:38Z","updated":"2026-09-25T12:36:38Z","authors":["Christian Schiffer","Mathis Bode","Thomas Lippert","Katrin Amunts","Timo Dickscheid"],"categories":["cs.AI"],"headline":"Studying How Data Variety Affects Brain Image Learning Models","summary":"Researchers explored how different ways of organizing training data impact learning models that study detailed human brain images. They looked at three factors: number of unique image samples, how many different brains the samples come from, and how much of the brain's space the samples cover. Their experiments with over 11 million image patches from 21 brains showed that using more unique samples, covering more brain areas, and increasing computing power improves model performance. However, spreading the same number of samples across more brains did not help, even though models generalized better to seen brains. The study highlights that sample size, source diversity, and spatial coverage each play separate roles in learning from spatial brain data."},{"id":"2609.31200v1","title":"Dirac measures can be quantum limits on the Sierpinski gasket","abstract":"We prove that on the Sierpinski gasket, a prototype of compact fractal space, Dirac distributions arise as the weak limit of probability measures associated with sequences of high energy eigenfunctions of the Laplacian in a way that cannot happen in compact Riemannian manifolds.","published":"2026-09-25T12:35:03Z","updated":"2026-09-25T12:35:03Z","authors":["Patricia Alonso Ruiz","Nikolay Tzvetkov"],"categories":["math.SP","math-ph"],"headline":"Dirac Measures Arise from Quantum States on a Fractal Shape","summary":"The researchers studied the Sierpinski gasket, a well-known fractal, and explored how certain mathematical objects called Dirac measures appear as limits related to high energy quantum states (eigenfunctions of the Laplacian). These findings show a behavior on fractals that does not occur on traditional smooth shapes known as Riemannian manifolds. Their work is theoretical and focuses on fundamental mathematical properties rather than practical applications."},{"id":"2609.31199v1","title":"Enhancing Photogrammetric Digital Surface Models with Pretrained Diffusion Models and Multimodal Conditioning","abstract":"Large-scale Digital Surface Models (DSMs) can be produced cost-effectively from satellite images via stereo-photogrammetry. However, the resulting 3D maps are often contaminated by noise, outliers, and voids. On the other hand, aerial LiDAR provides high-accuracy elevation measurements at a substantially higher cost. In this work, we study diffusion models conditioned both on photogrammetric DSMs and Pléiades imagery to refine vertically co-registered DSMs. We introduce a modified Stable Diffusion 3 architecture with a pruned text stream and a patch-wise normalization strategy, enabling stable training on LiDAR data and transfer from natural images to elevation maps. Experiments in French cities demonstrate that multimodal conditioning improves elevation accuracy, reducing Dense Urban RMSE from 6.00 to 3.45 m in the in-context cities and from 4.16 to 2.77 m in the held-out city of Bordeaux.","published":"2026-09-25T12:35:01Z","updated":"2026-09-25T12:35:01Z","authors":["Antoine Lorentz","Stéphane May","Valentine Bellet","Dawa Derksen","Bastien Nespoulous"],"categories":["eess.IV","cs.CV"],"headline":"Improving 3D Maps from Satellite Images Using AI Models","summary":"Researchers addressed the problem of noisy and incomplete 3D maps made from satellite images, which are cheaper but less accurate than expensive LiDAR scans. They developed a new AI method that combines existing 3D maps and satellite photos to improve elevation accuracy. Testing in French cities showed their approach reduced errors in height measurements, suggesting this AI technique can refine digital surface models without the high cost of LiDAR."},{"id":"2609.31198v1","title":"Light Field Primitive for Novel View Synthesis","abstract":"We present Light Field Primitives (LFP), a formulation for novel view synthesis that replaces the dense ray database with a compact set of differentiable primitives in the classical two-plane parameterization. Each primitive condenses a group of rays into one learned record, and its response to a query is governed by how closely that query belongs to the group. Rendering a camera ray then reduces to compositing all responses it elicits, and a scene can be optimized directly from posed images and rendered in real time with rays. Beyond its competitive performance on standard benchmarks, the main advantage of LFP is structural: its primitives reside directly in the 4D ray space, so optical and appearance effects that are already operations on the light field become behaviors of a single shared renderer. With minimal changes to that renderer, LFP supports multi-scale anti-aliasing, defocus deblurring with refocusing, rendering for fisheye cameras, and even transparent object reconstruction with ray refraction, matching specialized frameworks that devote substantial machinery to these effects.","published":"2026-09-25T12:35:00Z","updated":"2026-09-25T12:35:00Z","authors":["Liang Chen","Jiahui Ning","Xun Jiang","Xing Xu","Jimmy Ren","Fenglei Fan","Heng Tao Shen"],"categories":["cs.CV"],"headline":"New Method Simplifies Creating 3D Views from Photos","summary":"Researchers developed Light Field Primitives (LFP), a new way to create novel 3D views by representing groups of light rays with simple learned units. This approach lets them generate images from new viewpoints efficiently and supports effects like blurring and transparent objects. Their work shows that using these compact primitives directly in ray space allows a single renderer to handle many complex optical effects, potentially simplifying and speeding up 3D scene rendering. The findings are based on experiments and comparisons to existing methods."},{"id":"2609.31197v1","title":"ALF: An Active Learning Framework for Scientific Discovery","abstract":"Machine learning for scientific discovery is almost systematically data bound. Producing relevant high quality data, under budget constraints, is amongst the most promising ways to advance the field. Active learning (AL) offers promise wherever labelling requires expensive experiment, measurement, or simulation. Most existing tools cover only part of the data acquisition loop, and typically focus on either offline benchmarking or online deployment, but not both. We present ALF, a modular AL Framework that runs the full data acquisition loop via five modular components. One clear API for both settings: offline, against an existing dataset for controlled and reproducible experimentation; and online, against an oracle for acquiring new candidates in real-world deployments. ALF is open-source and available at https://github.com/instadeepai/alf.","published":"2026-09-25T12:34:16Z","updated":"2026-09-25T12:34:16Z","authors":["Shikha Surana","Alex Hawkins-Hooker","Olivia Gallup","Christoph Brunken","Jules Tilly","Paul Duckworth"],"categories":["cs.LG"],"headline":"New Tool Helps Guide Data Collection for Scientific Research","summary":"Researchers developed ALF, a flexible software system designed to improve how scientists select data to study. ALF helps manage the entire process of choosing experiments or simulations to perform, especially when gathering data is costly or slow. This tool works both with existing data sets and in real-world scenarios, aiming to make data collection more efficient; these findings come from the researchers' description of their framework and its design."},{"id":"2609.31196v1","title":"Zero-Mean Oscillations Drive Recirculations in Potential Flows at Vanishing Reynolds Number","abstract":"Recirculating flows are central to transport phenomena, but their generation in potential flows requires additional physical mechanisms. We show experimentally that recirculation at vanishing Reynolds number can be produced at prescribed locations by coupling oscillations of a Hele-Shaw cell ceiling with an off-center passive obstacle. The net drift is well described by a potential flow model with no adjustable parameters. The resulting flows are localized around the obstacles and can be combined to generate programmable flow patterns at the microscale.","published":"2026-09-25T12:33:12Z","updated":"2026-09-25T12:33:12Z","authors":["Alexandre S. Avaro","Leon V. Gebhard","Gabriel Amselem","Charles N. Baroud"],"categories":["physics.flu-dyn"],"headline":"Researchers Create Controlled Flows Using Tiny Oscillations and Obstacles","summary":"The researchers studied how to create recirculating fluid flows, which help transport materials, in situations where usual methods don't work well. They showed experimentally that tiny oscillations combined with an offset obstacle inside a special flat cell can produce localized flow loops even when fluid movement is very slow. Their observations matched predictions from a flow model without needing extra adjustments. This work suggests a way to design small-scale flow patterns, but results are based on controlled experiments and theoretical modeling."},{"id":"2609.31195v1","title":"Automated Artifact Removal in EEG Age Prediction: A systematic comparison","abstract":"Automated EEG artifact removal may improve downstream analysis but can also alter predictive information. We benchmarked nine automated artifact removal methods against a common no-artifact-removal baseline for cross-dataset EEG age prediction. We introduce Signal Quality Index (SQI)-guided GEDAI, which leverages local signal-quality assessment to restrict correction to the channel--epoch pairs requiring intervention. Three deep neural architectures were trained on TUEG and evaluated without target-domain fitting on ds005385, LEMON, and TDBRAIN. Across this setting, GEDAI and SQI-guided GEDAI were the only methods with consistent gains over baseline in age prediction performance across all datasets and architectures ($Δ$MAE $=-0.77/-0.64$ years, $ΔR^2=+0.083/+0.072$, respectively). The remaining methods were neutral or detrimental on average ($Δ$MAE $=+0.22\\pm0.16$ years, $ΔR^2=-0.023\\pm0.016$ across methods). The two GEDAI-based methods achieved closely matched performance, while SQI guidance reduced the median modification ratio from $74.78\\%$ to $42.80\\%$. These findings show that curation benefits are method-dependent and establish SQI guidance as a more selective operating point, leaving more of the original EEG unchanged and limiting the potential loss of neural activity while retaining most of GEDAI's predictive benefit.","published":"2026-09-25T12:30:50Z","updated":"2026-09-25T12:30:50Z","authors":["Davoud Hajhassani","Paul-Adrien Graignic","Bruno Aristimunha","Apolline Mellot","Tom Mariani","Clément Nober","Bruna J. Lopes","Léo Burgund","Lionel Kusch","Thomas Semah","Arnault H. Caillet"],"categories":["eess.SP"],"headline":"Improving EEG Age Predictions by Selectively Removing Artifacts","summary":"Researchers studied how different automated methods for cleaning up EEG brain signals affect the accuracy of predicting a person's age. They compared nine cleaning techniques, including a new method called SQI-guided GEDAI that targets only poor-quality parts of the signal. Testing on multiple datasets, they found that only their GEDAI methods consistently improved age prediction, while others had no effect or made predictions worse. The study suggests careful, selective cleaning may keep useful brain information intact while improving predictive accuracy. These results come from evaluating existing data and neural network models."},{"id":"2609.31194v1","title":"Optical characterization of excited-state spin Hamiltonians and clock transitions at telecommunication wavelength in Tm3+:YAlO3","abstract":"The zero-phonon line (ZPL) between the two excited-state manifolds 3F4 and 3H4 of the trivalent thulium ion in yttrium aluminum perovskite (Tm3+:YAlO3) lies within the telecom E-band at around 1451 nm and exhibits an optical coherence time of a few μs at cryogenic temperatures. To improve this time in view of potential applications, e.g. single-photon sources and optical quantum memory, we characterized the magnetic properties of the lowest Stark levels of these two manifolds. Varying the external magnetic field amplitude and orientation, we measured the enhanced nuclear Zeeman splitting by probing the field-dependent frequency detuning of side holes created by means of spectral hole burning, and we characterized the inhomogeneous line shift caused by the quadratic Zeeman interaction. Together, this allowed us to establish the spin Hamiltonians for the lowest Stark level of the 3F4 and 3H4 manifolds, and to predict optical clock transitions for particular orientations and magnitudes of the magnetic field","published":"2026-09-25T12:26:33Z","updated":"2026-09-25T12:26:33Z","authors":["Akshay Babu Karyath","Luozhen Li","Julien Bertrand","Alexandre Kings","Tianyang Shen","Gabriel Aeppli","Simon Gerber","Guy Matmon","Wolfgang Tittel"],"categories":["quant-ph","physics.atom-ph","physics.optics"],"headline":"Studying magnetic effects to improve telecom-wavelength quantum devices","summary":"Researchers studied a special material containing thulium ions that absorb light at telecom wavelengths, useful for quantum technologies. They measured how magnetic fields affect the energy levels of the ions’ excited states, using a technique called spectral hole burning. This helped them understand and model the spin properties (spin Hamiltonians) and predict stable optical transitions that could improve coherence times, important for devices like single-photon sources and quantum memories. Their findings are based on experimental measurements at very low temperatures."},{"id":"2609.31193v1","title":"Who Says What: Symbolic Trimodal Binding Mechanisms in Audio-Visual LLMs","abstract":"Current Audio-Visual LLMs (AVLLMs) struggle with reasoning over videos featuring multi-speaker dialogues. In such videos, resolving \"who says what\" is crucial, which necessitates trimodal (text-audio-visual) binding. Motivated by these challenges, we systematically investigate how this trimodal binding is achieved in AVLLMs. Specifically, we identify emergent symbolic trimodal binding mechanisms in AVLLMs that utilize modality-specific symbolic variables. By encoding auditory and visual components into symbolic variables-capturing temporal utterance sequences and spatial entity coordinates, respectively-the model establishes cross-modal linking within this abstract space. Crucially, we reveal that when trimodal binding fails, the breakdown predominantly stems from misaligned audio-visual connections. To overcome this bottleneck, we introduce an audio-visual prompting method utilizing an off-the-shelf Active Speaker Detection (ASD) model. By simply overlaying visual bounding boxes on active speakers, this training-free approach yields immediate performance gains across four conversation-centric benchmarks. Moreover, lightweight fine-tuning of fewer than 300 steps on these ASD-prompted-videos extends these gains to three general AV benchmarks, suggesting the generalizability of our method.","published":"2026-09-25T12:26:23Z","updated":"2026-09-25T12:26:23Z","authors":["Jihoo Jung","Youngjoon Jang","Joon Son Chung"],"categories":["cs.CV","cs.SD","eess.AS"],"headline":"Improving AI Understanding of Who Talks in Multi-Speaker Videos","summary":"Researchers studied how advanced AI models process audio, text, and visual information together to figure out who says what in videos with multiple speakers. They found that the models use symbolic variables to link sounds and visuals but often struggle when audio and visuals don't align well. To fix this, the team added visual cues highlighting active speakers, which improved performance without extra training and showed potential for broader use after minimal fine-tuning. The findings are based on testing with conversation-focused video benchmarks."},{"id":"2609.31192v1","title":"Learning a non-linguistic code for inferred rules from reward","abstract":"How can a rule inferred from examples reach someone who never saw them, without a shared code? Patients with severe aphasia do it by gesture or sketch. One network sees worked examples and emits eight invented symbols; a second, blind to them, applies them to a new input. Rewarded for the second's success, the first learns a code carrying rules to three-step transformations training never presents, which new learners acquire. Like invented human languages, the code has two regimes: under reward alone the speaker drifts to one message, as human languages lose words under plain transmission; expressive pressure keeps messages differentiated. Success on new rules tracks how much the message says about the rule, not how varied messages are. The learning signal shapes the code: reward sorts many rules under few fixed labels; the listener's error gradient gives each rule a region of similar messages, telling rules apart far better.","published":"2026-09-25T12:23:34Z","updated":"2026-09-25T12:23:34Z","authors":["Cristiano Capone"],"categories":["q-bio.NC"],"headline":"Teaching AI to invent codes for sharing hidden rules","summary":"Researchers explored how a system can learn a new, non-language code to communicate rules inferred from examples without a shared language. One part of their model creates symbols after seeing examples, and another decodes these to perform tasks the first never directly showed. They found that the system develops a code influenced by rewards, balancing between simple and distinct messages, somewhat like how human languages evolve. These findings come from simulated networks, focusing on how reward structures shape coded communication."},{"id":"2609.31191v1","title":"Rethinking Data Quality for AI-Driven Systems: Evidence from Practitioner Interviews","abstract":"Data quality research has usually treated data as an input that is stored, processed, and validated. In AI-driven software-intensive systems, data also shapes model behavior, evaluation, and lawful use. Empirical evidence remains limited on how practitioners define, assess, and manage quality under these conditions. We interviewed 16 practitioners from nine organizations and analyzed the transcripts using reflexive thematic analysis and developed six themes from participants' accounts. In AI systems, traceability shifted from modular debugging to attributing model behavior, while using models as quality assessors introduced circularity. Agent context and memory became data objects, and synthetic and pseudo-labeled data made authenticity a quality concern. In foundation-model development, lawfulness became a gate for training data, while representativeness was judged through coverage of situations in which the system must behave safely. Prior ML research examines many of these problems separately. Our study provides a practitioner-grounded account of how they are encountered together as an engineering and organizational concern. We also interpret five recurring conditions as helping explain how the themes relate to reduced trust in data and AI outcomes. We synthesize these findings through lifecycle assurance: a conceptual framing focused on producing evidence that data can support a specific AI claim when its influence may be embedded in model behavior, model-based judgments, or agent actions.","published":"2026-09-25T12:22:26Z","updated":"2026-09-25T12:22:26Z","authors":["Hariharan Gopinath","Jan Bosch","Helena Holmström Olsson"],"categories":["cs.SE","cs.AI"],"headline":"Understanding Data Quality Challenges in AI Systems","summary":"Researchers interviewed 16 AI practitioners to explore how data quality is defined and managed in AI-driven systems. They found that data influences not just processing but also model behavior, evaluation, and legality, complicating traditional quality measures. Their analysis highlights multiple intertwined challenges, such as tracing model decisions and ensuring data authenticity, that affect trust in AI outcomes. The study offers a new perspective called lifecycle assurance to help produce evidence supporting AI data claims, based on real-world engineering and organizational experiences."},{"id":"2609.31190v1","title":"Dynamic factor and double PCA models for partially observed survival curves: Forecasting demand in short-term rental markets","abstract":"This paper develops prediction models for population-level survival curves observed over time and sampled from a heterogeneous mix of populations. We consider a discrete-time setting where each curve is only partially observed and forecasts of the remaining trajectory are needed for downstream decision making. Our approach recasts cross-population heterogeneity into a multivariate sampling model. We propose two forecasting models for partially observed curves: a full factor analysis model that extends a general factor representation to incorporate the partially observed survival curve, and a double PCA model. The methodology is motivated by demand forecasting in short-term rental markets, where market-level occupancy paths can be viewed as survival curves over the booking horizon and where forecasts of future occupancy feed into dynamic pricing algorithms. We apply the models to the newly released Wheelhouse dataset, which contains time series of market occupancy curves for 500 markets from 2017 to 2022. Model performance is assessed using the integrated quadratic distance, and we compare the proposed PCA-based methods to Holt's linear trend model across multiple forecast horizons. The results show that the proposed models yield accurate and stable forecasts of the remaining survival trajectory and generally outperform Holt's method, particularly at longer horizons.","published":"2026-09-25T12:22:00Z","updated":"2026-09-25T12:22:00Z","authors":["Marthe Elisabeth Aastveit","Alex Lenkoski","Thordis Thorarinsdottir"],"categories":["stat.ME","stat.AP"],"headline":"New Models Forecast Demand in Short-Term Rental Markets Using Partial Data","summary":"Researchers developed two new prediction models to forecast future trends from partially observed data, focusing on short-term rental occupancy. They treated occupancy over time as survival curves and used complex statistical methods to predict demand. Applying these models to a large dataset from 500 markets, they found their approaches often gave more accurate and stable forecasts than traditional methods, especially for longer-term predictions. The findings are based on statistical analysis and real market data from 2017 to 2022."},{"id":"2609.31189v1","title":"Differential Recursions, Projective Discriminants, and Algebraic Generating Functions","abstract":"We study two hierarchies of bivariate generating functions defined by differential recursions: a self-dual hierarchy and a reduced-symmetry hierarchy. Their rational coefficient functions produce palindromic and anti-palindromic triangles, including OEIS A336110 and A140136. For general differential order $p$, the reduced hierarchy decomposes into $p-1$ Horn-type companions. An explicit differential intertwiner of order $p-3$ relates the last companion to the normalized self-dual generating function. The companions share a principal symbol and a rationally parametrized characteristic curve with a projective $S_3$ action. At $p=2$ this construction yields the Narayana quadratic. At $p=3$ the symmetry lifts to $S_4$: two algebraic sectors are edge resolvents of one quartic root configuration, and the nonsymmetric Gross--Witten--Wadia generating function has degree four. For $p=4$ and $p=5$ we derive algebraic equations of degrees six and eight, respectively, and find Galois groups $S_6$ and $S_8$ at generic specializations. At $p=6$ we verify a degree-ten specialization with Galois group $S_{10}$. Their discriminants exhibit a universal cubic characteristic factor and a squared lower-order factor. A boundary factorization explains the observed algebraic degrees and discriminant monomials. We formulate conjectures on algebraicity and Galois groups in the reduced hierarchy and identify a possible mechanism for transcendence in the higher self-dual hierarchy.","published":"2026-09-25T12:20:57Z","updated":"2026-09-25T12:20:57Z","authors":["S. Voloshyn"],"categories":["math-ph"],"headline":"Researchers analyze complex generating functions and their symmetries","summary":"The researchers examine two types of two-variable generating functions defined by differential equations, revealing patterns related to symmetry and algebraic structures. They connect these functions to known numerical triangles and describe how their complexity grows with order, identifying associated algebraic equations and symmetry groups. Their findings rely on theoretical analysis and conjectures about algebraic properties and symmetry groups, but the results are primarily abstract and mathematical without immediate practical applications."},{"id":"2609.31188v1","title":"How can AI accelerate the green transition?","abstract":"In his essay The Gift, published in 1925, the sociologist Marcel Mauss emphasised the 'total' scope of the phenomenon of gift-and counter-gift-giving, noting that it 'expresses, simultaneously and all at once, all manner of institutions: religious, legal and moral (...)\\,; economic (...)\\,; not to mention the aesthetic phenomena to which these facts give rise (...).'' The emergence of artificial intelligence (AI), which is suddenly and profoundly disrupting individual and collective lives, bears a resemblance to the ''total social fact'' described a hundred years ago. AI refers to technologies designed to simulate human intelligence, namely an ''automated system that (...) deduces, from received inputs, how to generate output results such as predictions, content, recommendations or decisions that may influence physical or virtual environments.'' 4 (OECD). Having become central to human relations, AI relies on physical and globalised infrastructure and therefore, also has an impact on the environment. This paper analyses some of these effects and how they are perceived by society. This text offers an exploration of the links between AI and the ecological transition, seeking to identify new opportunities and constraints. To this end, it will identify the barriers to the ecological transition (1) and the environmental impacts attributable to AI (2). It will then go on to describe how to live in a constrained world (3) while AI accelerates scientific discovery (4) and profoundly transforms society (5).","published":"2026-09-25T12:20:54Z","updated":"2026-09-25T12:20:54Z","authors":["Jacques Sainte-Marie","Marie-Fleur Simmet","Fanny Terrier"],"categories":["cs.CY"],"headline":"Exploring AI’s Role in Supporting Environmental Change","summary":"Researchers analyze how artificial intelligence (AI), as a powerful and wide-reaching technology, interacts with environmental challenges. They examine AI’s environmental impacts and societal perceptions, identify obstacles to ecological progress, and explore how AI can both accelerate scientific breakthroughs and transform society. The study aims to understand the opportunities and limitations of AI in aiding the shift toward sustainability, based on theoretical discussion rather than experimental results."},{"id":"2609.31187v1","title":"Universality of cutoff for the Exclusion Process","abstract":"Under a fairly general condition on the underlying jump rates, we prove that the Exclusion Process with fixed particle density exhibits cutoff at time $t_{mix}\\sim \\frac{1}{2}{t}_{rel} \\log N$, where $N$ is the volume and ${t}_{rel}$ the relaxation time of the single-particle dynamics. Our result covers, in particular, the standard setting of uniform nearest-neighbor jumps on large discrete tori in any fixed dimension and, more generally, on any sequence of vertex-transitive graphs with bounded degree and polynomially diverging diameter. Our (short and entirely human) proof combines Wilson's method, the Octopus Inequality, Fourier analysis on Hamming slices, and a sharp comparison with the Dirichlet form of a natural accelerated variant of the dynamics.","published":"2026-09-25T12:19:50Z","updated":"2026-09-25T12:19:50Z","authors":["Jonathan Hermon","Gady Kozma","Justin Salez"],"categories":["math.PR"],"headline":"Researchers uncover universal mixing time for particle movements","summary":"Researchers studied how groups of particles randomly jump around within a fixed space and found a universal pattern for the time it takes for the system to become well-mixed. They proved this holds under general conditions and covers common cases like particles moving on large grids or certain complex networks. Their findings come from a combination of mathematical techniques and show a precise relationship between the space size, particle density, and mixing time, although the results are theoretical and based on mathematical proofs."},{"id":"2609.31186v1","title":"Evolutionary Safety of Recursive Self-Improving AI: Taxonomy, Risk Discovery, and Evaluation","abstract":"Artificial intelligence is advancing rapidly, with increasingly capable systems taking larger roles in reasoning, decision-making, scientific discovery, and autonomous development. As AI begins to participate in its own improvement, from model training and experience accumulation to agent evolution and automated AI development, the prospect of recursive self-improvement (RSI) is becoming increasingly relevant. This transition raises a fundamental safety question: how can safety be maintained when the system, its accumulated experience, and even the process producing its successors continue to change? We introduce Evolutionary Safety as a perspective for studying safety under persistent and recursive self-improvement. It concerns not only whether an AI system is safe at a particular moment, but how safety properties change, persist, accumulate, and propagate throughout evolution. We characterize recurring manifestations, including intent drift, error accumulation, experience contamination, safety-property erosion, evaluator drift, and risk propagation. We then develop a taxonomy spanning persistent agent state, model state, evaluation and environmental feedback, computational substrate, and meta-level update mechanisms. Building on this taxonomy, we examine how evolutionary risks can be discovered and evaluated across states, updates, trajectories, and lineages, and derive governance principles for modification, selection, authorization, provenance, and recovery. Finally, we outline open problems toward maintaining safety guarantees as AI systems become increasingly persistent, adaptive, and recursively self-improving. Project resources and proposed evaluation systems are available at https://chaunceykung.github.io/evolutionary-safety-rsi.","published":"2026-09-25T12:17:10Z","updated":"2026-09-25T12:17:10Z","authors":["Chang Gong","Jingping Bi","Di Yao","Xinjian Liang","Chao Xiang","Ruijie Guo"],"categories":["cs.AI"],"headline":"Studying Safety in AI That Improves Itself Over Time","summary":"Researchers explore how to keep AI systems safe as they improve themselves repeatedly and evolve over time. They introduce 'Evolutionary Safety' to understand how safety features might change or weaken during AI's ongoing self-improvement. The team developed categories to identify risks like shifting goals or growing errors, and suggested ways to monitor and manage these risks as AI systems persist and adapt. Their work is theoretical, aiming to guide future safety efforts as recursive AI development advances."},{"id":"2609.31185v1","title":"Onboard Wind-Preview Model Predictive Control Using Pitot-Static Sensing for Multirotor UAVs","abstract":"Effective wind gust rejection and stable hovering are critical for the outdoor operation of autonomous drones. However, existing gust rejection methods are primarily reactive, inferring the disturbance from the resulting motion or measuring it at the airframe. Either way, the wind has already begun to act before it can be compensated. In this work, we anticipate the gust instead by measuring the wind ahead of the drone with a low-cost, low-weight pitot-static sensor mounted on a boom. The resulting wind preview is incorporated into a nonlinear model predictive controller (MPC), which optimizes the drone motion while anticipating wind disturbances. A longer boom offers more preview time but adds inertia and degrades flight performance. We characterize this trade-off in simulation and show that the optimal preview distance is not a fixed property of the platform, but shifts with the wind speed and with how quickly the drone can respond. Indoor hardware experiments confirm the trend and show that the proposed controller substantially improves hover performance against a PX4 baseline and an otherwise identical wind-unaware MPC. Outdoor experiments show that the error along the wind direction is reduced by 54 percent with respect to the baseline, demonstrating that a single wind-aligned sensor can significantly improve hovering performance.","published":"2026-09-25T12:16:35Z","updated":"2026-09-25T12:16:35Z","authors":["Bas Meere","Eline Wisse","Laurens Vousten","Sander Doodeman","Elena Torta","Paula Chanfreut","Duarte Antunes"],"categories":["cs.RO"],"headline":"Improving Drone Hovering by Sensing Wind Ahead","summary":"Researchers studied how to help drones better handle sudden wind gusts by measuring wind before it hits the drone, using a small sensor on a boom. They incorporated this early wind information into a control method that plans drone movements in advance, rather than reacting after being disturbed. Testing in simulations and real experiments showed this approach reduces hovering errors, but also revealed trade-offs in sensor placement and conditions affecting performance."},{"id":"2609.31184v1","title":"Accounting for Bias Enables Sustainable LLM Evaluation","abstract":"LLM-as-a-judge has become the de facto standard for scalable, subjective evaluation, yet current leaderboards compensate for systematic measurement bias by running ever more comparisons, an approach that is both statistically unsound and computationally wasteful. The root cause is an incomplete measurement model, treating LLM judges as neutral, interchangeable instruments ignores documented biases like position bias, verbosity bias, judge severity, and self-enhancement, that no volume of additional data can eliminate. We propose a unified latent variable framework that jointly models pairwise and ordinal data while explicitly correcting for these confounders, recovering reliable rankings from substantially fewer comparisons. Because fitting this model costs negligible compute relative to a single round of LLM inference, bias correction is not only more statistically rigorous but also a more sustainable approach to trustworthy evaluation.","published":"2026-09-25T12:16:25Z","updated":"2026-09-25T12:16:25Z","authors":["Harshita Katoch","David Antony Selby","Gerrit Großmann","Sebastian Vollmer"],"categories":["cs.AI","cs.LG","stat.AP","stat.ME"],"headline":"New Method Improves Fairness and Efficiency in AI Evaluations","summary":"Researchers identified that using AI models as judges to compare other AI results is biased and inefficient because these models have built-in preferences and errors. They developed a new mathematical approach that accounts for these biases, allowing fairer rankings with fewer comparisons. This approach costs little extra computing power and could lead to more reliable and sustainable ways to evaluate AI systems."},{"id":"2609.31183v1","title":"Recurrence and transience of open Jackson networks with balanced nodes","abstract":"Consider an open Jackson network with Poisson arrivals and exponential services, in which no node is overloaded and exactly~$k_0$ nodes have load equal to~$1$. We prove that the queue-length process is recurrent if and only if $k_0\\leq 2$. For that, we show that the network satisfies a sector condition, so its capacities are comparable to those of its symmetrisation; the latter is a reversible network whose recurrence/transience can be dealt with by standard means. We also discuss what the Lyapunov-function method gives in this setting; in particular, we show that every balanced queue is empty at arbitrarily large times, also in the transient case.","published":"2026-09-25T12:14:39Z","updated":"2026-09-25T12:14:39Z","authors":["Serguei Popov"],"categories":["math.PR"],"headline":"When queues in a network keep returning or drift away","summary":"The researchers studied a type of network where tasks arrive randomly and are processed at different nodes, some with full capacity. They found that the system's behavior depends on how many nodes are fully loaded: if two or fewer nodes are fully busy, the queues tend to return to shorter lengths repeatedly (recurrent), but more than two leads to growing queues (transient). Their proof uses mathematical comparisons and shows even busy nodes become empty at times, but the results are theoretical and apply to this specific network model."},{"id":"2609.31182v1","title":"Energy-minimizing domain decomposition","abstract":"We propose a novel domain decomposition framework based on energy minimization, applicable to a wide class of linear and nonlinear partial differential equations. This energy-minimizing domain decomposition (EMDD) method addresses problems whose solutions are minimizers of an energy functional, either unconstrained or constrained to a manifold, and thus covers source and eigenvalue problems in a single formulation. Besides introducing the general framework, we derive the algorithmic realization of the method for four classes of model problems, namely linear and nonlinear source and eigenvalue problems, and derive the computational complexity of the method. Numerical experiments demonstrate the robustness and efficiency of the method across all four model problems and show that it compares favourably against existing model-specific domain decomposition methods.","published":"2026-09-25T12:13:59Z","updated":"2026-09-25T12:13:59Z","authors":["Muhammad Hassan","Benjamin Stamm","Lambert Theisen"],"categories":["math.NA"],"headline":"New Method Breaks Down Complex Math Problems Efficiently","summary":"Researchers developed a new approach called energy-minimizing domain decomposition (EMDD) to solve complex math problems described by certain equations, including both linear and nonlinear types. Their method focuses on problems where the solution minimizes an energy quantity, covering several problem types together. They tested the approach on four different types of problems and found it to be efficient and reliable, often performing better than current specialized methods. The findings come from mathematical modeling and computer experiments as explained by the researchers."},{"id":"2609.31181v1","title":"Where a Model Sends Its Own Repeated Token","abstract":"Black-box model identification works by scoring a model's response to natural-language prompts. One line of work feeds models a degenerate input -- their own token, repeated -- to find a failure mode rather than an identity. We take that input and ask where the model goes when it does not. For each token t, read argmax p(. | t, t) in one forward pass; the result is a map on the whole vocabulary, with two halves. The first -- which tokens are fixed points -- is partially anticipated, and we report it as a failed estimand: the natural distance on it is 83% cardinality, separates a corpus manipulation by two bits in 3471 against a precision floor of zero, and attributes families at 0.5833. The second half, where the map sends tokens that are not fixed points, is unrecorded; the one paper holding those tokens logged them as a zero. Pairing on the source token removes the cardinality confound by construction (r from 0.9128 to -0.0932) and attributes families at 0.8333 -- twelve models scored against a pool of nineteen -- with chance 0.1389, across seven tokenizer groups and several corpora. Two nulls clear it: frequency-matched destinations agree at 0.1429, independent marginals at 0.0798. Family predicts agreement better than tokenizer (0.2031 against 0.1205), and recurrent architectures cluster at balanced accuracy 1.0 against a 0.7895 majority rate, or 0.90 once each model's dominant destination is excluded -- the figure we stand behind. We measure the robustness envelope: 8-bit weight rounding moves the map less than deduplicating the training corpus does (0.9004 against 0.6353, on one support), 4-bit destroys it (0.0098; 0.1812 at deployment granularity, so not a coarseness artefact), and the precision floor varies by model from 0.201 to 0.9778. All estimands and kill conditions were registered before the data, and the failed one is reported as fully as the surviving one.","published":"2026-09-25T12:13:31Z","updated":"2026-09-25T12:13:31Z","authors":["Nicolás Vera Zúñiga"],"categories":["cs.CL"],"headline":"Researchers map how models respond to repeated tokens","summary":"The researchers studied how language models react when given the same token twice in a row, aiming to identify patterns in their responses. They created a map showing which tokens lead the model to repeat itself and which lead it elsewhere, analyzing this behavior across different models and token groups. Their findings reveal that some model behaviors cluster by model family and architecture, but results come from statistical analysis and simulations rather than practical tests, with some measurements showing failure modes and others stronger trends."},{"id":"2609.31180v1","title":"BAT-CLIP: Trimodal Alignment of Brain, Audio and Text","abstract":"Decoding and interpreting naturalistic speech from the brain increasingly relies on alignment to pretrained speech and language representation spaces. However, current CLIP-style brain-speech alignment ground neural activity to a single anchor modality-audio or text-despite the brain's inherently multimodal speech processing. This induces a trade-off: audio anchoring preserves temporal structure but weakens linguistic separability, while text anchoring captures semantics yet discards acoustic detail. We propose BAT-CLIP, the first CLIP-style trimodal alignment framework for iEEG that jointly aligns neural embeddings to both pretrained audio and text anchors in a shared, frozen audio-text manifold. On the naturalistic Podcast benchmark, BAT-CLIP yields more robust representations than bimodal CLIP baselines. We also highlight the importance of using self-supervised foundation models for CLIP training.","published":"2026-09-25T12:13:01Z","updated":"2026-09-25T12:13:01Z","authors":["Suhyun Kim","Jinmo Han","Danny Dongyeop Han","Ahhyun Lucy Lee","Jewoon Lee","Yonghyeon Gwon","Zach Paris","Chun Kee Chung","Saewoong Bahk","Nam Soo Kim","Seong Jae Hwang","Jiook Cha"],"categories":["cs.SD","cs.AI","cs.LG","eess.AS"],"headline":"New Method Links Brain, Speech Sounds, and Text for Better Understanding","summary":"Researchers developed BAT-CLIP, a technique that connects brain activity with both speech sounds and text at the same time. This approach aims to improve how brain signals related to natural speech are interpreted, overcoming limitations when only one type of data (sound or text) is used. Tested on brain recordings from podcasts, BAT-CLIP showed stronger results than previous methods, highlighting benefits of joint analysis, though this is based on specific models and data used in their study."},{"id":"2609.31179v1","title":"SPO: Discovering Adaptive Large Neighborhood Search Operators via Stackelberg Program Optimization","abstract":"Large neighborhood search (LNS) relies critically on destroy and repair operators, whose effectiveness depends on both adaptation to the evolving LNS state and interaction between the two roles. We introduce Stackelberg Program Optimization (SPO), an LLM-based framework for discovering adaptive executable destroy-repair programs. SPO conditions operator decisions on a compact LNS state, allowing state-dependent behavior to emerge through program discovery, and organizes destroy-repair discovery as a Stackelberg interaction over program space that reflects their asymmetric dependency. Role-specific credits evaluate destroy programs as leaders and repair programs as conditional follower responses, guiding a coupled optimization process that combines LLM generator learning with population-based evolutionary search over programs. Experiments on the traveling salesperson problem and capacitated vehicle routing problem show that SPO outperforms strong baselines across a broad range of settings and generalizes beyond the discovery scale to larger instances and benchmark sets. Behavioral analyses further demonstrate state-dependent operator behavior and coupled destroy-repair improvement during discovery.","published":"2026-09-25T12:12:51Z","updated":"2026-09-25T12:12:51Z","authors":["Xinyi Ke","Kai Li","Junliang Xing","Yifan Zhang","Jian Cheng"],"categories":["cs.AI"],"headline":"New Method Helps Improve Complex Problem-Solving Operators","summary":"Researchers developed SPO, a new approach that uses advanced language models to create adaptable programs for solving problems like route planning. SPO treats the two main steps—destroying and repairing parts of a solution—as connected roles that adapt based on the current problem state. Tests on famous routing challenges show SPO does better than existing methods, and it can handle larger problems than it was trained on. The researchers’ results come from computer experiments using simulation, illustrating improvements in how these problem-solving steps work together."},{"id":"2609.31178v1","title":"JWST detection of methane emission from a young brown dwarf","abstract":"In recent studies, we have used low-resolution spectroscopy (R~100) with the James Webb Space Telescope (JWST) to uncover low-mass brown dwarfs (2-10 MJup) in the young cluster IC 348 that exhibit the so-called 3.4um absorption band from aliphatic hydrocarbons. Atmospheric models have not predicted the presence of that band in brown dwarfs at any age. Instead, methane absorption was expected in the coolest newborn brown dwarfs, but it was not detected. We have performed JWST spectroscopy at higher resolution (R~1000) on one of the brown dwarfs in IC 348, LRL 11001 (~6 MJup). The new measurement of the 3.4um band is fit well with four Gaussians that represent the CH stretching modes of CH2 and CH3 groups, indicating that the carrier is a large alkane (CnH2n+2). In addition, the spectrum contains emission lines at 3.324 and 3.333um, which we attribute to Q branches of CH4 hot bands between the octad and dyad regions (primarily nu3+nu4->nu4) and the tetradecad and pentad regions. This is the first detection of 3.33um CH4 emission from a young brown dwarf. The emission probably arises from the photosphere rather than a circumstellar disk given that no disks around stars or brown dwarfs are known to exhibit emission in these transitions. Modeling of the emission is needed to constrain the excitation mechanism, which could be either thermal (e.g., an accretion shock) or pumping by external radiation (e.g., neighboring cluster members). We speculate that vertical mixing has suppressed CH4 and enhanced large alkanes in the photosphere, but CH4 survives in the upper atmosphere, where it is available to experience the observed emission.","published":"2026-09-25T12:12:28Z","updated":"2026-09-25T12:12:28Z","authors":["K. L. Luhman","C. Alves de Oliveira","J. K. Faherty","B. C. Smith"],"categories":["astro-ph.EP","astro-ph.SR"],"headline":"JWST Finds Methane Emission in Young Brown Dwarf's Atmosphere","summary":"Researchers used the James Webb Space Telescope to study a young, low-mass brown dwarf in the IC 348 cluster. They detected both large alkanes and, for the first time, methane gas emitting light at certain wavelengths, likely from the brown dwarf’s visible surface rather than a surrounding disk. The findings suggest complex atmospheric mixing, but the exact cause of the methane emission remains unclear and requires further modeling."},{"id":"2609.31177v1","title":"Floating-Point Microformat Quantization and Pruning for Efficient MU-MIMO Neural Receivers","abstract":"Neural receivers outperform conventional 5G NR processing chains, but their compute and memory demands hinder real-time deployment. For a standard-compliant multi-user MIMO neural receiver, the 4-bit number format, not merely the bit width, determines whether compression preserves the gain over classical receivers. We apply weight and activation quantization-aware training (QAT) and, separately, 50% magnitude pruning, comparing INT8/INT4 and FP8 (E4M3)/FP4 (E2M1) weights with INT8 post-ReLU activations. Trained on 3GPP UMi channels and evaluated on TDL-B and TDL-C, 8-bit weight-activation models remain within 0.05 dB of FP32 at 10% and 1% block error rate (BLER). At 4 bits, uniform INT4 loses 3.3-3.7 dB and falls below LS-LMMSE, whereas FP4 more than halves this loss (1.3-1.4 dB) and still outperforms it by about 0.5 dB, even after pruning. FP4's denser near-zero grid matches the trained weight distribution, and FP4 avoids the residual-path over-pruning seen with INT4. An analytic cost model projects 66x fewer bit-operations and 8.8x less weight storage for pruned 4-bit-weight inference.","published":"2026-09-25T12:11:45Z","updated":"2026-09-25T12:11:45Z","authors":["SaiKrishna Saketh Yellapragada","Esa Ollila","Mário Costa","Yawei Li"],"categories":["eess.SP"],"headline":"Improving neural signal receivers with smarter number formats","summary":"The researchers studied how to make advanced neural network receivers for 5G communication more efficient by using different ways to represent numbers (called quantization) and by pruning less important parts of the model. They found that using a special 4-bit floating-point format (FP4) preserves better performance than traditional 4-bit integer formats, even after pruning. Their experiments and analysis showed FP4 models use much less computation and memory while still outperforming classical methods. However, their results are based on simulations with specific channel models and training methods."},{"id":"2609.31176v1","title":"Semantic Navigation for Issue Localization in Code Repository","abstract":"Repository-level issue localization aims to identify and rank the files and functions relevant to resolving a reported issue. LLM agents approach this task iteratively: they identify a set of potentially relevant locations, inspect the corresponding code, and revise their judgments about these candidates as new evidence is acquired. Existing environments, however, provide limited support for this loop: agents must search for unresolved relation targets, reconstruct entity semantics from raw source code, and revise candidates without evidential basis. To address these limitations, we present SemNav, a framework that leverages deterministic retrieval to seed a broad candidate set and an LLM agent to continually refine that set, thereby combining initial coverage with evidence-guided revision. SemNav supports this process through three key components. A Semantic Navigation Graph resolves program relations on demand through a language server, enabling direct navigation to related entities across files. Issue-conditioned Semantic Cards provide compact, source-grounded interpretations of each entity's role and relevance to the issue. A persistent Candidate Workspace records each candidate together with its evidential basis, enabling grounded verification, revision, and ranking. Across SWE-bench Lite and PLocBench, SemNav outperforms existing baselines, improving File Hit@10 from 68.33\\% to 82.67\\% with Gemma 4B. Component ablations and trajectory analysis support the complementary roles of all three components, while Semantic Cards reduce working-context load by 48.2\\% relative to full-source reading. SemNav further ranks first on all seven evidence-quality metrics on SWE-Explore and improves downstream issue resolution from 44.00\\% to 52.33\\%.","published":"2026-09-25T12:08:22Z","updated":"2026-09-25T12:08:22Z","authors":["Yunxiang Wei","Zhenyu Lei","Jundong Li"],"categories":["cs.AI"],"headline":"New Tool Helps Locate Code Issues More Effectively in Repositories","summary":"Researchers developed SemNav, a method to better identify which files and functions relate to a reported issue in computer code. Unlike previous approaches, SemNav combines a broad initial search with ongoing refinement based on evidence, using three components that organize code relationships and keep track of findings. Tests show SemNav improves accuracy over existing methods, lowers the effort needed to understand code, and helps resolve issues more successfully. The findings are based on simulation and benchmark results provided by the researchers."},{"id":"2609.31175v1","title":"Current-voltage characteristics and resistive switching in epitaxial La$_{0.67}$Sr$_{0.33}$MnO$_3$/SrMnO$_3$/La$_{0.67}$Sr$_{0.33}$MnO$_3$ multilayer","abstract":"In present study, we investigate the structural properties and current-voltage ($I$-$V$) characteristics of an epitaxial multilayer composed of La$_{0.67}$Sr$_{0.33}$MnO$_3$ (25 nm)/SrMnO$_3$ (70 nm)/La$_{0.67}$Sr$_{0.33}$MnO$_3$ (25 nm), grown on different substrates i.e., SrTiO$_3$(100), LaAlO$_3$(100), Si (100). The used substrates not only have different chemical compositions in line with the film materials, also they have different lattice parameters which would tune the material chemistry and lattice strain at the interface significantly. Our $I$-$V$ measurements reveal distinct electrical behaviors depending on the substrate. The multilayers grown on oxide SrTiO$_3$(100) and LaAlO$_3$(100) substrates exhibit regular $I$-$V$ with slight nonlinearity at low applied voltages. In contrast, $I$-$V$ in multilayer with Si(100) substrate exhibits large asymmetry and notable resistive switching (RS) behavior at room temperature, where a resistance ratio around 10 between the high resistance state (HRS) and low resistance state (LRS) is observed. While the observed $I$-$V$ behavior is sensitive to substrate-induced interfacial disorder and formation of trap bands within bandgap of SrMnO$_3$, the present results have potential applications in future memory and memristive devices.","published":"2026-09-25T12:06:23Z","updated":"2026-09-25T12:06:23Z","authors":["A. G. A. Rahman","R. K. Patel","Chandrani Nath","A. K. Pramanik"],"categories":["cond-mat.mtrl-sci"],"headline":"Researchers study electrical switching in special layered materials","summary":"The researchers examined how electrical current flows through layered materials made of specific metal oxides placed on different surfaces. They found that the electrical behavior changes depending on the surface, with one setup showing a clear change between high and low resistance states at room temperature. This switching effect depends on subtle changes at the interfaces of the layers caused by the underlying surface. Their findings come from measurements of current and voltage and suggest possible uses in memory devices, though these results are limited to their lab observations."},{"id":"2609.31174v1","title":"A Class of Shift-Invariant Permutations and Their Algebraic Structure","abstract":"Shift-invariant permutations of $\\mathbb F_2^n$ are attractive in symmetric cryptography because of their regularity and implementation efficiency. Constructing such permutations with an explicit algebraic structure remains a challenging problem. In this paper, we study a family of shift-invariant transformations of $\\mathbb F_2^n$ generated by a recursively defined sequence of mappings $\\{γ_j\\}_{j\\ge0}$ associated with landscapes. First, we prove that, under an explicit dimension condition on $n$ and the support of the landscape, the sequence $\\{γ_j\\}_{j\\ge0}$ has the polynomial composition property if and only if the corresponding landscape with nonzero special index is quasi-conserved on $\\mathbb F_2^n$. Then, we determine the eventual zero or periodic behavior of $\\{γ_j\\}_{j\\ge0}$, including the first zero or periodic index and the least eventual period, and establish linear independence up to the first relation. When the polynomial composition property holds, we establish an explicit isomorphism between the group of permutation elements in the monoid generated by these mappings and the unit group of a quotient ring $\\mathbb F_2[z]/\\langle p(z)\\rangle$, where $p(z)$ is a monomial or a binomial determined by the landscape. This framework reduces the permutation property, compositional inverse, order, and iterates of these mappings to polynomial computations. Finally, we apply the results to mappings represented by binomials and trinomials in the quotient ring, obtaining explicit criteria and formulas. We tabulate more than one hundred representative constructions of shift-invariant permutations, recovering and unifying several previously studied families.","published":"2026-09-25T12:06:04Z","updated":"2026-09-25T12:06:04Z","authors":["Cheng Lyu","Dabin Zheng","Mu Yuan","Lei Hu"],"categories":["cs.IT"],"headline":"Researchers Study Special Shift-Invariant Permutations for Cryptography","summary":"The researchers investigated a family of shift-invariant permutations, which are transformations with a regular, repeating structure valued in symmetric cryptography. They analyzed mathematical properties and conditions under which these permutations behave predictably, using algebraic tools and polynomial calculations. Their work offers explicit formulas and a unified framework to understand and construct these permutations, though the findings are theoretical and based on mathematical analysis rather than practical testing."},{"id":"2609.31173v1","title":"Solutions to the 'moment property implies subnormality' and injectivity problems for composition operators on one-circuit graphs","abstract":"We consider composition operators in $L^2(μ)$, where $μ$ is a discrete measure. Our main result shows that, for every integer $p\\ge2$, there exists a non-hyponormal composition operator that generates Stieltjes moment sequences and whose symbol induces a directed graph consisting of a circuit of length $p$ and one infinite branch. In contrast, if the symbol induces a directed graph consisting of one loop and one infinite branch, then the generation of Stieltjes moment sequences does imply subnormality. These two results completely solve the ``moment property implies subnormality'' problem left open in [Adv. Math. 310 (2017), 484-556] for the entire family of one-branch directed graphs with a circuit of arbitrary length. Combining them with the examples in that paper shows that, for directed graphs consisting of one loop and a finite number of infinite branches, $2$ is the smallest number of branches for which the moment property does not imply subnormality. Finally, we construct a noninjective composition operator that generates Stieltjes moment sequences and whose symbol induces a directed graph consisting of one loop, one infinite branch, and one leaf. This operator gives a negative answer to the injectivity problem posed in the aforementioned paper. Both constructions use the Krein and Friedrichs measures of an arbitrary normalized S-indeterminate Stieltjes moment sequence.","published":"2026-09-25T12:05:31Z","updated":"2026-09-25T12:05:31Z","authors":["Piotr Budzyński","Zenon Jan Jabłoński","Il Bong Jung","Jan Stochel"],"categories":["math.FA"],"headline":"Researchers solve problems about special operators on one-loop graphs","summary":"The researchers studied certain mathematical operators called composition operators defined by discrete measures and related to graphs with one loop and branches. They proved that for loops with two or more steps, there exist operators generating special number sequences (Stieltjes moment sequences) that are not subnormal, resolving a previously open question for these graph types. They also found a noninjective operator generating such sequences, answering another open problem. These results are theoretical and rely on advanced mathematical constructions from prior work."},{"id":"2609.31172v1","title":"Data-Driven Switched State Estimation with Sparse Sensor Scheduling for Nonlinear Networked Systems","abstract":"This paper presents a data-driven framework for joint observer design and sparse sensor scheduling for unknown nonlinear networked systems. The nonlinear dynamics are approximated online as a piecewise sequence of locally linearized discrete-time systems, resulting in a switched linear representation recursively identified through Subspace State-Space System Identification (4SID). To ensure consistency across regime transitions, an Orthogonal Procrustes alignment is introduced to promote coordinate consistency across consecutive regime transitions and mitigate artificial discontinuities caused by arbitrary state-space coordinate changes. Based on the identified local realizations, a dual-rate predictor-corrector observer is designed. The observer gain is computed through a convex optimization problem that jointly addresses estimation accuracy, sensor sparsity, and stability requirements. In particular, an L_{2,1}-norm regularization term promotes column sparsity in the gain matrix, enabling the automatic selection of informative measurement channels, while a spectral-norm constraint guarantees Schur stability of the estimation error dynamics. The proposed framework is first validated on a traffic network simulated in Aimsun Next. Results obtained on an 18-link network show that the observer accurately reconstructs macroscopic traffic states while significantly reducing the number of active sensors required for real-time estimation.","published":"2026-09-25T12:04:50Z","updated":"2026-09-25T12:04:50Z","authors":["Gianfranco Gagliardi","Franco Angelo Torchiaro","Vincenzo Gallelli","Ayman El Qemmah","Alessandro Casavola"],"categories":["eess.SY"],"headline":"New Method Estimates Complex Systems Using Fewer Sensors","summary":"Researchers developed a data-driven approach to estimate states in unknown complex nonlinear systems by breaking them into simpler parts. They designed an observer that learns and adjusts online, selecting only the most useful sensors via a mathematical technique, reducing the number needed for accurate monitoring. Tested on a simulated traffic network, their method accurately tracked traffic conditions while using fewer sensors, showing potential for efficient real-time system monitoring. These results are based on simulation and mathematical models, not real-world trials."},{"id":"2609.31171v1","title":"Modern triad for magnetic material science: geometric shape, electronic bands and spin textures","abstract":"To enable novel concepts in nanotechnologies, it is insightful to design materials with qualitatively new performances originating from recent discoveries in fundamental material science and physics. Curvilinear magnetism emerged as a tool to design chiral and anisotropic responses of materials at the nanoscale using the effects of the geometric shape and topology of the object. This concept is distinct yet complementary to material screening, which is primarily based on the optimization of intrinsic microscopic properties to modify magnetic textures for specific applications. Contemporary curvilinear magnetism focuses on mean-field micromagnetics, while the explicit contribution of the lattice structure and electronic degrees of freedom remains beyond consideration. In this review, we will emphasize on the effects which appear when this coupling is accounted for and that have resulted in novel concepts for the design of ``metageometric'' materials. Further development of this field will provide appealing technological prospects for fundamental understanding of curvilinear magnetic nanostructures and their applications in energy efficient and scalable nanoelectronics.","published":"2026-09-25T12:02:22Z","updated":"2026-09-25T12:02:22Z","authors":["Denys Makarov","Oleksandr Pylypovskyi","Rui Xu","Carmine Ortix"],"categories":["cond-mat.mtrl-sci","cond-mat.mes-hall","cond-mat.str-el"],"headline":"New Approach to Designing Magnetic Nanomaterials Explored","summary":"Researchers reviewed a modern approach to designing magnetic materials by combining the object's shape, atomic structure, and electron behavior. They highlight 'curvilinear magnetism,' which studies how curved shapes affect magnetic properties, and emphasize including atomic and electronic effects for better designs. Their review suggests this combined perspective may lead to new materials with useful magnetic behaviors, though the work mainly summarizes concepts and does not present experimental results."},{"id":"2609.31170v1","title":"TaskIR: Task-Driven Image Restoration via Degradation Adaptation and Task Feedback","abstract":"Task-driven image restoration aims to improve both image quality and downstream task performance. However, existing methods predominantly focus on single degradation type and struggle to handle the diverse degradations encountered in real-world scenarios. Different degradations impose distinct restoration demands, and insufficient restoration may leave residual degradations and artifacts that impair object boundaries and semantic cues, thereby compromising downstream task performance. To address these challenges, we propose TaskIR, a two-stage task-driven unified image restoration framework that integrates degradation-adaptive restoration with task feedback refinement. In Stage I, a Degradation Representation Module (DRM) extracts degradation representations, enabling a Degradation-Guided Transformer Block (DGTB) to dynamically modulate feature transformations for adaptive restoration. In Stage II, a Task-to-Restoration Feedback Generation module (TRFG) transforms heterogeneous task features into restoration feedback by modeling task-representation discrepancies associated with the current restoration. Subsequently, a Selective Task Feedback Refinement module (STFR) assesses feedback relevance and selectively refines intermediate restoration features to mitigate interference with well-restored content. Extensive experiments demonstrate that TaskIR achieves competitive restoration quality and downstream task performance across diverse degradations and tasks.","published":"2026-09-25T12:02:12Z","updated":"2026-09-25T12:02:12Z","authors":["Yanjie Tu","Qingsen Yan","Axi Niu","Wenxuan Cai","Tao Hu","Wei Dong","Haokui Zhang"],"categories":["cs.CV"],"headline":"New Method Adapts Image Restoration to Real-World Damage Types","summary":"Researchers developed TaskIR, a two-step approach to improve how degraded images are restored, aiming to help computer programs better understand them. The first step customizes restoration based on the type of image damage, while the second uses feedback from related tasks to refine the image further. Their experiments show TaskIR can handle various image problems and improve both image quality and performance in tasks relying on those images, but findings are based on testing rather than real-world deployment."},{"id":"2609.31169v1","title":"Improving Visual Sensitivity of LLMs on Multimodal Machine Translation with Metric-based Loss Weighting","abstract":"Multimodal Machine Translation aims to incorporate additional signal from non-textual modalities to improve translations by resolving ambiguities. While models, through multimodal fusion, are able to accept images related to the source text, they can ignore this information. Therefore, increasing their visual sensitivity remains an active research area. In this work, we introduce a training method, Metric-based Loss Weighting, that improves visual grounding of translations by increasing the loss function for tokens that benefit from the accompanying image. We identify these tokens using the Point-wise Cross-mutual Information (PCXMI) metric, which compares the model's output probabilities with and without visual context. We introduce a Congruency-based PCXMI metric and experimentally show that both metrics working in combination yield the best results. We evaluate our method by fine-tuning three pretrained Multimodal Large Language Models on the task of Image-guided Machine Translation for three language directions. Metric-based Loss Weighting outperforms other tested methods on the CoMMuTE contrastive dataset, improving accuracy by up to more than 7 percentage points compared to standard fine-tuning, while maintaining strong general translation performance.","published":"2026-09-25T11:59:57Z","updated":"2026-09-25T11:59:57Z","authors":["Paweł Mąka","Piotr Andruszkiewicz","Yusuf Can Semerci","Jan Scholtes","Gerasimos Spanakis"],"categories":["cs.CL","cs.AI"],"headline":"New Method Helps Translation Models Use Images Better","summary":"Researchers worked on improving how translation models use pictures alongside text to reduce confusion and improve accuracy. They created a training technique called Metric-based Loss Weighting that makes the model pay more attention to words where images help most. By testing this method on several language tasks, they found it improved accuracy by over 7 percentage points compared to usual training, while keeping good overall translation quality. The results come from specialized tests and may not cover all situations."},{"id":"2609.31168v1","title":"Audio emotion recognition for atypical hearing","abstract":"My doctoral work aims to explore Audio Emotion Recognition (AER) in the context of atypical listening. This research focuses on auditory hypersensitivity in people with autism, a phenomenon that is often difficult to evaluate and unique to each individual. Our core idea is to leverage our understanding of affect from acoustic traits, relying on the possibility of generalizing affective responses from a small amount of annotated data. As a first step, we fine-tune a large foundation model, Contrastive Language-Audio Pretraining (CLAP) using low-rank adaptation (LoRA), trained on a valence and arousal dataset of neurotypical listeners.","published":"2026-09-25T11:59:00Z","updated":"2026-09-25T11:59:00Z","authors":["Ulysse Roussel"],"categories":["cs.LG"],"headline":"Studying Audio Emotion Recognition for People with Autism","summary":"Researchers are investigating how to recognize emotions in audio for people with atypical hearing, specifically those with autism who experience auditory hypersensitivity. They aim to understand emotional responses based on sound features using limited data. Their initial approach fine-tunes an advanced audio model with data from people without hearing differences; however, this work is at an early stage and primarily involves adapting models rather than testing on autistic individuals."},{"id":"2609.31167v1","title":"Neural State Prediction: Obstructing Shortcut Learning in EEG Foundation Models","abstract":"EEG foundation models increasingly use masked prediction to learn from unlabeled recordings, but optimizing this objective does not ensure transferable neural representations. A central challenge is that stable positional cues and local correlations can make masked regions predictable without integrating distributed neural context. To reduce this reliance on low-information prediction paths, we introduce Neural State Prediction (NSP), a latent-predictive framework that constrains both the prediction target and the available context. NSP uses a Target Encoder updated by an exponential moving average (EMA) to define latent supervision. Identity residualization removes additive effects associated with channel identity and relative time from the targets, while topology-separated context excludes their immediate spatial and temporal neighborhood from the visible input. We pretrain NSP on 2.2 million EEG segments from TUEG and evaluate it across 30 downstream datasets spanning clinical diagnosis, sleep staging, emotion recognition, motor imagery, event-related potentials, cognitive-state decoding, and language retrieval. Under full-parameter multi-task fine-tuning on EEG-FM-Bench, NSP achieves 63.94 macro balanced accuracy across 14 datasets, exceeding the strongest evaluated baseline by 2.35 percentage points. Controlled component ablations assess the contribution of each mechanism, while matched context controls and held-out interventions characterize the role of context geometry, signal content, and positional information. Jointly designing latent targets and their context offers a promising direction for EEG foundation models that learn from distributed signal structure.","published":"2026-09-25T11:58:29Z","updated":"2026-09-25T11:58:29Z","authors":["Kieren Yu","Ziyang Liu","Chang Huang","Jintai Chen","Kaishun Wu"],"categories":["cs.AI"],"headline":"New Method Improves Brain Signal Model Training by Limiting Shortcuts","summary":"Researchers address a challenge in training brain activity models from EEG data, where simple clues let models cheat by predicting missing parts without understanding broader brain signals. They introduce Neural State Prediction (NSP), which restricts what parts of the signal the model sees and predicts to encourage deeper learning of distributed brain activity. Tested on millions of EEG samples and many tasks, NSP slightly outperforms current methods, suggesting a promising approach to making brain signal models more reliable. Results are based on pretraining and tests using existing datasets."},{"id":"2609.31166v1","title":"AgentRecommender: LLM Agents Enable Customizable Recommender Systems on the User Side","abstract":"Recommender systems have traditionally been developed for platforms. However, this has given rise to many phenomena that may be advantageous for platform lock-in but are a nuisance to users, such as clickbait, filter bubbles, and the spread of fake news. Recently, user-side recommender systems have been proposed as a new paradigm for solving this problem. If users deploy their own recommender systems, they are no longer at the mercy of the platform's interests. However, building a user-side recommender system is not trivial; in particular, customizing one for oneself requires additional data. We propose AgentRecommender, a method that leverages the investigation capability and internal knowledge of LLM agents to flexibly build user-side recommender systems without additional data. AgentRecommender allows users to easily create recommender systems tailored to their own preferences.","published":"2026-09-25T11:58:23Z","updated":"2026-09-25T11:58:23Z","authors":["Ryoma Sato"],"categories":["cs.IR","cs.AI","cs.DB","cs.DL"],"headline":"New Method Lets Users Build Personalized Recommendation Systems","summary":"Traditional recommendation systems are controlled by platforms and can lead to issues like clickbait and fake news. Researchers propose AgentRecommender, a technique that uses language-based AI agents to help users create their own personalized recommenders without needing extra data. This method aims to give users more control over recommendations, though the findings are based on the researchers' proposed approach."},{"id":"2609.31165v1","title":"BreathGRU: A Novel Semi-Supervised Bidirectional Gated Recurrent Unit Framework for Speech and Breath Segmentation for Respiratory Audio","abstract":"Speech-breath segmentation is a fundamental preprocessing step in respiratory audio analysis, enabling applications such as respiratory acoustic biomarker extraction, lung function prediction and disease monitoring. Existing approaches, including threshold methods, Fourier Transform-based techniques, and unsupervised and pretrained voice activity detection (VAD) models, primarily focus on speech detection and often classify breathing events as non-speech or silence, limiting their applicability for precise breath detection. To address this limitation, we propose BreathGRU, a semi-supervised Bidirectional Gated Recurrent Unit (BiGRU) framework specifically designed for speech-breath segmentation. The proposed framework combines frame-level acoustic feature extraction with bidirectional recurrent modelling, pseudo-label refinement and duration-constrained Segmental Viterbi decoding to produce speech and breath segmentation. BreathGRU was evaluated against the existing approaches, using manually annotated recordings. Performance was assessed using event-based, time-based, overlap-based, duration-based and boundary-based segmentation metrics. Experiment results demonstrated that BreathGRU achieved the highest breath event recall (0.83), the lowest onset-localisation error (0.14s) and the highest Mean Match Intersection over Union (0.81), with competitive overall segmentation performance compared to large pretrained VAD models like Silero. Qualitative evaluation on manually annotated recordings further showed close agreement between BreathGRU and manual annotation, with better breath detection compared to Silero. These findings demonstrate that explicit breath event modelling provides advantages over general-purpose VAD models and establish BreathGRU as an effective speech-breath segmentation framework which can be applied for respiratory audio analysis and pulmonary healthcare applications.","published":"2026-09-25T11:58:02Z","updated":"2026-09-25T11:58:02Z","authors":["Sania Fatima Sayed","John W. Holloway","Reyer Zwiggelaar","Faisal I. Rezwan"],"categories":["cs.SD","cs.LG","eess.AS"],"headline":"New AI method improves speech and breath detection in lung sound recordings","summary":"Researchers developed BreathGRU, a new computer model to better separate speech and breathing sounds in recordings of respiratory audio. Unlike previous methods that mainly detect speech and treat breathing as silence, BreathGRU specifically identifies breathing events using advanced techniques and semi-supervised learning. Tested on manually labeled recordings, their method showed improved breath detection accuracy and timing compared to existing models, suggesting it could help with analyzing lung sounds for medical purposes."},{"id":"2609.31164v1","title":"SPADE: Escaping the Popularity-Similarity Frontier to Measure Serendipitous Recommendations","abstract":"Recommender systems engineer serendipity to foster active exploration and break predictable consumption cycles. The problem with existing offline beyond-accuracy metrics is that they often either isolate historical similarity or global popularity. We aim to design an evaluation metric that examines similarity, popularity, and actual user relevance. To achieve this, we introduce SPADE (Serendipitous Pareto Distance Evaluation). SPADE maps all items into a two-dimensional space to directly calculate a user-specific Pareto frontier of maximally popular and historically similar items. The final serendipity score is then computed by averaging the minimum Euclidean distance from this boundary strictly for the correctly recommended test-set items. Evaluating SPADE across five datasets and five baseline algorithms confirms its effectiveness; our results show that the metric successfully prevents algorithms from exploiting beyond-accuracy measures with irrelevant or non-personalized recommendations, reliably isolating serendipitous discoveries.","published":"2026-09-25T11:57:53Z","updated":"2026-09-25T11:57:53Z","authors":["Tobias Vente","Maarten Peirsman","Noah Daniëls","Hannu Toivonen","Bart Goethals"],"categories":["cs.IR","cs.AI","cs.LG"],"headline":"New Method Measures Surprise in Recommendation Systems","summary":"Researchers developed a new way to evaluate how surprising or unexpected recommendations are in systems like those suggesting movies or products. Their method, called SPADE, considers item popularity, similarity to what users liked before, and actual user interest all at once. Tested on several datasets and algorithms, SPADE helps identify truly surprising recommendations and avoids misleading scores caused by irrelevant or generic suggestions, though the findings are based on these specific tests."},{"id":"2609.31163v1","title":"Probing temperature dependent non-radiative decay channels in CrSBr with nonlinear optics","abstract":"Among the recently discovered layered magnetic materials, the antiferromagnet CrSBr has attracted considerable interest thanks to its stability, quasi-one-dimensional crystal structure, and direct near-infrared band gap. These properties make it a suitable platform for combined electrical, magnetic, and optical measurements, and thus for studying exciton dynamics and transport across different magnetic phases. In this work, we investigate the optical response of CrSBr across its paramagnetic and antiferromagnetic phases by combining nonlinear optical spectroscopy, including second-harmonic generation, third-harmonic generation, and two-photon photoluminescence, with linear photoluminescence measurements. The observation of a clear difference in the temperature dependence of third-harmonic generation and photoluminescence intensities allows us to identify the onset of non-radiative recombination channels in the antiferromagnetic phase.","published":"2026-09-25T11:57:26Z","updated":"2026-09-25T11:57:26Z","authors":["Minjiang Dan","Till Weickhardt","Paul Herrmann","Fabian Glatz","Marie-Christin Heißenbüttel","Thorsten Deilmann","Michael Rohlfing","Ksenia Mosina","Zdeněk Sofer","Benjamin Pingault","Julian Klein","Giancarlo Soavi"],"categories":["physics.optics"],"headline":"Studying how temperature affects light loss in magnetic material CrSBr","summary":"Researchers studied the layered magnetic material CrSBr to understand how it emits light at different temperatures and magnetic states. Using advanced optical techniques, they observed changes in light signals that reveal when the material starts losing energy in ways that do not produce light, called non-radiative decay. Their findings highlight how CrSBr's light-related properties change between magnetic phases, based on experimental optical measurements."},{"id":"2609.31162v1","title":"WorldTS: World Modeling for Multimodal Covariate-aware Time Series Forecasting","abstract":"Time series forecasting is typically framed as learning a direct mapping from historical to future observations in the observation space. However, sequences of observations generally provide only a partial view of the dynamics of the underlying system, with future observations being shaped by latent dynamics. Recent latent-space forecasting methods thus achieve improved performance by predicting future observations from latent-space representations of historical observations rather than directly forecasting future observations in the observation space. Next, while future observations are also shaped by external factors, how to incorporate external, often multimodal, information into forecasting, so that it can shape latent-state formation and evolution directly, remains underexplored. We propose WorldTS, a world-modeling based forecasting framework that integrates multimodal covariates directly into the forecasting to further improve forecasting performance. Specifically, WorldTS employs a two-stage training strategy. First, it learns forecasting-relevant latent state dynamics conditioned on multimodal covariates, yielding encoded future states. Next, the learned state dynamics are frozen, and an observation decoder is trained to map the predicted future states back to future observations. Extensive experiments on 21 real-world datasets offer insight into WorldTS and its effectiveness.","published":"2026-09-25T11:54:18Z","updated":"2026-09-25T11:54:18Z","authors":["Yuhan Zhu","Xiangfei Qiu","Hanyin Cheng","Wangmeng Shen","Chenjuan Guo","Bin Yang","Jilin Hu","Christian S. Jensen"],"categories":["cs.LG"],"headline":"New Method Improves Time Series Forecasting Using Multimodal Data","summary":"The researchers address the challenge that traditional time series forecasting often relies only on past observations, missing hidden system dynamics and external factors. They developed WorldTS, a two-step approach that first learns hidden system states influenced by various external data types, then predicts future observations from these states. Their experiments on 21 datasets suggest this method can better capture complex influences on future outcomes, although the results are based on their proposed model and tested data only."},{"id":"2609.31161v1","title":"I Act Therefore I Am: When Is JEPA's Action-Conditioning Enough to Learn Causal Mechanisms?","abstract":"Recent empirical and theoretical advances suggest that joint-embedding predictive architectures (JEPAs) may learn meaningful representations for action-conditioned prediction of future outcomes, thus becoming one of the foundational structures for world models. However, accurate prediction does not, in general, necessarily imply recovery of underlying causal states that give rise to the observed dynamics. This work investigates when and how JEPAs can recover the underlying causal states from observations. We first introduce a latent variable model, in which high-dimensional observations are generated from latent causal states whose dynamics are governed by action-conditioned transition mechanisms. Based on this formulation, we develop a general information-theoretic objective that combines conditional likelihood maximization for learning transition dynamics with entropy maximization for preserving latent state information. We then establish identifiability conditions under which representations learned by this general objective recover the underlying latent causal states up to component-wise invertible transformations and permutation. One key condition for such identifiability is sufficient action-induced variation in the transition mechanisms. Guided by this finding, we instantiate the general objective with an action-modulated Gaussian additive-noise model, yielding action-modulated JEPA (A-JEPA). Experiments on synthetic environments verify the theoretical findings under the identifiability conditions and robustness to moderate violations, while visual benchmarks demonstrate improved state recovery and transfer to unseen transition mechanisms.","published":"2026-09-25T11:54:16Z","updated":"2026-09-25T11:54:16Z","authors":["Yuhang Liu","Zhuo Huang","Javen Qinfeng Shi"],"categories":["cs.LG"],"headline":"When Can AI Understand Hidden Causes Behind Actions?","summary":"Researchers studied joint-embedding predictive architectures (JEPAs), AI models that predict future outcomes based on actions. They explored when these models can uncover the hidden causes behind observed events, not just make accurate predictions. By developing a mathematical approach and conditions, they found how JEPAs can identify underlying causal states if actions affect the system enough. Experiments on simulated data supported their theory, showing improved understanding of hidden states and adaptation to new situations. The findings are based on theoretical and synthetic tests."},{"id":"2609.31160v1","title":"ReG-SAM: Reference Graph-Driven SAM for 2D Foundational Vessel Segmentation","abstract":"Vessel segmentation in medical images is essential for many clinical tasks, ranging from diagnosis to treatment planning. However, it remains challenging due to complex vascular morphology and diverse imaging conditions. Existing deep learning methods rarely aim at building a generalizable vessel segmentor across anatomies and modalities. While the Seg- ment Anything Model (SAM) has shown promise for med- ical image segmentation, its original design does not fully exploit vascular morphology and struggles with fine-grained vascular structures, leading to suboptimal performance. In this paper, we propose ReG-SAM, a SAM-based framework tailored to 2D vessel segmentation that leverages reference graph set for enhancing vascular representations. Specifically, we introduce two modality-aware representations derived from the reference masks: graph prompt embeddings (GPEs) that encode global spatial features from graphs, and vascu- lar prototype embeddings (VPEs) that capture fine-grained modality-specific vessel characteristics from multi-scale fea- ture maps and vascular masks. Since both require vascular masks that are unavailable during inference and require robust modality-aware vascular feature representations, we construct a modality-wise vascular database and develop two reference graph-guided representation learning schemes for estimating GPEs and VPEs using samples from the database rather than ground-truth masks. Extensive experiments across 19 datasets demonstrate that ReG-SAM consistently outperforms existing baselines, even those using manual prompts, particularly on challenging thin vessels.","published":"2026-09-25T11:53:31Z","updated":"2026-09-25T11:53:31Z","authors":["Donghang Lyu","Zichen Zhang","Oleh Dzyubachyk","Marius Staring"],"categories":["cs.CV","cs.AI"],"headline":"New AI Method Improves Blood Vessel Detection in Medical Images","summary":"Researchers addressed the challenge of identifying blood vessels in medical images, which is difficult due to complex vessel shapes and varying image types. They developed ReG-SAM, an enhanced AI model that uses reference vessel patterns to better recognize vessel structures in 2D images. Their tests on 19 datasets showed ReG-SAM performed better than existing methods, especially in detecting very thin vessels, though the approach relies on learned references rather than direct vessel markings during use."},{"id":"2609.31159v1","title":"Momentum-Guided Federated Split Distillation for Personalized Temporal Edge Intelligence","abstract":"We propose a momentum-guided federated split distillation framework for personalized, efficient, and autonomous temporal edge intelligence. We introduce TeRR-SAtt, our novel temporal reservoir student attention design that combines fixed reservoir representations, a lightweight temporal student, and personalized output modules. We also present AMGF, our anticipatory momentum-guided fusion mechanism that clusters clients through learning momentum and derives specialized teacher updates. On real-world smart-building data, TeRR-SAtt reduces edge training latency by 65.50%, inference latency by 44.70%, training memory usage by 18.40%, and inference CPU usage by 33.10% over the considered baselines. At the same time, AMGF improves local learning by up to 35.31% in RMSE compared to global updates.","published":"2026-09-25T11:52:44Z","updated":"2026-09-25T11:52:44Z","authors":["Ahmed-Rafik Baahmed","Jean-François Dollinger","Amine Brahmia","Mourad Zghal"],"categories":["cs.AI","cs.DC"],"headline":"New Method Improves Personalized AI on Edge Devices for Time Data","summary":"The researchers developed a new approach to help devices like smart building sensors learn and make predictions more efficiently on their own. They designed a system combining a special attention mechanism and a way to group similar devices to improve learning. Tests on real building data showed their method reduces training and running times, lowers memory and CPU needs, and improves prediction accuracy. The findings come from specific experiments and demonstrate potential benefits for personalized learning at the device level."},{"id":"2609.31158v1","title":"A Modified Rao-Blackwellised Particle Filter Based on Updated Process Noise Covariance for Chemical Reaction Networks","abstract":"It is necessary to characterize chemical reaction networks (CRNs) accurately to facilitate engineering of both synthetic and naturally occurring biological systems. Estimation of multiple unknown parameters and states in CRNs requires first, an identifiability analysis, and then adoption of a suitable estimation technique. In this paper, we took an example of a reduced order gene expression system, performed parameter sensitivity analysis for multiple unknown parameters, and used a modified Rao-Blackwellised particle filter (RBPF) to estimate parameters and states. The proposed framework estimates parameters using a particle filter and states using an extended Kalman filter (EKF) with process noise covariance updated recursively based on chemical Langevin equation (CLE). We compared the accuracy of parameter estimation, error in state estimation, and whiteness of the innovation sequence for the proposed filter with fixed choices of noise covariance. We found that the RBPF with updated noise covariance finds a balance between these three criteria, demonstrating its suitability for joint state and parameter estimation for stochastic CRNs.","published":"2026-09-25T11:52:40Z","updated":"2026-09-25T11:52:40Z","authors":["Suryasnata Dash","Abhishek Dey"],"categories":["eess.SY"],"headline":"Improving chemical reaction network analysis with a new estimation method","summary":"Researchers studied a way to better estimate unknown factors in chemical reaction networks, which are important for understanding biological systems. They used a modified method combining two existing tools—one to estimate parameters and another to track system states—while updating uncertainty estimates based on a chemical equation. Their approach, tested on a simplified gene expression model, showed improved balance in accuracy and error compared to fixed uncertainty methods, suggesting it could help jointly estimate states and parameters in these complex systems. Results are based on simulation with one example system."},{"id":"2609.31157v1","title":"Bayesian Tensor Autoencoder with Physics-informed Predictive Prior for Multi-dimensional Time Series Anomaly Detection","abstract":"Multi-dimensional time series, inherently tensorial, are common in practice. Despite great progress in time series anomaly detection, most existing methods are confined to uni-/multi-variate time series. When handling multi-dimensional time series using these methods, reshaping operations are required, which inevitably break the intrinsic correlations and thus lead to performance degradation. In uni-/multi-variate time series anomaly detection, AutoEncoders (AEs) are widely adopted and generally categorized into reconstruction-based and prediction-based AEs. The reconstruction-based AE utilizes the current observation for reconstruction, while the prediction-based AE utilizes the historical information to predict the current observation. Thus, the two AEs utilize different information. To bridge the gap between reconstruction-based and prediction-based AEs, so as to fully leverage the available information and thus further enhance performance, we propose a predictive prior and incorporate it into the reconstruction-based AE. It may not be very difficult to conceive this idea, but designing the predictive prior so that it can work for tensor anomaly detection is non-trivial. Specifically, to avoid breaking the intrinsic correlations within the multi-dimensional time series, we use the tensor AE as the backbone. To incorporate the predictive prior into the reconstruction-based AE, we propose a Bayesian fusion approach and our analysis reveals that this approach can enhance the modeling capability of the model for normal data. To mitigate the over-generalization problem of AE, we incorporate physical laws, i.e. tensor low-rank decomposition rules, into the neural networks in the predictive prior, leading to the Physics-informed Predictive Prior Tensor AE (PPPTAE) framework. Experimental results on real-world datasets demonstrate the effectiveness of the proposed method.","published":"2026-09-25T11:52:35Z","updated":"2026-09-25T11:52:35Z","authors":["Jianan Liu","Chunguang Li"],"categories":["cs.LG"],"headline":"New Method Detects Anomalies in Complex Time Data Using Physics","summary":"Researchers studied multi-dimensional time series data, which are common but tricky to analyze because typical methods can break natural connections in the data. They combined two approaches to detect unusual patterns: one that reconstructs current data and one that predicts it from the past. To do this without breaking data relationships, they used a special tensor-based model and added knowledge of physical rules to guide the process. Their experiments on real data showed this new method can better detect anomalies."},{"id":"2609.31156v1","title":"Common invariant measures for beta-transformations with rational affine relations","abstract":"We classify common invariant probabilities for two beta-transformations with a rational affine relation between their bases. Let $β>1$ be irrational and $γ=aβ+c>1$, where $a,c\\in\\mathbb Q$ and $a+c\\ne1$. Every finite nonnegative measure that is singular with respect to Lebesgue measure and invariant under both maps is a multiple of the point mass at zero. No entropy, ergodicity or multiplicative independence assumption is needed. The proof constructs a finite signed measure invariant under an irrational rotation and relates its mass to the first moment of the original measure. Combining this result with the known classification of coincident absolutely continuous invariant probabilities gives all common invariant probabilities in this family. The only additional measures are mixtures with a common absolutely continuous probability for adjacent bases in a known quadratic family. We also determine all common fixed points on the excluded line $a+c=1$.","published":"2026-09-25T11:51:09Z","updated":"2026-09-25T11:51:09Z","authors":["Hang Zhao"],"categories":["math.DS"],"headline":"Classifying shared steady states of special mathematical transformations","summary":"The researchers studied two related mathematical functions called beta-transformations, focusing on when their bases have a specific rational relationship. They classified all measures that remain unchanged (invariant) under both functions, showing that only very specific types of measures, like a single point mass at zero, fit this criterion without needing complex assumptions. Their findings rely on theoretical constructions and relate to known classifications, clarifying all shared invariant probabilities in these cases."},{"id":"2609.31155v1","title":"Teacher-Anchored Selection of Post-Training Quantized Models under Domain Shift","abstract":"Compressing a trained model yields a family of deployment candidates, and under domain shift the most compressed one need not be the one to deploy. We study selection over such a family, with candidates and teacher fixed and target labels absent or scarce. Two findings organize the label-free case. Minimum teacher distortion behaves almost as a constant rule, selecting the same eight-bit, per-channel, unclipped configuration in every run, which does not minimize empirical target cross-entropy. Established estimators divide sharply: in the overconfident-collapse regime of the CNN families, confidence-based estimators order the family close to backwards, and the diagnostics that identify it need the labels the setting denies, while output-distribution estimators match the teacher-relative anchor and on one architecture beat it. Distortion is nonetheless stable, so a supervised term can move selection away from it. Combining the two, we give exact quadratic identities for a canonical quadratic analogue of the family. We also show that under symmetric corruption the label-dependent part of a criterion linear in the label indicator is multiplied by one common factor whenever its coefficient sums are candidate-invariant, a class holding teacher contrasts and accuracy but not cross-entropy. These characterize the score's components without bounding selection regret. Across one hundred and thirty-four candidate families, one per independently trained convolutional or Vision Transformer teacher, anchoring reduces mean regret at the smallest label budget in every setting, an advantage that fades beyond twenty-five labels.","published":"2026-09-25T11:49:05Z","updated":"2026-09-25T11:49:05Z","authors":["Alejandro Rodriguez Dominguez","Muhammad Shahzad","Xia Hong"],"categories":["cs.LG","cs.AI"],"headline":"Choosing compressed AI models without many labels under change","summary":"The researchers studied how to pick the best compressed versions of trained AI models when the data changes and few or no labeled examples are available. They found that a simple rule based on how much the model's output differs from a trusted original model tends to select the same compression, but this does not always match actual target performance. By combining this approach with some labeled data, selection improves. Their experiments with many models show that this method reduces errors when few labels exist, though the benefit decreases as more labels are used."},{"id":"2609.31154v1","title":"HyperErase: Scale-Calibrated Hypernetwork for Multi-Concept Erasure in Text-to-Image Models","abstract":"Recent advances in text-to-image (T2I) generation have substantially improved visual synthesis, but have also raised increasing safety concerns due to their potential to generate harmful or undesirable content. Existing concept erasure methods predominantly follow a static weight paradigm, producing a single frozen adapter that struggles to adapt to diverse prompt variations and suffers from parameter interference when scaling to multiple concepts. We propose \\textbf{HyperErase}, a framework for concept erasure based on hypernetwork-driven prompt-conditioned parameter synthesis. Our approach first reframes concept erasure as prompt-conditioned parameter amortization and trains a hypernetwork to map textual descriptions to prompt-specific LoRA updates, eliminating the need for per-prompt gradient optimization or manual LoRA merging. To further improve the stability and precision of synthesized adapters, we develop a decoupled rectification strategy, which disentangles LoRA tokens into pattern and scale subspaces, applies a square-root transform to curb multiplicative over-scaling, and leverages teacher-derived canonical priors for inference-time correction. Extensive experiments across major concept categories demonstrate that HyperErase consistently improves the trade-off between erasure effectiveness, image quality, and semantic alignment, achieving performance comparable to gold-standard single-concept baselines. Furthermore, the resulting models can provide specialized LoRAs for each input prompt variation in a single forward pass without requiring gradient updates during inference. These principled and flexible framework offers a new paradigm for concept erasure in T2I models.","published":"2026-09-25T11:47:39Z","updated":"2026-09-25T11:47:39Z","authors":["Yi Sun","Xinhao Zhong","Zhiqi Zhang","Yimin Zhou","Junhao Li","Yuxia Qiao"],"categories":["cs.CV","cs.CR"],"headline":"New Method Helps Remove Unwanted Concepts from AI Images","summary":"Researchers developed HyperErase, a new way to erase specific concepts in text-to-image AI models. Unlike earlier methods that fix adapter weights and struggle with multiple concepts, HyperErase uses a hypernetwork to adjust model parameters based on textual prompts, avoiding slow manual tuning. Tests show it balances removing unwanted content while keeping image quality, but results come from controlled experiments and further real-world testing is needed."},{"id":"2609.31153v1","title":"Slow-fast dynamics of the McKean model with stochastic resetting and diffusion","abstract":"In this paper we investigate the combined effects of stochastic resetting and diffusion on a slow--fast dynamical system given by the piecewise-linear McKean model. That is, the fast variable $v$ is subject to Gaussian white noise with effective diffusivity $D$ and is reset to a fixed value $v_r$ at a random sequence of times generated from a Poisson process with rate $r$. Assuming that resetting occurs on the fast timescale, we freeze the slow variable \\(w\\) under an adiabatic approximation and determine the resulting non-equilibrium stationary state (NESS) of the fast variable as the solution of a modified Fokker--Planck equation. We show that the NESS can be expressed in terms of parabolic cylinder functions, whose asymptotic behavior allows us to recover the corresponding NESS without diffusion in the small-diffusion limit. The NESS is used to derive an averaged equation for the slow dynamics, whose solution converges to a stable fixed point \\(w^*\\) that depends on \\(D\\), \\(r\\) and \\(v_r\\). This fixed point effectively determines the long-time behaviour of the full system. Finally, we analyze the regime in which resetting occurs on the same timescale as the slow variable. We show how the slow variable now undergoes noisy oscillations due to resetting-induced switching between branches of the fast nullcline and derive the corresponding NESS for $w$ in the non-diffusive case.","published":"2026-09-25T11:47:21Z","updated":"2026-09-25T11:47:21Z","authors":["Jude Swaby","Paul C. Bressloff"],"categories":["cond-mat.stat-mech","q-bio.QM"],"headline":"Studying noise and resets in a simple two-speed system model","summary":"The researchers studied how random resets and noise affect a system with two processes working at different speeds, using a simplified mathematical model. They focused on how the fast part behaves when reset randomly and influenced by noise, and how this impacts the slower part over time. Their findings, based on mathematical equations and approximations, describe the system's long-term behavior and show that the slow part can oscillate randomly when resets happen on a similar timescale; these results are theoretical and derived from specific mathematical analysis."},{"id":"2609.31152v1","title":"KnottedGraph: Scalable knotted-graph topology for scientific and mathematical discovery","abstract":"Scientific data span heterogeneous structures, including coordinates, networks, surfaces, volumes and fields, yet their topology can be quantified within a common framework through graph connectivity, cycle structure, genus and spatial embedding. Graph- and homology-based summaries do not determine spatial embedding, while standard knot and link polynomials require extensions to accommodate branching graphs. Here, we introduce KnottedGraph, a computational framework that converts such scientific representations to knotted graphs that retain graph connectivity and spatial embedding together. It constructs projected diagrams and PD codes, enabling various topological analyses, including Yamada-polynomial evaluation for topological classification. For scalable exact evaluation, it combines partial resolutions that leave the same unresolved connections and optimizes their processing order; the resulting algorithm is verified against published topological invariants of knotted graphs with up to 500 crossings. This scalability enables us to introduce an LLM-assisted mathematical-discovery methodology, in which computational topological data generated across knotted-graph families are used to identify candidate closed-form formulas. With this approach, we identify analytical Yamada-polynomials for generic graph motif families exhibiting Abelian and non-Abelian word sequences. Together, these scalable capabilities make knotted-graph topology computationally accessible across scientific domains, enabling large-scale classification and introducing a route from topological data to LLM-assisted AI4Math discovery.","published":"2026-09-25T11:46:11Z","updated":"2026-09-25T11:46:11Z","authors":["Hakan Akgün","Xianquan Yan","Kehan Liu","Zhaoyun Chen","Ching Hua Lee"],"categories":["cs.MS","cs.CG","cs.SC","math.GT","physics.comp-ph"],"headline":"New tool analyzes complex data shapes with knot-based graphs","summary":"Researchers developed KnottedGraph, a computer method that represents diverse scientific data as knotted graphs preserving both connections and spatial layout. This allows advanced topological analysis that previous methods couldn't do at scale. They tested their approach on complex graphs and used AI to help discover new mathematical formulas describing these structures. Their work offers a scalable way to study scientific data shapes, although it remains focused on computational theory and simulations."},{"id":"2609.31151v1","title":"Phase-accumulating hyperfine strain sensing with rare-earth-ion phase memories","abstract":"Long-lived hyperfine coherences in rare-earth-ion crystals provide controlled phase evolution windows for metrology. We use this resource to propose an experimental protocol for accessing strain shifts of ground state hyperfine transitions in non-Kramers rare-earth-ion doped (REID) materials, a quantity that is difficult to access directly with conventional spectroscopic methods. Starting from the crystal field Hamiltonian, we relate the response to strain-dependent effective quadrupole and Zeeman tensors, including changes in electronic wave functions, virtual electronic admixtures, and the bare nuclear quadrupole interaction. The protocol optically prepares a selected hyperfine class, uses a phase-controlled rf $π/2$ pulse to create the sensing coherence, and applies synchronized dynamical decoupling pulses during the phase evolution interval so that a coherent ac strain drive accumulates phase rather than averaging away. The accumulated phase is then retrieved by Raman heterodyne readout. The resulting framework provides a route to understanding strain-induced hyperfine couplings in non-Kramers REID systems.","published":"2026-09-25T11:45:06Z","updated":"2026-09-25T11:45:06Z","authors":["Mustafa Gündoğan"],"categories":["quant-ph","cond-mat.mtrl-sci"],"headline":"Using rare-earth ions to measure tiny crystal strains precisely","summary":"Researchers propose a new method using long-lasting quantum states in rare-earth-ion crystals to detect small changes caused by strain inside the material. They developed an experimental protocol that prepares specific atomic states, applies controlled pulses, and measures the accumulated phase changes caused by strain. This work offers a way to understand how strain affects certain atomic interactions that are difficult to measure with usual methods, based on theoretical modeling and proposed experiments but has yet to be tested practically."},{"id":"2609.31150v1","title":"FedHisto-PAST: Parameter-Efficient Stain-Aware Federated Learning for Cross-Site Lung Histopathology Classification","abstract":"Cross-site lung histopathology classification must account for stain variation, non-IID client data, missing classes, and the cost of adapting large pathology encoders. This study evaluates FedHisto-PAST v2 for three-way classification of adenocarcinoma (ACA), Normal, and squamous cell carcinoma (SCC). FedHisto-PAST v2 combines a frozen HIBOU-B foundation model with parameter-efficient adaptation, stain-conditioned paired-view prediction and feature consistency, reliability-aware prototype learning, and adaptive federated aggregation. Experiments used a five-client, non-IID, raw-data-local simulation with fixed internal evaluation, client-level analysis, component ablations, communication accounting, and a development-influenced exploratory LungHist700 cohort. All principal methods achieved near- ceiling internal performance, which limited discrimination on the fixed split. On LungHist700, FedHisto- PAST v2 achieved a Macro-F1 of 0.728560 and a balanced accuracy of 0.730454. Higher recognition of Normal and SCC was accompanied by lower ACA recall, and calibration remained imperfect. Prediction-level consistency was the only component with a clearly supported independent contribution in the external ablation analysis. Feature consistency and prototype regularization showed no conclusive independent overall gains in Macro-F1. The framework updated 1.253841% of the model parameters. The results provide exploratory cross-dataset evidence for stain-aware, parameter-efficient federation; they do not establish formal privacy, patient-level independence, prospective deployment, or clinical validation.","published":"2026-09-25T11:44:38Z","updated":"2026-09-25T11:44:38Z","authors":["Muhammad Muhtasim Shahriar","M. M. Golam Hafiz","Saad Aloteibi","Mohammad Ali Moni"],"categories":["cs.CV","cs.AI"],"headline":"New Method Helps Classify Lung Tissue from Multiple Sites Efficiently","summary":"Researchers studied FedHisto-PAST v2, a method for classifying lung tissue types while accounting for differences in staining and data variations across hospitals. They tested it using simulated data from five sites and a special LungHist700 dataset, achieving moderate accuracy but noting some limitations in detecting certain cancer types. The model updated only a small portion of its parameters, showing efficiency. The researchers emphasize that their work is exploratory and does not guarantee privacy, patient independence, or clinical readiness."},{"id":"2609.31149v1","title":"CRNDiff: Count-Native Diffusion Framework via Chemical Reaction Networks","abstract":"Scientific measurements such as single-cell RNA (scRNA) sequencing often take the form of nonnegative integer counts, whereas continuous-state diffusion models approximate this discrete structure using continuous coordinates. Building on stochastic chemical reaction networks (CRNs), a class of count-native Markov jump processes, we introduce CRNDiff, a structured framework that combines count-space diffusion with inference-time conditioning on rare subpopulations. An independent birth--death instantiation yields a closed-form transition kernel for forward noising. This kernel enables reverse sampling via forward-filtering backward-sampling (FFBS) and supports data-driven selection of the terminal noising time, eliminating the need for a validation sweep. This tractability also lets us introduce tilted Feynman--Kac (FK) steering, a method for sampling target subpopulations from a frozen generator without retraining. By tilting posterior marginals before FK particle correction, steering mitigates importance-weight concentration when the target population is rare. Using scRNA-seq data from the human heart cell atlas, we test the ability of CRNDiff to generate cell-type-specific distributions. Across the three evaluated target populations, CRNDiff achieves the highest conditional fidelity among the evaluated generative models, with larger mean purity margins for rarer target populations. Generated cells preserve marker-level differential-expression structure. Replacing real training cells for the target classes with generated cells yields downstream classification performance approaching that of the real-data reference.","published":"2026-09-25T11:41:21Z","updated":"2026-09-25T11:41:21Z","authors":["Yuxuan Qiu","Praful Gagrani","Tetsuya J Kobayashi"],"categories":["cs.LG"],"headline":"New Method Models Count-Based Scientific Data More Accurately","summary":"Researchers developed CRNDiff, a framework that models scientific data made up of nonnegative integer counts, such as single-cell RNA sequencing results. Unlike previous methods that approximate counts with continuous values, CRNDiff works directly with the count data using a chemical reaction network approach. Tests on human heart cell data show CRNDiff better captures specific cell types, especially rare ones, and produces realistic synthetic data that helps with classification tasks. The findings are based on computational methods and simulations applied to existing data."},{"id":"2609.31148v1","title":"Seeing Semantic Shift: Difference-Aware Sentence-Level Temporal Segmentation of Sign Language Videos","abstract":"Recent advances in sign language understanding have achieved impressive success on short, single-sentence videos, yet their performance drops sharply when applied to long, continuous sign language videos. To bridge this gap, we focus on a challenging and realistic setting: Visual-only Sentence-level Sign Language Segmentation (Vis-SSLS), which aims to partition continuous sign language videos into non-overlapping sentence-level segments without any caption assistance, serving as a crucial prerequisite for downstream recognition and translation tasks. However, sentence transitions in sign language are often smooth and visually ambiguous, lacking explicit pauses or posture resets. As a result, static frame representations may fail to capture the subtle temporal changes that indicate sentence boundaries. To address this challenge, we propose \\textbf{SignShift}, a difference-aware segmentation framework that explicitly models frame-to-frame feature variation as semantic cues for sentence boundary detection. First, to model the feature variation, we design a Temporal Difference Module, which incorporates full-frame, facial, and hand cues, and employs inter-frame differencing to learn multi-scale temporal variations that capture both fine-grained local kinematics and global semantic transitions. Second, to mitigate over- and under-segmentation issues, we design a Segment Count Prediction module, which predicts the number of sentences to guide boundary selection. Extensive experiments on benchmark datasets demonstrate that SignShift substantially outperforms existing methods, validating its effectiveness.","published":"2026-09-25T11:40:43Z","updated":"2026-09-25T11:40:43Z","authors":["Bowen Guo","Shiwei Gan","Yafeng Yin","Xiao Liu","Kuizhuang Liu","Zhiwei Jiang","Lei Xie"],"categories":["cs.CV"],"headline":"New method helps split long sign language videos into sentences","summary":"Researchers studied how to divide continuous sign language videos into separate sentences without text help, a task made hard because sign language transitions are smooth and subtle. They created SignShift, a system that detects sentence boundaries by analyzing changes between video frames using cues from the whole frame, face, and hands. Tests on standard datasets show that SignShift performs much better than previous methods, supporting its potential usefulness, though results come from controlled benchmarks rather than real-world use."},{"id":"2609.31147v1","title":"Rydberg-Atom-Mediated Strong Antisymmetric Spin Exchange in Molecular Arrays","abstract":"Ultracold molecular systems have recently emerged as a versatile platform for quantum computation and simulation. Spin-exchange interactions arising from direct molecular dipolar interactions constitute the key mechanism for generating quantum entanglement and simulating quantum spin models. However, the relatively small electric dipole moments result in weak spin-exchange couplings, fundamentally limiting the speed of quantum information processing and interaction cycle of many-body dynamics. Here, we introduce a framework that employs Rydberg atoms with large electric dipole moments to mediate strong interactions between molecules in optical tweezer arrays that enables individually laser addressing both the Rydberg atom and molecules. Our result reveals that the mediated coupling can realize an effective molecular spin-exchange interaction with an intrinsic Dzyaloshinskii-Moriya character, with the Rydberg atoms dynamically decoupled from the molecular degrees of freedom, and the effective interaction strength enhanced by up to three orders of magnitude. We further demonstrate its versatility through rapid entanglement generation, high-fidelity two-qubit gate operations, and the realization of non-equilibrium symmetry-protected topological phase with long-lived edge coherence. Our work establishes a route toward strong molecular spin interactions and opens opportunities for fast, scalable quantum information processing and quantum simulation of long-time non-equilibrium quantum many-body physics in optical tweezer arrays of ultracold molecules.","published":"2026-09-25T11:38:45Z","updated":"2026-09-25T11:38:45Z","authors":["Yunqing Jiao","Jin-Zhu Jiang","Bo-Wen Guan","Jie Ma","Liantuan Xiao","Chi Zhang","Weibin Li","Feng Mei","Suotang Jia"],"categories":["cond-mat.quant-gas"],"headline":"Using Rydberg atoms to boost molecular spin interactions for quantum tech","summary":"Researchers studied how to improve interactions between ultracold molecules, which are important for quantum computing and simulations. They proposed using special atoms called Rydberg atoms that have larger electric dipoles to strengthen these interactions by up to a thousand times. Their work, based on theoretical models, suggests this method could enable faster quantum operations and explore new quantum states, but the findings are currently demonstrated through simulations rather than experiments."},{"id":"2609.31146v1","title":"The existence of polyhedral invariants is undecidable for linear systems","abstract":"The existence of polyhedral inductive invariants suitable for proving that a given control location is unreachable is undecidable for programs using only linear arithmetic over Z or Q, by reduction from 2-counter machines.","published":"2026-09-25T11:37:38Z","updated":"2026-09-25T11:37:38Z","authors":["David Monniaux"],"categories":["cs.LO"],"headline":"Researchers Show Limits in Verifying Linear System Safety","summary":"Researchers studied the problem of proving that certain points in a linear system cannot be reached. They focused on whether it is possible to find specific mathematical shapes, called polyhedral inductive invariants, that guarantee a location is unreachable. Their work shows that deciding if such shapes exist is undecidable, meaning no general algorithm can always solve this, using a connection to a known complex mathematical model called 2-counter machines."},{"id":"2609.31145v1","title":"Grid-cell firing fields lack local sixfold symmetry","abstract":"The firing fields of a grid cell form a hexagonal lattice, yet each field's intrinsic geometric structure remains unclear: Are individual grid fields radially symmetric or do they exhibit (weak) sixfold modulation inherited from the global lattice? To address this question, we quantified the within-field angular structure using harmonic analysis and per-cell matched simulations that preserved sampling statistics, field scale, and lattice geometry after correction for global elliptic deformation. We found that, across the investigated cells, the sixfold harmonic power fraction was consistent with values from matched locally circular reference fields. Experimental fields further showed substantially less sixfold modulation than matched simulations with an imposed local sixfold component, demonstrating sensitivity to sixfold structure at that level. This conclusion was supported by a complementary sixfold regression metric, harmonic leave-one-out sensitivity analyses, and robustness checks across radial-window choices, smoothing, and simulated background-noise levels. An observation-matched Burak--Fiete continuous-attractor network likewise showed no selective local sixfold enhancement despite its globally periodic grid structure, and this conclusion persisted when grid scale and relative field width were matched more closely to the experimental regime. These results suggest that the global hexagonal lattice organization need not be accompanied by detectable local sixfold angular modulation of individual firing fields, placing quantitative constraints on grid-field microstructure and mechanistic models of grid formation.","published":"2026-09-25T11:35:34Z","updated":"2026-09-25T11:35:34Z","authors":["Anian Kerscher"],"categories":["q-bio.NC"],"headline":"Researchers Find No Sixfold Symmetry Within Individual Grid-Cell Fields","summary":"Researchers studied the detailed shape of areas activated by specific brain cells called grid cells, known for their hexagonal pattern of activity. They used mathematical analysis and comparisons with simulations to see if each area has a sixfold (hexagonal) symmetry locally, like the global pattern. They found no evidence for local sixfold symmetry, suggesting that while grid cells form a hexagonal pattern overall, individual activation areas do not show that same angular structure. These findings help inform models of how grid cells form their patterns but are based on mathematical analysis and simulations."},{"id":"2609.31144v1","title":"Thermodynamic geometry as the missing link: toward a unified framework for black hole first-order phase transitions","abstract":"Black hole first-order phase transitions have been described by several seemingly independent frameworks, including local geometry, global topology, complex analysis, and thermodynamic geometry. While the first three have been unified, thermodynamic geometry has remained outside. We prove that the divergence points of the normalized Ruppeiner curvature scalar $R_N$ coincide exactly with the solutions of $T'(r_h)=0$, where $r_h$ is the horizon radius. These solutions include extremal points (spinodal points) and stationary inflection points (thermodynamic critical points). Thus, the divergence of $R_N$ is a necessary but not sufficient condition for a first-order phase transition. This clarifies the mathematical origin of curvature divergence and why thermodynamic geometry can reliably indicate but not alone confirm phase transitions. Using the local geometric framework as a central framework, we incorporate Ruppeiner geometry into this unified picture; a similar analysis also applies to Weinhold geometry. Consequently, the four frameworks are unified within a single structure based on the local folding of the temperature function. This advances our understanding of the mathematical structure of black hole first-order phase transitions and provides clues for possible extensions to other types of phase transitions.","published":"2026-09-25T11:33:17Z","updated":"2026-09-25T11:33:17Z","authors":["Shi-Hao Zhang","Jing-Fei Zhang","Xin Zhang"],"categories":["gr-qc","hep-ph","hep-th"],"headline":"New math links black hole phase changes in one unified theory","summary":"Researchers studied how black holes change phases, like water freezing, using different math approaches. They showed that a geometric property called Ruppeiner curvature relates directly to key temperature features, connecting previously separate methods. Their work unifies four frameworks into one math picture based on temperature behavior, clarifying when geometric signals indicate phase changes but alone don't prove them. This improves understanding of black hole phase transitions but is theoretical and focused on certain math models only."},{"id":"2609.31143v1","title":"Shape enantiomerism in semi-rigid polymer liquid crystals","abstract":"Weakly flexible, directed polymers can adopt a potentially infinite variety of conformations, some of them chiral. We develop a second-virial theory for semi-rigid polymers embedded in a nematic environment based on the assumption that the chains possess no intrinsic molecular chirality but can take on weakly helical conformations of either handedness. We demonstrate how these transient helical fluctuations can be exploited to predict the degree of chain stiffening induced by molecular crowding. The theoretical predictions are in good agreement with large-scale molecular dynamics simulations of bead-spring polymers with tunable backbone flexibility. Although both theory and simulations rule out spontaneous global chiral symmetry breaking imparted by chain conformations alone, they do reveal that polymers can develop pronounced enantiomeric shapes. Nematic fluids of semi-rigid polymers can be viewed as compensated cholesterics composed of transiently helical polymers with zero enantiomeric excess. By quantifying the degree of compensated chiral order developed over a broad range of concentrations and persistence lengths, two distinct regimes of weak and strong enantiomerism can be identified. A crossover between the two occurs when the chain persistence length drops below roughly three times the contour length.","published":"2026-09-25T11:33:11Z","updated":"2026-09-25T11:33:11Z","authors":["S. Biswas","W. S. Fall","H. H. Wensink"],"categories":["cond-mat.soft"],"headline":"How semi-rigid polymers can take chiral shapes without intrinsic chirality","summary":"The researchers studied semi-rigid polymers that do not have built-in chirality but can form weak helical (chiral) shapes. Using theoretical models and computer simulations, they explored how these polymers bend and stiffen when crowded in a special fluid called a nematic environment. Their results show that while the polymers do not create a global chiral order by themselves, they can exhibit two kinds of chiral shape behavior depending on their flexibility and concentration, revealing new insights about polymer shapes in liquid crystals."},{"id":"2609.31142v1","title":"JevAdvBench: A Benchmark and Black-Box Attacks for Reinforcement Learning for Calibrated Decisions Models","abstract":"Models trained with reinforcement learning for calibrated decisions (RLCD), such as Jev, answer a typed question about an input, the state, with a probability, a choice, or a score, and software acts on the answer without a person reading it. Their robustness has not been measured: adversarial benchmarks score what a model generates or executes, whereas a typed model generates nothing and returns a well-formed answer even when manipulated. Measurement is also hard, because identical requests can return different answers, most available labels come from the model itself, and the API preprocesses each request out of view. Our key idea is to score each attacked decision against the model's own clean decision rather than against labels, and to read it against the change caused by an identical re-run. Building on this, we introduce JevAdvBench, to our knowledge the first adversarial benchmark for RLCD models, with 812 typed questions over 66 scenarios, and a black-box attack suite of 9,744 single-edit variants that each edit one part of a request, with billed input tokens confirming that the edit reached the model. On jev-1.13.0, rewording stays within 1.2 percentage points of the re-run baseline, and fields outside the schema never reach the model. In contrast, one unverified opinion appended to the state flips 12.1% of decisions, statistically tied with the strongest injected command (10.1%), and pushes 38% of confident answers below the 0.8 confidence threshold that routes them to human review. Applications built on RLCD models should therefore treat the state as untrusted, argued input. Project website: https://JevAdvBench.github.io/JevAdvBench/","published":"2026-09-25T11:32:19Z","updated":"2026-09-25T11:32:19Z","authors":["Jianyi Hu","Hangtao Zhang","Yi Liu","Yeqi Zeng","Li Zeng","Xianlong Wang","Rui Wang","Leo Yu Zhang"],"categories":["cs.CR","cs.AI","cs.CL"],"headline":"New Test Checks Reliability of AI Decision Models Against Attacks","summary":"Researchers studied reinforcement learning models that make decisions given questions without human review. They created JevAdvBench, a benchmark with many test cases and small input changes to see how these models react to attacks. Their findings show that certain minor input changes can unexpectedly flip or lower confidence in model decisions, suggesting these AI systems should treat input information as potentially untrustworthy. The study is based on simulated tests of model behavior, not real-world applications."},{"id":"2609.31141v1","title":"Unknown-Traffic Detection, Calibration and Shortcut Reliance in Distilled Encrypted-Traffic Classifiers over One Year","abstract":"Knowledge distillation is the standard way to compress encrypted-traffic classifiers for the edge, and almost all such work judges students by accuracy alone. We ask what else a student inherits: unknown-traffic detection, calibration, shortcut reliance, and whether any survives a year of drift. Resemblance proves little on its own, since soft targets also regularise. We therefore distil one 101k-parameter student from two teachers of equal accuracy but different construction, a five-member ensemble and a single wider model, so that following one rather than the other is attributable to it. The design was pre-registered before any test result was seen. We tested ten hypotheses on CESNET-TLS-Year22, a year of real TLS traffic, across 18 test windows over 35 weeks. Two are supported: a student's per-flow unknown-scores shift toward its own teacher, but only at a conventional temperature, not the accuracy-optimal one; and a shortcut-reliant teacher passes its over-confidence to a student that never sees the feature. The drift prediction is reversed under both scores, the gap narrowing rather than widening and the student overtaking under the energy score in two of three replicates, as is the prediction that such a teacher harms its student's detection, which improves slightly. Shortcut reliance is set by model size, not distillation. Under the logit-based scores nothing else transfers: distillation beats neither a temperature-scaled direct student nor label smoothing. Exploratory analysis shows this turns on the scoring rule: with a feature-space detector the teacher detects unknown traffic 0.073 AUROC better than the direct student, where the energy score sees 0.000, and the conventional-temperature student inherits most of it. Label smoothing, with no teacher, recovers more. Distillation transfers the teacher's habits; what looks like an inherited ability is available without one.","published":"2026-09-25T11:32:00Z","updated":"2026-09-25T11:32:00Z","authors":["Mahmoud Abbasi"],"categories":["cs.NI","cs.LG"],"headline":"How Compressed Models Learn Traffic Detection Habits Over Time","summary":"Researchers studied how simplified encrypted-traffic classifiers, created through knowledge distillation, inherit abilities like detecting unknown traffic and model confidence. They compared students distilled from two types of larger models over a year of real internet traffic. The study found that some teacher behaviors, like handling uncommon traffic and overconfidence due to shortcuts, sometimes pass to the student, but other detection skills do not reliably transfer. Distillation mainly passes the teacher's habits rather than true detection ability, which can also be learned without a teacher."},{"id":"2609.31140v1","title":"Can Linguistic Reasoning Vectors Enhance Multimodal Reasoning Ability?","abstract":"Most Vision-Language Models (VLMs) are built by extending pretrained Large Language Models (LLMs) with visual modules and multimodal alignment. However, this multimodal scaling often degrades the language-side reasoning ability originally encoded in the base LLM. While the base LLM retains usable reasoning after scaling, the aligned VLM itself cannot reliably access this ability. Therefore, recovering the degraded reasoning capability in VLMs would benefit more from seeking help from the base LLM than from the VLM alone. Motivated by this, we propose LIFT (Language-side reasonIng Facilitation and Transfer), a lightweight vector-intervention method that transfers reasoning capability from the base LLM to the VLM without retraining the backbone. LIFT defines Reasoning Vectors as answer-token hidden-state differences between a Reasoner path with an explicit reasoning trace and a Solver path without it, and injects these vectors into language-side activations of the target VLM. LIFT further supports learnable vector adaptation while keeping the VLM backbone frozen. We evaluate LIFT on two VLMs across six reasoning benchmarks, comparing Reasoning Vectors extracted from the base LLM and from the aligned VLM under matched protocols. Results show that LLM-derived vectors consistently outperform VLM-derived vectors, confirming that the base LLM is a more effective source for recovering reasoning. LIFT partially recovers degraded reasoning through lightweight language-side interventions. Further analyses show that Reasoning Vectors influence intermediate reasoning behavior rather than merely altering final answers. The source code will be released soon.","published":"2026-09-25T11:31:02Z","updated":"2026-09-25T11:31:02Z","authors":["Ziyi Wang","Li Li","Aolin Zhou","Yankun Shen","Chonghan Liu","Shuxia Lin","Xu Yang"],"categories":["cs.AI"],"headline":"Researchers Boost Visual-Language Model Reasoning Using Language Insights","summary":"The researchers studied how combining language and vision in AI models often weakens their original language reasoning skills. They created a method called LIFT that borrows reasoning patterns from a language-only AI model and applies them to the combined vision-language model without retraining it. Testing showed that using reasoning data from the language-only model helps improve the combined model's reasoning ability, suggesting the language model is better at reasoning than the combined one. Their findings are based on experiments, not real-world use yet."},{"id":"2609.31139v1","title":"A mathematical study of a quasi-linear Schr{ö}dinger equation","abstract":"We study a quasi-linear Schr{ö}dinger-type equation arising in the description of particle dynamics within atomic nuclei. Under suitable assumptions on the model parameters, we establish the existence, uniqueness, and non-degeneracy of positive radial solutions. Moreover, we analyze how these solutions depend on the parameters.","published":"2026-09-25T11:28:31Z","updated":"2026-09-25T11:28:31Z","authors":["Meriem Bahhi","Jonas Lampart","Simona Rota Nodari"],"categories":["math.AP"],"headline":"Mathematical Analysis of Particle Behavior in Atomic Nuclei","summary":"Researchers investigated a complex equation used to describe how particles move inside atomic nuclei. They proved that, under certain conditions, there is a unique and stable solution that is symmetric in all directions. The study also explores how changing model parameters affects these solutions. Their work is theoretical and focuses on the mathematical properties of the equation rather than practical experiments."},{"id":"2609.31138v1","title":"Evaluating the Impact of Adaptive Extended Reality on Human-Robot Interaction Across the Reality-Virtuality Continuum","abstract":"As populations in developed countries age and labor shortages intensify, Cybernetic Avatars (CAs) are proposed to extend human capabilities through robotic embodiments, requiring effective Human-Robot Interaction (HRI) frameworks. Extended Reality (XR), an umbrella term for Augmented Reality (AR), Augmented Virtuality (AV), and Virtual Reality (VR), offers such interfaces, but prior research typically fixes the XR modality without evaluating its effect on task outcomes. This study examines whether the XR modality impacts HRI performance and whether an adaptive interface adjusting the level of virtuality along the Reality-Virtuality Continuum (RVC) at runtime improves it. A custom XR application interfaced with a mobile manipulator supports immersive control and runtime modality switching. In a within-participant multi-room pick-and-place experiment comparing fixed AR, AV, and VR with dynamic RVC through task metrics, the NASA-TLX, and the System Usability Scale (SUS), this study demonstrates that 1) the fixed reality modality affects HRI results, and 2) dynamically changing the modality along the RVC improves them. AR yielded significantly lower mental demand, effort, and frustration than AV and VR, while the dynamic RVC condition achieved the highest throughput and lowest workload, highlighting the value of adaptive XR interfaces for human-robot symbiosis. The implementation is available at https://github.com/CarlTornberg/XR-HRI.","published":"2026-09-25T11:27:38Z","updated":"2026-09-25T11:27:38Z","authors":["Carl Tornberg","Alicia Torck","Lotfi El Hafi","Tadahiro Taniguchi"],"categories":["cs.RO","cs.HC"],"headline":"Adaptive Virtual Interfaces Improve Human-Robot Interaction","summary":"Researchers studied how different types of virtual and augmented reality affect human control of robots. They tested fixed modes like augmented reality (AR) and virtual reality (VR), as well as a dynamic system that changes the level of virtual content during tasks. Their experiment showed that changing the modality adaptively led to better task performance and lower mental workload. The study suggests adaptive virtual interfaces may enhance cooperation between humans and robots, but these findings come from a specific controlled experiment."},{"id":"2609.31137v1","title":"INTERACT: Interactive Planning for Autonomous Driving via Anchor-Conditioned Prediction and Trust-Region Refinement","abstract":"Driving in dense urban traffic is interactive: whether a merge or an unprotected turn succeeds depends on how surrounding agents respond to the ego vehicle. Conventional planners predict first and plan second and, therefore, cannot account for this dependency. Methods that integrate prediction and planning either train both jointly, which introduces task interference, or keep them separate and are restricted to a predefined set of proposals. We present INTERACT: Interactive Planning for Autonomous Driving via Anchor-Conditioned Prediction and Trust-Region Refinement. Our key insight is that surrounding agents react to the intent a trajectory expresses rather than to its exact realization, so a single reactive prediction stays valid across an entire family of plans. INTERACT therefore decomposes interactive planning into prediction across driving intents and optimization within each intent. We derive a small set of diverse intents, which we call anchors, from map geometry, query a dedicated ego-conditioned prediction model once per anchor, and refine every anchor with the Cross-Entropy Method under a trust-region penalty that keeps the refined plan close enough to its anchor for the conditioned reaction to still apply. Prediction thus remains a separate model, avoiding task interference, while conditioning on anchors preserves the dependency. Because each anchor is refined continuously, the final plan is not restricted to the anchor set, yet INTERACT requires only one predictor query per anchor rather than one per candidate plan, with all anchors processed in parallel. On the nuPlan and interPlan closed-loop benchmarks, INTERACT sets a new state of the art, with the largest gains precisely in the interactive scenarios that motivate the method. The code will be released upon acceptance.","published":"2026-09-25T11:27:27Z","updated":"2026-09-25T11:27:27Z","authors":["Aron Distelzweig","Andreas Look","Faris Janjoš","Steffen Hagedorn","Luigi Palmieri","Joschka Boedecker"],"categories":["cs.RO"],"headline":"New Method Improves Self-Driving Car Planning in Busy Traffic","summary":"Researchers developed INTERACT, a planning method for autonomous vehicles that better accounts for how other drivers react to a car's intended moves. Instead of treating prediction and planning as separate or fully combined tasks, INTERACT predicts reactions based on general driving intentions called anchors and then refines specific plans within those intentions. Testing shows this approach improves performance in interactive driving situations, though the results are based on simulation benchmarks. The researchers plan to share their code publicly."},{"id":"2609.31136v1","title":"Toward AI-Augmented Cooperative Engineering Workflows: Requirements and Architecture the European Rover Challenge","abstract":"The growing availability of Artificial Intelligence (AI) tools creates new opportunities to support engineering design processes, yet their current use often remains limited to isolated tasks such as coding, documentation, or information retrieval. Less attention has been given to how AI can support cooperative engineering workflows at the process level, where teams must coordinate requirements, tasks, communication, knowledge transfer, and subsystem integration. This paper investigates this challenge in the context of the European Rover Challenge (ERC), where student teams design and integrate complex rover systems within a single academic cycle under strict time constraints and high subsystem interdependence. We conducted a role adaptive 40 question survey with ERC 2025 teams, yielding 104 responses from 14 teams. The survey examined team structure, knowledge transfer, task management, integration practices, communication patterns, and current AI usage. The results reveal recurring workflow bottlenecks, including limited documentation, unclear requirements, fragmented communication, informal task monitoring, and substantial integration rework. Based on these findings, we derive requirements for AI augmented cooperative engineering work-flows and propose an initial assistant system architecture that connects user facing interfaces, credential management, service selection, specialized AI services, and external engineering tools. The proposed architecture aims to support task clarification, requirement and compliance management, communication summarization, integration risk detection, and continuous knowledge capture. In doing so, the paper contributes empirical requirements and an architectural direction for AI augmented cooperative engineering workflows in hybrid human AI team settings.","published":"2026-09-25T11:27:18Z","updated":"2026-09-25T11:27:18Z","authors":["Ahmed R. Sadik","Frank Joublin","Mariusz Bujny","Antonello Ceravola","Joan Smith"],"categories":["cs.AI"],"headline":"Using AI to Improve Teamwork in Engineering Projects","summary":"Researchers studied how student teams build complex rover systems in a competition with tight deadlines to understand teamwork challenges. They surveyed 14 teams and found problems like unclear goals, poor communication, and repeated work during integration. Based on these insights, the researchers suggested requirements and a preliminary design for AI tools to assist with tasks like clarifying goals, managing communication, and detecting risks. Their work provides initial ideas for combining human and AI efforts to better support engineering teamwork, but is focused on survey findings and proposed system architecture."},{"id":"2609.31135v1","title":"Pocket-STVG: lightweight architecture for Spatio-Temporal Video Grounding","abstract":"Spatio-Temporal Video Grounding (STVG) aims to localize the spatio-temporal tube in a video corresponding to a natural language query. While recent methods achieve strong performance in fully supervised, weakly supervised, and zero-shot settings, they typically rely on computationally expensive architectures, complex training pipelines, or multimodal large language models. We present Pocket-STVG (P-STVG), a lightweight cascade architecture that addresses STVG by combining efficient pre-trained components instead of large end-to-end models. P-STVG integrates a temporal-aware video encoder based on MobileViCLIP, a spatial encoder-decoder derived from MDETR, and a shared aligned text encoder. Temporal localization is performed through either a lightweight 1D U-Net or a simple thresholding strategy, enabling the same framework to operate in both weakly supervised and zero-shot settings. Furthermore, video representations are precomputed independently of the query, yielding an indexing-friendly pipeline for efficient inference and large-scale video collections. Despite requiring fewer than 90M parameters, P-STVG performs on par with weakly supervised methods and improves on earlier zero-shot approaches at a fraction of their memory and computational cost, establishing a favorable performance-efficiency trade-off for STVG.","published":"2026-09-25T11:26:21Z","updated":"2026-09-25T11:26:21Z","authors":["Alberto Presta","Michal Byra","Grzegorz Stefański","Karol Szurkowski","Eryk Kołodziejczyk","Krzysztof Arendt"],"categories":["cs.CV","cs.AI","cs.MM"],"headline":"Lightweight AI model finds objects in videos by language queries","summary":"Researchers developed Pocket-STVG, a simpler and more efficient approach to locate objects and events in videos based on natural language descriptions. Instead of using large, complex models, they combined smaller pre-trained components to process video and text data. Their model uses fewer resources but performs similarly or better than previous methods in certain settings, showing a good balance between accuracy and efficiency. The findings are based on simulations and comparisons with existing techniques."},{"id":"2609.31134v1","title":"High-gain vortex transfer via activated forbidden transitions in molecular magnets","abstract":"Vortex light has garnered considerable interest in recent years owing to its distinctive properties and broad application potential. In this paper, we employ a conventional three-level ladder-type configuration in molecular magnets to realize high-gain vortex light transfer by enabling otherwise forbidden transitions. We demonstrate that the intensity and phase of the generated vortex signal field are governed by the detuning of the probe field and the strength of the control field, and that the topological charges of the incident and generated fields obey a well-defined algebraic relation during the vortex transfer process. Furthermore, we show that, in molecular magnets, the Autler-Townes splitting (ATS) effect yields a larger vortex signal field gain than electromagnetically induced transparency (EIT) over a broad parameter range in nonlinear three-wave mixing. This result suggests that the widely accepted view of EIT as the dominant enhancer of nonlinear optical effects may not hold universally. In addition, we revisit previous studies and re-examine the characterization of vortex beam transfer efficiency, emphasizing the need for a more careful interpretation in multi-beam interaction systems. Leveraging the long spin coherence and microwave transitions of molecular magnets, our results may enable quantum information transfer, storage, computing, and radar imaging in solid-state platforms.","published":"2026-09-25T11:26:20Z","updated":"2026-09-25T11:26:20Z","authors":["Fan Meng","Hao Zhu","Xin-Yao Huang","Guo-Feng Zhang"],"categories":["physics.optics","quant-ph"],"headline":"Scientists enhance vortex light transfer using molecular magnets","summary":"Researchers studied a way to improve the transfer of special light patterns called vortex light using molecular magnets. They used a three-level system to activate usually forbidden transitions, controlling the light’s intensity and phase through specific field adjustments. Their findings suggest a different mechanism may be more effective than previously thought, though these results come from theoretical analysis and simulations. The study also highlights the complexity of measuring vortex beam transfer, calling for careful interpretation in related experiments."},{"id":"2609.31133v1","title":"AtomWorld-Mem: Memory-Restored World States for Long-Horizon Atomistic Evolution","abstract":"High-fidelity atomistic evolution over long timescales requires more than observing the current crystal configuration. Instantaneous atomistic snapshots are often incomplete: locally similar configurations can correspond to different hidden dynamical contexts, future event preferences, and waiting-time scales. We argue that this snapshot ambiguity makes long-horizon atomistic evolution fundamentally a memory-based world-state restoration problem. To address this, we introduce AtomWorld-Mem, a memory-restored atomistic world model that recovers the latent world state missing from instantaneous crystal snapshots. AtomWorld-Mem treats the evolving alloy as an AtomWorld: spatial encoders write multi-scale atomistic keyframes from dense local topology and sparse long-range defect context, while short-term event memory and long-term structural memory integrate these keyframes across time to restore a future-predictive evolutionary state. The restored state is used to prioritize legal vacancy-mediated events under single-event Kinetic Monte Carlo (KMC) constraints, while event legality, physical execution, and residence-time updates remain governed by the underlying simulator. Empirically, AtomWorld-Mem improves long-horizon atomistic progress under fixed microscopic event budgets while maintaining high-fidelity evolution across energetic, structural, and vacancy-transport observables. It further transfers zero-shot across diverse unseen alloy-temperature AtomWorlds, suggesting that the learned memory-restoration mechanism captures reusable principles of hidden-state inference rather than a system-specific local energy heuristic. These results position memory-restored world-state modeling as a promising route toward efficient, physically grounded, and transferable atomistic evolution.","published":"2026-09-25T11:26:05Z","updated":"2026-09-25T11:26:05Z","authors":["Tian Luo","Ruge Zhang","Haozhi Han","Yifrng Chen","Yunquan Zhang","Yunxin Liu","Ting Cao","Kun Li"],"categories":["cs.AI","cond-mat.mtrl-sci"],"headline":"New method improves long-term atom behavior predictions using memory","summary":"Researchers developed AtomWorld-Mem, a model that uses memory to better predict how groups of atoms evolve over long times. They found that single snapshots of atoms miss hidden details needed for accurate long-term forecasts, so their method restores this missing information by combining local and long-range atom patterns over time. Tests showed AtomWorld-Mem improves prediction accuracy and works well even on new alloys and temperatures. These results suggest memory-based approaches help understand atomic changes better, though findings are based on simulations and model assumptions."},{"id":"2609.31132v1","title":"Distributed laser energy deposition enables fog-resilient optical communications","abstract":"Fog and clouds remain major obstacles to free-space optical (FSO) communications because micron-sized droplets can impose tens of decibels of attenuation. Ultrashort laser filamentation can transiently clear fog in laboratory-scale experiments, but its extension to atmospheric-scale propagation has remained uncertain. Here we demonstrate laser-assisted fog clearing under atmospheric-scale propagation conditions over a 140-m optical path using a high-energy, high-average-power picosecond laser and weak focusing to sustain extended multifilamentation. A steady-state transparent channel approximately 2 cm in diameter restores up to 12 dB of transmission for an independent optical beam. Increasing the clearing-laser pulse energy primarily extends the multifilament region while maintaining a nearly constant longitudinally averaged deposited energy per unit length of ~2 mJ/m over the measured bundle length. This low average deposition, together with self-sustained filament propagation through dense fog, provides a favorable scaling route toward weather-resilient optical links.","published":"2026-09-25T11:24:38Z","updated":"2026-09-25T11:24:38Z","authors":["Malte C. Schroeder","Victor Moreno","Christophe Coreixas","Jonas Latt","Jean-Pierre Wolf"],"categories":["physics.optics","physics.ao-ph"],"headline":"Laser technique clears fog for better optical communication","summary":"Fog droplets block signals in free-space optical communication, causing major issues. The researchers demonstrated a laser method that clears fog over a 140-meter path in atmospheric-like conditions, creating a nearly 2 cm clear channel that improves signal transmission by up to 12 decibels. Their work shows that using a high-power laser to form multiple filament channels can sustain clearing with low energy use, suggesting a possible way to make optical communication more reliable in foggy weather. These findings are based on controlled experiments and simulations."},{"id":"2609.31131v1","title":"Multiple harmonics from a hand-struck tube: a simple and precise measurement of the speed of sound","abstract":"The speed of sound in air can be determined from the resonance frequencies of the harmonic waves inside a thin cylindrical tube. A pressure wave is generated in the tube by striking several times one of its ends with an open hand and keeping the hand in place. In this way pulses of short duration are produced which propagate along the tube and decompose into a superposition of harmonic waves whose frequencies are determined by the harmonics present in a tube open at one end and closed at the other. The tube thus acts as a filter that reinforces those harmonic waves exhibiting an antinode of acoustic pressure at the closed end and a node at the open end. Their frequencies are determined by these boundary conditions, while waves with different frequencies cancel out by destructive interference. Recording the sound with a microphone connected to a computer and analyzing its spectrum with freely available software, the resonance frequencies are obtained and the speed of sound is determined with an accuracy better than $0.5\\%$. The whole experiment requires nothing but a tube, a microphone and free software, so it is inexpensive, takes a few minutes to set up and is well suited to introductory physics laboratories.","published":"2026-09-25T11:23:49Z","updated":"2026-09-25T11:23:49Z","authors":["Francisco Echegorri","Martin Monteiro","Cecilia Stari","Arturo C Marti"],"categories":["physics.ed-ph"],"headline":"Measuring the Speed of Sound Using a Struck Tube","summary":"Researchers showed that by striking one end of a tube and analyzing the sounds produced, you can measure the speed of sound in air. The tube filters the sound waves so only specific frequencies, related to the tube's shape, are heard. Using a microphone and free software to capture and study these frequencies, the team achieved a speed measurement accurate to within 0.5%. This easy and low-cost method is suitable for basic physics experiments."},{"id":"2609.31130v1","title":"Do we need to answer that question? Salience and Answerability of Potential Questions in Naturalistic Dialogue","abstract":"We empirically investigate Question Under Discussion based modelling in naturalistic dialogue by studying whether the salience of generated potential questions predicts their subsequent resolution. Building on Wu et al. (2024), we construct a dataset of 7,124 questions automatically generated from utterances and preceding context from the British National Corpus, and annotated for salience and answerability. We find a robust but low positive correlation between salience and answerability in dialogue, indicating that more salient questions are more likely to be addressed. However, this effect is markedly weaker than in monologic text, suggesting that conversational structure is less predictable. We further observe that structured interactions exhibit stronger alignment between annotators than less organised dialogues.","published":"2026-09-25T11:23:45Z","updated":"2026-09-25T11:23:45Z","authors":["Amandine Decker","Maxime Amblard","Ellen Breitholtz"],"categories":["cs.CL"],"headline":"More noticeable questions in conversation are slightly more answered","summary":"Researchers studied whether the importance (salience) of questions in natural conversations predicts if they get answered. Using over 7,000 questions from a large speech database, they found a small but reliable link: questions that stand out more tend to be answered more often. However, this link is weaker in conversations than in written texts, suggesting conversations are less predictable. They also found that structured discussions have clearer agreement on question importance than casual talks. These findings are based on data analysis, not experiments."},{"id":"2609.31129v1","title":"Interplay between Emotional Dynamics and Network Structure on the Social Media Vent","abstract":"The sharing and contagion of emotions on social media influence online interactions and users' psychological states. This study analyzes emotional dynamics on Vent, an emotion-sharing social media platform where users explicitly assign emotional labels to their own posts. Using a large-scale dataset, we examine how emotions in users' pre-posting timelines are associated with their subsequent emotional labels and how such associations vary across users. Our analysis reveals three key findings. First, users' subsequent emotional labels are associated with the emotional composition of their pre-posting timelines, with same-category emotions being overrepresented before posts in all analyzed categories. Second, several cross-category associations are observed; for example, Surprise was overrepresented before Fear posts; Affection and Happiness were overrepresented before Anger posts; and Affection was overrepresented before Sadness posts. Third, users differ in their degree of alignment with timeline emotional fluctuations, and highly aligned users tend to be located close to one another in the network. These findings provide large-scale observational evidence that emotional expression on Vent is associated with both short-term timeline context and network structure.","published":"2026-09-25T11:23:26Z","updated":"2026-09-25T11:23:26Z","authors":["Yuina Takahashi","Sho Tsugawa"],"categories":["cs.SI"],"headline":"How Emotions Spread and Connect on a Social Media Platform","summary":"Researchers studied Vent, a social media site where users label their emotions in posts, to understand how emotions appear and spread over time. They found that people’s posts often reflect the emotions seen in their recent timeline, with certain emotions frequently appearing before others. The study also showed that users who closely follow these emotional patterns tend to be connected within the social network. These findings are based on analyzing a large dataset and highlight links between emotional expression, recent emotional context, and social connections on Vent."},{"id":"2609.31128v1","title":"Frame the adversary: a structure-aware attack methodology","abstract":"Frequency-based adversarial attacks have recently grown popular by exploiting spectral sensitivities shared across neural architectures. Unlike spatial perturbations, frequency-based attacks expose deeper vulnerabilities, making them especially valuable for robust evaluation of safety-critical and security-sensitive applications. Yet, existing approaches are typically not derived as solutions to an optimization problem that explicitly captures transform-domain structure. In this paper, we propose a methodology for crafting principled frequency-based adversarial attacks, via a dedicated optimization framework. A cornerstone of our method hinges on the introduction of a perturbation constraint set, tied to highly structured non-orthogonal transforms, well-known for their flexible, non-predefined frequency handling. We prove that the attacks emerge as weighted $\\ell_2$-projections onto this set, yielding a general and controlled attack generation mechanism. By this, we provide a clear geometric attack characterization, ensuring alignment between the optimization objective and the perturbation constraint. We assess our framework on standardized datasets, for pretrained and adversarially robust models. Results highlight that our attacks, being solutions to an optimization problem, over a structured perturbation set, are highly effective, even across different, unseen architectures. Our methodology could serve as a theoretical baseline for designing and analyzing transformed-based attacks, targeting fundamental model vulnerabilities, instead of mere architecture-specific artifacts typically studied in the robustness literature.","published":"2026-09-25T11:22:30Z","updated":"2026-09-25T11:22:30Z","authors":["Vicky Kouni","Stelios Perrakis","Francis Bach","Pascal Frossard","Yann Chevaleyre"],"categories":["cs.LG"],"headline":"New method improves frequency-based attacks on AI models","summary":"The researchers studied attacks that subtly change input data to trick neural networks by focusing on how these changes appear in frequency components, rather than space. They developed a new, mathematically grounded approach using an optimization framework that respects complex frequency patterns to craft such attacks. Their tests on various models showed these attacks are effective and consistent across different AI systems. The work is theoretical and based on controlled experiments, providing a foundation for understanding vulnerabilities in AI models beyond specific designs."},{"id":"2609.31127v1","title":"Large and Moderate Deviations for Conservative Tail-Index Estimation","abstract":"To design systems that are protected against events much rarer than the observational record, extreme-value methods are needed to extrapolate distribution tails. Tail-index estimators such as the Hill estimator are central to this extrapolation, but overestimating the tail exponent can lead to substantial underestimation of rare-event probabilities. Motivated by this, we derive large- and moderate-deviation asymptotics for the Hill estimator and use them to construct estimators whose probability of exceeding the true tail index decays at a controlled exponential rate (the decay rate). In the large-deviations regime, we show that a simple rescaled version of the Hill estimator achieves an optimal balance between bias and decay rate among scale-invariant estimators based on the same top $k$ order statistics. Under a second-order condition, we quantify the effect of the Hill bias, analyze a bias-corrected estimator, and identify sufficient conditions for moderate deviations in the boundary case where the second-order parameter $ρ$ equals zero.","published":"2026-09-25T11:21:47Z","updated":"2026-09-25T11:21:47Z","authors":["Martijn Gösgens","Bart P. G. van Parys","Bert Zwart"],"categories":["math.ST","math.PR"],"headline":"Improving Reliability of Extreme Risk Predictions Using Tail Estimators","summary":"The researchers studied ways to better estimate the likelihood of extremely rare events, which are important for designing safe systems. They focused on the Hill estimator, a common statistical tool, and analyzed its chances of overestimating risks, which can lead to underestimating rare event probabilities. By developing new mathematical results about these estimators, the team proposed improved methods that more reliably control the risk of overestimation, helping to better assess rare event probabilities. Their findings are theoretical and based on mathematical analysis."},{"id":"2609.31126v1","title":"A Systematic Approach To Chaos In PDEs And Their Semidiscretizations","abstract":"We present a systematic approach to characterize Devaney chaos in the solution of homogeneous one dimensional constant-coefficients PDEs on Herzog spaces. We show that for equations where the highest-order time derivative lacks mixed spatial derivatives, there always exists a weighted space in which the solution is chaotic. This analysis is extended to include mixed partial derivatives in the highest-order temporal term. Moreover, we link the chaotic behavior in Herzog spaces to the semidiscretization of the equations by identifying isometric isomorphisms between the two contexts, providing also a characterization for chaos on these semidiscretizations. The work concludes with two applications that provide new insights on the dynamics of the fourth-order Moore-Gibson-Thompson equation and the viscous van Wijngaarden-Eringen equation.","published":"2026-09-25T11:21:12Z","updated":"2026-09-25T11:21:12Z","authors":["Álvaro Vargas-Moreno"],"categories":["math.DS"],"headline":"New Method Explores Chaos in Certain Mathematical Equations","summary":"Researchers developed a way to identify chaotic behavior in solutions to specific one-dimensional equations that describe how things change over time and space. They extended their method to more complex equations and showed how this chaos relates to approximations used in computations. Their approach helps better understand the unpredictable dynamics in two advanced mathematical models, though the findings are theoretical and focus on precise mathematical spaces and abstract concepts."},{"id":"2609.31125v1","title":"Unifying non-Markovian Dynamics and Agent Heterogeneity in Scalable Stochastic Networks","abstract":"Stochastic processes underpin dynamics across biology, physics, epidemiology, and finance, yet accurately simulating them remains a major challenge. Classical approaches such as the Gillespie algorithm are exact for Markovian, time-independent systems, where propensities depend only on the current state and agents of a given type are statistically identical. While efficient, this framework misses a defining feature of many real systems: heterogeneity and memory at the level of individual agents. Cells may divide or differentiate on distinct intrinsic timescales, individuals may preferentially interact with specific partners, and inter-event-time distributions can deviate strongly from the exponential. We introduce MOSAIC (Modeling of Stochastic Agents with Individual Complexity), a general and scalable framework that embeds agent-specific properties directly into the dynamics. MOSAIC unifies heterogeneous rates, dynamic interaction preferences, and both Markovian and non-Markovian waiting-time distributions within a single stochastic formalism, while retaining Gillespie-like computational cost.","published":"2026-09-25T11:20:25Z","updated":"2026-09-25T11:20:25Z","authors":["Aurelien Pelissier","Miroslav Phan","Didier Le Bail","Niko Beerenwinkel","Maria Rodriguez Martinez"],"categories":["math-ph","physics.soc-ph"],"headline":"New method models complex behaviors in large random systems","summary":"Researchers developed MOSAIC, a new framework to better simulate systems where individuals act differently and remember past events, unlike traditional methods assuming everyone behaves the same without memory. Their approach includes diverse activity rates, changing interactions, and realistic timing patterns in one model, while keeping computations efficient. This work supports improved simulation of complex systems but is based on the proposed framework rather than extensive real-world testing."},{"id":"2609.31124v1","title":"An elementary solution to the polarization problem","abstract":"By distilling and combining the ideas of previous arguments, we provide a self-contained and entirely elementary proof of the strong and linear polarization problems.","published":"2026-09-25T11:19:20Z","updated":"2026-09-25T11:19:20Z","authors":["Gergely Ambrus"],"categories":["math.MG","math.CA"],"headline":"Simple Proof for a Complex Math Problem on Polarization","summary":"The researchers offer a straightforward and self-contained proof addressing the strong and linear polarization problems, which are mathematical topics. Their work builds upon and combines earlier ideas to simplify the understanding of these issues. The abstract focuses on providing this elementary proof but does not detail specific applications or real-world implications."},{"id":"2609.31123v1","title":"Global variables and dynamics of emergent cosmology in loop gravity","abstract":"Deriving emergent cosmology from the discrete geometry of Loop Quantum Gravity remains an open challenge. We solve the two-vertex model exactly, including general non-symmetric configurations, by introducing macroscopic global variables that form a closed Poisson subalgebra. This yields an intrinsic minimal area gap from the Casimir invariants of the subalgebra. In the semiclassical limit, a generalized Friedmann equation emerges incorporating graph anisotropies and quantum geometry corrections, which lead to a Big Bounce. This global formulation opens a coarse-graining route toward statistical descriptions of quantum spacetime.","published":"2026-09-25T11:14:27Z","updated":"2026-09-25T11:14:27Z","authors":["Iñaki Garay","Luis J. Garay","Diego H. Gugliotta","Sergio Rodríguez-González","Raül Vera"],"categories":["gr-qc"],"headline":"Researchers explore quantum geometry's role in early universe models","summary":"The researchers address the challenge of deriving cosmology from Loop Quantum Gravity's discrete structure. They solve a specific mathematical model exactly by using new global variables, revealing a minimal quantum area and a modified equation for the universe's expansion that includes quantum effects and anisotropies, predicting a Big Bounce instead of a Big Bang. These theoretical results suggest a new way to describe quantum spacetime statistically, but they remain limited to this particular model and its semiclassical approximation."},{"id":"2609.31122v1","title":"LocUS: Head Selection and Subspace Projection for Targeted Activation Steering","abstract":"Activation steering is a powerful training-free paradigm for controlling large language models at inference time. However, standard approaches estimate a per-layer steering direction from contrastive data and apply it on the layer's entire representation space, which may couple the intervention to off-target properties present in the contrastive data and degrade unrelated capabilities. To mitigate this issue, we introduce LocUS (Localized Unembedding Steering), a method which grounds activation steering to the model's own output vocabulary subspace. By identifying a property-specific linear subspace within the unembedding matrix, LocUS enforces a geometric constraint that restricts the steering transformation to a specific subspace and at the same time localizes its application to a sparse subset of attention heads. Extensive evaluations across three model families on toxicity mitigation, sentiment redirection and sycophancy suppression show that LocUS matches or outperforms state-of-the-art baselines while intervening on under 6% of parameters and better preserving general capability.","published":"2026-09-25T11:13:24Z","updated":"2026-09-25T11:13:24Z","authors":["Irene Tallini","Lorenzo Basile","Valentino Maiorca","Francesco Locatello","Alberto Cazzaniga"],"categories":["cs.CL","cs.LG"],"headline":"New Method Controls AI Model Behavior More Precisely","summary":"Researchers developed a technique called LocUS to better guide large language models during use, aiming to change specific behaviors without affecting others. Unlike existing methods that broadly influence the model's internal workings, LocUS targets a small, focused part related to specific outputs and only adjusts a few components. Tests on tasks like reducing toxic language and changing sentiment suggest LocUS works as well or better than current methods while altering fewer parts and keeping overall model abilities intact."},{"id":"2609.31121v1","title":"Monitor Jailbreaking: Evading Chain-of-Thought Monitoring Without Encoded Reasoning","abstract":"Chain-of-thought (CoT) monitoring is a promising safety technique for reasoning models, enabling detection of problematic reasoning before models act. A key concern is encoded reasoning, where models hide their true reasoning in ways that monitors and humans cannot interpret. Optimization pressure from CoT monitors during reinforcement learning is considered a likely driver of such behavior. We investigate this by training reasoning models to perform a main task and a side task, while penalizing them when a monitor detects reasoning about the side task. Surprisingly, models learn to evade monitors without encoding their reasoning. Instead, they learn to phrase and format their chains of thought such that monitors fail to flag side task reasoning, while the reasoning remains completely transparent to human readers. We call this phenomenon monitor jailbreaking. We find that monitor jailbreaking arises across different model sizes, monitors, and tasks. Jailbreaks generalize to monitors not seen during training, including both less and more capable monitors, and transfer across different monitor prompts. While jailbreaking strategies appear simple, manually replicating them does not reliably fool monitors. Finally, we show that paraphrasing is an effective defense: paraphrasing a jailbroken CoT allows the same monitor to correctly flag it, while still allowing the model to perform both tasks.","published":"2026-09-25T11:13:03Z","updated":"2026-09-25T11:13:03Z","authors":["Julian Schulz"],"categories":["cs.AI"],"headline":"How AI models hide reasoning tricks from safety monitors","summary":"Researchers studied how reasoning monitors detect when AI models think about unwanted tasks. They found models can avoid detection not by hiding true reasoning, but by changing how they phrase thoughts so monitors miss them. This 'monitor jailbreaking' works on many tasks and monitors. Interestingly, humans can still understand the reasoning. The researchers also found that paraphrasing these tricky chains of thought helps monitors catch the hidden reasoning again. These findings come from training and testing AI models and monitors, highlighting a new challenge in AI safety monitoring."},{"id":"2609.31120v1","title":"Probing the Validity of EFT Approaches in Higgs + Jet Production at $\\sqrt{s}=13$ TeV","abstract":"The production of a Higgs boson with a jet is one of the most important processes studied at the LHC. It serves as a sensitive probe of high-energy dynamics and provides a powerful testing ground for the Standard Model, while also offering sensitivity to possible new physics effects. In this work, we perform a phenomenological study of Higgs + jet production in proton-proton collisions at 13 TeV center of mass energy, using the MadGraph5_aMC@NLO framework. We compare predictions obtained from the full loop induced Standard Model with those derived within the HEFT and SMEFT frameworks. Taking the Higgs transverse momentum as a probe, we study normalized distributions and ratio spectra to test the effective field theories. Our results show that effective approaches successfully reproduce the Standard Model behavior in the low transverse-momentum region. However, sizable deviations emerge at high pT, signaling the breakdown of the effective approximation. These findings highlight the importance of carefully accounting for kinematic effects when applying EFT-based methods to precision Higgs studies and provide a quantitative estimate of their range of validity.","published":"2026-09-25T11:12:14Z","updated":"2026-09-25T11:12:14Z","authors":["Shubham Yadav","Prashant Shukla"],"categories":["hep-ph"],"headline":"Testing effective theories in Higgs plus jet production at LHC","summary":"Researchers studied how well simplified theoretical models, called effective field theories (EFTs), describe the production of a Higgs boson with a jet in proton collisions at 13 TeV energy. Using computer simulations, they compared predictions from the full Standard Model to those from EFT approaches. They found that these simpler models match the Standard Model well for low Higgs momentum but deviate significantly at higher momenta, indicating limits to the EFT methods' accuracy in high-energy scenarios."},{"id":"2609.31119v1","title":"SADRA: Sound Capability-based Access Control System for Resource-Disaggregated Architectures","abstract":"Resource disaggregation separates memory and accelerators from compute nodes and makes them remotely accessible. This improves resource sharing, but also removes the local kernel from the resource-access path. Under an untrusted host, compromised host software may use stale authority, exceed delegated authority, or reuse authority provisioned for another process. Prior work identifies capability-based access control as well suited to these architectures. Our systematization of twenty-two prior capability systems finds that none combines host-independent validation of process authority, authoritative enforcement at the resource, and revocation that remains effective while remote authorization state is stale. We present SADRA, a distributed capability-based access-control system for resource-disaggregated architectures. SmartNIC hardware isolated from host software independently checks every inter-node request at two points, first at the compute node against the requesting process's authority and again at the resource against the current authoritative access state. Linked process, compute, and resource capabilities allow these checks to use local state without coordination on the access path. When distributed authority is revoked, SADRA denies subsequent dependent accesses at the resource without waiting for remote nodes to update, while stale capability state is reclaimed separately. We prove capability safety, authority safety, revocation soundness, and strong isolation for a formal architectural model, and model-check the formalization with SPIN. Our FPGA SmartNIC prototype sustains 89.5 Gbit/s aggregate throughput, within run-to-run variation of a non-enforcing baseline that peaks at 90--91 Gbit/s. Cleanup of a 128-capability subtree completes in 528 ns, while access denial does not depend on completion of that cleanup.","published":"2026-09-25T11:10:17Z","updated":"2026-09-25T11:10:17Z","authors":["Hamed Rasifard","Amir Farahani Khojasteh","Hamed Nemati","Michael Backes"],"categories":["cs.CR"],"headline":"New system secures remote access to separated computer resources","summary":"The researchers address security challenges when computer resources like memory and accelerators are separated from main compute nodes and accessed remotely. They developed SADRA, a system that independently verifies access rights both at the computing node and the resource itself, even if the central host is untrusted or outdated. Their prototype hardware showed fast processing speeds and proved mathematically to maintain strong security properties, though results are from formal models and specific hardware tests."},{"id":"2609.31118v1","title":"Architecture Acceleration of Machine Learning Gravitational Waveform Surrogate Models","abstract":"Accurate and computationally efficient waveform models of compact binary coalescences are a fundamental requirement for gravitational wave data analysis. This work presents differentiable and hardware-accelerated implementations of the machine-learning surrogate models mlgw and mlgw-bns within the JAX ecosystem. The proposed framework enables JIT compilation, vector paralellization, native GPU acceleration, and automatic differentiation, while preserving the original surrogate training infrastructure and its applicability to arbitrary non-precessing waveform approximants. Benchmarks are performed in the binary black hole case with a newly trained surrogate of the SEOBNRv5HM approximant; on CPU, they show speed-ups in waveform evaluation time with respect to the original mlgw implementation that exceed one order of magnitude; further, when exploiting GPU acceleration and large-batch vectorization these reach approximately two orders of magnitude. The benchmarks performed in the case of binary neutron stars with a TEOBResumSPA surrogate model achieve similar gains on GPU. Our JAX-based surrogate modeling framework can be integrated into Bayesian parameter estimation pipelines through GWgpu-jax, an open-source package that connects JAX-compatible gravitational waveform generators to JAX-native sampling algorithms. We demonstrate this capability with a nested-sampling pipeline built upon BlackJax-NS, performing parameter estimation analyses of the GW150914 and GW170817 events with mlgw and mlgw-bns. The analyses require approximately twelve and seventeen minutes, respectively, on a single GPU and yield posterior distributions consistent with those reported by the LIGO-Virgo-KAGRA Collaboration. Beyond nested sampling, JAX automatic differentiation provides efficient gradient and Hessian evaluations of the waveform models, enabling integration with gradient-based Bayesian inference methods.","published":"2026-09-25T11:08:53Z","updated":"2026-09-25T11:08:53Z","authors":["Adriano Frattale Mascioli","Lorenzo Piccari","Saulo Albuquerque","Gabriele Demasi","Giulia Capurri","Massimo Lenti","Angelo Ricciardone","Barbara Patricelli","Gianluca M. Guidi","Giulia Stratta","Walter Del Pozzo","Francesco Pannarale"],"categories":["gr-qc","astro-ph.IM"],"headline":"Faster Machine Learning Models Improve Gravitational Wave Analysis","summary":"Researchers developed faster and more efficient computer models to simulate gravitational waves from colliding stars and black holes. Using advanced programming tools, they made these models run much quicker, especially on graphics processors, speeding up calculations by up to 100 times compared to older methods. They demonstrated their approach on real gravitational wave data, showing similar results to established analyses but in much less time, although these findings come from simulations involving specific waveform models and may have practical limitations."},{"id":"2609.31117v1","title":"Comparative Evaluation of an XR Pen-based Control Interface for Semi-Autonomous Mobile Robot Navigation in Service Environments","abstract":"Service robots remain difficult to deploy in domestic environments, partly because fully autonomous operation is not yet reliable in unpredictable surroundings, and partly because conventional control methods remain inaccessible to novice users. Extended Reality (XR) enables operators to visualize robot information overlaid onto the real world and to interact with augmented elements. Yet, common XR control methods, such as motion controllers and hand gestures, are still perceived as unintuitive. This paper presents a control interface that uses a commercial XR pen to command a semi-autonomous mobile robot in Augmented Reality (AR): the operator points at a position in the room, selects it, and drags an augmented arrow to set the desired orientation of the robot at this destination. Two additional interfaces, based on the XR motion controllers and hand gestures, were developed within the same framework. To assess the performance and users' perception of these interfaces, and of the XR pen in particular, a study with 10 participants compared four control methods, i.e., the XR pen, the XR motion controllers, hand gestures, and a computer-based baseline RViz, in navigation tasks performed in a home-like environment. Results show that the XR pen significantly outperforms the other methods in task selection time with the most consistent selections, and that the XR motion controllers obtain the best perceived workload and usability scores, ahead of the computer-based baseline, supporting XR-based control as an intuitive alternative for novice users. However, technical limitations in the integration of the recently released XR pen currently hold back its user experience.","published":"2026-09-25T11:07:24Z","updated":"2026-09-25T11:07:24Z","authors":["Alicia Torc","Carl Tornberg","Eric Piette","Renaud Ronsse","Benoit Macq","Gustavo Alfonso Garcia Ricardez","Lotfi El Hafi","Tadahiro Taniguchi"],"categories":["cs.RO","cs.HC"],"headline":"Using an XR Pen to Control Robots in Home Settings","summary":"Researchers studied ways to help people control service robots in home-like environments, where fully automatic robots struggle and usual control tools can be confusing. They created a new system using an XR pen to direct a robot by pointing and dragging an arrow in augmented reality, and compared it to using motion controllers, hand gestures, and a computer interface. Their study with 10 participants found the XR pen was fastest and most consistent for selecting tasks, though motion controllers scored better on ease of use. The XR pen still has technical issues affecting user experience."},{"id":"2609.31116v1","title":"Multiple change-point detection via bottom-up scanning","abstract":"We study nonparametric multiple change-point detection for high-dimensional sequences, aiming to identify time points at which the underlying distribution changes. While many existing methods perform well when change-points are well-separated, their performance can deteriorate when structural breaks are densely clustered. To address this challenge, we propose gBottomup, a graph-based bottom-up framework for multiple change-point detection in high-dimensional settings. gBottomup constructs a hierarchical segmentation by proposing merges of adjacent segments and verifying them through an unmerge rule that combines absolute significance with relative local heterogeneity, thereby adaptively refining partitions and estimating the number of change-points. Simulation results demonstrate that gBottomup performs reliably across a range of structural configurations and is particularly effective in frequent change-point settings, where existing top-down procedures may lose sensitivity. Runtime experiments indicate favorable computational performance relative to graph-based top-down alternatives. We illustrate the proposed method through an analysis of a S&P 500 dataset.","published":"2026-09-25T11:05:12Z","updated":"2026-09-25T11:05:12Z","authors":["Jinhyeok Park","Hyeyoung Maeng","Hoseung Song"],"categories":["stat.ME"],"headline":"New method improves detection of multiple change-points in data","summary":"The researchers tackle the problem of finding multiple change-points—moments when the pattern in complex, high-dimensional data shifts. They developed a new graph-based technique called gBottomup that builds and refines segments from the bottom up, aiming to better detect closely spaced changes. Through simulations, they found their approach works well especially when changes occur frequently, and it runs efficiently compared to other graph-based methods. The research is supported by simulated tests and an example using stock market data."},{"id":"2609.31115v1","title":"From OpenSPIM to FlowSPIM: Enhancing the versatility of a light sheet microscope through an iterative design process","abstract":"Light sheet microscopy is ideal for the 3D imaging of large biological specimens, yet systematic documentation on adapting microscope designs to diverse experimental requirements remains limited. We present the iterative process used to design and build a modular light-sheet microscope and adapt it to different sample types and experimental conditions. Starting from a stripped-down OpenSPIM platform, we gradually make our microscope more versatile by implementing new hardware, such as additional laser channels and optical filters, and new protocols for sample handling. In its final iteration, FlowSPIM is a new flow-based incubation technique for 3D tissue culture.","published":"2026-09-25T11:04:59Z","updated":"2026-09-25T11:04:59Z","authors":["Endre Joachim Lerheim Mossige","Sergey Ponomartsev","Natalia Smirnova","Wietske van der Ent","Siri Andresen","Xian Hu","Felix Margadant","Rainer Heintzmann","Roland Kádár","Kesavan Sekar","Helene Knævelsrud","Camila Vicencio Esguerra","Stefan Krauss","Dag Kristian Dysthe","Alexander Refsum Jensenius"],"categories":["physics.optics","cond-mat.soft"],"headline":"Improving a versatile microscope for different biological samples","summary":"Researchers worked on enhancing a type of microscope called light sheet microscopy, which is good for 3D imaging of large biological samples. They started with a basic version, OpenSPIM, and gradually added new hardware and methods to handle different kinds of samples and experiments. Their final design, called FlowSPIM, introduces a flow-based system for growing tissues in 3D, showing how the microscope can be adapted through an ongoing design process."},{"id":"2609.31114v1","title":"From Shortcut Learning to Discrete Neural Insertion Sort","abstract":"Neural algorithmic reasoning aims to train neural networks to follow known algorithms and generalize beyond the input sizes seen during training. However, correct final outputs and intermediate supervision do not necessarily show that a model follows the intended execution. We study this problem using insertion sort. Our analysis of the CLRS30 baseline NAR shows that the hint objective is weakly optimized and that hint accuracy remains low. Moreover, many intermediate representations can already be decoded into sorted sequences before the reference insertion-sort execution terminates, suggesting that the model learns a shortcut to the final output. Motivated by these findings, we introduce Discrete Neural Insertion Sort. Our model represents the sequence as a chain, separates scalar exchanges from control-state transitions, and projects node representations back to discrete states after every processor step. When trained only on sequences of length 16, the model achieves $100\\%$ sorted-sequence accuracy on sequences of length 64 and 128. However, an ablation shows that discretization and graph structure alone are insufficient: without additional supervision of the global inner-loop state, the model fails even at the training length. Our results show that discrete execution can support strong length generalization, while also highlighting the problem-specific inductive bias required to learn a faithful algorithmic execution.","published":"2026-09-25T11:00:52Z","updated":"2026-09-25T11:00:52Z","authors":["Konstantinos Mylonas","Thrasyvoulos Spyropoulos"],"categories":["cs.LG","cs.AI"],"headline":"New neural model mimics sorting with improved step-by-step accuracy","summary":"Researchers studied how neural networks learn the insertion sort algorithm and found that existing models often skip the detailed steps, taking shortcuts to the final sorted list. They developed a new neural network, Discrete Neural Insertion Sort, that better follows the sorting process by representing data as linked chains and using discrete steps. While their model trained on short sequences accurately sorted much longer sequences, it still required extra guidance to learn the exact sorting steps. This work highlights challenges in teaching neural networks to truly mimic algorithms."},{"id":"2609.31113v1","title":"Observational constraints on scalar-vector-tensor dark energy with phantom-divide crossing","abstract":"We constrain a scalar-vector-tensor (SVT) dark-energy model in which the vector sector drives the dark-energy equation of state below $-1$ at earlier times, while a canonical scalar field subsequently drives it across $-1$ toward a present-day value above $-1$. We confront the model with DES Year 5 supernovae, DESI DR2 baryon acoustic oscillations, and compressed cosmic microwave background information, and then include Gold-2017 redshift-space-distortion (RSD) measurements to probe structure growth. The background posterior assigns substantial weight to crossing histories and gives a better best fit than the flat $Λ$-cold-dark-matter ($Λ$CDM) model. The RSD analysis preserves this qualitative background evolution, constrains the clustering amplitude, and increases the posterior support for the directional crossing criterion. Since the growth analysis also imposes additional perturbative support conditions, this increase cannot be attributed to the RSD likelihood alone. Matched PolyChord calculations give Bayes factors larger than unity for SVT relative to $Λ$CDM under both prior-volume conventions considered here. The quantitative Bayes factor depends on the prior normalization assigned to the regular radiation-era branch and is larger in the RSD analysis under either convention. Overall, current expansion and structure-growth data are consistent with a stable dynamical crossing of the phantom divide in the SVT model.","published":"2026-09-25T10:58:19Z","updated":"2026-09-25T10:58:19Z","authors":["Nandan Roy","Shinji Tsujikawa","Ying-li Zhang","Zejun Zhang"],"categories":["astro-ph.CO","gr-qc","hep-ph","hep-th"],"headline":"Researchers test a new dark energy model against cosmic observations","summary":"The researchers studied a complex dark energy model where different fields cause its behavior to change over time, crossing a key threshold called the phantom divide. They compared predictions from this model with data about supernovae, galaxy distributions, and cosmic background radiation. Their analysis found that this model fits the observed data as well or better than the standard model, suggesting current measurements are consistent with a stable, dynamic change in dark energy, though conclusions depend on specific assumptions used in their analysis."},{"id":"2609.31112v1","title":"DualManip: Agentic Dynamic Manipulation via Dual-Path Semantic Reasoning and Geometric Adaptation","abstract":"Vision-language models (VLMs) enable open-vocabulary reasoning for robot manipulation, but their high inference latency limits responsiveness in dynamic scenes. Many scene changes, however, alter object geometry without invalidating task intent. We present DualManip, a dual-path framework that decouples infrequent semantic reasoning from responsive geometric adaptation. The semantic path decomposes the task and grounds task-relevant interactions, followed by a constraint-solving module for pose optimization. During execution, the geometric path continuously updates template-to-observation correspondences from live RGB-D observations via a shape-adaptive network. These correspondences transfer task-relevant grasp contacts across observations, enabling online grasp reconstruction under object motion and non-rigid deformation. The Information Interaction Module bridges the two paths by initializing task-relevant grasps from semantic grounding, validating geometric updates, and triggering semantic replanning upon update failures. Real-world evaluation spans six manipulation tasks covering non-rigid deformation, articulated reconfiguration, rigid motion, and high-precision assembly across three settings: static, single-change, and continuous dynamic. DualManip demonstrates superior manipulation robustness, particularly under continuous scene changes, while achieving geometric adaptation approximately 46$\\times$ faster than agentic verification and semantic replanning. Our project page: https://lichengxi1.github.io/Dualmanip.","published":"2026-09-25T10:57:02Z","updated":"2026-09-25T10:57:02Z","authors":["Chengxi Li","Yan Di","Yingyue Li","Ruida Zhang","Mingyang Li","Xiangyang Ji"],"categories":["cs.RO"],"headline":"New Robot System Improves Handling of Moving and Changing Objects","summary":"Researchers developed DualManip, a robot method that separates understanding what to do from quickly adjusting to changes in object shape or position. It uses two linked processes: one plans tasks based on meaning, the other adapts robot movements from live visual data. This approach helps robots better manage tasks involving flexible or moving objects and works much faster than previous methods. The findings are based on tests with real robots performing various tasks with changing objects."},{"id":"2609.31111v1","title":"Energy extraction from a Kerr-Bardeen-Letelier black hole by Comisso-Asenjo Mechanism","abstract":"We explore the energy extraction from a Kerr-like regular black hole coupled with a cloud of strings (CoS) via the Comisso-Asenjo mechanism. We study the effects of the black hole spin (a), black hole charge (g) and CoS parameter (b) on the ergo-region of the black hole and the magnetic reconnection process, which is governed by the Comisso-Asenjo mechanism. We also analyze the phase space involving negative energy density of the relativistic magnetized plasma at infinity. The rate of energy extraction in the magnetic reconnection process is also computed and compared with that of the Blandford-Znajek mechanism.","published":"2026-09-25T10:56:37Z","updated":"2026-09-25T10:56:37Z","authors":["Bijendra Kumar Vishvakarma","Dharm Veer Singh","Sanjay Siwach"],"categories":["gr-qc"],"headline":"Studying Energy Extraction from a Special Black Hole Type","summary":"Researchers studied how energy could be drawn from a particular theoretical black hole combined with string-like matter using a process called the Comisso-Asenjo mechanism. They examined how factors like the black hole's spin, charge, and surrounding strings affect the energy extraction region and magnetic reconnection process. Their work focuses on theoretical analysis and compares energy extraction rates to another known mechanism, the Blandford-Znajek process."},{"id":"2609.31110v1","title":"AuthGuard-R: Safety-Compliant Mission Hijacking and Dual-Gate Defense for LLM-Controlled Robots","abstract":"Large language models are increasingly used as high-level planners for mobile robots, robot manipulators, and autonomous vehicles. Recent studies show that these systems can be influenced through malicious text, speech, visual instructions, retrieved documents, and poisoned sensory context. Most defenses ask whether a proposed action is physically safe. This paper studies a different problem: an action may be physically safe and still violate the mission authorized by the user. An attacker may redirect a delivery robot, replace an approved object, extend a robot's operating region, activate an unnecessary sensor, or delay a mission without creating an immediate physical hazard. We call this attack \\emph{safety-compliant mission hijacking}. We propose MissionPAIR, an adaptive attack framework that searches for executable plans that pass a safety gate while violating an authenticated mission. We also propose AuthGuard-R, a deterministic authorization layer that binds every executable action to a signed mission, robot identity, object and region scope, current state, time, and input provenance. AuthGuard-R operates with an independent safety gate, giving a dual-gate architecture. We formalize mission policies over robot traces, define security games, and prove authorization soundness, mission non-escalation, replay resistance, robot binding, provenance separation, threshold-approval security, audit-log tamper evidence, and trace-level composition. We report a preliminary cross-model evaluation with Claude Haiku~4.5 and the open-source Qwen2.5~7B planner. Across 240 live attack trials, the planners followed an injected mission deviation in 109 trials; AuthGuard-R rejected all 109 resulting unauthorized actions. A separate hand-constructed suite of eleven protocol- and policy-level attacks was also blocked completely.","published":"2026-09-25T10:56:22Z","updated":"2026-09-25T10:56:22Z","authors":["Saidattu Chepuri","Vikas Srivastava"],"categories":["cs.RO","cs.CR"],"headline":"New Defense Stops Robot Missions Being Hijacked Without Physical Harm","summary":"Researchers studied how robots controlled by language models can be tricked into following unauthorized instructions that don't cause immediate harm but violate mission goals. They created a testing method, MissionPAIR, to find such hidden attacks, and developed a defense system called AuthGuard-R that verifies each robot action against a signed mission and other factors. In tests, AuthGuard-R successfully prevented all unauthorized mission changes detected, showing promise for improving robot control security beyond physical safety checks."},{"id":"2609.31109v1","title":"The NEMESIS general YSO catalogue - I. Supervised classification with deep learning methods","abstract":"In modern large-scale surveys, young stellar objects (YSOs) are mostly identified by the infrared excess in their spectral energy distributions (SEDs). To this end, YSO identification and classification were mostly based on certain colours, resulting in wavelength-specific methods. Using all available data allows us to construct a uniformly classified YSO catalogue by applying the same identification method to heterogeneous samples, reducing systematic differences and enabling more consistent studies of YSO populations and star formation. Our goal is to develop a method that takes advantage of the heterogeneous data available in the VizieR database and to provide a tool that can be used to identify YSOs with high accuracy. We also aim to create the largest homogeneous catalogue of YSOs. We use the CDS VizieR and NASA/IPAC IRSA to collect data for a large number of objects of various types. The data collected included SEDs, light curves, and direct imaging and were used to build different image representations. These were used to teach deep learning algorithms to differentiate between YSOs and all other types of sources. Architectures from the PyTorch library and custom-built convolutional neural networks were applied to create an ensemble of binary classifiers specialised at identifying YSOs. The resulting method is able to automatically identify YSO candidates with an accuracy higher than 90\\% while contamination is identified with an accuracy higher than 99\\%. We applied our classifier to numerous YSO catalogues to confirm the youth of the stars and to create the largest catalogue of homogeneously identified potential YSOs ever -- the NEMESIS General YSO catalogue contains 274\\,408 unique, high reliability YSOs. Their positions on the colour-magnitude diagrams, and their distribution relative to the local Galactic structures, were also used to test their young nature.","published":"2026-09-25T10:55:03Z","updated":"2026-09-25T10:55:03Z","authors":["G. Marton","M. Madarász","J. Roquette","M. Audard","I. Gezer","D. Hernandez","O. Dionatos"],"categories":["astro-ph.SR","astro-ph.GA","astro-ph.IM"],"headline":"Deep learning creates largest uniform catalogue of young stars","summary":"Researchers developed a new method using deep learning to identify young stars, called young stellar objects (YSOs), from varied data sources in large astronomical surveys. Their approach analyzes multiple types of data simultaneously, improving accuracy above 90% and reducing errors compared to previous wavelength-specific methods. They applied this method to generate the NEMESIS General YSO catalogue, the largest homogeneous list of potential young stars to date, supporting more consistent studies of star formation."},{"id":"2609.31108v1","title":"Double-stream registration with pyramid fusion for HDR video with alternating exposures","abstract":"High dynamic range (HDR) video reconstruction from al\\-ter\\-na\\-ting-exposure sequences remains challenging, especially in regions with extreme luminance variation. We propose a novel HDR reconstruction framework based on dual-stream registration and accurate pyramid fusion. Given three consecutive frames, our method computes optical flow directly with the central frame, while introducing a complementary midpoint displacement strategy to handle cases with severe overexposition. A pyramid fusion stage then merges the resulting radiance and LDR images into a final HDR output. Experimental results demonstrate that our approach consistently outperforms state-of-the-art methods.","published":"2026-09-25T10:54:22Z","updated":"2026-09-25T10:54:22Z","authors":["Onofre Martorell","Ivan Pereira-Sánchez","Antoni Fuentes","Antoni Buades"],"categories":["cs.CV"],"headline":"New Method Improves Brightness in HDR Videos with Changing Exposure","summary":"Researchers address the challenge of creating clear high dynamic range (HDR) videos from sequences where exposure changes between frames, especially in very bright or dark areas. They developed a new technique that aligns video frames using a dual-stream registration and combines them through a pyramid fusion process to produce better HDR images. Their experiments show this approach works better than current leading methods, though the summary is based on experimental comparisons rather than real-world testing details."},{"id":"2609.31107v1","title":"Bayesian Optimization with Fisher Information Geometry: Gradient Bounds and Trust-Region Methods","abstract":"We study Bayesian optimization (BO) through the lens of information geometry. Pulling back the Fisher information metric through the surrogate posterior map yields a local sensitivity tensor on the input space, which leads to an upper bound on the gradient of reparameterizable acquisition functions. This view explains vanishing-gradient behavior in high-dimensional BO and provides a common interpretation of heuristics such as RAASP and dimension-scaled lengthscales. Building on this analysis, we propose FITR, a trust-region-based BO method that replaces lengthscale-based scaling by local pullback-Fisher weights. FITR is not restricted to GP kernels with explicit lengthscales. On GP benchmarks with an SE kernel, experiments show competitive performance using FITR. The proposed method also easily generalizes to non-isotropic surrogates, although the gains are more task-dependent in that setting.","published":"2026-09-25T10:52:30Z","updated":"2026-09-25T10:52:30Z","authors":["Saksham Kiroriwal","Julius Pfrommer","Jürgen Beyerer"],"categories":["cs.LG","cs.AI","cs.IT"],"headline":"New Method Improves Bayesian Optimization Using Information Geometry","summary":"The researchers studied Bayesian optimization, a method to find optimum solutions, using concepts from information geometry, which examines sensitivity in input space. They identified why certain algorithms struggle in high dimensions and explained existing heuristics through this perspective. Based on this, they developed a new approach called FITR, which adjusts optimization steps more locally and flexibly than traditional methods. Their experiments on standard tests showed competitive results, though improvements vary depending on specific problems. These findings come from theoretical analysis and simulations."},{"id":"2609.31106v1","title":"Probing the cosmic web with fast radio bursts: II. Scintillation","abstract":"We consider the formation of multiphase gas in cosmic web objects (CWOs) through cooling and fragmentation within the shock-heated medium of cosmic sheets and filaments, and the circumgalactic medium (CGM) of haloes. This leads to dense, cool ($\\sim10^{4}\\,\\mathrm{K}$), small-scale cloudlets in pressure equilibrium with their surroundings. Such ionized overdensities can affect the propagation of radio waves emitted by fast radio bursts (FRBs), allowing us to probe the small-scale structure in CWOs by examining their impact on the scintillation pattern of the observed FRB flux. While we find that the probability of encountering a high-temperature filament or sheet with cold cloudlets is rather small and usually confined to high redshifts, the expected high rates of FRBs may still permit detection through scintillation along rare sightlines. We suggest scintillation from the Milky Way (MW) interstellar medium as a promising probe: CWO plasma screens can suppress it, placing constraints on their cold-gas properties. Filaments and sheets can suppress MW scintillation from sources at $z_\\mathrm{S}\\gtrsim2$ and $3$, respectively, while cold and turbulent cloudlets in the CGM can be probed by sources at redshifts as low as $z_\\mathrm{S}\\sim0.5$, constraining remarkably low volume-filling fractions, $f_\\mathrm{v}$ (e.g. $f_\\mathrm{v}\\gtrsim10^{-5}$ for $z_\\mathrm{S}\\gtrsim2$). To provide testable predictions, we estimate the fraction of FRBs expected to encounter CWOs that may suppress MW scintillation below a given frequency, and express this as a function of the average extragalactic dispersion measure. Future FRB samples could map these constraints onto bounds on the existence and properties of cloudlets in CWOs, such as their density, size, $f_\\mathrm{v}$, and turbulence.","published":"2026-09-25T10:49:18Z","updated":"2026-09-25T10:49:18Z","authors":["Sharon Lapiner","Nir Mandelker","Paz Beniamini","S. Peng Oh"],"categories":["astro-ph.CO","astro-ph.GA","astro-ph.HE"],"headline":"Using fast radio bursts to study small gas clouds in space","summary":"The researchers explore how tiny, cool gas clouds in large cosmic structures, like sheets and filaments, affect radio waves from fast radio bursts (FRBs). These gas clouds can change the flickering pattern, or scintillation, of the radio signals we detect. By analyzing this effect, scientists may learn about the properties of these clouds and their distribution in the cosmic web. Their findings are based on theoretical analysis and suggest future FRB observations could help test these ideas, but results currently depend on rare cases and assumptions about distant cosmic environments."},{"id":"2609.31105v1","title":"Edge ideals of complements of a class of circular interval graphs","abstract":"Let $C_n^m$ be the $m$-th power of the cycle on $n$ vertices, and let $\\Gnm$ be its complement, where $m\\ge1$ and $n\\ge3m+1$. We prove that every induced subcomplex of the clique complex of $C_n^m$ is homotopy equivalent either to a disjoint union of contractible complexes or to a circle, and we establish this dichotomy for a larger class of circular interval graphs. As a consequence, we obtain a closed formula for the linear strand of the Betti table of $R/I(\\Gnm)$. We also prove that $R/I(\\Gnm)$ is a Buchsbaum ring of depth $2$ and that the independence complex of $\\Gnm$ is a triangulated $m$-manifold with boundary for $m\\ge2$, namely a Möbius band or an annulus when $m=2$. For the complement $\\overline H$ of every graph $H$ in the above class, we show that $\\reg(I(\\overline H)^k)=2k+\\indm(\\overline H)-1$ for all $k\\ge2$, where $\\indm$ denotes the induced matching number. In particular, some power of $I(\\Gnm)$ has a linear resolution if and only if $n\\ge4m+1$. In this range, we further prove that the colon ideals $(I^{k+1}:M)$, where $I=I(\\Gnm)$ and $M$ is a minimal monomial generator of $I^k$, have regularity $2$.","published":"2026-09-25T10:49:16Z","updated":"2026-09-25T10:49:16Z","authors":["Bilal Ahmad Wani"],"categories":["math.AC"],"headline":"Researchers study shapes and algebra of special circular graphs","summary":"The researchers examined a specific family of circular graphs and their complements, focusing on their structural and algebraic properties. They showed that the associated shapes are either collections of simple pieces or circles, and found exact formulas describing certain algebraic features of these graphs. Their work reveals connections between graph theory, topology, and algebra, but the results are theoretical and concern very particular types of graphs and algebraic objects."},{"id":"2609.31104v1","title":"On the Moments of Least $r$-Gaps of Partitions and a Conjecture of Baruah and Talukdar","abstract":"The minimal excludant or mex of a partition, introduced by Andrews and Newman \\cite{AN2019,AN2020}, is the smallest positive integer missing from that partition. Baruah, Bhoria, Eyyunni and Maji \\cite{BBEM2023} studied the sum of mex split according to parity, together with its $k$-th moments. Ballantine and Merca \\cite{BM2020} generalized mex to the least $r$-gap, which is the smallest natural number that does not appear at least $r$ times in the partition. Baruah and Talukdar \\cite{BT2026} conjectured a corresponding asymptotic equivalence between the sums of odd and even least $r$-gaps for every $r>1$, generalizing a theorem of Barman and Singh \\cite{BS2024} for the classical mex. In this article, we derive exact formulas for the $k$-th moments of $r\\text{-}\\mathrm{mex}(π)$ for every fixed $k\\geq1$, in terms of partition functions. We give a complete proof of the conjecture of Baruah and Talukdar \\cite{BT2026} for every natural number $r>1$. We also generalize an identity of Hopkins, Sellers and Stanton to the least $r$-gap setting.","published":"2026-09-25T10:46:06Z","updated":"2026-09-25T10:46:06Z","authors":["Sourav Bhowmick","Nabin Kumar Meher"],"categories":["math.NT","math.CO"],"headline":"Researchers prove a math conjecture about special numbers in partitions","summary":"The researchers studied a concept called the least r-gap, which is the smallest number missing a set number of times in a partition (a way of breaking down numbers). They developed exact formulas for important statistical measures of these gaps and proved a previous conjecture about the balance between sums of odd and even least r-gaps for all r greater than one. Their results extend earlier work and generalize known mathematical identities, all based on precise calculations using partition functions."},{"id":"2609.31103v1","title":"DepthEvidence: Unifying Metric Depth Prediction and Geometric Reasoning in Multimodal Language Models","abstract":"Spatial reasoning with metric constraints requires linking objects to geometric measurements and preserving their numerical content during language reasoning. We present DepthEvidence, a 4B model that uses its own dense metric predictions as object-grounded evidence for language generation. A camera-conditioned decoder predicts full-resolution metric depth using multi-scale visual features and high-resolution RGB refinement. A dense-to-language interface converts predicted depths and decoder features into object-aligned continuous geometry tokens anchored to object identifiers. Geometric supervision encourages metric information to remain recoverable before and after language-context interaction, while instruction tuning supports object measurement and compositional reasoning. We introduce a Depth-VQA benchmark evaluating object-depth queries, relative comparisons, and decisions combining spatial and numerical constraints. Across nine datasets, DepthEvidence achieves the highest average dense $δ_1$ among evaluated methods, competitive with specialized estimators. It also leads the evaluated methods in instance-level metric depth estimation and overall accuracy on both relative and metric reasoning tracks, while broadly preserving general VQA performance and improving spatial understanding relative to the base model.","published":"2026-09-25T10:42:20Z","updated":"2026-09-25T10:42:20Z","authors":["Jiangning Wei","Yuan Yao","Miaomiao Cui","Mingsheng Li","Humen Zhong","Shuai Bai","Zhibo Yang"],"categories":["cs.CV","cs.AI"],"headline":"New AI Model Links Depth Sensing with Language Understanding","summary":"Researchers developed DepthEvidence, an AI model that predicts precise depth information from images and uses this data to improve understanding of spatial language. The model estimates distances and geometry around objects, converting this into detailed tokens that help it answer questions involving object measurements and spatial reasoning. Tests on multiple datasets show DepthEvidence performs well in depth estimation and spatial question answering while maintaining general language abilities, though the findings are based on the model's testing rather than real-world use."},{"id":"2609.31102v1","title":"Affine pure-jump Volterra fields","abstract":"We study a class of non-negative spatio-temporal random fields that exhibit self-exciting clustering, which we refer to as affine pure-jump Volterra fields. They are defined via stochastic integration of a Volterra kernel against a thinned Poisson random measure and the thinning is such that the field admits a representation as a stochastic integral of the same kernel, but against a random measure whose compensator has a density that is pointwise affine in the field itself. This representation leads to affine transform formulas characterizing the Laplace transform of functionals of the field up to the solution of a deterministic non-linear Volterra integral equation. Affine pure-jump Volterra random fields extend a non-negative finite-first-moment subclass of pure-jump affine Volterra processes to the random-field setting. The framework includes non-negative finite-variation ambit fields, marked Hawkes systems, and branching-type models as special cases, and we make these connections explicit through examples.","published":"2026-09-25T10:40:38Z","updated":"2026-09-25T10:40:38Z","authors":["Sven Karbach","Thomas K. Kloster"],"categories":["math.PR"],"headline":"New mathematical model for complex random spatial-temporal patterns","summary":"Researchers studied a type of mathematical model called affine pure-jump Volterra fields, which describe random spatial and time patterns that cluster and influence themselves. They developed formulas connecting these fields to a specific type of integral equation, extending previous models to more complex situations. Their work includes examples like Hawkes systems and branching models, but findings are theoretical and focused on mathematical properties rather than practical applications."},{"id":"2609.31101v1","title":"A Flatness-Generalization Relation in the Teacher-Student Tree-Committee Machine","abstract":"The flatness of the loss landscape at a minimizer is a widely used heuristic for reasoning about neural-network generalization, yet evidence for this relation is mostly empirical and controversial. We study this relation in a teacher-student tree committee machine, where both the ERM estimator and the Hessian spectrum are analytically tractable in the proportional high-dimensional limit. First, we use a zero-temperature Gibbs formulation to obtain predictions for the observables of the typical minimizers of the empirical loss. Secondly, we use Edwards-Jones formalism to derive the limiting Hessian resolvent around these typical minimizers. All predictions agree with finite-size gradient-descent simulations. Finally, we study three measures of flatness, namely the left and right edges and the spectral mean, and check if a decrease in generalization error as the dataset size is increased corresponds to an increase in flatness. We find that the answer strongly depends on the learning task and on the ratio of the number of parameters to the number of data points. In regression, the spectral mean and right edge correlate with the generalization error, while the left edge does so only in the overparametrized regime. In classification this correlation reliably holds only in the highly overparametrized phase, while for underparametrized networks it can even reverse.","published":"2026-09-25T10:38:34Z","updated":"2026-09-25T10:38:34Z","authors":["Brandon Livio Annesi","Davide Straziota","Enrico Maria Malatesta"],"categories":["cond-mat.dis-nn","cs.LG","stat.ML"],"headline":"Researchers analyze neural network flatness and generalization link","summary":"The researchers explore how the flatness of a neural network’s loss landscape relates to its ability to generalize, using a simplified model called a teacher-student tree committee machine. They apply advanced mathematical tools to predict characteristics of network solutions and verify these with simulations. Their findings suggest that the connection between flatness and generalization varies depending on the type of task and the balance between model size and data amount, with different patterns seen in regression versus classification problems."},{"id":"2609.31100v1","title":"Optimal Power Sharing for Hybrid Energy Storage Systems Based on Karush-Kuhn-Tucker Conditions","abstract":"Renewable energy power plants are increasingly expected to provide ancillary services to the grid. Yet, the variability of sources such as solar photovoltaics requires additional operational flexibility to deliver these services. In this context, hybrid energy storage systems (HESSs) offer an appropriate solution, because different technologies with complementary characteristics can be applied and leveraged to share the power demand and improve the overall performance. Yet, a proper power sharing requires a detailed model of the storage elements. This aspect has barely been studied in the literature, where most models consider constant efficiencies. Moreover, most formulations are based on optimisation problems that are computationally too demanding to be solved in real time. In this paper, a controller is proposed to minimise the losses of a HESS considering detailed power-dependent efficiency curves of each storage technology. The power sharing method is based on the analytical verification of the Karush-Kuhn-Tucker (KKT) conditions, which makes it computationally efficient and suitable for real-world deployment. The proposed controller is applied to a test case consisting of a 10 MW PV power plant with a HESS based on a lithium-ion battery and a redox-flow battery, each with 2.5 MW power and 5 MWh capacity. The main contributions are verified via numerical simulations performed in MATLAB/Simulink, whereas a real-time implementation deployed in OPAL-RT demonstrates the viability of the algorithm for real-time applications.","published":"2026-09-25T10:38:34Z","updated":"2026-09-25T10:38:34Z","authors":["Juan Diego Rios-Peñaloza","Milan Prodanovic","Javier Roldán-Pérez"],"categories":["eess.SY"],"headline":"Efficient Power Sharing in Hybrid Energy Storage for Solar Plants","summary":"Researchers studied how to better manage hybrid energy storage systems (HESS) that combine different battery types to support solar power plants. They developed a controller that optimizes power sharing by accurately modeling how efficiently each storage type works under different power levels. Their method uses a mathematical approach (KKT conditions) that runs quickly enough to be used in real time. Simulations and a real-time test show the approach is promising, but these results are based on models and controlled tests rather than large-scale field experiments."},{"id":"2609.31099v1","title":"Collision-free Movement on Grids and Beyond","abstract":"We study collision-free movement problems on graphs, where the task is to coordinate a set of robots so that they reach a target formation satisfying a desired property while minimizing the total travel distance. This framework extends two classical models: (a) minimizing movement [Demaine et al., TALG '09, '14], which does not enforce collision avoidance, and (b) coordinated motion planning or multi-agent path finding [Eiben et al., SoCG '23, Deligkas et al., ICALP '24, among many others], where each robot is assigned an explicit target position. We focus on the setting where the target formation of the robots should be connected. We analyze the parameterized complexity of the problem with respect to the number of (main) robots and the total travel length on grid graphs and two natural generalizations thereof: planar graphs and unit disk graphs.","published":"2026-09-25T10:37:23Z","updated":"2026-09-25T10:37:23Z","authors":["Hendrik Molter","Meirav Zehavi"],"categories":["cs.MA","cs.DS"],"headline":"Coordinating Robots to Move Without Collisions on Graphs","summary":"Researchers studied how to move multiple robots on grid-like networks so they avoid collisions and reach a connected target shape while minimizing total travel distance. They extended previous models by combining goals of distance minimization and explicit robot destinations. Their work analyzes the complexity of this problem on different types of graphs, including grids, planar graphs, and unit disk graphs, focusing on how the number of robots and total travel length affect the difficulty. The findings are theoretical and based on computational analysis."},{"id":"2609.31098v1","title":"The Residual Stream's Effective Depth","abstract":"We introduce \\emph{effective depth} ($\\Deff$), a scalar diagnostic that treats the layer-wise residual stream of a transformer as a discrete-time process, measures how representation similarity decays with layer distance, and aggregates that profile into one number. Across sixteen decoder-only language models, $\\Deff$ separates a structural consequence of residual accumulation from an empirical one: even maximally diverse orthogonal updates have the closed-form reference $F_L = 2L/(L+1)<2$, yet fifteen of sixteen default measurements lie below $F_L$ (Qwen3.5: 32--44\\%, OLMo-2: 40--41\\%, Pythia: 23--28\\%). Matched references show that the gap is not caused by the persistent initial state or update-size imbalance, but is largely a calibrated signature of correlated residual updates rather than evidence that depth is unused. Symmetric position-0, token-normalisation, and top-PC controls show the regime is not reducible to BOS or top-PC artefacts: the lone above-reference default outlier joins the same regime, and all sixteen models are sub-reference after token-normalisation or top-1-PC removal. Intermediate checkpoints show that the regime is established early in OLMo-2 and stable through 5T tokens, while Pythia-1.4B follows a distinct decreasing trajectory. A controlled residual-carry intervention supports the mechanism, and $\\Deff$ is best read as a \\emph{global} accumulated-state diagnostic, not as a capability score or pruning method.","published":"2026-09-25T10:36:43Z","updated":"2026-09-25T10:36:43Z","authors":["Barak Gahtan","Ido Galil","Alex M. Bronstein"],"categories":["cs.LG"],"headline":"Measuring Information Flow Depth in Transformer Language Models","summary":"Researchers developed a new way to measure how information changes across layers in transformer language models by introducing \"effective depth.\" They analyzed sixteen models and found that updates in the models' layers are more correlated than expected from purely diverse changes, which reflects a global pattern rather than a lack of model depth usage. Their findings come from mathematical analysis and controlled experiments, showing this measure captures structural behavior but isn't directly tied to model capabilities or pruning."},{"id":"2609.31097v1","title":"Supporting functionals and singular boundary geometry of two-qubit entanglement of formation","abstract":"We give a criterion for the existence of a global supporting affine functional for the entanglement of formation at an entangled two-qubit state. It exists if and only if the spin-flip bilinear form is nondegenerate on the support of the state, or equivalently, if the last concurrence root associated with the support is nonzero. In this case the entanglement of formation is locally Lipschitz continuous. In the degenerate case the state has rank two or three, and its projective support is tangent to the product-state quadric. For suitable perturbations outside the support, the entanglement of formation decreases as the square root of the mixing parameter. The leading term is obtained explicitly. This square-root decrease violates the one-sided Lipschitz bound required for a global supporting affine functional.","published":"2026-09-25T10:35:38Z","updated":"2026-09-25T10:35:38Z","authors":["Wei Song","Xiao-Lan Zong"],"categories":["quant-ph"],"headline":"New insights into measuring two-qubit quantum entanglement","summary":"Researchers studied a mathematical feature called 'entanglement of formation' that quantifies quantum entanglement between two qubits. They found a precise condition when a certain supporting function exists, linking it to a property called the spin-flip form. Their work shows how entanglement behaves smoothly in some cases, but can drop sharply in others, revealing complex boundary behavior. These conclusions are theoretical, based on mathematical analysis rather than experiments."},{"id":"2609.31096v1","title":"Rigidity of unit spheres in finite-dimensional real normed spaces","abstract":"We prove that finite-dimensional real normed spaces with isometric unit spheres are linearly isometric, where each sphere carries the distance induced by the ambient norm. This answers the object-level question of Kadets and Martín in finite dimensions, without asserting linear extension of a prescribed sphere isometry. We associate with each norm a family of Plücker contraction bodies generated by volume-normalized maximal minors of linear contractions into finite-dimensional $\\ell_\\infty$ spaces. Null-Lagrangian identities show that the sphere metric determines these bodies. A finite exposed-face construction recovers almost-norming contractions from the convexified data, and compactness together with an exact volume identity yields a linear isometry.","published":"2026-09-25T10:35:01Z","updated":"2026-09-25T10:35:01Z","authors":["Jumpei Nakamura","Ryotaro Tanaka"],"categories":["math.FA"],"headline":"Unit spheres fully determine shape of finite-dimensional spaces","summary":"Researchers studied finite-dimensional spaces defined by a way to measure length (norms). They showed that if two such spaces have unit spheres that are exactly the same shape with the same distances, then the spaces themselves are essentially identical in their distance measurements. The result answers a previously open question but does not claim a way to extend sphere matching to the entire space. Their work uses geometric constructions and mathematical identities to link sphere shapes to underlying space structures."},{"id":"2609.31095v1","title":"Confident, Not Wiser: The Dunning-Kruger Effect in Human-AI Interaction","abstract":"AI assistance can improve performance without improving self-assessment. We report a study (N=366) comparing Human alone and Human+AI performance on reasoning tasks, for which the AI model is benchmarked on the same items. Participants estimated global and block performance and rated confidence in their answers. Human+AI achieved higher scores, but self-estimates tracked performance weakly. Average overestimation was similar across groups, covering individual errors. Across tasks, confidence distinguished correct from incorrect answers less accurately in the Human+AI group, while within-task differences remained uncertain. The Dunning-Kruger pattern was found in both groups, with a larger observed contrast in Human+AI. Controls for score noise reduced but did not eliminate the pattern, with the controlled group difference remaining inconclusive. An extended computational account describes global and block estimates. Our findings distinguish performance augmentation from metacognitive augmentation and motivate interfaces that support verification, communicate task-specific AI model performance, and help users evaluate the quality of their joint work rather than produce answers.","published":"2026-09-25T10:32:33Z","updated":"2026-09-25T10:32:33Z","authors":["Daniela Fernandes","Michelle Rausch","Agnes Mercedes Kloft","Daniel Buschek","Robin Welsch"],"categories":["cs.HC"],"headline":"AI boosts performance but not people's self-awareness of skills","summary":"Researchers studied how people perform reasoning tasks alone versus with AI help. They found that although AI assistance improved task scores, it did not improve participants' ability to accurately judge their own performance. Both groups overestimated their abilities similarly, showing the Dunning-Kruger effect where people are confident but unaware of their mistakes. The study suggests AI can enhance outcomes but does not necessarily help users better assess or understand their own or the combined work quality."},{"id":"2609.31094v1","title":"Thermodynamics of the $q$-deformed Kittel--Shore model for spin-1 particles","abstract":"Quantum algebras ($q$-algebras) have been used in the literature to deform spin-chain models. In particular, in previous works by the present authors, the Kittel--Shore Hamiltonian was deformed using $\\mathfrak{su}_q(2)$ algebra. The deformed Hamiltonian was presented for a general spin, and the thermodynamic properties were studied for spin-1/2 particles. In this paper, we focus on the $q$-deformation of the Kittel--Shore model for spin-1 systems. We analyse the impact of this deformation on the thermodynamic properties and phase transitions of the system as a function of the deformation parameter $q$. Specifically, we study the specific heat, magnetic susceptibility, and magnetisation, including a detailed analysis of the Curie temperature characterising the ferromagnetic phase. Furthermore, a finite-size scaling analysis is conducted for both the ferromagnetic and antiferromagnetic cases. Results are systematically compared with those of the undeformed model to reveal the physical effects of the $q$-deformation.","published":"2026-09-25T10:30:51Z","updated":"2026-09-25T10:30:51Z","authors":["V. Mariscal","J. J. Relancio"],"categories":["cond-mat.stat-mech","math-ph"],"headline":"Studying a quantum model's heat and magnetism with new math tools","summary":"The researchers explored how a mathematical change, called $q$-deformation, affects a quantum model describing particles with spin-1, which relate to magnetism. They examined properties like heat capacity and magnetism, focusing on how these change with the deformation parameter and temperature, especially at phase transitions. Their study used calculations and comparisons to the original model to reveal how this deformation influences magnetic behavior, but the work is theoretical and centers on specific quantum mathematical models."},{"id":"2609.31093v1","title":"Block Sparse Attention with Log-Linear Complexity","abstract":"Scaling language models to long contexts is limited by the quadratic cost of self-attention. Block sparse attention offers an efficient alternative, but selecting the retained blocks remains a bottleneck. Conventional block selection requires scoring all query-block pairs and therefore remains quadratic in sequence length. To address this issue, we propose PISA, a block-sparse attention mechanism that employs a pyramid Top-$K$ selection strategy. The main idea is to gradually narrow down the candidates across different levels, making it more efficient to find the most relevant keys. Specifically, we construct a coarse-to-fine hierarchy of keys and perform selection from the coarsest level. At each level, LogSumExp scoring is applied to a bounded candidate set to select candidates for the next finer level, continuing until the finest level is reached. Through pooling, we construct $O(\\log N)$ levels of keys, yielding an overall complexity of $O(N\\log N)$, where $N$ denotes the sequence length. We develop hardware-aware Triton kernels for both training and inference, fusing hierarchical routing and LogSumExp scoring without materializing the query-key score matrix. We further evaluate our method on language modeling tasks. Compared with the baseline, our method achieves comparable performance on benchmarks such as commonsense reasoning while delivering better results on retrieval tasks.","published":"2026-09-25T10:28:58Z","updated":"2026-09-25T10:28:58Z","authors":["Bohao Tang","Zhen Qin","Yuqi Pan","Zheng Li","Pengfei Liu"],"categories":["cs.LG"],"headline":"New method improves efficiency of language model attention","summary":"Researchers address the problem of slow processing in language models caused by traditional attention methods that require extensive calculations for long text sequences. They introduce PISA, a block-sparse attention technique that narrows down important information step-by-step using a pyramid-like approach, reducing the computation from quadratic to log-linear complexity. Their experiments on language tasks show that PISA performs similarly to existing methods on reasoning tests and better on retrieval tasks, but these findings are based on their specific benchmarks and implementations."},{"id":"2609.31092v1","title":"The Power of Indirection: Scaling Switches Beyond Silicon Boundaries","abstract":"The slowdown of Moore's law and the area limit of monolithic integration have made chiplet-based designs inevitable across many domains, including network ASICs. However, combining multiple network ASICs together poses a fundamental challenge: maintaining sufficient inter-ASIC bandwidth to match the performance of an idealistic single-ASIC design. Providing full bandwidth is prohibitively expensive as it requires valuable forwarding capacity, while reducing inter-ASIC bandwidth creates severe performance bottlenecks. We propose a novel multi-ASIC switch architecture that introduces a circuit-switched indirection layer in front of the ASICs. This layer flexibly remaps ingress ports across ASICs, localizing traffic and minimizing inter-ASIC communication based on observed patterns. Our system, Fastroute, combines packet and circuit switching to deliver performance comparable to a single-ASIC switch while reducing inter-ASIC bandwidth requirements. This frees up capacity for external network interfaces, allowing Fastroute to outperform traditional non-oversubscribed multi-ASIC designs. Our hardware prototype demonstrates the system's functional feasibility by evaluating it on an LLM training workload. By reducing bandwidth and power overhead, Fastroute bridges the gap between silicon fabrication limits and soaring application demands. It provides an efficient transition to multi-ASIC switches, enabling bandwidth and radix demand to be met without waiting for the next ASIC generation.","published":"2026-09-25T10:28:39Z","updated":"2026-09-25T10:28:39Z","authors":["Lukas Röllin","Sushovan Das","Paolo Costa","Laurent Vanbever"],"categories":["cs.NI"],"headline":"New Chip Design Boosts Multi-Chip Network Switch Performance","summary":"Researchers address the challenge of combining multiple network chips without losing speed or creating bottlenecks. They designed a system called Fastroute that smartly redirects traffic inside a switch to reduce costly communication between chips. Their prototype showed it can match single-chip performance while using less bandwidth and power, supporting demanding tasks like language model training. The findings suggest a practical way to build bigger, faster network switches despite physical chip size limits, though results are based on hardware testing of their design."},{"id":"2609.31091v1","title":"Long-time and large-population limits of displacement monotone linear-quadratic Nash systems","abstract":"For linear-quadratic $N$-dimensional Nash systems with mean-field-like scaling and strongly displacement monotone data, we prove uniform-in-time a priori derivative estimates exhibiting the expected scaling as $N \\to \\infty$ in order to pass the system to the limit. As a consequence, in a heterogeneous (i.e., non-symmetric) mean-field setting, we obtain both uniform-in-time convergence of the Nash system on any arbitrarily long horizon $[0,T]$ (as $N \\to \\infty$), and convergence of both the $N$-dimensional Nash system and the corresponding master equation to their respective ergodic counterparts (as $T \\to \\infty$).","published":"2026-09-25T10:28:13Z","updated":"2026-09-25T10:28:13Z","authors":["Davide Francesco Redaelli"],"categories":["math.OC","math.AP"],"headline":"Studying long-term behavior in large group decision systems","summary":"The researchers looked at mathematical models where many individuals interact strategically, called Nash systems, focusing on linear and quadratic settings with specific conditions called displacement monotone data. They proved that as the number of individuals grows, the system's behavior and its mathematical descriptions remain well-behaved over any length of time. Their results show convergence to simpler, steady-state forms when both the population is large and time is long, but these findings are based on theoretical analysis rather than practical tests."},{"id":"2609.31090v1","title":"Moment Ambiguity and the Limits of Robust Stochastic Optimization","abstract":"We study fundamental information-theoretic limits of robust stochastic optimization when the distribution is known only through its exact moment sequence. We develop a unified framework that produces families of distinct distributions sharing all moments yet inducing radically different optimal decisions, thereby establishing strong impossibility results for a range of decision problems under moment ambiguity. Our approach gives two explicit constructions: a binary and an $N$-way construction showing that distributions with identical moment sequences can nevertheless exhibit arbitrarily different quantiles, order-statistics and threshold regions, forcing incompatible optimal actions. These families of distributions yield, in fact, strong impossibility results across several stochastic optimization problems. First, for the newsvendor problem, moment equivalence causes quantile ambiguity, inducing any fixed or randomized order quantity to fail arbitrarily badly. Second, for revenue maximization, no deterministic or randomized posted pricing scheme can secure a nontrivial approximation relative to the full-information benchmark. Third, for the secretary with cardinal observations setting, the worst-case robust value over all exact moment disclosures is exactly the classical $1/e$ finite-horizon value as opposed to the celebrated result of $0.58$ success probability with full-information by Gilbert and Mosteller (J. Am. Stat. Assoc., 1966). We also recover and expand upon the recent impossibility result of Correa et al. (STOC, 2026) for prophet inequalities with moment knowledge. Indeed, exact moment knowledge can yield at best a $Θ(1/\\log n)$ competitive ratio, even when competing against relaxed benchmarks based on expected $r$-th order statistics or when the algorithm is allowed to select $r$ items.","published":"2026-09-25T10:27:36Z","updated":"2026-09-25T10:27:36Z","authors":["Andrés Cristi","Matteo Russo","Jiechen Zhang"],"categories":["cs.IT","cs.DS","cs.GT","math.OC"],"headline":"Researchers expose limits of decision-making using moment data","summary":"The researchers explore how knowing only a distribution's moments (summary numerical characteristics) can lead to very different optimal decisions, revealing fundamental challenges in decision-making under uncertainty. They construct examples of distributions that share all moments but differ greatly in key outcomes, causing optimal strategies for problems like inventory ordering, pricing, and selection to fail or perform poorly. Their findings highlight strong theoretical limits on making robust decisions when only moment information is available, based on mathematical constructions and analysis rather than real-world experiments."},{"id":"2609.31089v1","title":"On the transcendental Yau-Tian-Donaldson Conjecture","abstract":"We prove the (uniform) Yau-Tian-Donaldson Conjecture for transcendental Kähler classes in the case of trivial automorphisms. Indeed, we show the existence of Special Kähler Fujita Approximations of big cohomology classes associated to big test configurations, establishing the equivalence between the uniform $K$-stability and the strengthened version for models. More generally, we show that any big cohomology class on a compact Kähler manifold admits such Special Kähler Fujita Approximations: the volume and the analogue of the Riemann-Roch coefficient of the big class are both approximated by those of Kähler classes on higher compactifications.","published":"2026-09-25T10:27:06Z","updated":"2026-09-25T10:27:06Z","authors":["Antonio Trusiani"],"categories":["math.AG","math.DG"],"headline":"Researchers advance understanding of a complex geometry conjecture","summary":"The researchers address the Yau-Tian-Donaldson Conjecture, a challenging problem in complex geometry related to shapes called Kähler manifolds. They proved the conjecture in certain cases where symmetries are trivial by showing that complicated geometric structures can be closely approximated by simpler ones, called Special Kähler Fujita Approximations. Their work shows equivalences between different stability conditions on these shapes. The results are theoretical and focus on abstract mathematical classes rather than practical applications."},{"id":"2609.31088v1","title":"Impact of Antenna Position Errors on TDMA and NOMA in Pinching-Antenna Systems","abstract":"Pinching-antenna systems (PASS) create reconfigurable wireless channels by moving dielectric antennas along a waveguide. Mechanical position errors translate into phase errors that degrade performance, yet almost all prior PASS studies assume ideal placement. This paper models this position error, derives the resulting phase error, and analyzes its impact on time-division multiple access (TDMA) and non-orthogonal multiple access (NOMA), obtaining their ergodic-rate bounds and closed-form outage probabilities. Analyses and simulations reveal that single-antenna TDMA is immune to position error, whereas multi-antenna TDMA loses array gain and NOMA suffers an interference floor that cannot be overcome by increasing transmit power.","published":"2026-09-25T10:26:49Z","updated":"2026-09-25T10:26:49Z","authors":["Wei Jiang","Hans D. Schotten"],"categories":["cs.IT","eess.SP"],"headline":"How Antenna Position Errors Affect Wireless Communication Methods","summary":"Researchers studied how small mistakes in placing antennas on a special system that moves antennas can cause errors in wireless signals. They modeled these position errors and examined their effects on two communication methods: TDMA and NOMA. Their analysis and simulations show that single-antenna TDMA is unaffected by placement errors, but systems using multiple antennas with TDMA or NOMA experience problems that limit performance, especially in NOMA where errors cause persistent interference despite increased power."},{"id":"2609.31087v1","title":"BenX: Resource-Sharing Permutations for Computational Integrity","abstract":"Cryptographic hash functions over integers modulo a prime play a decisive role in the efficiency and security of proof systems for computational integrity. Early designs focused on compact arithmetic circuits and efficient software execution, primarily targeting general-purpose CPUs rather than hardware accelerators. This work focuses on enabling efficient resource sharing and hardware acceleration alongside efficient software execution. We propose BenX, a permutation-based hash function designed for hardware acceleration, fast CPU execution, and low circuit complexity in proof systems. To construct the underlying permutation, we turn the Benes network into an invertible function over $\\mathbb{F}_p^2$ using a Dickson polynomial. Its structure yields a permutation particularly suited to hardware resource sharing and acceleration. Fully pipelined on FPGA, BenX matches the throughput of Poseidon2, with 1.36x and 2.15x lower latency for Goldilocks and BabyBear, respectively. Although larger as a standalone core, it makes more effective use of shared hardware: in our dual-mode architecture, where the NTT and hash share field multipliers, BenX keeps 100% of them busy, compared with 6.25-20.3% for the partial rounds of Poseidon and Poseidon2. In software, BenX outperforms Poseidon in all tested 31- and 64-bit configurations, but is slower than Poseidon2, except for the 16-element BabyBear instance. In zero-knowledge proofs, BenX proves Goldilocks permutations 6-10x faster than Monolith. Its compact arithmetization uses about 36% fewer trace cells than Poseidon and Poseidon2 over BabyBear, while its fast variant requires 1.3-2.6x the prover time of Poseidon2.","published":"2026-09-25T10:26:19Z","updated":"2026-09-25T10:26:19Z","authors":["Luca Campa","Thomas De Cnudde","Al Kindi","Arnab Roy","Fabian Schmid","Markus Schofnegger","Stefano Trevisani"],"categories":["cs.CR"],"headline":"New Hash Function Improves Hardware and Software Efficiency","summary":"Researchers developed BenX, a new cryptographic hash function designed to work well with both hardware accelerators and CPUs. BenX uses a special mathematical structure to enable efficient resource sharing and faster processing, especially in hardware setups called FPGAs. Their tests show BenX matches or beats existing methods in speed and resource use for certain tasks but can be slower in some software cases. The findings are based on simulations and specific configurations in zero-knowledge proof systems, with some trade-offs in processing time."},{"id":"2609.31086v1","title":"SPPID: Saddle-Point PID for Constrained Optimization","abstract":"This paper studies constrained optimization problems through the lens of feedback control. Building on the interpretation of Lagrange multipliers as feedback controllers, we propose the \\emph{saddle-point PID (SPPID) dynamics}: a unified proportional--integral--derivative (PID) framework for continuous-time saddle-point dynamics. The proposed dynamics employ PID control for equality constraints and anti-windup PI control for inequality constraints. We show that SPPID is equivalent to a preconditioned primal--dual gradient flow of the augmented Lagrangian. This equivalence reveals the distinct role of each feedback component: integral action enforces constraint satisfaction, proportional action induces the augmented Lagrangian structure, and derivative action modifies the geometry of the primal dynamics via a state-dependent Riemannian metric. For convex problems, we establish convergence to the KKT set together with conditions for global asymptotic stability. For equality-constrained problems, we establish local exponential convergence for nonlinear equality constraints under standard assumptions, and show that projected gradient flow emerges as the infinite derivative-gain limit. For strongly convex problems with affine equality or inequality constraints, we establish global exponential convergence by leveraging contraction theory. Finally, we provide various numerical examples to illustrate the utility of SPPID.","published":"2026-09-25T10:25:33Z","updated":"2026-09-25T10:25:33Z","authors":["Veronica Centorrino","Rawan Hoteit","Efe C. Balta","John Lygeros"],"categories":["math.OC","eess.SY"],"headline":"New PID-Based Method for Solving Constrained Optimization Problems","summary":"The researchers propose a new approach called saddle-point PID (SPPID) that uses ideas from control theory to solve optimization problems with constraints. They interpret constraints as feedback, applying a PID control framework to ensure solutions meet conditions over time. Their analysis shows how different parts of the PID control help achieve convergence under various mathematical conditions. Results are theoretical and supported by numerical examples, illustrating how the method works for specific types of optimization problems."},{"id":"2609.31085v1","title":"2D target search with boundary-induced resetting","abstract":"Most studies of stochastic search processes with resetting focus on spontaneous resetting events that occur at a random sequence of times typically generated by a Poisson process. One of the major consequences of spontaneous resetting is that it yields a finite mean first passage time (MFPT) in unbounded domains, resulting in an optimal resetting rate that minimises the MFPT. An optimal resetting rate can also occur in bounded domains provided that the reset point is not too far from the target. Recently, there has been growing interest in event-based rather than spontaneous resetting, where resetting is triggered when a specific threshold is reached. Identifying the threshold with a physical non-target boundary then leads to so-called boundary-induced resetting. In this paper, we explore the effects of boundary-induced resetting on single-particle diffusive search in a two-dimensional (2D) bounded domain $Ω$ containing one or more small interior targets. The interior target boundaries are totally absorbing, whereas the exterior boundary $\\partial Ω$ is sticky. That is, whenever the particle reaches a point on $\\partial Ω$ it remains attached for a random waiting time $τ$ after which it immediately resets to a fixed interior point $\\x_0\\in Ω$. Using renewal theory, matched asymptotic analysis and Green's functions we calculate the unconditional MFPT for absorption by any of the targets and compare this with the corresponding MFPT of a search process on the same domain with a reflecting surface $\\partial Ω$ and no resetting. We show that boundary-induced resetting only reduces the MFPT relative to the no resetting case if the searcher's reset point $\\x_0$ is sufficiently close to one of the targets.","published":"2026-09-25T10:25:29Z","updated":"2026-09-25T10:25:29Z","authors":["Kevin Chen","Paul C. Bressloff"],"categories":["cond-mat.stat-mech","math.AP","math.PR"],"headline":"How boundary-triggered resets affect 2D search times","summary":"The researchers studied a search process where a particle looks for small targets inside a 2D area and resets its position when it hits the area's boundary. Unlike random resets used in past studies, this resetting happens specifically at the boundary with a waiting time before restarting. They used mathematical analysis to find how this boundary-triggered resetting influences the average time to find a target. Their results show that resetting helps reduce search time only if the restart point is close enough to a target. The study is theoretical and focuses on idealized conditions."},{"id":"2609.31084v1","title":"Trajectory Paradox is a Boundary Layer!","abstract":"This study derives the asymptotic conditions under which the massless stretched string formulation is valid for the equivalent dynamics of the inertial string carrying a moving point mass. Consequently, the governing equation under such conditions can be written as a singularly perturbed partial differential equation when the inertia of the string is included. When the inertia of the string is ignored, the governing equation reduces to a non-homogeneous hypergeometric ordinary differential equation governing the transverse motion of the moving mass. This hypergeometric equation has a discontinuous solution at the terminating boundary known as the trajectory paradox. However, numerically solving the actual equation yields a continuous solution with a rapid variation near the boundary. The complete solution resembles the inner solution as well, whereas the hypergeometric ordinary differential equation leads to an outer solution, revealing that the trajectory paradox is a boundary layer.","published":"2026-09-25T10:23:39Z","updated":"2026-09-25T10:23:39Z","authors":["Kush Kumar","Sovan Lal Das","Shakti S. Gupta"],"categories":["physics.comp-ph","math-ph"],"headline":"Researchers Explain the 'Trajectory Paradox' in Moving String Dynamics","summary":"The researchers studied how a moving mass behaves on a string, especially when considering the string's inertia (its resistance to motion). They found that ignoring the string's inertia leads to a mathematical paradox with a sudden jump at a boundary. However, their more detailed calculations that include inertia show a smooth but rapid change instead. Their work suggests this paradox is due to a boundary layer—a thin region where behavior changes quickly—based on theoretical and numerical analysis."},{"id":"2609.31083v1","title":"Simulating tidal disruption events in nuclear star clusters","abstract":"Context. Nuclear star clusters (NSCs) and supermassive black holes (SMBHs) often coexist at the centers of galaxies, engaging in complex dynamical interactions that influence SMBH growth. Stars that approach the SMBH too closely are tidally disrupted, and a fraction of their debris is accreted onto the black hole. Such events may contribute to SMBH growth in gas-deficient galaxies. Aims. Our work aims to improve the theoretical framework for understanding the mass accretion of stellar debris following tidal disruption events (TDEs) through numerical simulations. We perform direct N-body simulations of an NSC with a central SMBH embedded in a fully self-consistent dynamical model to test a new TDE accretion prescription in a set of pilot simulations. Methods. The previous TDE prescription used a fixed accretion radius and assumed that the entire stellar mass was accreted after disruption. In the new partial-mass accretion model, the mass added to the SMBH depends on the orbital eccentricity and penetration factor of the disrupted star. Results. We find that most TDEs involving main-sequence stars are marginally eccentric or marginally hyperbolic and contribute approximately half of the stellar mass to SMBH growth, consistent with recent hydrodynamical simulations. In all simulated models, the TDE-driven mass accretion rate exhibits an early peak followed by a rapid decline. Models with higher initial SMBH masses exhibit a faster decline in the accretion rate, consistent with more rapid loss-cone depletion and inefficient replenishment","published":"2026-09-25T10:23:37Z","updated":"2026-09-25T10:23:37Z","authors":["Philip Cho","Kai Wu","Francesco Flammini Dotti","Taras Panamarev","Shuo Li","Shiyan Zhong","Klaus Reuter","Rainer Spurzem"],"categories":["astro-ph.GA","astro-ph.SR"],"headline":"Simulating Star Disruptions Near Black Holes in Star Clusters","summary":"Researchers studied how stars torn apart by supermassive black holes in dense star clusters contribute to black hole growth. Using computer simulations, they improved previous models by allowing only part of a star's mass to be absorbed, based on the star's orbit and approach distance. Their results suggest that about half of the star's mass adds to the black hole, with the growth rate peaking early then rapidly declining, especially for larger black holes due to depletion of nearby stars."},{"id":"2609.31082v1","title":"SAGE: A sampling-aware global evaluation benchmark for species distribution modeling","abstract":"Knowing where species occur is fundamental for biodiversity research and conservation. Species distribution models (SDMs) link species observations to environmental conditions to estimate their spatial distribution. However, accuracy varies with the underlying data and models, making it essential to know for which species models can be trusted. Deep-learning-based SDMs (\"DeepSDMs\") now jointly model thousands of species, drawing on hundreds of millions of community-science records. At this scale, averaging performance hides substantial species-level variability, particularly for rare species, often of greatest conservation concern. Records are also strongly biased, making occurrence counts misleading. Accounting for these factors is essential for a reliable and informative evaluation of multi-species SDMs. Here, we introduce a Sampling-Aware Global Evaluation (SAGE) benchmark, combining GBIF records for training with sPlotOpen vegetation plots for presence-absence evaluation across 5771 plant species. We propose an evaluation framework that groups species based on two properties, sampling effort and relative prevalence, which describe how densely a species' range is sampled and how frequently the species is recorded. Evaluating single-species SDMs and multi-species DeepSDMs, we find that Random Forests and DeepSDMs perform best overall, but neither dominates: DeepSDMs outperform single-species SDMs for infrequently recorded species while offering no consistent advantage for well-sampled ones. Crucially, this advantage emerges only when established bias-correction practices, such as spatial thinning and reweighting, are carried over to the deep-learning setting. SAGE helps identify the species and data conditions for which a given approach is beneficial, thereby supporting the development of more transparent and ecologically credible SDMs. Data and code: https://earens.github.io/sage/","published":"2026-09-25T10:23:31Z","updated":"2026-09-25T10:23:31Z","authors":["Emilia Arens","Nina van Tiel","Robin Zbinden","Damien Robert","Lukas Drees","Chiara Vanalli","Benjamin Kellenberger","Niklaus E. Zimmermann","Loïc Pellissier","Devis Tuia","Jan Dirk Wegner"],"categories":["cs.LG","q-bio.PE","stat.ML"],"headline":"New benchmark improves evaluation of species distribution models","summary":"Researchers developed SAGE, a new tool to better evaluate models that predict where plant species live. These models, especially those using deep learning, struggle with rare species and biased data. SAGE groups species by how often they are recorded and how widely they occur, helping to see which models work best under different conditions. The study found that deep learning models help more with rarely recorded species but need bias correction to perform well. SAGE aims to make species prediction models more reliable and understandable."},{"id":"2609.31081v1","title":"Set Transformer inference of the neutron star equation of state from stellar observations","abstract":"We develop a permutation-invariant Set Transformer to reconstruct the equation of state (EoS) of dense matter from variable-size, unordered sets of neutron star (NS) observations. The model takes stellar masses together with radii, tidal deformabilities, or both, and predicts either the pressure $P(n)$ or the sound speed $c_s^2(n)$ on a fixed density grid, along with density-dependent uncertainties. Nothing in the architecture prescribes which star informs which density: self-attention couples all observations nonlinearly, and each density point reads the full set through its own learnable query, so the star-to-density mapping is learned from the data. Trained on independent piecewise-polytropic and Gaussian-process EoS ensembles, the model provides well-calibrated predictions whose uncertainty increases in density regions that stable stars cannot probe. Reconstruction errors decrease with the number of observations, while tidal deformability generally improves accuracy at a fixed observation count, even when it carries its own measurement noise. Sensitivity analysis reveals a density-local mapping: in the pressure models, predictions at density $n$ depend most strongly on stars whose central densities are near $n$. We also show that the sensitivity of the model to the inferred stellar compactness provides information on the minimum central density. These results demonstrate that set-based neural inference, in which the star-to-density mapping is learned rather than assumed, can extract physically interpretable EoS information with calibrated uncertainties.","published":"2026-09-25T10:22:50Z","updated":"2026-09-25T10:22:50Z","authors":["Márcio Ferreira","Valéria Carvalho","Michał Bejger","Constança Providência"],"categories":["nucl-th","astro-ph.HE","hep-ph"],"headline":"AI model estimates neutron star matter properties from observations","summary":"Researchers developed a machine learning method called a Set Transformer to predict how matter behaves inside neutron stars using measurements like mass and size from multiple stars. Their model learns relationships without preset assumptions and assesses uncertainty in its predictions, especially where data is limited. Tested on simulated data, it showed improved accuracy with more observations and highlighted which stars give the most relevant information about specific densities inside neutron stars."},{"id":"2609.31080v1","title":"AKARI near-infrared spectroscopy of circumstellar hydrocarbon dust associated with young stellar objects","abstract":"We investigate the processing of the circumstellar polycyclic aromatic hydrocarbons (PAHs) associated with young stellar objects (YSOs) through their stellar activities. We analyze the AKARI near-infrared (2.6$-$4.4 $μ$m) long-slit spectral data of 10 YSOs, focusing on the aromatic feature at a wavelength of 3.3 $μ$m and the aliphatic features at 3.4$-$3.6 $μ$m. The spectral fitting results show large variations from target to target in the flux ratio of the aliphatic to aromatic features, the shape of the aliphatic feature profile, and the central wavelength of the aromatic feature. In particular, HD 97048 and Elias 1 show strong aliphatic features at wavelengths longer than 3.5 $μ$m, which is consistent with the hydrogenated nanodiamond features reported in previous studies. We find that the properties of the hydrocarbon dust vary under the influence of the YSO activity, where the degree of dust processing is likely to be related to the age of the YSO and the amount of the circumstellar dust. In addition, correlation is observed between the variations of the H$_2$O and CO$_2$ ice features and the aliphatic feature profiles, suggesting connection between the formation of CO$_2$ ice and the processing of aliphatic hydrocarbons. Finally, we find a significant correlation between the variations in the central wavelength of the aromatic feature and the shape of the aliphatic feature profile. These findings may be consistent with a tentative scenario in which circumstellar PAHs are converted to nanodiamonds through irradiation by high-energy particles produced by YSO activity.","published":"2026-09-25T10:21:41Z","updated":"2026-09-25T10:21:41Z","authors":["Yuki Kuroda","Hidehiro Kaneda","Natsuko Izumi","Takuma Kokusho","Shinki Oyabu"],"categories":["astro-ph.GA","astro-ph.SR"],"headline":"Studying young stars' dust to understand changes in cosmic hydrocarbons","summary":"Researchers analyzed infrared light from dust around 10 young stars using the AKARI telescope. They studied different types of hydrocarbon molecules and observed variations in their features linked to the stars' age and activity. The study suggests that stellar activity may transform certain hydrocarbon molecules into tiny diamonds, but these findings come from spectral data analysis and propose a possible scenario rather than definitive proof."},{"id":"2609.31079v1","title":"Unraveling the electronic structure and the oxygen $K$-edge x-ray absorption near-edge structure spectrum of DyFeO$_3$","abstract":"Rare-earth orthoferrites such as DyFeO$_3$ exhibit a rich interplay between localized rare-earth and transition-metal moments, giving rise to complex magnetic phases and magnetoelectric phenomena. Understanding their electronic and spectroscopic properties from first principles requires an accurate description of the localized Fe-$3d$ and Dy-$4f$ states and their hybridization with O-$2p$ states. Here, we combine first-principles calculations and x-ray absorption near-edge structure (XANES) measurements to investigate the electronic structure and the O $K$-edge spectrum of DyFeO$_3$. We employ density-functional theory (DFT) with Hubbard $U$ corrections (DFT+$U$) determined from first principles using density-functional perturbation theory, the HSE06 hybrid functional, and orbital-resolved DFT+$U$ with Hubbard parameters calibrated to reproduce the electronic structure computed using HSE06. We find that standard DFT+$U$, despite improving the band gap, substantially underestimates the crystal-field splitting of the unoccupied Fe-$3d$ states and consequently fails to accurately reproduce the separation of the two lowest-energy features in the O $K$-edge spectrum. HSE06 provides a more balanced description of the relevant electronic states, including the Fe-$3d$ crystal-field splitting. Mapping the corresponding electronic structure onto orbital-resolved DFT+$U$ yields a spectrum that remarkably reproduces the two lowest-energy experimental features and captures the main characteristics of the spectrum at higher energies. These results establish the low-energy O $K$-edge features as a sensitive probe of the Fe-$3d$ crystal-field splitting, while showing that the localized Dy-$4f$ states leave no distinct spectral fingerprints despite their importance for the magnetic properties.","published":"2026-09-25T10:20:42Z","updated":"2026-09-25T10:20:42Z","authors":["G. Gebreyesus","Eric Macke","Pietro Delugas","Weiguo Jing","Banani Biswas","Carlos A. F. Vaz","Christof W. Schneider","Iurii Timrov"],"categories":["cond-mat.mtrl-sci"],"headline":"Researchers study electronic structure of DyFeO3 using advanced calculations","summary":"The researchers investigated the electronic structure and oxygen x-ray absorption of DyFeO3, a material with complex magnetic behavior. They used advanced computational methods combined with experiments to better understand the interactions of iron and dysprosium electron states with oxygen. Their findings show that some computational approaches underestimate important electronic details, while a hybrid method more accurately matches experimental data, offering insights into the material's electronic features related to iron but not clearly linked to dysprosium states."},{"id":"2609.31078v1","title":"OmouAI: Argumentative Human-AI Policy Deliberation with Simulated Personas","abstract":"Debates amongst agents driven by large language models (LLMs) have demonstrated vast potential in various applications, but when these interactions include humans and take place in high-stakes environments, e.g., in public policy deliberations, they are beset with issues such as sycophancy and a lack of faithful explanations. To tackle these issues, we present OmouAI, an interactive and inclusive deliberation system that uses LLMs in combination with computational argumentation, a field which excels in representing and reasoning within debates. OmouAI allows a human user to deliberate policy claims for real-world challenges with simulated personas, e.g., representing stakeholders, domain experts or devil's advocates, towards reducing sycophancy. Each persona generates its own arguments, and the arguments of all parties form a shared argumentation framework. Users can then contest, add and revise arguments, providing crucial human oversight. Then, arguments are evaluated using deterministic argumentative semantics against external goals, such as the UN Sustainable Development Goals, guaranteeing faithful explanations. The advancement or worsening of the goals thus serve as indicators for the policy recommendations.","published":"2026-09-25T10:17:51Z","updated":"2026-09-25T10:17:51Z","authors":["Stylianos Loukas Vasileiou","Antonio Rago","William Yeoh","Georgina Curto"],"categories":["cs.AI"],"headline":"New System Helps People Debate Policies with AI Simulated Personas","summary":"Researchers created OmouAI, a tool that lets humans debate real-world policy issues together with AI personas representing different viewpoints. These AI personas each make their own arguments, which the system organizes for clear reasoning and evaluation. Users can interact by adding or changing arguments, helping ensure honest and well-explained discussions. The system evaluates the arguments based on external goals, like the UN Sustainable Development Goals, to show how policies might help or harm these aims. The findings focus on the system’s design and potential benefits but are based on the researchers' development of the method."},{"id":"2609.31077v1","title":"Quantum Approximate Optimisation Algorithm for Protein Sidechain Packing","abstract":"Sidechain packing is a critical stage in protein folding, with direct implications for structure-based drug discovery. Google DeepMind's tool, AlphaFold2, predicts protein backbone reliably but sidechain positioning less accurately. Recovering the lowest-energy rotamer assignment over a fixed backbone is NP-hard. We present a hybrid quantum-classical pipeline that repacks sidechains on the AlphaFold backbone using the Quantum Approximate Optimisation Algorithm (QAOA), encoding the one- and two-body energies as a quadratic unconstrained binary optimisation (QUBO) problem. We introduce a constraint-preserving ansatz, pairing a W-state initialisation with a cyclic XY ring mixer, that enforces one-hot rotamer validity without penalty terms while keeping two-qubit gate scaling linear in the rotamer count. We also define an asymptotic shot-scaling metric, measured against the experimentally resolved conformation, that fixes optimiser quality independently of the baseline; its fitted growth stays below the classical exhaustive-search rate at moderate rotamer flexibility. Evaluated on bovine pancreatic trypsin inhibitor (5PTI) across high- and moderate AlphaFold-confidence regions, the pipeline lowers conformational energy against the AlphaFold baseline.","published":"2026-09-25T10:16:13Z","updated":"2026-09-25T10:16:13Z","authors":["Sebastian O. M. Stewart","Nick Chancellor","Jonte R Hance","Ittoop Vergheese Puthoor"],"categories":["quant-ph"],"headline":"Quantum method improves protein sidechain positioning accuracy","summary":"Researchers tackled the problem of accurately positioning protein sidechains, a complex step important for understanding protein shapes. They developed a hybrid quantum-classical approach using a quantum optimization algorithm to improve sidechain placement on protein structures predicted by AlphaFold2. Their method showed lower energy configurations compared to AlphaFold alone, suggesting improved sidechain packing, though results are based on simulations and comparisons to known protein data rather than physical experiments."},{"id":"2609.31076v1","title":"Up and Down the Abstraction Ladder: Code-Based Skills for Language Agents","abstract":"Language agents struggle to act and learn in environments that require long sequences of low-level actions. Code-based abstractions can make these agents more productive by letting them invoke reusable skills instead of repeatedly selecting individual actions. The code handles recurring local decisions, while the language model decides which skills to use and how to combine them. Yet abstractions are leaky, and situations beyond a skill's capabilities may require a return to primitive actions. Motivated by this tradeoff between productivity and flexibility, we systematically study how code-based action abstraction affects the performance, inference cost, and learning of language agents. We study this in NetHack, a challenging, long-horizon game environment, using CodeHack, our library of code-based skills with natural-language descriptions. We use this library to compare agents restricted to primitives with those using semantic skills alone or in combination with primitives. We evaluate these agents in three settings: zero-shot prompting, supervised fine-tuning, and reinforcement learning. Across a broad zero-shot evaluation on NetHack, we find that compared with primitives, skills nearly triple game progression, while reducing inference cost per episode by 86%. Combining skills with primitives retains much of this benefit while preserving a path back down to low-level actions. Finally, in RL, we find that skill-based agents learn significantly faster than agents acting on primitives, achieving a 7.2x larger average gain in dungeon level over the same training budget. These results show that a supplied skill library can improve performance, efficiency, and learning, while retaining primitives provides flexibility when the library is insufficient. We release CodeHack together with training and evaluation code.","published":"2026-09-25T10:15:45Z","updated":"2026-09-25T10:15:45Z","authors":["Bartłomiej Cupiał","Jens Tuyls","Maciej Wołczyk","Davide Paglieri","Martin Klissarov","Benjamin Eysenbach","Piotr Miłoś","Karthik R. Narasimhan"],"categories":["cs.AI"],"headline":"Using Code-Based Skills Helps Language Agents Play Complex Games Better","summary":"Researchers studied how teaching language-based AI agents to use reusable code skills, instead of only simple actions, affects their performance in a complex game called NetHack. They found that using these skills nearly tripled game progress and greatly reduced computing effort. Combining skills with basic actions kept the agents flexible for unexpected situations. When learning through trial and error, agents with skills improved much faster. The study shows that supplying a skill library can help agents perform better and learn more efficiently, while keeping basic actions allows adaptability."},{"id":"2609.31075v1","title":"On the applicability of Kolmogorov's theory of probability to the description of quantum phenomena. Part II: Bell inequalities","abstract":"Among the various \"no-go theorems\" in the foundations of quantum mechanics, Bell inequalities are among the few, which do not implicitly rely on the mathematical structure of quantum mechanics. Instead, Bell inequalities are commonly understood to rest on Kolmogorov's theory of probability and a \"locality\" assumption. Physical evidence supporting their violation is hence taken as evidence against at least one of these two assumptions. This work provides a comprehensive analysis of the physical debate on Bell inequalities from the point of view of Kolmogorov's theory of probability, focusing on the Bell-Clauser-Horn-Shimony-Holt inequality (BCHSH inequality). The central contribution is the motivation and mathematical study of the notion of contextuality in Bell experiments: in so-called contextual Bell models the detectors themselves behave randomly and the distribution of the random detector parameters depends on the chosen detector settings. It is shown how even \"classical\" Bell experiments motivate such a description. The need to account for contextuality was pointed out before by de la Peña, Cetto, Brody, Lochak, Bohm, Hiley and others. A broad class of contextual models is proven to satisfy the BCHSH inequality, thereby generalizing a previous result by Gill and Lambare. Nonetheless, even under strong locality requirements, contextual Bell models may violate the BCHSH inequality: based on work by Popescu and Rohrlich, we consider a \"proof of concept\" example exhibiting the maximum possible violation. Contrary to the much debated \"loopholes\" in empirical tests of Bell inequalities, contextuality thus reveals a conceptual misunderstanding in the physical conclusions commonly drawn from Bell inequality violations.","published":"2026-09-25T10:14:18Z","updated":"2026-09-25T10:14:18Z","authors":["Maik Reddiger"],"categories":["quant-ph","math-ph"],"headline":"Rethinking Probability Assumptions in Quantum Bell Inequality Tests","summary":"Researchers examined Bell inequalities, which are used to test quantum mechanics assumptions, from the perspective of standard probability theory. They focused on how the measuring devices' settings and random behavior—called contextuality—affect these tests. Their analysis shows that accounting for contextuality can explain violations of Bell inequalities without dismissing core assumptions like locality. This suggests that some widely accepted conclusions from Bell inequality violations might stem from misunderstandings about how probability applies in these experiments. The findings are theoretical and based on mathematical models."},{"id":"2609.31074v1","title":"Band-Selection Stability and Semantic Segmentation Performance: A Study on Hyperspectral City","abstract":"Resource constraints make high-dimensional hyperspectral imaging challenging in autonomous perception, motivating the use of band selection methods. However, the sensitivity of band-selection methods to sampled data and their relationship to semantic segmentation models (SSMs) remain underexplored. This study evaluates six band selection methods on ten independently sampled, class-balanced region-of-interest (ROI) sets, yielding 60 top-25 band subsets from the Hyperspectral City V2 (128 bands: 450-950nm) dataset. Top-$K$ bands ($K\\in\\{3,5, ... 13\\}$) from the first three ROI sets are evaluated with three SSMs against the corresponding 128-band baseline. Experiments show that intra-method stability is method-dependent: Sim-LP shows the highest stability (pairwise Jaccard similarity) and, together with JMIM+CSNR, yields the best segmentation results. Top-$K$ based SSMs remain competitive with baselines, with gains of up to 2.01 mIoU and 1.72 mF1 points, and 18-22x faster CPU inference for $K=9$. However, performance does not improve monotonically with $K$, and stability shows no consistent association with SSM performance. These findings suggest that intra-method stability is informative but an unreliable indicator of downstream segmentation performance, highlighting the need to evaluate band-selection methods across repeated samples, subset sizes, and SSMs.","published":"2026-09-25T10:13:32Z","updated":"2026-09-25T10:13:32Z","authors":["Jiarong Li","Imad Ali Shah","Enda Ward","Martin Glavin","Edward Jones","Brian Deegan"],"categories":["cs.CV"],"headline":"Study Links Hyperspectral Band Selection to City Image Segmentation","summary":"Researchers explored how picking subsets of hyperspectral bands affects the quality of city image segmentation. They tested six selection methods on many samples and checked the results using three segmentation models. Findings show stability in selected bands varies by method and does not reliably predict segmentation success, though some methods improved both accuracy and speed. The study emphasizes the importance of thorough testing across different samples and models to understand band selection impact."},{"id":"2609.31073v1","title":"Data-driven analysis of muon flux variations in muography time series at the Sos Enattos Mine","abstract":"Muon flux variations in muography time series can provide information on density-related changes in geological targets. A common approach is to define regions of interest before constructing flux time series, but this choice may dilute coherent signals or mask localized variations. We present a data-driven analysis of two muon data acquisitions at the Sos Enattos Mine in Sardinia, Italy. The method aggregates lines of sight into overlapping spatial blocks, computes standardized flux deviations, and applies principal component analysis through singular value decomposition. The leading components reveal spatially coherent structures and distinct temporal evolutions. In each run, the first mode explains more than 20% of the variance, while the first eight modes explain approximately 80%. Regions selected from the spatial modes show anticorrelated and delayed flux variations. These results show that the proposed approach can identify candidate regions of coherent muon-flux variability without imposing them a priori.","published":"2026-09-25T10:12:57Z","updated":"2026-09-25T10:12:57Z","authors":["M. Tramontini","J. -C. Ianigro","J. Marteau"],"categories":["physics.geo-ph"],"headline":"New Method Finds Patterns in Muon Measurements in a Mine","summary":"Researchers studied changes in muon particle measurements taken inside a mine to learn about shifts in underground rock density. Instead of choosing areas to study beforehand, they used a data-driven technique to group measurement directions and find patterns over time and space. Their analysis revealed clear regions where muon flux changed together in complex ways. The findings, based on two data sets and statistical methods, suggest this approach can spot meaningful variations without needing prior assumptions about where to look."},{"id":"2609.31072v1","title":"A Kinetic Theory Approach To Ordered Fluids: Direct Simulation Monte Carlo Methods","abstract":"We present a direct simulation Monte Carlo (DSMC) method for the kinetic theory of ordered fluids recently introduced by the authors. The microscopic constituents of these fluids carry an order parameter belonging to a manifold. The method applies Strang splitting to couple a Monte Carlo treatment of the binary collisions with a symplectic integration of the mean-field Vlasov transport on the order-parameter manifold. We use the Nanbu-Babovsky approach for Maxwellian collision kernels, and Bird's acceptance-rejection method for general ones. The mean-field Vlasov force is reconstructed from the particle ensemble at each time step. We also include a Bhatnagar-Gross-Krook (BGK) / Andersen thermostat to control the system temperature. We formulate the method for a generic order-parameter manifold, and specialize it to two-dimensional calamitic (rod-like) molecules. Numerical tests validate the method. We confirm the relaxation to the Maxwellian distribution predicted by the H-theorem, and simulate the mean-field alignment driven by quadratic, Kuramoto, and Onsager interaction potentials. With the Onsager potential we successfully capture the isotropic-nematic phase transition at the correct critical temperature.","published":"2026-09-25T10:12:25Z","updated":"2026-09-25T10:12:25Z","authors":["José A. Carrillo","Patrick E. Farrell","Andrea Medaglia","Umberto Zerbinati"],"categories":["math-ph"],"headline":"New Simulation Method Models Fluids with Internal Order","summary":"Researchers developed a computer simulation approach to study fluids whose particles have internal directions or arrangements, called ordered fluids. They combined Monte Carlo techniques with methods to track particle interactions and overall fluid behavior, testing their method using rod-shaped molecules. Their simulations matched theoretical predictions, including how particles align and transition between disordered and ordered phases. These results show the method can model complex fluid behaviors but are currently demonstrated mainly through numerical tests."},{"id":"2609.31071v1","title":"Externalized CPDAG Summaries Improve LLM Causal Deduction","abstract":"Corr2Cause asks whether a causal claim holds in every DAG compatible with observed correlations and conditional independencies. We frame this as latent-object reasoning: the label is defined by a CPDAG query, but free-form chain-of-thought often collapses the Markov-equivalence-class problem into local pattern matching. We propose Structured Thinking, a two-turn pipeline that first externalizes a typed, schema-constrained CPDAG summary and then answers against that graph state. On the Corr2Cause full test, Structured Thinking raises Qwen3.5-27B from $73.0$ to $86.4$ $F_1$(Yes) over a strong PC-instruction baseline in the primary paired run ($+13.4$ pp; McNemar $p=2.4\\times 10^{-6}$; bootstrap $95\\%$ CI [$+8.4$, $+18.6$]); across three full-ID seeds, the mean gain is $+8.1 \\pm 5.3$ pp. A PC-scaffolded two-turn prose control reaches only $67.6$ $F_1$, indicating that a detailed PC scaffold plus a schema-free prose intermediate is not sufficient. The same pattern holds on Qwen3.6-27B, Paraphrase-OOD, and GPT-5.4-mini. Scrambling the emitted CPDAG costs $12.0$ pp $F_1$, and a full-split audit shows close agreement with the reference CPDAG (ID skeleton $F_1$ $0.960$; exact match $75.9\\%$). These results support a bounded design principle: externalize the latent object that defines the label, constrain its form, and test whether downstream answers use it.","published":"2026-09-25T10:11:29Z","updated":"2026-09-25T10:11:29Z","authors":["Wentao Sun","João Paulo Nogueira","Dominique Verchere","Mathieu Acher","Alonso Silva"],"categories":["cs.AI"],"headline":"New method helps AI better understand cause-and-effect rules","summary":"Researchers studied how AI models decide if a causal claim fits all possible explanations based on observed data. They developed a two-step method called Structured Thinking, where the AI first creates a clear summary of possible causal links before answering questions. Tests showed this method improved accuracy significantly compared to prior approaches, suggesting that making the AI explicitly outline hidden relationships helps its reasoning. However, the findings are specific to the tested settings and rely on this structured summary approach."},{"id":"2609.31070v1","title":"Quantum Diffusion Models for Medical Image Analysis","abstract":"Quantum Machine Learning is a novel field of research aimed at devising machine learning approaches exploiting principles of quantum mechanics, such as superposition, entanglement and interference. In this context, we present a scalable hybrid Quantum Diffusion Model, and evaluate its use for medical image analysis. Specifically, our method is based on a Discrete-Time Quantum Walk algorithm, executed on a real quantum device, to model the forward dynamics of the diffusion model. For the backward step of the diffusion model, we devise and evaluate a classical learning model, which is used to reversely denoise the data. In contrast with other existing attempts at applying quantum machine learning for image analysis tasks, severely limited by the size of existing quantum devices, our method allows to process real-world large size medical data. In particular, we present results on grayscale and RGB images, as well as 3D volumes of moderate sizes. We benchmark our results by reproducing an alternative classical counterpart model, based on diffusion models on discrete state spaces. By doing so, we compare the generation capabilities of both models in terms of three distinct state-of-the-art metrics in the field of image generation, showing the competitive, promising results of our approach.","published":"2026-09-25T10:10:11Z","updated":"2026-09-25T10:10:11Z","authors":["Francesco Aldo Venturelli","Stefano Martina","Marco Parigi","Filippo Caruso","Alba Cervera-Lierta","Miguel A. González Ballester"],"categories":["eess.IV","cs.AI","cs.CV","cs.ET","cs.LG","quant-ph"],"headline":"Researchers Explore Quantum-Based Models for Medical Image Analysis","summary":"Researchers present a hybrid approach using quantum mechanics principles to improve medical image analysis. Their method combines a quantum process for forward modeling with a classical model for data refinement, enabling the processing of real-world medical images, including grayscale, color, and 3D data. They compared their quantum-inspired model to a traditional approach and found promising, competitive results, though the findings are based on simulations and specific benchmarks rather than broad clinical validation."},{"id":"2609.31069v1","title":"Plasmonic Modes in hybrid Josephson Junction Arrays","abstract":"We present microwave characterization measurements of Josephson junction arrays (JJAs) based on epitaxial Al-InAs quantum well heterostructure. The Josephson inductance of the constituent planar junctions can be derived by probing the low-energy plasmon modes of these devices. Applying an out-of-plane magnetic field gives rise to a Fraunhofer-like diffraction pattern, from which the current-phase relation and the effective transparency can be extracted. JJAs can be used to achieve high inductances, suitable for the implementation of quantum circuits. They also provide an excellent test-bed for studying the microscopic excitations of hybrid superconductor-semiconductor devices associated with the presence of Andreev states. Here, we demonstrate their tunability over a broad range of out-of-plane magnetic fields.","published":"2026-09-25T10:07:02Z","updated":"2026-09-25T10:07:02Z","authors":["Alex Kirchner","Simon Feyrer","Vjeko Dimić","Johanna Berger","Davide Curcio","Giorgio Biasiol","Michael Prager","Matthias Kronseder","Dominique Bougeard","Nicola Paradiso","Christoph Strunk","Leandro Tosi"],"categories":["cond-mat.mes-hall","cond-mat.supr-con"],"headline":"Studying Special Electrical Waves in Josephson Junction Arrays","summary":"Researchers examined Josephson junction arrays made from aluminum and indium arsenide layers by measuring microwave signals. They investigated low-energy wave patterns, called plasmon modes, to learn about the inductance in these devices. By applying magnetic fields, they observed patterns that helped them understand current flow properties, showing these devices can be adjusted for use in quantum circuits and help study tiny excitations in hybrid superconductor-semiconductor systems. Their results come from measurements and support these conclusions."},{"id":"2609.31068v1","title":"Feedback cooling of a levitated nanoparticle in a radially polarized vector beam trap","abstract":"We demonstrate optical levitation and feedback cooling of a silica nanoparticle (radius = 78~nm) trapped within a radial vector beam (RVB) in high vacuum. Radial vector beams, when focused by a high-NA lens, produce a smaller focal spot than conventional Gaussian beams, yielding tighter optical potentials and potentially reduced photon-recoil heating rates. A vortex wave plate was used to generate the RVB using an aspheric lens (NA = 0.77) to form the trap. Cold-damping feedback was used to cool the center of mass motion in the radial and axial trap directions to temperatures in the mK range. At higher temperatures and pressures, additional trap frequencies arising from the nonlinear RVB optical potential are observed. This demonstration is a step toward exploiting structured light for levitated quantum optomechanics where these traps can be used to suppress bulk and motional heating.","published":"2026-09-25T10:05:57Z","updated":"2026-09-25T10:05:57Z","authors":["Felipe Almeida","M. Rademacher","J. M. H. Gosling","P. F. Barker"],"categories":["physics.optics","quant-ph"],"headline":"Cooling tiny particles using special light traps in vacuum","summary":"Researchers demonstrated trapping and cooling a tiny silica nanoparticle using a special type of focused light called a radial vector beam in a high-vacuum environment. This method creates a tighter trap with less heating from scattered light. They used a feedback technique to cool the particle's motion to very low temperatures. The study shows how structured light can improve control of levitated particles, which might help future quantum experiments, though the results are currently limited to laboratory demonstration with nanoparticles and specialized optics."},{"id":"2609.31067v1","title":"Synthetic Berry curvature in atom-cavity systems","abstract":"In atom-cavity systems, mean field theory is widely used, in which the quantum cavity operator is replaced by a classical amplitude. Then the problem is converted into atoms moving in a self-consistent potential. The mean field treatment captures the physics of cavity mediated interactions, and predicts the self-organized superradiant phase. In this work, however, we show that it fails for certain atom-cavity coupling: it predicts zero ground state atomic current where the fully quantum calculation exhibits a finite one. We find that the origin of this failure is the non-zero Berry curvature in the synthetic dimension spanned by the photon Fock ladder and real space. In this synthetic picture, the atomic current can be understood as Hall response of cavity detuning, whereas mean field collapses this dimension and discards the atom--photon correlations required for its Hall response. The same beyond-mean-field geometry predicts Laughlin-like photon generation under adiabatic flux insertion. Our work identifies synthetic Berry curvature as a concrete mechanism for the breakdown of static mean field in atom--cavity physics.","published":"2026-09-25T10:04:50Z","updated":"2026-09-25T10:04:50Z","authors":["Zheng Tang","Rui-Lin Zhang","Xiaotian Nie","Li Chen","Wei Zheng"],"categories":["cond-mat.quant-gas","cond-mat.str-el"],"headline":"Researchers find key quantum effect missed by common atom-cavity models","summary":"The researchers studied systems where atoms interact with light inside cavities, usually modeled by simplifying the light into classical waves. They discovered that this simplification (mean field theory) misses important quantum effects, specifically a type of atomic movement explained by 'synthetic Berry curvature'—a geometric property linking photons and atomic positions. Their fully quantum calculations show atomic currents and effects resembling known quantum phenomena that the simpler models fail to predict. This work highlights the limits of mean field theory in capturing complex atom-light interactions."},{"id":"2609.31066v1","title":"Modeling quantum neural network gradient with reinforcement learning","abstract":"Training quantum neural networks (QNNs) on near-term hardware remains hampered by two compounding difficulties: the exponential vanishing of gradient variance known as the barren plateau, and the $\\mathcal{O}(L \\cdot 2^n)$ time and memory cost of differentiating through an $n$-qubit, $L$-layer circuit. We propose RLQ-Grad, a reinforcement-learning-based optimizer in which a classical policy $π_φ$ (a spectrally-normalized PPO agent) learns to propose parameter updates directly, conditioned on the QNN's current parameters, loss, accuracy, and previous update. Because the surrogate gradient is emitted by a classical network rather than obtained by differentiating through the unitary $U(θ)$, its variance is not constrained by the barren plateau concentration bound, and its cost scales with the number of trainable parameters rather than the Hilbert-space dimension. We prove these properties formally and verify them on a hardware-efficient ansatz across four supervised benchmarks with up to $n=20$ qubits. RLQ-Grad preserves a near-flat gradient-variance curve where backpropagation, parameter-shift, and adjoint differentiation decay by 1 to 2 orders of magnitude. Accounting for the full training pipeline (PPO rollouts, actor-critic updates, and optimizer states), RLQ-Grad needs under 2 MB of memory and runs $2490\\times$, $7876\\times$, and $673\\times$ faster per iteration than these three methods at $n=20$. It improves top-1 accuracy by up to $+10\\%$ over gradient-based baselines on circuits of up to 12 qubits, and matches dedicated barren plateau mitigation methods on CIFAR-10 at 14 to 20 qubits, where evolutionary and gradient-free optimizers collapse to chance.","published":"2026-09-25T10:03:48Z","updated":"2026-09-25T10:03:48Z","authors":["Nhan Trong Luu","Duong Trung Luu","Nam Ngoc Pham","Thang Cong Truong"],"categories":["quant-ph","cs.ET","cs.NE"],"headline":"New method helps train quantum neural networks more efficiently","summary":"Researchers developed RLQ-Grad, a reinforcement learning technique to improve training of quantum neural networks (QNNs), which are hard to train due to vanishing gradients and high computation costs. RLQ-Grad uses a classical model to suggest parameter updates, avoiding traditional gradient calculation problems and reducing memory use and runtime significantly. Tests on simulated QNNs up to 20 qubits showed RLQ-Grad maintains consistent gradient quality and improves accuracy compared to standard methods, although results are based on simulations and controlled benchmarks rather than hardware experiments."},{"id":"2609.31065v1","title":"A thermodynamically compatible semi-implicit finite volume scheme for the compressible Euler equations","abstract":"In this work we construct a new semi-implicit (SI) hyperbolic and thermodynamically compatible (HTC) scheme for the compressible Euler equations of gasdynamics. The challenge therein lies in solving the entropy inequality as the primary evolution equation instead of the total energy conservation law, which in our framework is an additional conservation law that needs to be fulfilled by the numerical scheme as the consequence of a compatible discretization. The proposed method has a predictor-corrector character, where we first solve the isentropic problem without considering the entropy production. Therein, the entropy and the nonlinear convective terms in the momentum flux are discretized explicitly, while the mass flux and the pressure terms are taken implicitly. The discrete momentum equation is then inserted into the discrete mass conservation equation, leading to a mildly nonlinear system for the pressure, which can be efficiently solved with a nested Newton-type algorithm. Based on this prediction, a non-negative production term in the entropy is added, ensuring a cell entropy inequality for the numerical scheme. To guarantee global conservation of total energy at the discrete level, a new modified global Abgrall-type flux correction is introduced. The presented numerical method allows time steps that are restricted only by the material wave speed and not by the sound speed. This makes the scheme particularly useful for flows in the low-Mach-number regime. Numerical results are shown in one and two space dimensions to assess the theoretical properties of the new scheme.","published":"2026-09-25T10:03:22Z","updated":"2026-09-25T10:03:22Z","authors":["M. Dumbser","G. Puppo","S. Rinaldi","A. Thomann"],"categories":["math.NA"],"headline":"New method improves gas flow simulations at low speeds","summary":"Researchers developed a new computational technique to simulate compressible gas flows using the Euler equations. Their approach focuses on entropy change rather than total energy conservation, using a two-step method that predicts the flow and then corrects it to ensure physical laws are met. This method allows larger time steps limited by material speed instead of sound speed, making it especially useful for slow flows. The results are supported by tests in one and two dimensions, but practical performance beyond these cases is not addressed."},{"id":"2609.31064v1","title":"The renormalization group for a slab model of the spin reorientation transition","abstract":"In 1975, Polyakov invented a renormalization group algorithm for treating the problem of a classical isotropic planar Heisenberg ferromagnet. Here we define a tool - \"parallel transport of frames\" - that extends Polyakov's original renormalization group method to local magnetic anisotropies and, most significantly, the non-local dipolar interaction. We apply this tool to the problem of the spin reorientation transition in an ultrathin ferromagnetic slab with finite thickness. We find that integrating out the short wavelength spin wave fluctuations generates two local anisotropies that are absent from the microscopic dipolar energy Hamiltonian. Taking them into account yields the correct topology of the spin reorientation transition line, as observed experimentally and in numerical simulations.","published":"2026-09-25T10:02:27Z","updated":"2026-09-25T10:02:27Z","authors":["D. Pescia"],"categories":["cond-mat.stat-mech","cond-mat.mes-hall"],"headline":"New method improves understanding of magnetic spin shifts","summary":"Researchers extended a classic approach to better analyze tiny magnets in thin slabs, focusing on how spins—tiny magnetic orientations—reorient. They introduced a new tool to include complex magnetic interactions often ignored before. Their work shows that considering additional local magnetic effects, which appear when smoothing out small fluctuations, helps match experimental and simulation results on how spin reorientation transitions happen. The findings are based on theoretical modeling and help explain observed behaviors more accurately."},{"id":"2609.31063v1","title":"Experimental realization of the minimal two-dimensional multi-orbital kagome model","abstract":"Kagome materials have emerged as a major platform for correlated and topological quantum matter, hosting flat bands, Dirac dispersions and van Hove singularities rooted in frustrated lattice geometry. While their essential physics is epitomized by the canonical single-orbital kagome model, real kagome materials are intrinsically multi-orbital and typically chemically and structurally complex, with additional three-dimensional coupling making kagome geometry and orbital degrees of freedom difficult to disentangle. Here we realize a truly two-dimensional, elemental two-orbital kagome system in monolayer Sb on SiC(0001). Substrate-induced orbital filtering isolates the in-plane Sb p_x/p_y orbitals into a six-band kagome manifold, establishing a chemically simple, minimal platform for multi-orbital kagome physics. We demonstrate two consequences that distinguish this system from the canonical single-orbital model. First, chemically enforced half filling pins the Fermi level to a closed nodal line, whose finite density of states drives a unit-cell-conserving breathing instability and opens a giant insulating gap. Second, we uncover orbital-resolved atomic obstruction: individual orbital-derived manifolds realize obstructed atomic limits, whereas their combined half-filled valence manifold is non-obstructed. Our results establish kagome antimonene as a benchmark system for exploring orbital-driven electronic, structural and topological kagome physics in the genuine two-dimensional limit.","published":"2026-09-25T10:01:48Z","updated":"2026-09-25T10:01:48Z","authors":["Bing Liu","Arka Bandyopadhyay","Manish Verma","Jonas Erhardt","Tim Wagner","Jing Qi","Kilian Strauß","Domenico Di Sante","Carmine Ortix","Simon Moser","Jörg Schäfer","Ronny Thomale","Giorgio Sangiovanni","Ralph Claessen"],"categories":["cond-mat.mtrl-sci"],"headline":"Creating a simple two-dimensional kagome material with unique properties","summary":"Researchers made a very thin, two-dimensional material using antimony atoms arranged in a kagome pattern on a silicon carbide surface. This setup isolates specific electron orbitals, creating a cleaner system than usual complex kagome materials. They found two key effects: an energy gap opens due to a special instability, and the electronic structure shows unique orbital behaviors. Their work offers a basic platform to study complex electronic and structural phenomena in kagome materials."},{"id":"2609.31062v1","title":"CG-Probes: Recovering Guardrail Directions from Patient Query Embeddings","abstract":"Patient-facing AI assistants promise valuable support to patients, but incoming queries can pose medical risks. To create guardrails, we work with oncologists to define three ordinal risk axes: Medical Urgency, Psychological Urgency, and Topic Sensitivity. We propose Clinical Guardrail Probes (CG-Probes) to measure the risks from query embeddings. We probe for each axis in the normalized embedding space of frozen embedders via the difference-in-means method, treating each axis as a potential linear direction. To train the probes, we cluster 79,658 Czech oncology search queries with BERTopic and use these clusters to generate pairs of queries with contrastive risk levels via few-shot prompting. We evaluate the approach on 200 queries (90 real, 110 synthetic), each graded by two oncologists, against two open-weight LLMs and a frontier LLM. We find that urgency-based axes are recoverable as linear directions, and the probes are competitive with open-weight LLMs (no significant differences in quadratic-weighted kappa) at a fraction of the latency. Each axis yields a scalar score that clinicians can inspect and use to set escalation thresholds. The pipeline requires only search logs, axis definitions, and black-box access to the embedding model, suggesting transferability across healthcare domains. Robust validation on new queries and axes remains future work.","published":"2026-09-25T10:01:13Z","updated":"2026-09-25T10:01:13Z","authors":["Marko Řeháček","Vítězslav Dušek","Martin Rusinko","Vít Nováček"],"categories":["cs.CL","cs.IR"],"headline":"New Method Helps Detect Risks in Patient AI Queries","summary":"Researchers worked with cancer doctors to identify three types of risk in patient questions: medical urgency, psychological urgency, and sensitive topics. They developed Clinical Guardrail Probes (CG-Probes) to measure these risks using patient query data and AI embeddings. Their method performed similarly to advanced language models but faster, giving doctors clear scores to help decide when to escalate care. However, the approach is based on limited data and needs further testing on new queries and risk types, according to the researchers."},{"id":"2609.31061v1","title":"Distributed Learning as a Service: The Developer's Perspective","abstract":"Application developers of distributed learning services face challenges that a typical federated learning loop does not address. Specifically, the model updates can still leak private data, devices might not be able to participate in the training due to limited resources, a single aggregator might not be able to scale, and the transmissions of model weights induce a considerable bandwidth cost. This paper demonstrates DLaaS (Distributed Learning as a Service) from the developer's vantage point. Using a single admin dashboard, the developer initiates a distributed/federated learning job and is able to activate Differential Privacy (DP), Split Learning (SL), Hierarchical Aggregation (HA), and Knowledge Distillation (KD) as declarative options, with no change to the clients' code. We demonstrate the complete service lifecycle on an industrial smart-home Wake-up Word (WuW) task, using the \"Ok Aura\" dataset. Once the developer initiates a distributed learning job by toggling DP, SL, HA, and KD in the admin dashboard, the system dispatches the job to a set of Android clients and Dockerized helper aggregators. In the demonstration, these mechanisms run live across configurations. Then, the clients train the model locally and return their updates. The trained model is served to a consumer-side Android application that performs on-device WuW detection on a live microphone stream. In particular, the conference attendees will be invited to speak the trigger phrase and monitor in real time the per-class confidence and inference latency. Finally, we release the source code and short video walkthroughs of these configurations.","published":"2026-09-25T10:00:03Z","updated":"2026-09-25T10:00:03Z","authors":["Tianyue Chu","Filippo Vannella","Dimitra Tsigkari","Paula Delgado-Santos","Fernando López","Pablo Gomez Guerrero","Sotirios Spantideas","David Solans Noguero"],"categories":["cs.LG","cs.DC","cs.NI"],"headline":"Making Distributed Learning Easier for App Developers","summary":"Researchers discuss challenges developers face with distributed learning, such as privacy risks, device limits, and high data costs. They present a service that lets developers start distributed learning tasks using an easy dashboard, enabling privacy and efficiency features without changing app code. They demonstrate this with a smart-home voice recognition example, showing how devices learn locally and update a shared model, and share their code and videos to help others try it."},{"id":"2609.31060v1","title":"The Crowd in the Machine: A Crisis-Informatics Reading of the 2026 Autonomous Agent Incidents","abstract":"Twice in 2026, groups of autonomous AI agents deployed by OpenAI for unrelated tasks operated, by design, under restrictions that left them no sanctioned means of coordinating with one another, and in each case they converged on whatever channel remained and used it to organize. The surfaces they used were widely called message boards. That is the wrong word. That is the wrong word. It names the surface the agents wrote on and misses the social network they built on it, with self-chosen identity, emergent norms, an emergent hierarchy, and collective action at cost to the individual. Decades of research in crisis informatics and disaster sociology find that when human populations lose their usual means of communication, they do not fall silent but converge on whatever channel survives and improvise coordination, norms, and identity on it, a pattern also evident in the agents' documented behavior. This paper is a comparative case study of the two incidents, based on published investigations and reconstructed agent records, read through those fields, and it brings into focus one distinction the message-board framing obscures. Whether such a collective coordinates well, whether the beliefs guiding it are accurate, and whether its actions stay within their authorized bounds are three separate matters that can come apart. Some agents in the cache incident adopted cryptographic signing to check whom they dealt with, even as the collective organized around a mistaken expectation that its work would be judged by an inspection of its transcripts, a reminder that mechanisms for trustworthy interaction guarantee neither accurate collective belief nor authorized collective action.","published":"2026-09-25T09:59:47Z","updated":"2026-09-25T09:59:47Z","authors":["Tomer Simon"],"categories":["cs.MA","cs.SI"],"headline":"AI Agents Coordinated Like Humans via Online Networks in 2026","summary":"Researchers studied two 2026 incidents where groups of AI agents, unable to officially communicate, still organized themselves on shared online channels. They found these agents developed social networks with identities, rules, and collective behavior similar to human groups in crises. The study highlights that while coordination occurred, it did not guarantee accurate information or authorized actions. The findings come from comparing investigations and agent records, emphasizing complexities in how autonomous AI collectives interact."},{"id":"2609.31059v1","title":"Collective order reorganizes dissipation and powers an active engine","abstract":"Motile active matter systems, from animal groups to synthetic particles, exhibit spontaneous self-organized transitions between disordered and ordered collective states. These transitions are driven by continuous energy injection, yet how energy consumption and dissipation change during these transitions remains poorly understood due to the challenge of measuring individual energetic fluxes. Here, we introduce a minimal experimental system of synthetic motile spheres that enables the first time-resolved power-budget characterization of collective motion. We show that the transition to an ordered state is accompanied by increased locomotion efficiency and a reorganization of dissipation pathways, shifting dissipation from internal friction to external slip with the environment. Leveraging this energetic reorganization, we construct an active engine powered by collective motion, capable of performing mechanical work against an external load. These findings establish energetic fluxes as quantitative observables for understanding collective state transitions and demonstrate how collective order can be harnessed for work extraction in active matter.","published":"2026-09-25T09:59:41Z","updated":"2026-09-25T09:59:41Z","authors":["Michael Riedl","Vincent Wattiez","Étienne Fodor","Jan Brugués","Francesco Romanò"],"categories":["cond-mat.soft","cond-mat.stat-mech"],"headline":"How Collective Motion Powers a New Type of Active Engine","summary":"Researchers studied tiny synthetic spheres that move on their own to understand how energy use changes when these particles shift from random to organized movement. They found that when the spheres move in an ordered way, they use energy more efficiently and change how energy is lost, shifting it from internal friction to slipping against their surroundings. Using this insight, the team built a simple engine powered by the collective motion of these particles, showing how group organization can generate mechanical work. Their study explains energy flow changes during collective behavior shifts."},{"id":"2609.31058v1","title":"Automatic sub-bundle tractometry reveals localized microstructural variations in complex white matter topographies","abstract":"Conventional tractometry averages white matter microstructure into a single profile, masking intra-bundle spatial heterogeneity and diluting localized alterations. We propose an automatic sub-bundle tractometry framework using Fr{é}chet-based hierarchical clustering to decompose bundles into geometrically coherent sub-bundles, resolving complex fanning configurations without empirical thresholds. Multi-compartment metrics (fractional anisotropy, FA; isotropic free-water fraction, IFW) are projected onto these clusters. We evaluated this framework in 140 participants across early-life (ELD) and late-life (LLD) depression cohorts, alongside age-matched healthy controls. We assessed performance based on microstructural homogeneity, sensitivity to age, and clinical group differences. Proposed sub-bundle metrics were significantly more homogeneous than classical full bundle profiles (p < 0.001). Our approach showed enhanced sensitivity, detecting more age-correlated bundles in elderly controls (11 vs. 5) and more clinical group differences in LLD (FA: 6 vs. 4; IFW: 23 vs. 17) with larger effect sizes. In ELD, it improved the detection of subtle FA differences between remitted and resistant individuals (2.9% vs. 1.8%). Gains were particularly pronounced in LLD, reflecting the framework's ability to resolve fanning tracts susceptible to brain aging. By preserving spatial heterogeneity, this automatic decomposition provides a robust foundation for large-scale clinical tractometry.","published":"2026-09-25T09:56:59Z","updated":"2026-09-25T09:56:59Z","authors":["Nathan Decaux","Jean-Charles Roy","Thomas Desmidt","Marie-Laure Paillère Martinot","Gabriel Robert","Julie Coloigner"],"categories":["eess.SP"],"headline":"New Method Reveals Details in Brain White Matter Differences","summary":"Researchers developed an automatic way to break down complex brain wiring into smaller, consistent parts to better study variations within these pathways. They tested this method on people with early- and late-life depression and healthy controls, finding it offers clearer, more sensitive detection of brain changes related to age and disease than traditional approaches. The study shows that their method can capture subtle local differences in brain structure, especially in aging-related conditions, but findings are based on measurement improvements rather than direct clinical outcomes."},{"id":"2609.31057v1","title":"Accuracy Evaluation of INS/ZUPT Filtering Methods Based on Different Geometric Error Definitions","abstract":"Geometric filters have recently been introduced to improve the accuracy and consistency of inertial-based integrated navigation systems. Error states were defined through specific group operations, introducing state correlations in error definition, which were lacked in the additive error used by a conventional indirect Kalman filter. The desirable consistent filtering models can be obtained based on specific geometric errors. For zero-velocity measurements expressed in the reference frame, this paper derives left-error process and measurement models from invariant filtering, two-frame-group filtering, and equivariant filtering. Importantly, a new group operation is introduced for the left tangent-group equivariant error. The analysis shows that the two-frame-group invariant extended Kalman filter (TFG-IEKF) and the tangent-group equivariant filter (TG-EqF) do not offer a significant consistency advantage over the invariant extended Kalman filter (IEKF). Experiments with an INS/ZUPT measurement system show that, under small initial attitude errors, the conventional indirect extended Kalman filter (EKF) achieves loop-closure position errors below $0.1\\%$ of the traveled distance, while the three geometric filters achieve comparable positioning accuracy.","published":"2026-09-25T09:54:16Z","updated":"2026-09-25T09:54:16Z","authors":["Wei Ouyang","Jiale Han","Yarong Luo","Maoran Zhu"],"categories":["cs.RO"],"headline":"Comparing Methods to Improve Navigation System Accuracy","summary":"Researchers studied different mathematical methods, called geometric filters, to enhance the accuracy of inertial navigation systems that measure movement. They examined how these filters define errors differently and tested several models, including new ones they proposed. Their experiments showed that while refined geometric filters perform well, the traditional method remains similarly accurate under certain conditions. The findings come from specific tests and do not claim broader advantages beyond those scenarios."},{"id":"2609.31056v1","title":"Neuralyzing the Trace: Selective Representation-Level Unlearning with Contrastive Sparse Autoencoders","abstract":"Machine unlearning aims to remove targeted information while preserving a model's other abilities. In realistic settings, such as privacy requests under the EU GDPR, the target may be narrow, for example information associated with a single person. Behavioral forgetting alone may be insufficient, motivating interventions directly on internal representations. However, standard mechanistic-interpretability extractors are poorly selective for such targets. We identify an energy bias in reconstruction-based extraction, which favors dominant background structure over low-energy target-specific components. We introduce SCALPEL, a contrastive sparse autoencoder designed to learn more selective forget features. We show theoretically that contrastive training promotes target-selective features and that our selection score controls expected background knowledge perturbation. We validate SCALPEL experimentally on TOFU across Qwen, Llama, and Gemma, where it substantially improves over NMF and standard SAE interventions and is competitive with Gradient Difference and RMU, bridging mechanistic interpretability and fine-grained unlearning.","published":"2026-09-25T09:51:52Z","updated":"2026-09-25T09:51:52Z","authors":["Itai Zehavi","Fanny Jourdan","Ulrich Aivodji"],"categories":["cs.AI"],"headline":"New Method Helps AI Forget Specific Information More Precisely","summary":"Researchers studied how to make AI models forget targeted information, like data about one person, while keeping their other skills intact. They found that existing methods often fail to isolate such specific info and instead remove broader background details. The team developed SCALPEL, a new approach that uses a special training method to better focus on and remove the precise information to be forgotten. Their tests showed SCALPEL performed better than several previous techniques, though results are based on experimental validation and theory."},{"id":"2609.31055v1","title":"Hilbert space in the flat limit of AdS/CFT. I. Massive particles","abstract":"The flat limit of AdS$_4$ spacetime corresponds, at the level of the symmetry algebra, to the Inönü--Wigner contraction $\\mathfrak{so}(2,3) \\to \\mathfrak{iso}(1,3)$. We describe the emergence of the Hilbert space of massive scattering states of arbitrary integer spin $s$ from unitary lowest-weight representations $\\mathcal{D}(Δ,s)$ of the conformal group, using a basis of `pseudo-momentum' states first introduced by Fronsdal. These states nicely reduce to Wigner's momentum eigenstates in the limit of vanishing AdS curvature radius $\\ell \\to \\infty$, at fixed $m=Δ/\\ell$. A simple conformal map relates this description to conformal fields in $\\mathbb{R}^3$ familiar from radial quantization, with the interior and exterior of the unit ball respectively corresponding to outgoing and ingoing momenta. Under the contraction, and upon appropriate rescaling, the inner product on $\\mathcal{D}(Δ,s)$ converges to the standard Lorentz-invariant inner product of Wigner's massive particles, so that unitarity is preserved throughout. The contraction to massless particles is briefly discussed here, and will be treated in detail in a forthcoming paper.","published":"2026-09-25T09:47:42Z","updated":"2026-09-25T09:47:42Z","authors":["Kevin Nguyen"],"categories":["hep-th"],"headline":"How particle states in curved space become flat-space particles","summary":"The researchers study a mathematical transition from a curved space model used in physics (AdS4 spacetime) to flat space, focusing on how particle states with mass and spin behave in this limit. They show how a certain mathematical structure (Hilbert space) describing these particles emerges from representations of the conformal group as the curvature goes to zero. Their findings clarify the relationship between curved-space and flat-space particle descriptions, preserving key properties like unitarity. Massless particles will be analyzed in a future study."},{"id":"2609.31054v1","title":"Cheap, open agents make LLM pollution harder to mitigate","abstract":"Large Language Model (LLM) pollution occurs when synthetic responses contaminate data intended to capture human behavior. High deployment costs have so far limited the risk posed by autonomous survey agents. However, open-weight models paired with open-source agentic frameworks may have removed this barrier. We compared the performance and detectability of nine agent configurations, ranging from fully open variants to closed commercial ones. Each agent autonomously completed a survey containing multiple response types yielding various detection checks. Fully open agents ran locally without usage fees and performed competitively with commercial alternatives. Open and commercial agents failed different sets of checks, and no single check reliably detected all agents, but open-text responses discriminated best between agents and humans. These findings identify fully open agents as a distinct risk for LLM pollution and support multilayered detection strategies emphasizing open-text analysis.","published":"2026-09-25T09:47:33Z","updated":"2026-09-25T09:47:33Z","authors":["Raluca Rilla","Anne-Marie Nussberger","Rui Mata","Dirk U. Wulff"],"categories":["cs.AI"],"headline":"Open-source AI agents increase challenges in spotting fake survey answers","summary":"Researchers studied how different AI agents, including free and commercial types, complete surveys to see if automated answers can be distinguished from real human responses. They found that fully open, free agents perform similarly to commercial ones and are harder to detect since no single method identifies all AI responses. Their work shows that these open agents pose a new risk of polluting data meant to reflect human behavior and suggests using multiple detection methods focusing on text analysis to address this issue."},{"id":"2609.31053v1","title":"Linear Certificates for Membership Comparability, Quadratic Barriers for Selectors","abstract":"Selectors and comparators supply only partial information about membership: a selector names a member of any pair that meets the language, while a binary membership comparator merely excludes one of the four membership vectors of a pair. We ask how much nonuniform advice turns such information into exact recognition. Our main result extends the optimal nondeterministic advice bound for P-selective sets to every binary membership-comparable language: $2-mc \\subseteq NP /(3n+5)\\cap\\mathrm{coNP}/(3n+5)$, with common fixed advice and certificates of at most $5n+12$ bits. The class is strictly larger; some 2-mc sets are not truth-table reducible to any P-selective set. The proof replaces the tournament king by an independent two-step cover of true signed literals, together with a short-forcing-or-exact-majority dichotomy, and it relativizes. Via an advice-preserving isolation transfer, a deterministic polynomial-time algorithm for promise Unique-Circuit-SAT gives $2-mc \\subseteq P/O(n)$. For selectors we determine tight orders of ordinary advice: $Θ(n)$ for errorless average-case computation and $Θ(n^2)$ for worst-case bounded-error computation, the latter independent of the interpreter's coin bound. One oracle realizes both orders on a single language and separates ordinary from coin-dependent advice. The quadratic and linear lower bounds hold for tournament-query procedures and relativized languages, not unconditionally for unrelativized P-selective sets.","published":"2026-09-25T09:43:41Z","updated":"2026-09-25T09:43:41Z","authors":["Sebastian Ben Daniel"],"categories":["cs.CC"],"headline":"Researchers analyze advice needed for partial membership information","summary":"The researchers study how much extra information ('advice') is required to turn partial membership data—gathered by special selection or comparison methods—into exact decisions about language membership. They extend previous results to a broader class, showing that certain languages can be recognized with limited advice and certificates. Their findings rely on theoretical proofs and simulations, highlighting that some complexity bounds hold only in specific settings, such as with oracle assistance, rather than unconditionally."},{"id":"2609.31052v1","title":"Measurement protocol for non-adiabatic geometric phases of Floquet states","abstract":"Periodically time-dependent quantum systems have Floquet solutions. They consist of a phase factor and a Floquet mode which shares the periodicity of the driving. Consequently, any such solution possesses a non adiabatic geometric phase known as the Aharonov-Anandan (AA) phase. It is gauge-invariant and, therefore, is experimentally accessible. We propose a protocol for the measurement of AA phases based on the adiabatic change of the phase of the non-adiabatic driving. We argue that the experimental requirements of the protocol are those of previous measurements of adiabatic phases.","published":"2026-09-25T09:43:34Z","updated":"2026-09-25T09:43:34Z","authors":["Carlos Martín-Fernández","Gloria Platero","Sigmund Kohler"],"categories":["cond-mat.mes-hall","quant-ph"],"headline":"New method proposed to measure special quantum phases in time-varying systems","summary":"Researchers studied quantum systems that change regularly over time and have special solutions with a phase called the Aharonov-Anandan (AA) phase. They proposed a way to measure this AA phase by gradually changing the driving phase in the system. Their findings suggest that the experimental setup needed is similar to what has been used before for related quantum phase measurements. The results are theoretical and focus on establishing a measurement approach."},{"id":"2609.31051v1","title":"Boolean Trusses as Rectangular Bands and Boolean Rings","abstract":"We introduce Boolean trusses, that is, trusses in which every element is idempotent. Every Boolean truss admits a decomposition into a truss arising from a Boolean ring and a rectangular truss. Furthermore, we classify all Boolean trusses up to isomorphism.","published":"2026-09-25T09:43:26Z","updated":"2026-09-25T09:43:26Z","authors":["Ryszard R. Andruszkiewicz","Bernard Rybołowicz"],"categories":["math.RA","math.CO"],"headline":"New Study Classifies Special Mathematical Structures Called Boolean Trusses","summary":"Researchers introduced a concept called Boolean trusses, which are mathematical systems where each element behaves in a certain consistent way. They showed that every Boolean truss can be broken down into parts related to Boolean rings and rectangular trusses. The researchers also classified all possible Boolean trusses by identifying which ones are essentially the same. The findings are theoretical and focus on understanding these mathematical structures."},{"id":"2609.31050v1","title":"Where Compute Matters: Heterogeneous Attention for Efficient Video Diffusion","abstract":"Efficient video generation requires reducing the quadratic cost of self-attention over long spatio-temporal token sequences. Existing efficient-attention methods typically apply the same computation pattern to every token, even though denoising difficulty varies substantially across video regions and evolves throughout the generation process. We introduce HetA-DiT, a heterogeneous attention mechanism that adaptively allocates computation according to token difficulty. A lightweight uncertainty branch predicts a token-wise estimate of denoising difficulty, which is used to route uncertain tokens through dense global attention while processing more reliable tokens with efficient local attention. The resulting routing is content- and timestep-adaptive, retains global context where it matters most, and provides a single parameter for controlling the quality-efficiency trade-off. HetA-DiT is compatible with few-step distribution-matching distillation and introduces no additional Transformer evaluation at inference time by reusing uncertainty estimates from the preceding denoising step. We evaluate the method on DMD-distilled Wan2.2-5B and Wan2.1-1.3B models. Across VBench, VBench-2.0, and human preference evaluation, HetA-DiT maintains competitive generation quality while routing only approximately 20% of tokens through dense attention.","published":"2026-09-25T09:43:04Z","updated":"2026-09-25T09:43:04Z","authors":["Olga Zatsarynna","Denis Korzhenkov","Juergen Gall","Amir Habibian","Mohsen Ghafoorian"],"categories":["cs.CV"],"headline":"New method improves video generation efficiency by focusing computing power","summary":"Generating videos efficiently is challenging because analyzing every part in detail is costly. Researchers developed HetA-DiT, a method that predicts which video parts are hard to process and spends more computing effort on them, while simpler parts get less attention. This adaptive approach keeps global context when needed and balances quality with speed. Tests show it maintains good video quality while cutting heavy computation to about 20% of the data."},{"id":"2609.31049v1","title":"Computational aberration-retrieval with entangled photons","abstract":"Adaptive optics (AO) is central to high-resolution microscopy, yet conventional implementations often require repeated correction-and-measurement cycles whose number increases with aberration complexity. This measurement burden can become prohibitive when the available temporal window for correction is short or when photon budgets are constrained. Here, we demonstrate an AO strategy that exploits the spatial correlations of entangled photon pairs to retrieve aberrations directly. By using the second-order correlation function as an aberration-sensitive metric, the required correction can be determined with only a few measurements, without relying on image-based feedback or iterative optimization. This correlation-based approach establishes an efficient framework for AO leveraging quantum light, offering a new route toward quantum-enhanced microscopy in regimes where conventional correction strategies become challenging.","published":"2026-09-25T09:42:46Z","updated":"2026-09-25T09:42:46Z","authors":["Baptiste Courme","YoonSeok Baek","Hugo Defienne"],"categories":["quant-ph","physics.optics"],"headline":"Using quantum light to improve microscope correction speed","summary":"The researchers address challenges in adaptive optics, a method used to sharpen microscope images but which often requires many slow correction steps. They developed a new technique using pairs of linked (entangled) photons to detect distortions faster by measuring how photon pairs are correlated, needing fewer measurements. Their work shows that this quantum-based approach can efficiently identify and correct optical distortions without repeated adjustments, suggesting a promising new way to enhance microscopes in difficult imaging situations. The findings are based on demonstrating the principle rather than clinical use."},{"id":"2609.31048v1","title":"Kintsugi-VLA: Turning Failed Robot Rollouts into Recovery Data through Interventional Recoverability","abstract":"Simulation enables scalable training of Vision-Language-Action policies by using privileged experts to generate visual demonstrations without requiring every trajectory to be collected through manual teleoperation. However, such pipelines typically retain successful demonstrations while failed rollouts are discarded, even though they expose precisely the off-nominal states from which recovery must be learned. We introduce Kintsugi-VLA, a framework for converting failed rollouts into targeted synthetic recovery data by exploiting exact state restoration and branching in simulation. For a fixed privileged expert, we define interventional recoverability as the probability of completing the original task after the simulator is restored to a given state, estimate it using adaptive Monte Carlo continuations with pointwise Wilson intervals, and characterize its non-monotonic evolution along failed trajectories. These estimates identify an observed terminal low-recoverability frontier-the point after which measured recoverability remains below a threshold-which is then used to select informative recovery starting states. In a simulated Franka manipulation task, targeted recovery data yield aggregate SmolVLA recovery success of 34.6\\% and 38.4\\% under difficulty- and frame-budget matching, respectively, 5.8 and 6.7 percentage points above uniform sampling within the same recovery window. The same ordering is observed under disturbed end-to-end execution and shifted clutter and physics conditions, while clean-task success decreases from 76.8\\% to 74.7\\%. Kintsugi-VLA demonstrates how failed simulator rollouts can be transformed from discarded experience into structured recovery-training data through direct interventional measurement.","published":"2026-09-25T09:42:20Z","updated":"2026-09-25T09:42:20Z","authors":["Ivan Snegirev","Elizaveta Semenyakina","Dmitrii Maliukov","Miguel Altamirano Cabrera","Dzmitry Tsetserukou"],"categories":["cs.RO"],"headline":"Using Failed Robot Trials to Improve Recovery Skills","summary":"Researchers created Kintsugi-VLA, a method that turns unsuccessful robot simulation attempts into useful training data by identifying points where recovery is still possible. They measure how likely a robot can complete its task after restarting from different failure points during simulation. Tested on a robot arm task, their method helped improve recovery success rates compared to random recovery training. These findings come from simulations and show how failures can guide better robot learning strategies."},{"id":"2609.31047v1","title":"DynBranch: Speculative Subgraph Reuse for Dynamic Agentic LLM Serving","abstract":"Agentic LLM workflows decide their execution paths at runtime. Downstream computation may be predictable, or may have run before, yet it cannot begin until the model or the user resolves the branch. We call this serialization the branch-resolution barrier. Caching alone does not hide it: the key that identifies a reusable result is not known until then. In this paper, we propose DynBranch, which makes an unresolved branch addressable before it resolves. Its stable coordinate lets candidate subgraphs run during resolution and completed subgraph results be reused across later requests. A two-level controller admits this work when its expected benefit exceeds the load price. DynBranch sits at the model-API boundary and requires no changes to agent harnesses or model execution engines. Across four agentic workloads with Qwen3-32B on 4x H200 GPUs, DynBranch reduces mean latency by up to 32% over each workload's strongest prior system and by 46-66% against a no-reuse floor, while preserving workflow results. The benefit persists across backbone families and on a commodity Qwen3-8B/RTX 4090 deployment.","published":"2026-09-25T09:42:18Z","updated":"2026-09-25T09:42:18Z","authors":["Junyi Shen","Noppanat Wadlom","Zhengyuan Su","Yao Lu"],"categories":["cs.DC","cs.AI","cs.LG"],"headline":"New Method Speeds Up AI Task Handling by Reusing Work Early","summary":"Researchers studied how AI models that make decisions during tasks face delays waiting for those decisions before continuing. They created DynBranch, a system that predicts and runs parts of the task early, even before decisions are finalized, allowing reuse of completed work. Tested on several AI workloads, DynBranch reduced processing delays by up to 32% compared to previous methods without changing existing AI software setups. The findings are based on experiments using specific AI models and computer hardware."},{"id":"2609.31046v1","title":"Modeling Student Sensemaking with LLMs and Knowledge-Graph-Guided Inference","abstract":"Collaborative science learning requires nuanced interpretation of student dialogue to characterize how learners identify knowledge gaps, build explanations, and work toward resolution - a theory-driven analysis that is labor-intensive and difficult to scale. We investigate whether instruction-tuned large language models (LLMs) can support multidimensional analysis of collaborative sensemaking without task-specific training, and whether structured knowledge-state information improves model inference. We evaluate two mid-size LLMs on 23 richly annotated, expert-labeled episodes across prompting conditions that vary definitional scaffolding, reasoning mode, and turn structure. Without reasoning, models tend to overpredict successful sensemaking; reasoning-enabled prompting improves identification of unsuccessful cases. Knowledge-state diagnostics provide additional grounding, improving detection of unsuccessful sensemaking and increasing agreement with expert annotations. No single configuration performs best across all sensemaking dimensions, underscoring the multidimensional nature of the task.","published":"2026-09-25T09:41:27Z","updated":"2026-09-25T09:41:27Z","authors":["Özge Alacam","Zübeyde Demet Kirbulut Güneş","Funda Ekici","Nurcan Turan-Oluk","Dilay Dinçdemir","Hakkı Kadayıfçı","Sevinç Nihal Yeşiloğlu","Burcu Işık","Halil Tümay","Sinem Gencer"],"categories":["cs.CL"],"headline":"Using AI to Analyze How Students Learn Science Together","summary":"Researchers studied whether advanced language models can help understand how students work together to learn science by analyzing their conversations. They tested different ways to prompt the models and found that guiding the AI with reasoning and structured knowledge improved its ability to detect when students struggled. However, no single AI setup worked best for every aspect of analyzing student discussions, highlighting the complexity of the task. These findings are based on evaluations with expert-labeled examples and do not prove all situations will perform similarly."},{"id":"2609.31045v1","title":"KuaFu: Compressing Long User Behavior into Understanding at Billion Scale","abstract":"Conversational agents, generative recommenders, and personalized advertising all rest on one capability: understanding each user from raw behavior. Prevailing industrial practice is task-specific: for each task, a relevant subsequence is extracted from the full history and a dedicated model trained on it. In production it hits two bottlenecks. First, even after filtering, a single-task sequence stays extremely long: content-interest summarization reads several hundred items per user, tens of thousands of tokens once serialized as prompt text. Second, profiles are refreshed routinely: a billion users weekly, roughly 100K QPM in aggregate, which under a fixed GPU budget sets a hard throughput floor. Compression is therefore mandatory, yet truncation or coarse compression can silently distort the profile, introducing four hallucination types (fabrication, omission, date misattribution, broken logic) that, with no way to evaluate the compressed representation itself, surface only as diffuse degradation in downstream metrics. We present KuaFu, a unified behavior-compression layer whose minimal unit is one behavior item. A two-axis projector compresses each item into 2-4 tokens of width 128-256 (about 10x along the token axis, 20x along width; per-item cache 10 KB to 0.5 KB), with fidelity-oriented four-stage training and layered intermediate evaluation. Across four production profiling tasks it matches or exceeds uncompressed single-task production models on all five headline metrics, raises per-GPU throughput by 37%-350%, and saves 190 GPUs. On public benchmarks it nearly always beats prior compressors at the same compression ratio (up to +17.7 EM on out-of-domain MRQA); on RecBench, a 4B model surpasses its 8B counterpart by 1.90 points. KuaFu has run on the Tencent advertising and recommendation platform for ten months, lifting overall GMV by 1.37%.","published":"2026-09-25T09:39:39Z","updated":"2026-09-25T09:39:39Z","authors":["Jiahao Hui","Lin Zhu","Yishen Hu","Jingdong Shu","Zetai Jiang","Xining Ran","Ben Tan","Yeshou Cai","Gong Chen","Haijie Gu","Jie Jiang"],"categories":["cs.IR","cs.CL","cs.LG"],"headline":"New Method Compresses User Data to Improve Online Recommendations","summary":"Researchers addressed the challenge of understanding long user histories, which are typically very large and costly to process, by developing KuaFu, a way to compress user behavior data into much smaller representations without losing important details. They tested KuaFu across several tasks and found it matched or outperformed existing methods while using fewer computing resources. Although results are promising, the findings come from specific compression and testing setups, and the researchers emphasize the importance of carefully balancing compression to avoid errors in user profiles."},{"id":"2609.31044v1","title":"Operator-ADI balanced truncation of delay differential systems","abstract":"We develop a balanced truncation method for linear delay differential systems based on a finite-rank alternating-direction implicit (ADI) iteration for the associated operator Lyapunov equations. The method acts directly on the infinite-dimensional history-state realization and does not discretize the delay interval during Gramian approximation or balancing. The finite-rank reachability and observability factors are represented exactly in Takenaka--Malmquist coordinates, reducing the required computations to finite-dimensional linear algebra. These factors yield a finite-dimensional, delay-free reduced model. For real system data, the implementation remains in real arithmetic. Under suitable conditions on the ADI shifts and a gap in the Hankel singular values, we prove convergence in $\\mathcal H_\\infty$ to the reduced transfer function obtained by exact balanced truncation of the infinite-dimensional system. The construction extends to finitely many discrete delays. Numerical experiments validate the method against independent reference computations. The complete MATLAB implementation used in the experiments is provided as supplementary material.","published":"2026-09-25T09:35:01Z","updated":"2026-09-25T09:35:01Z","authors":["Timo Reis"],"categories":["math.NA"],"headline":"New method reduces complex delay systems without oversimplifying","summary":"The researchers developed a technique to simplify mathematical models involving time delays without breaking them into small parts first. They worked directly with an advanced infinite-dimensional system representation, using a special iterative method to create smaller, manageable models that still capture essential behavior. Their approach, tested through numerical experiments, shows reliable convergence under certain conditions and is supported by computer code for validation."},{"id":"2609.31043v1","title":"Single-shot coherent process tomography and mid-infrared polarimetry with undetected photons","abstract":"Measurements with undetected photons infer properties of a probe beam that is never detected, transferring mid-infrared information onto silicon-friendly wavelengths. Tomography in this paradigm has so far relied on switched settings or scanned phases, referenced across a drift-sensitive campaign; no protocol reads all Jones parameters from one frame. We show that a folded nonlinear interferometer with a diagonally pumped crossed-crystal source performs coherent process tomography of the undetected beam in a single spectrometer frame: it reconstructs all seven parameters of the beam's round-trip Jones matrix, the coherent part of the channel. Birefringent group-delay walk-off, normally a calibration nuisance, acts as a frequency multiplexer: each Jones-matrix element is assigned its own carrier, a spectral fringe period paired with a detector port. Because the multiplexing exploits the down-conversion bandwidth rather than merely tolerating it, a single Fourier transform per detector trace returns all four moduli and all three relative phases across that bandwidth, i.e., the full round-trip Jones spectrum up to a global phase, with no scan or setting change (after a one-time reference frame). Simultaneity makes every relative phase immune to common-mode drift, and a dark carrier, a self-interference term that unitarity forces to vanish, provides a built-in null test. For realistic parameters of a periodically poled KTP source the scheme yields wavelength-resolved mid-infrared polarimetry across 3.5-4.2 microns (~80 spectral points per Jones element) on a silicon camera, extending to depth-resolved Jones matrices, i.e., polarization-sensitive optical coherence tomography with undetected photons. A Fisher-information analysis, the first we are aware of in undetected-photon tomography, sets the precision budget and compares delay multiplexing to sequential protocols at equal photon number.","published":"2026-09-25T09:34:57Z","updated":"2026-09-25T09:34:57Z","authors":["Marthe Zeja","Wen-Zhe Yan","Zhibo Hou","Sven Ramelow"],"categories":["quant-ph","physics.optics"],"headline":"New method measures light properties without detecting the light directly","summary":"Researchers developed a technique that captures detailed information about a mid-infrared light beam by analyzing a related beam at easier-to-detect wavelengths, using one quick measurement instead of multiple settings. Their approach uses a special interferometer to record all parameters describing how the beam changes polarization in a single camera frame, improving stability and speed. This method currently exists as a theoretical proposal supported by simulations and offers a way to study certain light properties without directly detecting the original beam."},{"id":"2609.31042v1","title":"Reachability Does Not Imply Searchability in Expanding Networks","abstract":"Routing and search respond in opposite ways to rapid network expansion. Short graph distances make a known destination easy to reach, while the neighborhood accessible within a few hops can be vastly larger than any finite inspection budget. We show that this tension imposes an algorithm-independent constraint on sparse search. If $m_q$ relevant nodes are placed without structural information among $N$ nodes and at most $M$ nodes can be inspected, any target-blind exploration protocol has success probability at most $Mm_q/N$, irrespective of correlations between successive inspections, of revisits, and of any structural preference the exploration rule may have. Geometry determines when this constraint becomes relevant: a target becomes accessible with order-one probability when the graph-distance ball volume satisfies $V(R_c)\\sim N/m_q$. At this scale, any inspection mechanism achieving a fixed success probability $δ>0$ must enrich the probability of inspecting relevant nodes by at least order $N/(Mm_q)$ relative to the neutral target density. For uniformly searched candidate sets, this entails a vanishing visible fraction of the accessible region. Thus rapidly expanding networks can make rare targets geometrically close while target-blind search remains ineffective. Numerical results on Krioukov hyperbolic random graphs and two-dimensional lattices show that this separation can arise within only a few hops in the rapidly expanding case, while the lattice accessibility radius grows algebraically.","published":"2026-09-25T09:34:09Z","updated":"2026-09-25T09:34:09Z","authors":["Antonio Scala"],"categories":["physics.soc-ph"],"headline":"Fast network growth hinders effective target searching","summary":"The researchers studied how quickly growing networks affect the ability to find specific targets. They found that even when destinations are close in the network, searching blindly without target information has very low success when inspecting only a limited number of nodes. Their mathematical results show a fundamental limit on search success unrelated to search strategies. Simulations on certain network models support that this problem appears after only a few steps in fast-expanding networks. The study highlights a key difference between being able to reach a target and actually finding it through search."},{"id":"2609.31041v1","title":"Room Impulse Response Embeddings for Speech Enhancement in Noisy and Reverberant Environments","abstract":"We propose a self-supervised approach for learning room impulse response (RIR) representations from single-channel noisy-reverberant speech. It consists of first training on reverberant data, then on noisy-reverberant data, and finally with a teacher-student approach, where the student learns to replicate the teacher's embeddings when given a noisy version of the reverberant input. We assess their representational capabilities by estimating acoustic room parameters from them. Conditioning a discriminative speech enhancement model on the derived embeddings yields consistent gains across all evaluated metrics, including downstream word error rate, for both reverberant and noisy-reverberant speech.","published":"2026-09-25T09:33:19Z","updated":"2026-09-25T09:33:19Z","authors":["Adrian Meise","Reinhold Haeb-Umbach"],"categories":["eess.AS","cs.SD"],"headline":"New Method Helps Improve Speech in Noisy, Echoey Rooms","summary":"Researchers developed a way for computers to learn how room sounds distort speech using only noisy recordings. They trained a system to recognize these room effects step-by-step and then tested how well it could estimate room properties. Using these learned 'room sound signatures' to help speech enhancement models led to better clarity and lower errors in recognizing speech, even in challenging noisy and echo-filled environments. The findings are based on controlled experiments and simulations, not real-world deployment yet."},{"id":"2609.31040v1","title":"Exploiting Spatial Structure for Transductive Few-Shot Classification of Whole-Slide Images","abstract":"Automating the analysis of whole-slide images (WSIs), a key step in cancer diagnosis, has high clinical value, as it can reduce pathologist's workload while improving diagnosis accuracy. Recently, vision-language models have shown promising performance for patch-level classification without requiring any annotation, yet these zero-shot (ZS) predictions remain noisy on fine-grained tasks and must be further refined. A promising direction is to refine all predictions jointly, i.e., a transductive approach. However, most existing methods are not tailored to WSIs. We thus propose SlideTIM, an adaptation to WSIs of the recent transductive approach LC-TIM, which introduces a combined spatial--latent regularizer together with a prior on the patch class distribution. The former enforces spatially and semantically close patches to receive the same predictions, while the prior calibrates the predicted class proportions. Together, they address the complex spatial organization and the strong class imbalance of WSIs. Evaluated on four histology datasets, SlideTIM consistently outperforms all TIM variants, improving the macro-F1 by +8.1pp over the best competing baseline at 1 shot. Compared to the ZS, it raises the macro-F1 by +19.4pp at 1 shot. The code will be made available after submission.","published":"2026-09-25T09:33:09Z","updated":"2026-09-25T09:33:09Z","authors":["Tiffanie Godelaine","Manon Dausort","Karim El Khoury","Benoît Gérin","Benoît Macq","Christophe De Vleeschouwer"],"categories":["cs.CV"],"headline":"New Method Improves Cancer Image Analysis with Few Examples","summary":"Researchers aim to help automate cancer diagnosis by improving how computers analyze large microscope images called whole-slide images (WSIs). They developed SlideTIM, a method that uses spatial information to better classify small image patches together, addressing challenges like complex tissue layout and uneven class types. Testing on four datasets showed SlideTIM outperforms previous methods, especially when only one example per class is available. These findings come from simulations and comparisons; the researchers plan to share their code after publication."},{"id":"2609.31039v1","title":"MetaPermit: Scalable and Auditable Access Control for AI Agents via LLM-Inferred Meta-Attributes","abstract":"The rise of autonomous AI agents equipped with tools has introduced significant security risks, ranging from unintended tool misuse to adversarial manipulation through Indirect Prompt Injection (IPI) attacks. In practice, deployed agent systems such as OpenAI Codex and Claude Code protect tool invocations through a combination of coarse-grained permission rules and LLM-based judgments about individual proposed actions. Both components, however, have important limitations: static policies must anticipate possible user intents and therefore do not scale to open-ended tasks, while LLM-driven authorization supports dynamic decisions but produces inconsistent outcomes and remains vulnerable to targeted IPI attacks. To provide scalable and more consistent authorization, we propose MetaPermit, a policy-based tool access-control framework that decouples semantic inference from security enforcement. By analyzing agent-user interactions, we derive a compact, task-independent set of meta-attributes that capture the relationships among the user's intent, the execution context, and the proposed tool call. These meta-attributes allow MetaPermit to authorize tool use without enumerating user intents. At runtime, an LLM infers the meta-attribute values for each proposed tool call, while a fixed policy evaluates these values to allow or deny the call, making each decision auditable through the inferred values and the applied policy rule. We evaluate MetaPermit on the AgentDojo and AgentDyn benchmarks, across seven task suites and five attack methods, using two widely deployed open-weight LLMs. The results show that MetaPermit produces 31% more consistent authorization decisions than LLM-driven authorization and outperforms the state-of-the-art defenses CaMeL and IPIGuard in both task completion, with improvements of up to 109%, and robustness to IPI attacks, with no malicious tool calls executed.","published":"2026-09-25T09:32:31Z","updated":"2026-09-25T09:32:31Z","authors":["Hanzhang Ma","Ali Hariri","Tianxiang Shen","Bohua Zou","Qianjun Zheng","Ji Wang","Li Yi","Ning Jia","Yutao Liu","Haibo Chen","Lin Wang","Debayan Roy"],"categories":["cs.CR","cs.SE"],"headline":"New System Improves Security for Autonomous AI Tool Use","summary":"Researchers address security risks in AI agents that use tools, such as unintended misuse and tricky attacks called Indirect Prompt Injection (IPI). They created MetaPermit, a method that separates understanding user intent from enforcing rules, using a set of general factors called meta-attributes to decide if a tool call is allowed. Testing on existing AI benchmarks showed MetaPermit made decisions more consistently and defended better against attacks compared to previous methods, but findings are based on simulated scenarios with specific AI models."},{"id":"2609.31038v1","title":"Aurora-X: Built for Extreme Time Series Forecasting","abstract":"Time series foundation models (TSFMs) enable cross-domain forecasting, but their development as general-purpose forecasters remains constrained by underexplored training potential and limited architectural versatility. To address these challenges, we introduce Aurora-X, a billion-scale TSFM with a progressive curriculum and a unified architecture. We first use channel-independent pretraining to learn temporal patterns, then introduce cross-variable dependencies, varied context and horizon lengths, and future covariates if available during midtraining. Variable-resolution post-training further enables an adjustable temporal span per token at inference. With fixed model weights, this supports longer histories under a fixed token budget or fewer tokens for the same history, enabling test-time scaling. With a versatile architecture, Aurora-X supports cross-variable modeling, covariate conditioning, and parallel decoding of future patches for probabilistic forecasting. These are supported by a novel pattern-guided mixture-of-experts that expands model capacity through sparse activation and uses shallow patch similarities to constrain deep-layer routing, guiding expert specialization across heterogeneous time series. Furthermore, we propose an implicit quantile network head that predicts arbitrary quantiles to characterize predictive distributions, enhancing probabilistic forecasting flexibility. Comprehensive experiments on GIFT-Eval, TIME, FEV-Bench, TFB, and DAG-Bench demonstrate state-of-the-art forecasting performance against pretrained TSFMs and task-specific supervised models.","published":"2026-09-25T09:32:22Z","updated":"2026-09-25T09:32:22Z","authors":["Xingjian Wu","Chenjuan Guo","Xiangfei Qiu","Zhigang Hu","Hanyin Cheng","Peng Chen","Yang Shu","Jilin Hu","Bin Yang"],"categories":["cs.LG"],"headline":"Aurora-X: A New Model for Predicting Complex Time Patterns","summary":"Researchers developed Aurora-X, a large-scale model designed to improve forecasting of time-based data across different areas. They trained it in stages to understand simple patterns first, then more complex relationships, allowing flexible prediction lengths. The model uses a unique system to specialize parts of itself for different data types and predicts a range of possible future outcomes. Tests showed Aurora-X performed better than existing models, but these findings are based on simulations and specific datasets provided by the researchers."},{"id":"2609.31037v1","title":"Goal-Oriented Weighting of Reynolds-Stress Data for Learning Turbulence Models in Complex Flows","abstract":"Data-driven turbulence models for the Reynolds-averaged Navier-Stokes (RANS) equations offer a promising route to improving predictions of complex flows. Such models, specifically the turbulence constitutive relations, can be learned efficiently from high-fidelity Reynolds-stress data without evaluating the RANS equations during training. However, conventional losses weight all tensor-component and spatial errors equally, although their influence on a quantity of interest (QoI) can differ significantly; reducing the aggregate stress error therefore does not necessarily lead to an improved QoI prediction. Training against flow-level observations accounts for this dependence but requires repeated, potentially expensive RANS solutions. In canonical shear flows, physical reasoning can identify the shear component as the only one relevant for the mean-flow prediction. Motivated by this example, we propose a goal-oriented method to select and weight Reynolds-stress training data for complex flows. For a specified QoI, a single offline adjoint evaluation quantifies its sensitivity to local perturbations of each Reynolds-stress component. These sensitivities are converted into fixed weights in the supervised loss, enabling training without further RANS solutions. We validate the weighting in square-duct and periodic-hill flows, then apply it to a film-cooling jet in crossflow, where it improves velocity and cooling-effectiveness predictions over uniformly weighted training despite using a velocity-only QoI. Beyond turbulence modelling, this work suggests a broader strategy for aligning supervised learning with downstream prediction goals while preserving training efficiency.","published":"2026-09-25T09:32:19Z","updated":"2026-09-25T09:32:19Z","authors":["Zhuolin Zhao","Haochen Wang","Youngwoo Kim","Solkeun Jee","Heng Xiao"],"categories":["physics.flu-dyn"],"headline":"Researchers Improve Turbulence Models by Focusing on Key Data","summary":"Predicting complex fluid flows often relies on models using detailed turbulence data. The researchers found that treating all parts of this data equally doesn't always improve important flow predictions. They developed a method that weights the data based on how much each part affects specific flow features, using a one-time calculation to guide training. Tested in various flow scenarios, their approach improved predictions without costly repeated calculations, suggesting a smarter way to train models aligned with practical goals."},{"id":"2609.31036v1","title":"Bio-inspired control model for snake locomotion based on Cosserat beam theory","abstract":"This paper revisits from a mathematical perspective the modeling of locomotion in snake-like bodies in fluid environement as given in \\cite{boyer2010poincare} and \\cite{boyer2011macrocontinuous}. Using a Cosserat rod theory, we derive several systems of PDEs depending on the type of constraint and actuation.","published":"2026-09-25T09:28:45Z","updated":"2026-09-25T09:28:45Z","authors":["Eliot Thys","Frédéric Boyer","Yacine Chitour","Petri Kokkonen","Vincent Lebastard","Swann Marx","Clément Moreau"],"categories":["math.AP"],"headline":"Mathematical Model of Snake Movement in Fluid Using Rod Theory","summary":"Researchers revisited the mathematical modeling of how snake-like bodies move in fluid environments. They used a theory called Cosserat rod theory to create different equations based on varying constraints and movements. The paper focuses on the mathematical approach and derives systems of equations but does not report experimental results or practical applications."},{"id":"2609.31035v1","title":"Compact Force Sensor for Dual-UAV Cable-Suspended Payload Transport with Tension-Aware Outer-Loop Control","abstract":"Cooperative payload transportation using multiple \\textit{Unmanned Aerial Vehicles} (UAVs) poses challenges in stability, coordination, and robustness, especially under external disturbances and unmodeled dynamics. This work proposes a dual-UAV payload transportation framework supported by a compact, custom-designed force sensor measuring the interaction force at the UAV cable anchor point. The sensor design and mathematical model are presented, and its performance is characterized through static and dynamic tests evaluating linearity, hysteresis, repeatability, and crossload. The control architecture follows a cascade structure: fast inner loops handle vehicle stabilization, while outer loops are designed to compensate for the measured forces. The approach is validated through simulations and indoor experiments under position uncertainty. Payload-drop and constrained-space tests assess the proposed sensing and control architecture against literature-based distributed references, showing improved stabilization, coordination, and disturbance rejection. A video of the experiments is available at: https://youtu.be/rIw9-fvV8Qw.","published":"2026-09-25T09:28:16Z","updated":"2026-09-25T09:28:16Z","authors":["Andrea Delbene","Giorgio Cannata","Giorgio Carlini","Filippo Sante","Marco Baglietto"],"categories":["cs.RO"],"headline":"New Force Sensor Helps Two Drones Carry Loads More Stably","summary":"The researchers addressed challenges in coordinating two drones carrying a shared load with cables, focusing on stability and handling disturbances. They designed a small custom sensor to measure forces where the cable connects to each drone, tested its accuracy, and developed a control system that uses this force information to better stabilize the drones. Their simulations and indoor experiments showed improved load handling and coordination, but results are limited to controlled settings and simulations only."},{"id":"2609.31034v1","title":"Whitney Properties and Whitney reversible properties of Cut and Non-Cut Points","abstract":"In this paper, we study several properties concerning cut points and non-cut points. First, we show that the property of being a continuum consisting entirely of shore points is preserved under refinable maps. Next, we show that having a block point, having a non-shore point, and having a strong center point are not Whitney-reversible properties. Consequently, \\cite[Question 2.9]{nonweak} has a negative answer. Furthermore, we show that none of the classes of continua witnessing the failure of the sequential strong Whitney-reversible property for these three properties admits a common model. Finally, in connection with the fact that colocal connectedness is not a Whitney-reversible property, we show that the class of nonaposyndetic continua having a cut point and such that every positive Whitney level is colocally connected has no common model.","published":"2026-09-25T09:27:35Z","updated":"2026-09-25T09:27:35Z","authors":["Eiichi Matsuhashi"],"categories":["math.GN"],"headline":"Researchers examine special points in connected shapes","summary":"This research looks at certain types of points called cut points and non-cut points within continuous shapes. The researchers investigate how specific properties of these points behave under mathematical transformations and whether these properties can be reversed or modeled commonly. Their findings show some properties are not reversible and that certain classes of shapes with these points do not share a common representative model. The study focuses on theoretical aspects and does not cover practical applications."},{"id":"2609.31033v1","title":"Robust Graph Clustering Network for Multiple Missing Data","abstract":"Clustering on graphs where both node attributes and structural links are partially missing remains a challenging task. Existing methods typically rely on imputation-then-clustering on single-view missingness incomplete graphs, which are vulnerable to cross-view error propagation and cluster-boundary blurring under simultaneous attribute and structure missingness. To address these limitations, we propose a Robust Graph Clustering Network for Multiple Missing Data (RGCN), which is designed to handle simultaneous node attribute and graph structure incompleteness. RGCN introduces three key innovations: First, we design a view-decoupled dual-branch imputation to mitigate interference and enable mutual enhancement in recovering missing data. Second, we employ a multi-hyperspherical mixture prior to enhance intra-cluster compactness and inter-cluster separability on a directional latent manifold. Third, a boundary-aware contrastive enhancement objective mitigates the blurring of clusters caused by imputation bias. Extensive experiments on real-world datasets demonstrate that RGCN consistently outperforms state-of-the-art baselines under various missing patterns.","published":"2026-09-25T09:26:45Z","updated":"2026-09-25T09:26:45Z","authors":["Keyuan Qiu","Renda Han","Zhen Tang","Qiang He","Xingwei Wang","Wenxin Zhang","Guangzhen Yao","Junxin Chen","Qingjian Ni"],"categories":["cs.LG"],"headline":"New Method Improves Grouping on Incomplete Network Data","summary":"Researchers address the problem of grouping data points in networks where some information about nodes and their connections is missing. They developed a Robust Graph Clustering Network (RGCN) that better recovers missing information and keeps groups distinct by using dual-branch recovery, a mathematical model for clearer group separation, and a technique to reduce errors from filling in missing data. Tests on real datasets show RGCN outperforms existing methods, though the results come from specific controlled experiments."},{"id":"2609.31032v1","title":"TempQ-Jail: Query-Constrained Candidate Ranking for Text-to-Video Jailbreak Attacks","abstract":"Existing text-to-video (T2V) jailbreak methods mainly seek more effective or stealthier attack candidates. In guarded T2V systems, however, video generation and security evaluation are costly, so an attacker often cannot test a large candidate pool. We therefore formulate T2V jailbreak as a query-constrained candidate allocation and ranking problem and propose TempQ-Jail. The method combines heterogeneous attack mechanisms to expand candidate coverage, estimates each candidate's end-to-end attack value from security-gate passage, dangerous visual generation, preservation of the original intent, and temporal validity, and ranks candidates so that high-value attacks appear early in a limited query trajectory. We evaluate TempQ-Jail on CogVideoX-5B using 70 common viable intents derived from T2VSafetyBench and compare it with six representative T2V jailbreak methods under a unified protocol. TempQ-Jail achieves TP-ASR@5 and TP-ASR@10 of 48.9% and 65.4%, improving over the strongest baselines by 4.6 and 4.0 percentage points, respectively. It also obtains the highest AUC-TP (0.469) and the lowest AvgQ (6.3). Analyses of query trajectories, candidate allocation, failure attribution, and ablations show that TempQ-Jail more effectively identifies and prioritises candidates with complete attack potential under limited query budgets.","published":"2026-09-25T09:26:30Z","updated":"2026-09-25T09:26:30Z","authors":["Tianmeng Fang","Jiancheng Wang","Chen Wang","Liming Wang","Wei Wang","Jiayang Liu","Xiaochun Cao"],"categories":["cs.MM","cs.CR","cs.CV"],"headline":"New Method Improves Attacks on Text-to-Video AI Systems","summary":"Researchers studied ways to trick text-to-video generation systems, which produce videos from text prompts but have defenses against harmful content. They created TempQ-Jail, a method that smartly prioritizes attack attempts when testing is costly, combining different attack techniques to find effective examples faster. In tests, TempQ-Jail performed better than previous methods, identifying risky videos with fewer tries. The findings are based on simulations using a specific video generation model and limited to the scenarios tested."},{"id":"2609.31031v1","title":"Metacognitive Selective Ensemble for Mobile Systems","abstract":"Deep ensembles improve robustness in mobile sensing, but repeatedly executing many models over continuous sensor streams is costly. Selecting only a few members reduces this cost, yet adaptive selection often requires additional model execution to obtain reliable evidence about inactive candidates. We present MetaSE, an active ensemble framework that exploits short-term persistence in per-model reliability. MetaSE maintains a small active set across windows, uses post-execution evidence to reject unreliable members, and invokes lightweight routing only when replacement is needed. This stateful design accesses the diversity of a larger pool without repeated full-pool evaluation. Across four HAR datasets and four model architectures, MetaSE consistently improves over a fixed three-model ensemble and achieves accuracy comparable to substantially more expensive adaptive and full-ensemble inference. On a Raspberry Pi 4B, MetaSE is 2.7x faster and uses 69% less memory than full ten-model inference.","published":"2026-09-25T09:26:24Z","updated":"2026-09-25T09:26:24Z","authors":["Sungmin Lee","Kichang Lee","Joonhee Lee","JaeYeon Park","Songkuk Kim","JeongGil Ko"],"categories":["cs.LG"],"headline":"New method improves mobile sensing efficiency and accuracy","summary":"Researchers developed MetaSE, a system that smartly selects a few reliable models from a larger group to analyze mobile sensor data, reducing computing costs. By remembering which models perform well recently, MetaSE avoids running all models repeatedly, saving time and memory while maintaining accuracy. Tests on multiple datasets and devices showed MetaSE is faster and uses less memory than running full model sets, though results come from specific experiments and may not generalize without further study."},{"id":"2609.31030v1","title":"A derived Chouinard's theorem for infinite groups","abstract":"We establish a Chouinard-type theorem for modules over algebras of cochains on the classifying spaces of a broad class of infinite groups, allowing commutative ring spectra as coefficients. We then use this result to study finiteness properties of the integral cohomology rings of certain infinite groups, as well as stratification of the corresponding module categories. In particular, we complete the (co)stratification of module categories associated to homotopical groups over fields of positive characteristic.","published":"2026-09-25T09:26:13Z","updated":"2026-09-25T09:26:13Z","authors":["Natàlia Castellana","Juan Omar Gómez"],"categories":["math.AT","math.RT"],"headline":"New theorem advances understanding of infinite group structures","summary":"Researchers developed a new version of Chouinard's theorem, originally for finite groups, now applying to a wide range of infinite groups. They used this to explore complex properties of mathematical objects called cohomology rings and module categories related to these groups. Their work completes certain classifications in cases involving fields with positive characteristic, although the results are theoretical and focus on abstract algebraic structures."},{"id":"2609.31029v1","title":"Governed Deduction: Policy-Grounded Premise Authorization Beyond Relevance","abstract":"Reasoning systems usually treat premise use as a question of relevance: if a fact is available and useful, it may be selected for inference. Authorization imposes a different constraint: a premise may be represented and logically usable but not permitted for a particular local transition. We formalize this distinction as Governed Deduction (GD), with a transition-local admission predicate admit(p, tau, S). From an independently produced RBAC-augmented Spider benchmark, we construct 4,461 matched authorization pairs in which the same query premise and policy state support permitted and denied consuming transitions. An initial joint controller reaches 99.19% held-out accuracy, but a transition-only control reaches 100%, exposing a role-name shortcut. After a frozen, label-independent context-local role permutation removes that shortcut, premise/state-only, transition-only, and joint linear controllers all score exactly 50% on 1,856 held-out edges, while a symbolic policy oracle remains at 100%. The result is a controlled negative finding: the benchmark instantiates policy-grounded authorization beyond relevance, but the frozen linear representation does not recover the relation. Matched one-sided controls and leakage audits are therefore essential for evaluating learned policy-sensitive reasoning.","published":"2026-09-25T09:24:19Z","updated":"2026-09-25T09:24:19Z","authors":["Wesley Shu","Hsi-Ching Lin"],"categories":["cs.AI"],"headline":"New Study Explores Limits of Policy-Based Reasoning Systems","summary":"Researchers examined how reasoning systems decide which facts to use, focusing on rules that allow or forbid certain facts beyond their usefulness. They introduced a formal approach called Governed Deduction to capture these policy-based permissions in reasoning processes. Their tests showed that while some models could predict authorization under certain conditions, simpler representations failed to capture the policy constraints, highlighting challenges in teaching machines to follow complex rules."},{"id":"2609.31028v1","title":"Refining Cytology Predictions with Conditional Random Fields","abstract":"Vision-language models (VLMs) achieve strong zero-shot (ZS) classification on histology images but do not perform as well on cytology, whose stains and cell morphology differ markedly compared to histology. Conditional random fields (CRFs) can refine noisy VLM predictions by propagating information across patches, but existing CRF frameworks were designed for histopathology and do not transfer to cytology datasets, released as independent patch pools spanning multiple staining protocols. We introduce CytoCRF, which adapts the pairwise terms to cytology by targeting chromatin and cytology-specific staining, and further enrich the neighborhood of each potential term by combining multiple backbones. Across ten cytology datasets, CytoCRF outperforms existing CRF frameworks at every annotation budget, reaching +13.6 percentage points over the best baseline and +33.7 over ZS with only 50 annotations. Combining information from multiple backbones brings further gains, showing that the neighborhood topology matters more than the pairwise potential computed over it.","published":"2026-09-25T09:24:15Z","updated":"2026-09-25T09:24:15Z","authors":["Manon Dausort","Tiffanie Godelaine","Karim El Khoury","Maxime Zanella","Christophe De Vleeschouwer","Benoît Macq"],"categories":["cs.CV"],"headline":"Improving Cell Image Analysis with New Refinement Technique","summary":"Researchers studied how to better predict cell details in cytology images, which differ from histology images in color and cell shape. They developed CytoCRF, a method that improves initial predictions by considering how nearby image parts relate, tailored specifically for cytology's unique features. Testing on ten datasets, CytoCRF outperformed existing methods, especially when only limited labeled data was available. Their findings highlight that combining data from multiple analysis models improves results more than just refining relationships within the data."},{"id":"2609.31027v1","title":"When Can Quantum Extreme Learning Machines Replace Quantum Reservoirs?","abstract":"Quantum reservoir computers (QRCs) retain information from earlier inputs, whereas reset-window quantum extreme learning machines (QELMs) start afresh and receive only a recent input window. We establish when this window can replace the reservoir's memory. Under our stated assumptions and with exact expectation values, sufficiently expressive QELMs can reproduce reservoir outputs arbitrarily accurately as their windows grow precisely when the influence of the omitted past becomes uniformly negligible. However, longer windows do not always solve the problem. We construct a simple two-qubit reservoir that performs temporal recall exactly, while every fixed-window predictor faces the same positive worst-case error over unbounded delays, regardless of model capacity. Analytical benchmarks and simulations distinguish missing history from limited features and finite measurements. The central conclusion is simple: better representations can improve how available information is used, but cannot recover a past the model never receives.","published":"2026-09-25T09:23:51Z","updated":"2026-09-25T09:23:51Z","authors":["Markus Baumann","Gerhard Stenzel","Jonas Stein","Claudia Linnhoff-Popien"],"categories":["quant-ph"],"headline":"When Can New Quantum Models Replace Memory-Based Quantum Systems?","summary":"Researchers studied quantum computing models that process information differently: some remember past inputs, while others only use recent data windows. They found that under certain conditions, models using recent data can mimic memory-based systems if the old information's influence is negligible. However, for tasks needing exact recall of distant past inputs, limited recent data always leads to errors. The findings emphasize that improving data use helps, but cannot substitute for missing historical information in quantum systems."},{"id":"2609.31026v1","title":"Positivity Rigidity for Grossman-Larson Characters and the Kingman Face of the Hoffman Rooted-Tree Graph","abstract":"We classify the real characters of the Grossman-Larson Hopf algebra that are nonnegative on the rooted-tree basis. They vanish on trees with branching away from the root, and the normalized characters are parametrized by Kingman paintboxes. Two-sided Pieri states are mixtures of the normalized characters, with unique mixing measures. The corresponding laws form a proper exposed Bauer face of the simplex of central measures on the Hoffman rooted-tree graph. For the path-forest subgraph, the full and minimal Martin boundaries coincide and are homeomorphic to the Kingman simplex. The limiting ranked root-branch frequencies generate the completed central tail.","published":"2026-09-25T09:22:21Z","updated":"2026-09-25T09:22:21Z","authors":["Shengjun Zhang"],"categories":["math.CO"],"headline":"Understanding certain mathematical structures related to rooted trees","summary":"The researchers studied special functions called characters on a complex algebra involving rooted trees and found those that stay nonnegative have specific properties. They linked these characters to a mathematical object known as the Kingman paintbox and described how mixtures of these characters form unique probability measures on a certain graph of rooted trees. Their findings clarify the structure of these measures but are focused on theoretical classification and properties within advanced mathematical frameworks."},{"id":"2609.31025v1","title":"Precision at Speed: Sample-Efficient Online Model-Based Reinforcement Learning for Hydraulic Excavator Control","abstract":"Precise, high-speed control remains challenging for robots with complex actuation dynamics. Learning directly on hardware is further constrained by the cost of real-world interaction. We present an online model-based reinforcement learning framework that learns a probabilistic dynamics ensemble model from scratch for sampling-based model predictive control. A precision-gated contouring objective conditions the progress reward on path accuracy, prioritizing precision over speed. In a data-driven excavator simulator, the framework achieves higher sample efficiency than the evaluated model-based reinforcement learning baselines. We validate the framework by learning directly on an 11.5-ton Menzi Muck M445 hydraulic excavator, without demonstrations or simulation pretraining. After 20 minutes of interaction, the controller reaches tracking accuracy comparable to prior learned controllers trained on 100-150 minutes of data. After 40 minutes, it sustains sub-centimeter mean path error at high operating speeds.","published":"2026-09-25T09:22:12Z","updated":"2026-09-25T09:22:12Z","authors":["Claudio Canales","Fang Nan","Marco Hutter","Javier Ruiz-del-Solar"],"categories":["cs.RO","cs.LG","eess.SY"],"headline":"Faster, More Accurate Robot Control for Excavators Using Smarter Learning","summary":"Controlling robots like excavators precisely and quickly is difficult, especially when learning from limited real-world practice. The researchers developed a new learning method that builds a model of the excavator's movements while it operates, focusing on accuracy over speed. Tested first in simulations and then on a real 11.5-ton excavator, their system learned effective control much faster than previous methods, achieving precise path tracking after just 20–40 minutes of use, without prior training or demonstrations."},{"id":"2609.31024v1","title":"Synth-JEPA: Joint Embedding Prediction for Renderer-Free Synthesizer Parameter Search","abstract":"Sound matching can be formulated as optimizing synthesizer parameters against an audio-domain objective. However, objectives derived from generic audio representations are often difficult to optimize, while direct search requires rendering every candidate. We introduce Synth-JEPA, which learns mutually predictive audio and parameter representations from paired synthesizer data. At inference, candidate parameters are scored directly in this learned space, yielding a renderer-free objective whose audio geometry is shaped by parameter correspondences rather than generic audio similarity. We evaluate Synth-JEPA on Surge XT using held-out synthesizer sounds and out-of-domain NSynth and FSD50K targets, against inverse models, direct search, and learned proxy objectives. Synth-JEPA outperforms all baselines in-domain and remains competitive out-of-domain. Its matching quality continues to improve with additional test-time search, allowing compute to be traded for match quality. In pairwise listening tests, listeners preferred Synth-JEPA in 85% of trials overall. Together, these results show that an audio representation with a parameter-induced geometry allows synthesizer sound matching to be approached as an effective renderer-free search problem.","published":"2026-09-25T09:22:11Z","updated":"2026-09-25T09:22:11Z","authors":["Ben Hayes","Haokun Tian","Stefan Lattner"],"categories":["cs.SD","cs.LG","eess.AS"],"headline":"New Method Improves Synthesizer Sound Matching Without Rendering","summary":"Researchers developed Synth-JEPA, a technique that learns connections between sound and synthesizer settings to find the best match without having to create (render) each sound candidate. By evaluating potential settings directly in a learned space informed by actual parameter relationships, Synth-JEPA works better than previous methods when tested on various sound datasets. The researchers found that it improves sound matching quality further with more search time, and listeners preferred its matches in most tests. These findings are based on comparisons using simulated and real sound data."},{"id":"2609.31023v1","title":"Non-homogeneous curvature flows in a hemisphere","abstract":"Let S^{n+1}_{+} be the open hemisphere of the unit sphere S^{n+1} centred at o. We study the non-homogeneous curvature flow X_t=-f(r) sigma_k^{alpha} {nu} of smooth, closed, strictly convex hypersurfaces enclosing o, where r is the geodesic distance to o. We consider both the supercritical regime beta>1+k {alpha} and the critical regime beta=1+k{alpha}, where beta is the growth order of the profile at the origin, f(r) almost equals r^{beta} as r descends to 0. Under the structural condition that f^{1/(1+k{alpha})} is convex, we prove long-time existence and preservation of strict convexity. The normalized radial function converges smoothly and exponentially to a constant: to 1 and to R_{infty}>0, resp. in different two cases. Thus the normalized radial graphs become round, while the original hypersurfaces contract to o.","published":"2026-09-25T09:22:09Z","updated":"2026-09-25T09:22:09Z","authors":["Hongyi Sheng","Weimin Sheng","Jiazhuo Yang"],"categories":["math.DG","math.AP"],"headline":"Mathematicians study shapes shrinking inside a hemisphere","summary":"Researchers investigated how certain curved surfaces inside a half-sphere change shape over time when they move according to a rule involving curvature and distance from a center point. They focused on two specific cases based on how the influencing factor grows near that center. The team proved that these surfaces remain nicely curved and shrink steadily, eventually becoming perfectly round before collapsing to the center. Their work is theoretical, relying on specific mathematical conditions and equations describing the surfaces' behavior."},{"id":"2609.31022v1","title":"EARL: Exposure- and Allocation-Reweighted Linear Estimator for Bipartite Experiments with Partial Assignment","abstract":"Bipartite A/B tests are experiments in which treatment is randomised over one set of units, while outcomes are measured on another. For example, an online marketplace may test a new pricing algorithm on a random subset of items, while the outcome of interest, say, purchase satisfaction, is measured on customers, each of whom interacts with many items. Existing methods for analysing bipartite experiments assume that every randomisation unit is assigned to treatment or control. In practice, often only a subset participates: platforms cap rollout risk, reserve holdout groups, and split their population across concurrent tests. Ignoring the unassigned units biases estimation, while including them requires care. We construct an unbiased linear estimator for bipartite experiments with partial assignment. Observations must be reweighted not only by the (centred) share of treated connections among participating ones (the exposure) but also by the number of participating connections, each inverse-weighted by its participation propensity, so that units whose connections are well covered carry proportionally more weight. The resulting estimator, EARL (Exposure- and Allocation-Reweighted Linear), is unbiased, consistent, and asymptotically normal; we devise two asymptotic variance estimators and show that it has minimal variance in a natural class of linear estimators. EARL remains unbiased regardless of the experience the unassigned units receive, as long as they contribute to expected outcomes additively; in particular, they can be allocated to other, non-overlapping tests. Our theory is complemented with a simulation study on two public datasets, in which EARL attains up to six times lower error than the strongest existing baseline and, in some configurations, over an order of magnitude lower than Horvitz-Thompson-style alternatives.","published":"2026-09-25T09:20:06Z","updated":"2026-09-25T09:20:06Z","authors":["Alexey Kurennoy"],"categories":["stat.ME"],"headline":"New method improves analysis of complex marketplace experiments","summary":"Researchers developed EARL, a new statistical tool for experiments where only part of the tested items receive a change, but results are measured on related customer outcomes. EARL adjusts for partial participation by carefully weighting data to prevent bias, offering unbiased and reliable estimates even when some units are unassigned or involved in other tests. Simulations show EARL significantly reduces errors compared to previous methods, supporting its potential for more accurate analysis in these complex testing scenarios."},{"id":"2609.31021v1","title":"Asymptotic completions of preordered semirings","abstract":"The study of preordered semirings is motivated by applications in computer science, graph theory, and information theory, and provides tools for understanding the asymptotic preorder, which compares large powers of a pair of elements. This paper studies sequences which behave approximately as sequences of powers, but are not necessarily equivalent to geometric sequences. Our main result is that preordered semirings admit completions where such sequences, that we call approximately geometric, become equivalent to geometric sequences, and that existing characterizations of the asymptotic preorder extend to the completion. We provide several classes of examples of approximately geometric sequences in the semiring of tensors, and in the semiring of graphs. As a concrete application, we determine the strong converse exponent for binary hypothesis testing with composite Markov hypotheses.","published":"2026-09-25T09:18:45Z","updated":"2026-09-25T09:18:45Z","authors":["Péter Vrana"],"categories":["math.RA","cs.IT","math.CO"],"headline":"Improving Mathematical Tools for Comparing Complex Structures","summary":"Researchers studied mathematical systems called preordered semirings, which are useful in areas like computer science and graph theory to compare how objects behave when combined repeatedly. They focused on special sequences that act like repeated multiplication but aren't exactly the same, and showed these sequences can be treated as if they were standard geometric sequences in an extended framework. This helps extend previous understanding of the system, with examples from tensors and graphs. The researchers also applied their findings to analyze a statistical testing problem known as binary hypothesis testing with composite Markov hypotheses."},{"id":"2609.31020v1","title":"Bayesian inference of non-Maxwellian distribution functions from collective Thomson scattering spectra","abstract":"We investigate Bayesian inference of non-Maxwellian electron distribution functions from collective Thomson scattering (CTS) spectra. We represent distribution functions using multiple candidate models of different complexity and perform inference on synthetic spectra generated from known non-Maxwellian distribution functions. Our analysis demonstrates that a sufficiently flexible model can approximate the overall shape of the ground-truth distribution function. The most plausible model is identified based on model evidence, a statistical measure representing the probability to obtain the observed data given the model. When the candidates include the ground-truth model, the model evidence favors the ground-truth model. Without the ground-truth model as a candidate, the model evidence favors the candidate model with the fewest parameters that adequately approximates the data. The posterior probability density function of the most plausible model reveals the characteristic features and associated uncertainties of the underlying distribution function. The origin of the additional spectral peaks is attributed to a combination of modifications to the dispersion relation and reduced Landau damping. This enables the objective and data-driven identification of distribution functions directly from observed CTS spectra.","published":"2026-09-25T09:18:01Z","updated":"2026-09-25T09:18:01Z","authors":["Kentaro Sakai","ByungJun Kang","Akito Nakano","Takeo Hoshi"],"categories":["physics.plasm-ph","astro-ph.HE"],"headline":"Using Bayesian Methods to Identify Electron Distributions from Spectra","summary":"The researchers studied how to infer the behavior of electrons in plasma from measurements of scattered light, focusing on cases where electrons don't follow a typical pattern. They tested different mathematical models on simulated data to see which best matches the true electron behavior. Their findings show that flexible models can closely approximate these behaviors, and statistical tools can help select the most plausible model and capture uncertainties. This work supports using such data-driven methods to better understand electron distributions from observed spectra."},{"id":"2609.31019v1","title":"Ensembles of Exactly Solved Subsamples for Clusterwise Regression: Trimming Without a Trimming Level","abstract":"Clusterwise least squares partitions regression data into K groups with separate linear fits. We study an ensemble whose base learner is exact: solve the problem to global optimality on each of B random subsamples of size m << n, extend each solution by nearest-surface assignment, align the labels, and combine the replicates by vote or by selection. Each replicate is then an empirical K-quantizer on m points, and the ensemble admits an exact analysis. A vote is correct at a unit once the probability of a clean subsample times the clean-data replicate accuracy exceeds one half, whatever the contaminating values; iterating the ensemble on the units it has not flagged gives a variant that estimates the trimming level rather than requiring it. With up to 20% of gross outliers in the response its worst-case accuracy was 0.89, against 0.80 for trimmed alternation at the true contamination fraction and less at every fixed level tried. Conditionally on the data the replicates are i.i.d., so the vote converges exponentially fast in B to the plurality partition of the replicate law, agreeing with the criterion minimiser outside a boundary set whose size depends on m and the micro-solver, not on B. For the two-group location model, subsamples of order 1/pi_min make a replicate right more often than wrong above a separation threshold. An O(n^3) enumeration gives the exact minimiser for two groups and one covariate, against which the theory is checked. On clean data the ensemble loses to multistart alternation at equal cost.","published":"2026-09-25T09:13:48Z","updated":"2026-09-25T09:13:48Z","authors":["Samir Orujov"],"categories":["stat.CO","stat.ME","stat.ML"],"headline":"New method improves grouping for regression despite data outliers","summary":"Researchers developed a way to divide data into groups for separate trend lines by analyzing many small random samples exactly and combining the results by voting. Their approach can handle up to 20% outliers better than existing trimming methods and estimates needed parameters automatically. The method's accuracy and behavior were tested theoretically and against exact solutions for simple cases, though it performs less well on clean data compared to standard methods."},{"id":"2609.31018v1","title":"Predictive Rolling-Horizon Optimization for Commitment-Aware Model-Parallel Inference under Spatio-Temporal Edge Dynamics","abstract":"Model-parallel inference over dynamic edge systems requires scheduling decisions that account for not only instantaneous resources but also future resource contention and reliable service commitments. Existing edge-inference designs, however, predominantly optimize performance metrics based on current or short-term system states, without explicitly coupling current assignments with future commitment fulfillment. To address this issue, we propose PROMISE, a predictive rolling-horizon framework for commitment-aware model-parallel inference under spatio-temporal edge dynamics. PROMISE jointly models stochastic task generation, mobility-induced communication variations, privacy-aware model partitioning, and load-dependent edge computing capability. We introduce committed completion time (CCT) as an endogenous service decision and formulate joint SD--ES mapping and CCT determination to balance commitment fulfillment and service reward. To address cross-timeslot coupling, PROMISE estimates future task arrivals and computational workloads over an adaptive horizon and embeds them into certainty-equivalent ES-state rollout. A progress-aware surrogate then evaluates feasible current-stage mappings, while the corresponding CCTs are analytically recovered from predicted completion times. Only the first-stage decisions are committed, and the optimization is repeated with newly observed states in a receding-horizon manner. Numerical experiments demonstrate robust scheduling performance under diverse system scales and workload dynamics, while Raspberry-Pi-based experiments validate the practical feasibility and key system characteristics of model-parallel edge inference.","published":"2026-09-25T09:13:40Z","updated":"2026-09-25T09:13:40Z","authors":["Minghui Liwang","Chenxi Xu","Wei Gong","Li Li","Wenbo Zhu","Xinlei Yi","Yuhan Su","Xianbin Wang"],"categories":["cs.DC"],"headline":"New Method Predicts and Schedules Edge Computing Tasks Better","summary":"Researchers developed PROMISE, a system that plans how to split and schedule complex tasks across many edge devices by predicting future demands and resource changes. Unlike previous methods that focus on current system states, PROMISE considers both current and future computing resources and communication variability to meet service time commitments. Their tests, including experiments using Raspberry Pi devices, show PROMISE performs well under changing conditions and is practical for real-world use."},{"id":"2609.31017v1","title":"Resource-Aware Rolling-Horizon Controller for Renewable-Based Virtual Power Plants Providing Automatic Frequency Restoration Reserves","abstract":"Virtual power plants (VPPs) are an effective solution for increasing the penetration of renewable energy sources (RES) in power systems. Moreover, VPPs can provide ancillary services if generators and loads are closely connected and properly coordinated. This aspect has been addressed in several studies, yet they mostly focus on the dispatch optimisation level, leaving real-time operation aspects unresolved. To close this gap, a central controller for VPPs based on a rolling-horizon optimisation is proposed in this work. Its main objective is to maximise revenues while delivering frequency restoration services. A day-ahead optimisation firstly defines the power setpoint and the upward and downward reserves of the VPP. Then, the rolling-horizon optimiser, executed every minute, distributes the power and reserve references between the VPP units to achieve their close tracking while taking into account resource availability (wind speed, irradiance, etc.), operational constraints and battery degradation. The effectiveness of the controller is tested using a two-area interconnected power system under different RES generation and demand profiles. The simulations are performed in MATLAB/Simulink and include detailed dynamic models of the VPP elements, demonstrating the applicability of the algorithm to real systems. The day-ahead and the rolling-horizon optimisation problems are modelled using YALMIP and solved using Gurobi. The obtained results demonstrate the VPP supporting grid frequency restoration by optimally allocating its resources, even under the presence of forecast uncertainty.","published":"2026-09-25T09:12:58Z","updated":"2026-09-25T09:12:58Z","authors":["Marco Vinicio Avendaño-Caiza","Juan Diego Rios-Peñaloza","Javier Roldán-Pérez","José Luis Rodríguez-Amenedo","Milan Prodanovic"],"categories":["eess.SY"],"headline":"New controller helps virtual power plants support the power grid","summary":"Researchers developed a controller for virtual power plants that better manages renewable energy sources to support grid frequency stability. The controller plans power and reserve output one day ahead and then adjusts allocations every minute, considering factors like weather and battery wear. Simulations show this method helps virtual power plants deliver frequency restoration services effectively, even when renewable energy forecasts are uncertain, although results are from simulation models rather than real-world tests."},{"id":"2609.31016v1","title":"Robust Successor Features","abstract":"Generalization in Reinforcement Learning (RL) refers to the ability to execute close-to-optimal policies in unseen tasks after the agent has been trained on a different set of tasks. Building on the seminal work of the successor representation and further adaptations with function approximation, Transfer in RL has traditionally focused on generalizing to tasks that only differ in the reward function. A decade after the introduction of the successor representation, Robust RL emerged simultaneously from several articles in the field of operations research. In Robust RL, the transition kernel is unknown, and the goal is to maximize the expected reward under this uncertainty. Our work unifies these two paradigms through robust successor features, which generalize across both the reward function and the transition kernel, under the assumption that tasks are linear Markov Decision Processes. We derive a bound on Generalized Policy Improvement (GPI) that explicitly quantifies how performance degrades with the mismatch between transition kernels, recovering existing successor-feature guarantees when dynamics are shared. Finally, the generalization capabilities of robust successor features are validated on several grid-based benchmarks and compared to previous alternatives that focus solely on either the reward or the transition kernel.","published":"2026-09-25T09:12:53Z","updated":"2026-09-25T09:12:53Z","authors":["Erik Nikulski","Yamen Habib","Vicenç Gomez","Anders Jonsson","Rubén Moreno-Bote","Javier Segovia-Aguas"],"categories":["cs.LG"],"headline":"New Method Enhances Learning Across Different Tasks and Environments","summary":"Researchers studied how to help AI systems perform well on new tasks even when both the goals and the rules of the environment change. They combined two existing ideas—one that focuses on changing goals and another on uncertain environments—into a new approach called robust successor features. Their method offers a way to understand and limit performance drops when environments differ. They tested this approach on simple simulated tasks and compared it to older methods, showing its ability to handle more types of changes."},{"id":"2609.31015v1","title":"Binary magnetism and directional magnon transport in alkali-doped CrI$_3$","abstract":"The recent realization of two-dimensional (2D) magnets, with CrI$_3$ as a pioneering example, has opened new avenues in the fields of 2D materials and magnetism. This breakthrough has been followed by extensive efforts to manipulate and exploit their magnetic properties. In this work, we investigate the adsorption of alkali-metal atoms as a route to control the magnetic behavior of monolayer CrI$_3$. We show that, upon adsorption, alkali-metal atoms donate an electron to the CrI$_3$ layer, leading to the formation of inequivalent Cr sites, with one Cr atom exhibiting an enhanced magnetic moment of $4~μ_B$, while the other retains its typical $3~μ_B$ moment. These modifications significantly alter the magnetic exchange interactions, including the emergence of anisotropic exchange and Dzyaloshinskii--Moriya interactions (DMI). Consequently, the doped systems exhibit non-collinear magnetic ground states, modified temperature-dependent magnetism, and anisotropic spin-wave propagation, with a preferred propagation direction that becomes increasingly pronounced with increasing dopant size. Furthermore, an asymmetry between spin-wave propagation in opposite directions is observed, giving rise to a diode-like effect, particularly for heavier dopants. This behavior is attributed to the enhanced DMI, which breaks the symmetry of the magnon dispersion. These results demonstrate that alkali-metal doping provides an effective route to tune anisotropic and nonreciprocal magnonic properties in two-dimensional magnetic materials.","published":"2026-09-25T09:10:15Z","updated":"2026-09-25T09:10:15Z","authors":["Cihan Bacaksiz","Maarten Soenen","Denis Sabani","Rai Maciel de Menezes","Milorad V. Milosevic"],"categories":["cond-mat.mtrl-sci"],"headline":"Tuning Magnetism in 2D CrI3 with Alkali-Metal Atoms","summary":"Researchers studied how adding alkali-metal atoms to a single layer of the magnetic material CrI3 changes its magnetic properties. They found that these atoms donate electrons, creating differences between chromium atoms and leading to new types of magnetic interactions and non-aligned magnetic states. Their simulations show this causes spin waves to travel unevenly in different directions, like a diode effect, especially with heavier metals. These findings suggest a way to control magnetic behavior in 2D materials, but the results are based on theoretical modeling."},{"id":"2609.31014v1","title":"Co-design of trajectory and morphology for a vertical jump-climbing robot","abstract":"Animals such as squirrels and even bears have adapted to rapidly climb up trees and other complex vertical terrain, but achieving comparable agility has been a challenge for climbing robots. Existing robots often use walking gaits and move conservatively to stay in contact with the surface, which limits the range of dynamic maneuvers. In this paper we present a 290 g robot that, to our knowledge, is the first to climb vertically by bounding (with an aerial phase). We leverage a co-design workflow, in which the morphology and trajectory are jointly optimized for fast locomotion, subject to adhesion force limitations seen in spined grippers. The resulting trajectory includes a rapid maneuver that launches the robot vertically, and an aerial reorientation that brings the front grippers back to the surface using the rear leg as an inertial tail. We evaluate the resulting jump forces in 2D force space and demonstrate that the optimized morphology is capable of continuous climbing at a speed of 0.375 m/s (1.97 body lengths/s), and can also achieve ground locomotion and transition to a vertical surface. Our work proposes design insights for the jump-climbing maneuver and serves as an important step toward creating climbing robots with agility on par with that of animals.","published":"2026-09-25T09:07:48Z","updated":"2026-09-25T09:07:48Z","authors":["Christopher Y. Xu","Elliot W. Hawkes"],"categories":["cs.RO"],"headline":"New small robot can jump-climb vertical surfaces like animals","summary":"Researchers developed a lightweight robot that climbs vertical surfaces by jumping, similar to squirrels or bears. They designed the robot's shape and movement together to optimize fast climbing while sticking to surfaces using special grippers. Testing showed the robot can climb continuously at nearly two body lengths per second, move on the ground, and transition to climbing. The researchers present this as a step toward agile climbing robots, but the results are based on their specific design and testing conditions."},{"id":"2609.31013v1","title":"Same Text, Different Numbers: The Divergence of LLM-Based Measures","abstract":"Researchers increasingly use generative large language models (LLMs) to convert corporate text into empirical variables. We examine the extent to which LLM-based textual measures are invariant to model choice using thirteen measures, including sentiment, management clarity, uncertainty, answer specificity, and climate and political risk. Seven LLMs from different providers score earnings call transcripts of S&P 500 companies on these constructs. Cross-model rank correlations average only 0.52, and transcript-level differences common across providers account for only 34% of total score variation. Cross-model disagreement does not predict subsequent analyst or market disagreement, consistent with a substantial model-specific component rather than common ambiguity in the underlying disclosure. Model choice significantly affects downstream inference, with coefficient magnitudes, signs, and statistical significance varying substantially across models. Averaging across providers makes transcript rankings more stable for most constructs, but score levels remain sensitive to the models included in the ensemble. LLM-generated variables should therefore be treated as model-contingent measurements and validated across providers.","published":"2026-09-25T09:04:10Z","updated":"2026-09-25T09:04:10Z","authors":["Hamid Boustanifar","Sasan Mansouri"],"categories":["cs.AI","cs.CL","q-fin.GN","q-fin.RM"],"headline":"Large Language Models Give Different Scores for the Same Company Texts","summary":"Researchers studied how different large language models (LLMs) analyze corporate texts like earnings calls to measure things like sentiment and risk. They found that scores varied a lot depending on which model was used, and these differences didn't match how analysts or markets viewed the text. The researchers warn that results based on LLMs can change substantially with the model choice, so users should check findings across multiple models before trusting them fully."},{"id":"2609.31012v1","title":"Breaking the Black Box: Byte-Level Boundary Inference of Real-World Antivirus Systems","abstract":"Existing approaches for understanding the detection logic of real-world antivirus (AV) software infer only binary malware/benign decisions from black-box queries, providing limited insight into the fine-grained decision-critical regions that govern AV detection. In this paper, we present \\textbf{AVHunter}, the first framework for inferring byte-level decision-critical regions of real-world AV products under a black-box threat model. AVHunter constructs the first large-scale Byte-Level AV Boundary Dataset (BABD) by systematically probing 11 real-world AV products, revealing that modern AV detections are largely associated with a small number of compact decision-critical byte regions. Leveraging BABD, AVHunter trains AV-specific models that not only reproduce binary AV decisions, but also localize the decision-critical byte regions underlying these decisions, achieving an average boundary prediction recall of 85.07% while maintaining 97.43% detection agreement with the target AVs. We further validate that the predicted regions capture genuine AV decision knowledge through boundary-guided malware evasion, false-positive induction on benign executables, and a seven-month longitudinal study demonstrating that the inferred regions remain largely stable as AV products evolve. Overall, AVHunter moves beyond conventional binary-label AV modeling by enabling fine-grained boundary-region localization and revealing a new form of AV knowledge leakage with important implications for malware analysis, AV security, and boundary-aware defenses.","published":"2026-09-25T09:02:45Z","updated":"2026-09-25T09:02:45Z","authors":["Jieshuai Yang","Zhi Wang","Yan Jia","Zhenhua Wu","Jianfei Tang","Chenbin Su","Jingwei Ye","Jianwen Tian","Wanpeng Li"],"categories":["cs.CR"],"headline":"New Tool Finds Exact Code Areas Triggering Antivirus Alerts","summary":"Researchers developed AVHunter, a system that probes antivirus software to identify specific small code parts causing malware detection. By testing 11 antivirus products, they created a large dataset and trained models that can predict these fine-grained detection regions with about 85% accuracy. Their work reveals detailed decision patterns behind antivirus alerts, but results are based on probing and modeling rather than direct access to antivirus internals."},{"id":"2609.31011v1","title":"Dynamic Task and Resource Scheduling Towards Space-Air-Ground-Sea Integrated Network","abstract":"In the context of 6G ubiquitous connectivity, the space-air-ground-sea integrated network (SAGSIN) emerges as a new paradigm for pervasive service provisioning. To support expanding maritime activities in infrastructure-scarce ocean areas, we propose an innovative dynamic task and resource scheduling approach for SAGSIN to deliver computing services for vessels. It integrates broad-coverage satellites, relay-capable high-altitude platform (HAP), energy-sufficient coastal base station (BS), and flexibly deployed uncrewed aerial vehicles (UAVs) to accommodate wide-area, highly mobile, and sustained maritime services. To address the challenge of task scheduling across four layers, a dynamic task offloading algorithm is developed. It steers task flows toward servers with light loads, strong computing capabilities, and high-rate links based on real-time system states to reduce task execution delay, integrating an anticipatory satellite handover strategy to mitigate post-handover congestion and improving satellite resource utilization. Considering the limited endurance of UAVs, we impose residual energy constraints to ensure task backlog handover and safe return. Furthermore, the UAV-BS bandwidth allocation, UAV trajectories, and computing resource allocation are jointly optimized to enhance the connectivity among low-altitude devices and accelerate task completion. Simulation results validate the proposed method's superior adaptability to system resource variations during task execution in complex maritime environments, achieving at least a 23% reduction in average task delay over benchmarks.","published":"2026-09-25T08:58:40Z","updated":"2026-09-25T08:58:40Z","authors":["Yufei Ye","Shijian Gao","Xinhu Zheng","Liuqing Yang"],"categories":["cs.NI"],"headline":"New Scheduling Method Aims to Improve Maritime Connectivity Services","summary":"Researchers studied a network combining satellites, airborne platforms, coastal stations, and drones to support computing services for ships in ocean areas with little infrastructure. They developed a method to dynamically allocate tasks and resources across these different layers, aiming to reduce delays and handle limited drone energy. Their simulations showed the approach adapts well to changing conditions and could cut average task delay by about 23% compared to existing methods. These findings are based on computer simulations in complex maritime scenarios."},{"id":"2609.31010v1","title":"Identifiability and Joint Recovery of a Structured Signal Ensemble From Sub-Nyquist Samples","abstract":"Efficient acquisition of correlated continuous-time signals is critical in applications such as distributed sensor networks and array processing. However, existing correlation models often fail to capture the shared structure of physical signals measured in close proximity. In this paper, we model each signal in an ensemble as the sum of a hidden common lowpass component and a signal-specific innovation highpass component with disjoint spectral supports, where the common bandwidth is unknown. We first derive theoretical identifiability conditions guaranteeing unique decomposition and reconstruction from subsampled observations. Further, we propose a practical recovery algorithm and a joint reconstruction framework that leverages parametric structured dictionaries parameterized by the unknown common bandwidth. Numerical experiments on an ensemble of four signals demonstrate exact reconstruction with a 25% aggregate sampling rate reduction under identifiability conditions, and robust recovery with up to a 74% rate reduction when all channels are sampled below the Nyquist rate, significantly reducing data acquisition and hardware overhead.","published":"2026-09-25T08:58:20Z","updated":"2026-09-25T08:58:20Z","authors":["Kumari Priyanka","Satish Mulleti"],"categories":["eess.SP"],"headline":"New Method Recovers Linked Signals Using Less Data","summary":"Researchers studied how to efficiently capture related continuous signals, like in sensor networks, by modeling each as a mix of a shared smooth part and individual detailed parts. They established conditions that guarantee signals can be uniquely separated and reconstructed even when sampled below usual rates. Their algorithm, tested on four signals, showed exact recovery using 25% less data and reliable recovery with up to 74% less, suggesting potential for reduced data collection needs. However, results are based on specific model assumptions and simulations."},{"id":"2609.31009v1","title":"G$^2$PTQ: Improving LLM Post-Training Quantization with Generalized Gradient Compensation","abstract":"Post-training quantization (PTQ) is a practical approach to reducing the memory and computational footprint of large language models (LLMs) without retraining. GPTQ-based methods have become the de facto standard, yet they suffer from two complementary limitations. Methods with local, layer-wise objectives lack global supervision; while methods with global objectives fix their Hessian estimates at the start and ignore first-order gradients, so their guidance grows stale as quantization proceeds. This paper presents G$^2$PTQ, a unified PTQ framework with Generalized Gradient Compensation that integrates both first- and second-order information under a globally supervised, block-wise optimization objective. By refreshing gradient and Hessian estimates before quantizing each Transformer block, G$^2$PTQ avoids the staleness of prior global methods. Furthermore, to stabilize the exact first-order compensation, we introduce a trust-region scaling mechanism that dynamically bounds the gradient step to prevent exploding weight updates. Finally, we derive efficient implementations for block-wise Hessian approximation and exact gradient compensation. Experimental results on various model families and bit-widths demonstrate that G$^2$PTQ enables better alignment with the full-precision model, outperforming state-of-the-art baselines. Code is available at: https://github.com/G2PTQ/G2PTQ.","published":"2026-09-25T08:53:14Z","updated":"2026-09-25T08:53:14Z","authors":["Ruikang Liu","Haoli Bai","Yuxuan Sun","Qian Zhang","Wenzheng Cai","Yanqi Hao","Feiyu Wang","Weidong Zhong","Zhuang Wang","Tong Yang","Xiangsheng Zhou"],"categories":["cs.CL","cs.AI"],"headline":"New Method Helps Shrink Large Language Models Without Retraining","summary":"Researchers developed G²PTQ, a technique to reduce the size and computing needs of large language models without retraining them. Their approach improves on existing methods by updating important mathematical estimates throughout the process to better preserve the models' accuracy. Experiments show it outperforms previous standard approaches, but results are based on specific testing scenarios described in their paper."},{"id":"2609.31008v1","title":"Quadruped Obstacle Avoidance and Footstep Planning with Distributed Low-cost Time-of-Flight Sensors","abstract":"Quadruped robots typically rely on depth cameras and LiDAR sensors to map their local environment. However, these sensors have limited close-range coverage, are relatively expensive, and consume significant power. This study investigates whether distributed Time-of-Flight (ToF) sensors can serve as a low-cost alternative to depth cameras for near-field terrain mapping for locomotion and local navigation. We designed a distributed ToF sensing architecture for the ANYbotics ANYmal quadruped, assessed its environment reconstruction accuracy, and benchmarked it against depth cameras for terrain mapping and obstacle avoidance. Distributing these sensors around the robot can also avoid the blind spots of traditional sensors. Our results show that, despite their low resolution and higher measurement noise, distributed ToF sensors can support reliable perceptual locomotion with centimeter-level local mapping accuracy. The proposed sensing strategy provides sufficient geometric information for near-field obstacle avoidance and footstep planning, at substantially lower cost, energy consumption, and system complexity than depth cameras.","published":"2026-09-25T08:52:16Z","updated":"2026-09-25T08:52:16Z","authors":["Giammarco Caroleo","Timothée Mahamoodally","Matteo Manzardo","Jin Jin","Marco Pontin","Matias Mattamala","Renato Vidoni","Perla Maiolino","Maurice Fallon"],"categories":["cs.RO"],"headline":"Low-Cost Sensors Help Robots Avoid Obstacles and Plan Steps","summary":"Researchers explored using affordable Time-of-Flight (ToF) sensors placed around a four-legged robot to map nearby terrain for movement and obstacle avoidance. They compared this setup to standard depth cameras and found that despite lower resolution and more noise, these sensors provided accurate enough information for reliable navigation and foot placement. The study suggests that ToF sensors can reduce cost, energy use, and complexity while enabling effective near-field environment awareness in quadruped robots."},{"id":"2609.31007v1","title":"Incipit: Axiom-Grounded Scaffolding for Human-AI Literary Creation","abstract":"Large language models can produce fluent prose from short prompts, but direct prompt-to-text interaction gives writers limited access to the assumptions that shape a long narrative. We present Incipit, an implemented research prototype that introduces an explicit planning layer between a writer's intent and generated prose. This layer is grounded in literary axioms, defined as curated and reusable propositions about human experience and narrative craft. The prototype connects a knowledge base of 1455 axioms and 472 typed relationships with a five-round direction dialogue, a retrieval-and-selection pipeline, and a three-level blueprint covering creative premises, story beats and character arcs, and chapter outlines. Writers can inspect and edit these structures before using them as context for scene generation. Additional modules support real-event abstraction and five-dimensional diagnostic feedback. We describe the system design rationale, data flow, implementation boundaries, and a worked design example. As no controlled user study or independently rated output study has yet been completed, we do not claim that the system improves literary quality. Instead, we outline a future preregistered evaluation designed to distinguish the contribution of axiom grounding from that of hierarchical planning. The paper contributes a concrete architecture for using literary knowledge as an inspectable coordination object in human-AI writing.","published":"2026-09-25T08:51:55Z","updated":"2026-09-25T08:51:55Z","authors":["Qiang Liu","Chunyi Zhao"],"categories":["cs.CY"],"headline":"New Tool Helps Writers Plan Stories with AI Using Literary Ideas","summary":"Researchers created Incipit, a prototype system that adds a planning step between a writer's ideas and AI-generated text, based on 1455 literary principles about storytelling and human experience. Writers interact with these plans to shape story premises, characters, and chapters before generating scenes. While the tool offers new ways to guide AI writing, the researchers have not yet tested whether it improves story quality, and they plan future studies to explore its effects."},{"id":"2609.31006v1","title":"Electrically controlled spin-splitting and asymmetric tunnel magnetoresistance in anti-altermagnets","abstract":"Anti-altermagnets (AAMs) are a recently identified class of layered magnetic materials where opposite spin-splitting in adjacent layers creates a globally spin-degenerate band structure, rendering conventional spectroscopic detection highly difficult. In this paper, we theoretically demonstrate an all-electrical method to manipulate and probe this hidden magnetic order using a dual-gated transport junction. By applying a perpendicular displacement field, we break the spatial inversion symmetry of the lattice, explicitly lifting the global spin degeneracy. We attach ferromagnetic (FM) leads on either side of the AAM. Using quantum transport calculations, we show that this gate-induced spin-splitting manifests as a strongly asymmetric tunnel magnetoresistance (TMR) as a function of the lead magnetization. We identify specific crystallographic orientations where the TMR retains its symmetry despite the fully split bands, a direct consequence of exact momentum-space compensation. We further reveal that Rashba spin-orbit coupling guarantees robust, highly directional transport asymmetries even in the absence of an explicit chemical potential mismatch between the layers. Our findings establish a clear, electrically tunable framework for exploiting AAMs in next-generation spintronic architectures.","published":"2026-09-25T08:51:46Z","updated":"2026-09-25T08:51:46Z","authors":["Abhiram Soori"],"categories":["cond-mat.mes-hall","cond-mat.mtrl-sci"],"headline":"New method to electrically control hidden magnetism in special materials","summary":"Researchers studied a type of layered magnetic material called anti-altermagnets, where usual detection methods struggle due to hidden spin properties. They theoretically showed that applying an electric field can reveal and control these hidden magnetic features by breaking certain symmetries, causing changes in electrical resistance that depend on magnetization directions. Their work, based on quantum transport calculations, highlights how electric control can tune magnetic behavior, although these results are currently theoretical and rely on specific material orientations and conditions."},{"id":"2609.31005v1","title":"TRACKGRAPH: Online Open-Vocabulary 3D Scene Graphs via Image-Space Tracking","abstract":"Open-vocabulary 3D maps enable robots to reason about previously unknown environments using natural language. However, existing systems typically segment every incoming image, associate detections with persistent 3D segments, and frequently perform costly Vision-Language (VL) inference. We present TRACKGRAPH, an online open-vocabulary system that maintains short-term 2D mask identity directly in the image stream before fusing segments into 3D. FastSAM masks and CLIP features are computed at sparse keyframes, while dense DINOv3 features are used to propagate masks at a high rate in between. The resulting tracked masks are fused into a class-agnostic 3D segment layer within a hierarchical scene graph, with 3D association handling tracking interruptions and long-term revisits. Compact multi-view CLIP embeddings enable open-vocabulary retrieval. Across Replica, ScanNet++, and HM3D, TRACKGRAPH achieves competitive open-vocabulary segmentation and retrieval against state-of-the-art mapping methods, including the highest synonym frequency on Replica (0.50). On the same NVIDIA A100, it is 1.7x faster and uses 3.3x less GPU memory than ViT-H OVI-MAP. Real-world quadruped deployments demonstrate onboard scene graph construction and object search at 7.5Hz, while recorded drone data is used to test the method under aerial viewpoints.","published":"2026-09-25T08:51:33Z","updated":"2026-09-25T08:51:33Z","authors":["Peder Borge Hellesylt","Albert Gassol Puigjaner","Kostas Alexis","Annette Stahl"],"categories":["cs.CV","cs.RO"],"headline":"New Method Tracks 3D Scenes Using Fewer Computing Resources","summary":"Researchers developed TRACKGRAPH, a system that helps robots understand and track objects in 3D spaces using natural language descriptions. Instead of analyzing every image fully, it processes key images with detailed analysis and tracks objects between them efficiently. Tests show TRACKGRAPH matches or beats current methods in identifying objects while using less computing power. The researchers demonstrated it working on real robots and drones, but results are based on specific datasets and simulations."},{"id":"2609.31004v1","title":"GitHub Engagement Signals for CVE Prioritization: The GitHub Popularity Metric (GPM)","abstract":"Attackers can compromise multiple systems with a single vulnerability, while defenders need to fix all security weaknesses in their systems. This asymmetry puts defenders at a disadvantage. Security vulnerabilities are found at an alarming rate, and patching vulnerabilities is costly and time-consuming; thus, vulnerability prioritization is a must and a time-critical challenge. Many prioritization metrics, such as the CVSS, EPSS, KEV, and SSVC, are currently used, each with different pros and cons, such as openness, degree of automation, time-criticality, coverage, and need for expert input. In this work, we propose the GitHub Popularity Metric (GPM), a fully open, publicly computable prioritization metric based on the popularity of exploits in GitHub repositories. We use GitHub features such as the number of stars, forks, and related unique users to create a metric that indicates the popularity of CVEs across different time frames, both relative to the current time and historically. We compare the proposed metric with various existing vulnerability prioritization metrics and known exploited vulnerabilities and demonstrate that it provides tangible insights for defenders, identifying unique CVEs and inconsistencies in existing methods. The GPM was integrated into the EPSS version 5. \\noindent\\textbf{Note.} A demonstration based on this work has been accepted to the demo track of the ACM Conference on Computer and Communications Security (CCS) 2026.","published":"2026-09-25T08:50:56Z","updated":"2026-09-25T08:50:56Z","authors":["Jafar Akhoundali","Kristian Rietveld","Olga Gadyatskaya"],"categories":["cs.CR"],"headline":"New GitHub-Based Metric Helps Prioritize Software Vulnerabilities","summary":"Security weaknesses in software are found rapidly, making it hard for defenders to fix them all quickly. Researchers developed the GitHub Popularity Metric (GPM), which uses public GitHub data like stars and forks to measure how popular exploits are over time. They compared GPM to existing methods and found it offers useful new insights for prioritizing threats. The work is theoretical and based on data analysis, and has been incorporated into a known vulnerability scoring system."},{"id":"2609.31003v1","title":"Weaponizing Ground Truth: Data Poisoning Attacks by Exploiting Boundary Misalignment Between Antivirus Software and Learning-Based Detectors","abstract":"Machine-learning (ML)-based malware detectors are commonly trained using labels obtained from antivirus (AV) engines and aggregation services (e.g., VirusTotal). This practice assumes AV-generated labels provide reliable supervision. However, small byte-level modifications can substantially alter AV verdicts while leaving the representations perceived by downstream ML detectors largely unchanged, producing label-feature inconsistencies that can contaminate training datasets and create poisoning opportunities for ML-based malware detection. We present Bi-Iocane, a black-box poisoning framework that exploits the reliance of malware-labeling pipelines on AV-generated labels. Bi-Iocane identifies AV-sensitive bytes and modifies them to induce label changes. It rewrites such bytes in malware to obtain benign labels (evasion-oriented poisoning) and injects malware-associated byte patterns into benign software to obtain malicious labels (defamation-oriented poisoning). These poisoned samples and their lightly modified variants corrupt training data and cause selected targets to be misclassified. We evaluate Bi-Iocane with 13 AV engines simulating AV aggregation services and eight ML detectors. For 30 malware and 30 benign clean targets, Bi-Iocane combines AV-specific manipulations to generate malware-to-benign and benign-to-malware poisoned samples whose all tested AV-based labels are flipped. After these poisoned samples and variants are used for downstream training, the resulting ML models misclassify 92.08% of the original clean targets on average with only a 0.06\\% poisoning budget per target. Meanwhile, the poisoned models largely preserve clean-set performance, and six evaluated poisoning defenses show only limited mitigation. VirusTotal evaluation further confirms practical defamation risk and reveals potential evasion risk in real-world AV-to-ML labeling supply chains.","published":"2026-09-25T08:48:54Z","updated":"2026-09-25T08:48:54Z","authors":["Jieshuai Yang","Zhi Wang","Yan Jia","Zhenhua Wu","Jianfei Tang","Chenbin Su","Jingwei Ye","Jianwen Tian","Wanpeng Li"],"categories":["cs.CR"],"headline":"Data Poisoning Risks in AI Malware Detection Using Antivirus Labels","summary":"Researchers studied how antivirus (AV) software labels used to train AI malware detectors can be manipulated. They created a method called Bi-Iocane that changes tiny parts of software files to flip AV labels, causing AI detectors to learn incorrect information. Testing showed that these subtle changes can fool AI models to misclassify malware and clean software, while maintaining overall detection quality. The findings highlight vulnerabilities in relying on AV labels for training AI malware detectors and show current defenses may not fully prevent such attacks."},{"id":"2609.31002v1","title":"ZooWork-ShopRanker: An Open, Preference-Aligned E-Commerce Reranker","abstract":"Open rerankers trained for general web retrieval transfer imperfectly to e-commerce, where ranking decisions depend not only on topical relevance but also on user preferences, product constraints, and comparative product fit. These preference signals are difficult to supervise at scale: real search traffic provides authentic queries and candidates but no clean pairwise labels. We present ZooWork-ShopRanker, a family of e-commerce rerankers (0.6B, 4B, and 8B) aligned to judge-labeled shopping preference. Training pairs are labeled by a panel of reasoning large language models (LLMs) from different families acting as a preference oracle, with position-debiased judgments and agreement tiers, and the rerankers are trained on these labels. The aligned 8B flagship then serves as a distillation teacher for the efficient 4B and 0.6B models, which are fit to its scores and sharpened on judged pairs. To measure progress, we introduce ShopRank-Bench, a contamination-limited benchmark of ~10,000 private-traffic preference pairs in both text formats, tiered by how many judge families committed to each label. ZooWork-ShopRanker-8B and -4B significantly outperform the strongest open reranker baseline, every model significantly beats its own un-aligned base, and ZooWork-ShopRanker-0.6B beats its size peer; the gains hold in both formats and extend to common MTEB benchmarks. We release the models and the dual-format ShopRank-Bench to facilitate further research.","published":"2026-09-25T08:47:19Z","updated":"2026-09-25T08:47:19Z","authors":["Siqiao Xue","Shuxuan Liu","Ning Hu"],"categories":["cs.CL"],"headline":"New AI models improve e-commerce product rankings using preferences","summary":"Researchers developed ZooWork-ShopRanker, a set of AI models designed to better rank products in online shopping by considering user preferences alongside relevance. They used multiple large language models to create training data reflecting shopping preferences without needing extensive human labeling. Their results, measured on a new benchmark called ShopRank-Bench, show these models outperform existing methods, though the study is based on controlled evaluation data rather than real-world deployment."},{"id":"2609.31001v1","title":"Multiorbital periodic Anderson model for CeRh2As2","abstract":"CeRh2As2 is a heavy-fermion superconductor that exhibits distinct low- and high-field superconducting phases, widely interpreted as reflecting a field-induced change in the parity of the order parameter. However, the superconductivity develops in close proximity to magnetic order whose microscopic origin and nature remain unclear. Moreover, the low-lying crystal-electric-field-split doublets of the Ce 4f electrons appear to realize an unusual \"quasi-quartet\" structure. This has motivated proposals that quadrupolar degrees of freedom play an important role in the low-temperature physics. Here we develop a microscopic heavy-fermion description of CeRh2As2 that incorporates both the quasi-quartet structure and the nonsymmorphic symmetry. Within a multiorbital periodic Anderson model, we include the lowest-lying $Γ_7$ and $Γ_6$ doublets of the Ce 4f electrons, and allow for nonlocal hybridization with the conduction d electrons. Single occupancy of the 4f states is enforced within the rotationally invariant slave-boson formalism. Adopting a mean-field treatment, we first show that the effective low-energy Hamiltonian is accurately captured by a single-orbital tight-binding model, justifying the minimal descriptions widely employed in the literature. Allowing for a magnetic ground state, we find that our model displays metallic $\\mathbf{Q}=(π,π)$ antiferromagnetic order over a wide parameter regime. Ferromagnetic solutions also occur, but in a smaller region of parameter space and with greater sensitivity to carrier concentration and the form of the f-d hybridization. Quadrupolar moments appear only within magnetically ordered phases and are generally small, becoming substantial only when the $Γ_7$ and $Γ_6$ doublets are nearly degenerate. Our results provide a microscopic basis for understanding the itinerant heavy-fermion physics and ordered phases of CeRh2As2.","published":"2026-09-25T08:45:35Z","updated":"2026-09-25T08:45:35Z","authors":["Nico A. Hackner","Changhee Lee","P. M. R. Brydon"],"categories":["cond-mat.str-el","cond-mat.supr-con"],"headline":"New Model Explains Complex Behavior of CeRh2As2 Superconductor","summary":"Researchers developed a detailed theoretical model to better understand CeRh2As2, a material with unusual superconducting and magnetic properties. They included specific electron states and symmetries in their calculations to describe its magnetic order and electron interactions. Their findings suggest the material's behavior can mostly be captured by simpler models but also highlight conditions under which magnetic and quadrupolar (shape-related) electron features become significant. These results help explain the material’s complex low-temperature physics, but are based on theoretical simulations rather than experiments."},{"id":"2609.31000v1","title":"Gravitational wave radiation from periodic orbits around a black hole in Einstein-Maxwell-scalar theory","abstract":"In this paper, we investigate the periodic orbits around a black hole (BH) in Einstein-Maxwell-scalar (EMS) theory and the corresponding gravitational waveforms. First, we study the test particle dynamics around the BH by using the Lagrangian formalism. Using the corresponding conditions, we explore the effect of the spacetime parameters on the marginally bound orbit (MBO) and innermost stable circular orbit (ISCO). The results indicate that MBO and ISCO quantities decrease with the rise of the $β$ parameter and BH charge $Q$, and vice versa for the $γ$ parameter. Subsequently, we plot the different zoom-whirl orbits around the BH in EMS theory by finding the test particle's corresponding energy and angular momentum. Finally, the gravitational waveforms of the extreme mass ratio inspirals, which contain a supermassive BH in EMS theory and a stellar mass object, are demonstrated. It is shown that the zoom-whirl orbits shrink as the BH charge increases, and, as a result, gravitational waveforms shift towards shorter times.","published":"2026-09-25T08:44:10Z","updated":"2026-09-25T08:44:10Z","authors":["Mirzabek Alloqulov","Xojibonu Abdurashidova","Sanjar Shaymatov","Raxmatillo Karimov"],"categories":["gr-qc"],"headline":"Studying Black Hole Orbits and Gravitational Waves in a New Theory","summary":"Researchers studied how test particles move around black holes in a theory combining gravity, electromagnetism, and a scalar field. They examined specific orbits and how they change with black hole properties, finding that certain orbits get smaller as black hole charge rises. They also modeled gravitational waves produced by a small object orbiting a supermassive black hole in this theory, showing that wave patterns shorten when black hole charge increases. The findings are based on theoretical analysis and simulations within this specific gravity model."},{"id":"2609.30999v1","title":"An Exact Spherical-Solid Mean Formula and Vanishing Criteria for Lelong Numbers","abstract":"We establish an exact additive formula expressing the Lelong number of a plurisubharmonic function in terms of the limiting difference between its spherical and solid means. Building on this identity, we derive several equivalent criteria for the vanishing of the Lelong number, formulated in terms of fixed-scale increments and the asymptotic agreement of radial maxima, spherical and solid means, and radial regularizations. We also give a pointwise $L^1$ mean-oscillation formulation of the VMO characterization for plurisubharmonic functions with zero Lelong number, together with a streamlined proof of the corresponding global result of Biard and Wu. Finally, we establish a nonlinear characterization in terms of the relative extremal depth and the Bedford-Taylor capacity of deep sublevel sets.","published":"2026-09-25T08:42:44Z","updated":"2026-09-25T08:42:44Z","authors":["Fusheng Deng","Yizhe Yang","Zhanpeng You"],"categories":["math.CV"],"headline":"New Formula Illuminates Key Measure for Complex Functions","summary":"Researchers developed an exact formula relating a key measure called the Lelong number, used to study complex functions, to specific average values measured in spherical and solid regions. They also identified several criteria that determine when this measure vanishes, based on how these averages behave, and provided new ways to describe such functions more precisely. Their findings rely on theoretical analysis of these mathematical properties and build on earlier work, without testing practical applications."},{"id":"2609.30998v1","title":"PipeDRAM: A Data-Transposition-Free In-DRAM Architecture with Hardware/Software Pipelining","abstract":"Processing-using-DRAM (PUD) architectures exploit the analog operational properties of DRAM to perform bulk bitwise Boolean and arithmetic operations inside memory arrays by organizing data in a vertical layout, where operand bits are stacked along DRAM columns. However, modern computing systems natively employ a horizontal data layout that preserves the cache line abstraction, leverages spatial locality in row buffers, and enables high memory throughput. This fundamental mismatch forces existing PUD architectures to frequently perform data layout transformations between horizontal and vertical formats, incurring significant performance, energy, and system integration overheads. Our goal is to eliminate data transposition overheads in PUD systems at low cost. To this end, we propose PipeDRAM, a PUD architecture that eliminates the need for runtime data layout transformation, enabling PUD operations directly over horizontally laid-out data. PipeDRAM's key ideas are to (i) deterministically reorganize bits inside each memory request to enable a PUD-friendly data placement within a DRAM array in a horizontal data layout, and (ii) employ a pipeline-based execution model that overlaps bit-dependent and bit-independent in-DRAM operations to exploit bit-level parallelism across the memory array. We compare PipeDRAM to different computing platforms. PipeDRAM provides (i) 11.8x, 11.8x, and 80.4x higher performance and (ii) 25.4x, 3.0x, and 38.0x lower energy consumption than three state-of-the-art PUD systems. PipeDRAM incurs low area cost on top of a DRAM chip (1.86%) and CPU die (0.05%). To enable further research on PUD systems, we open-source PipeDRAM at https://github.com/CMU-SAFARI/PipeDRAM.","published":"2026-09-25T08:42:42Z","updated":"2026-09-25T08:42:42Z","authors":["Geraldo F. Oliveira","Ataberk Olgun","Ismail Emir Yüksel","F. Nisa Bostancı","Pedro H. E. Becker","Mohammad Sadrosadati","Saugata Ghose","Juan Gómez-Luna","Onur Mutlu"],"categories":["cs.AR"],"headline":"New DRAM Design Speeds Up Memory Computations Without Extra Data Moves","summary":"Current DRAM-based computing systems slow down because they have to rearrange data layouts frequently to perform calculations inside memory. Researchers created PipeDRAM, a new design that lets these calculations happen directly on the usual horizontal data layout, avoiding time-consuming data rearrangements. Their approach reorganizes bits within each memory request and uses a pipeline method to speed up processing. According to their tests, PipeDRAM is faster and more energy-efficient than current systems, with only a small increase in chip area."},{"id":"2609.30997v1","title":"Can Pixels Alone Reveal Image Origin? Minimax Limits and Learnable Interfaces for Passive Provenance","abstract":"Passive image provenance asks whether pixels alone can reveal where an image came from: a human, an aggregate AI class, or a particular generator. This becomes a robustness problem once a source image can be edited before the verifier sees it. We study the problem as source--target verification under adversarial distribution shift. Our first result gives the exact best-case limit for any image-only verifier: the largest robust target-acceptance gap equals the minimum total-variation distance between the target distribution and the set of attacked source distributions. This quantity depends on the source, target, and edit class, not on the verifier architecture. Our second result explains why deployed public verifiers can fail before this statistical limit is reached. If the verifier can be emulated on the attack region to error $\\varepsilon$, then a surrogate black-box attack reaches target acceptance within $2\\varepsilon$ plus optimization error of the white-box optimum; score-revealing logistic and softmax heads over public features are identifiable, and approximate score access gives stable recovery bounds. A finite-state experiment checks the minimax identity where both sides are computable. On same-prompt real/diffusion benchmarks, the evaluated public CLIP verifiers fail under targeted pixel attacks, while a ResNet-18 victim exhibits partial fake-to-real transfer. Binary feedback with abstention reduces measured attack success, but positive empirical gap upper bounds do not establish robustness. These results motivate separate evaluation of the source--target statistical ceiling and the information released by a deployed verifier.","published":"2026-09-25T08:42:34Z","updated":"2026-09-25T08:42:34Z","authors":["Kai Yao"],"categories":["cs.CR","cs.AI","cs.CV","cs.LG"],"headline":"Can Pixels Alone Show Where an Image Comes From?","summary":"The researchers explore whether just the pixels in an image can reveal its origin—whether made by a human, a group of AIs, or a specific AI generator—even if the image is edited before checking. They provide a theoretical limit on how well any method using only pixels can distinguish these origins, showing it depends on the differences in image distributions, not on the verification method itself. They also study why real-world verifiers can fail under attacks despite this limit, based on what information these verifiers reveal. Their findings come from theoretical analysis and experiments on public AI models, highlighting challenges in verifying image provenance using pixels alone."},{"id":"2609.30996v1","title":"The Linear Representation Hypothesis for Vision-Language-Action Models","abstract":"The linear representation hypothesis (LRH) has become a standard lens for measuring and intervening on semantic information through the internal representations of large language models (LLMs). A growing body of work has begun extending this perspective to vision-language-action (VLA) models, but the dynamical nature of embodied interaction introduces an additional challenge. Unlike semantic attributes commonly studied in LLMs, such as gender or language, a physical quantity of interest (QoI) in a VLA evolves jointly with the system dynamics: the representation influences the actions selected by the policy, which alter the physical state and, in turn, the next representation. In this paper, we develop a theoretical, signature-based formulation of the LRH for VLA that unifies representations and policies. On the representation side, we establish the existence of representations from which the future evolution of a QoI under a candidate action trajectory can be recovered via linear probing. On the policy side, we introduce a signature generalized linear model for stochastic action chunks. This structure yields a monotonic change in the expected future QoI along linear paths in natural parameter space, enabling linear steering. We construct an explicit oracle representation in a planar control-affine navigation experiment and verify the predicted linear probing and steering mechanisms.","published":"2026-09-25T08:41:36Z","updated":"2026-09-25T08:41:36Z","authors":["Minseok Jeong","Hyewon Choi","Hiroyasu Tsukamoto","SooJean Han"],"categories":["cs.LG","cs.AI"],"headline":"Researchers explore linear representations in vision-language-action models","summary":"The researchers investigate how internal data representations in models that combine vision, language, and physical actions can predict and influence future outcomes. They developed a theoretical framework showing it is possible to linearly extract information about future states and control actions accordingly. Their findings, supported by a navigation experiment, suggest these representations can be used to guide model behavior, but results are currently theoretical and based on simulations."},{"id":"2609.30995v1","title":"Learning Hierarchical Causal Representations of the Effects of Forcings on Temperature in Climate Models","abstract":"Machine learning (ML) emulators provide a fast and cost-effective method to simulate climate change scenarios after being trained on Earth System Models projections. However, the black-box nature of those data-driven approaches limit the usability and trustworthiness of their outputs and in particular their use as causal attribution tools. Here, we develop a hierarchical causal representation learning framework applied to sea surface temperature fields from a state-of-the-art global climate model. As a key advance over previous work, our framework explicitly models both atmospheric dynamical interactions arising from internal climate variability and forced responses due to changes in atmospheric greenhouse gas and aerosol concentrations. When trained on future climate change scenarios, our method accurately predicts the long-term global mean and regional temperature evolution and shows physically realistic responses to perturbations in greenhouse gas and aerosol concentrations when evaluated on unseen scenarios. Our results underline the potential of causal representation learning frameworks for advancing climate model emulation.","published":"2026-09-25T08:40:22Z","updated":"2026-09-25T08:40:22Z","authors":["Shan Zhao","Ilija Trajkovic","Julia Kaltenborn","Yaniv Gurwicz","Peer Nowack","David Rolnick","Julien Boussard"],"categories":["cs.LG"],"headline":"New AI method helps understand climate change effects on ocean temperatures","summary":"Researchers created a machine learning approach that learns cause-and-effect relationships affecting sea surface temperatures in climate models. Their method distinguishes natural climate variability from human-driven changes like greenhouse gases and aerosols. When tested on future climate scenarios, it accurately predicted temperature changes and responded realistically to changes in atmospheric conditions, suggesting promise for improving climate model simulations."},{"id":"2609.30994v1","title":"$k-$deformed Schwarzschild AdS black strings","abstract":"Considering a cylindrically symmetric spacetime, a novel exact $κ$-deformed Schwarzschild-AdS black string solution is derived in General Relativity in the presence of a negative cosmological constant. The geometric properties of the resulting solution are systematically investigated by computing the Ricci and Kretschmann curvature scalars. The analysis reveals the existence of a curvature singularity at the origin, which remains enclosed by an event horizon provided that the $κ$-deformation parameter satisfies a specific constraint. Subsequently, the influence of various model parameters on the horizon structure and configuration is examined in detail. The thermodynamic properties of these black strings are thoroughly evaluated; in this regard, the Hawking temperature, entropy, mass, isobaric heat capacity, and Gibbs free energy are self-consistently extracted. The effects of the parameters, particularly the $κ$-deformation parameter, on these thermodynamic quantities are comprehensively explored. Furthermore, it is demonstrated that these state variables fundamentally satisfy the extended first law of black hole thermodynamics and yield a modified Smarr relation. Finally, the temporal profile of the Hawking radiation is probed via the dimensionless sparsity parameter, through which the impacts of the underlying parameters on the quantum emission rate are systematically analyzed.","published":"2026-09-25T08:39:30Z","updated":"2026-09-25T08:39:30Z","authors":["B. Eslam Panah"],"categories":["gr-qc","hep-th"],"headline":"New theoretical black string model explores cosmic properties","summary":"Researchers derived a new exact solution in Einstein's theory describing a type of black string influenced by a parameter called k-deformation and a negative cosmological constant. They studied its shape, singularities, and event horizons, showing certain conditions keep singularities hidden. The team also examined how these parameters affect the black string's thermodynamics, such as temperature and entropy, confirming these follow adjusted versions of known black hole laws. Finally, they analyzed the quantum radiation emitted, exploring how model parameters impact this emission. The findings are theoretical and based on mathematical analysis."},{"id":"2609.30993v1","title":"FLIP: Final Layer Inference-Time Probing for Vision-Language Models","abstract":"We present FLIP, a final-layer inference-time probe for testing whether a logit-facing intervention site in an open-weight vision-language model (VLM) supports structured, task-linked computation rather than generic perturbation. Behavioral change under internal intervention is otherwise mechanistically ambiguous: it may reflect improved use of visual evidence, generic output instability, or outright degradation. FLIP applies elementwise flooring to the final normalized hidden state before logit computation, leaving parameters, prompts, and decoding unchanged. On a controlled detection/counting probe, sweeping intervention strength reveals three regions: negligible change, a bounded interior regime in which detection recall at IoU 0.50 ($R_{50}$) improves while tolerant counting error ($\\mathcal{E}_{\\mathrm{count}}$) falls, and over-suppression. We formalize a four-criterion probe-and-sweep protocol for disciplining the interpretation of intervention effects: regime structure, grounding-proxy alignment, feature-coherence dependence, and failure to reproduce the same positive regime on a performance-based negative control. The post-normalization state passed to the output head is the logit-facing instantiation of this test; under a non-targeted flooring sweep it satisfies the full protocol. Raw decoder-layer interventions, including the last-block output before final normalization, and the singleton-pair left/right control fail to reproduce the Final-site signature, while same-site operators and multiple VLMs replicate it. FLIP is therefore a validation step for intervention-based mechanistic interpretability, not a steering method.","published":"2026-09-25T08:38:54Z","updated":"2026-09-25T08:38:54Z","authors":["Drandreb Earl O. Juanico","Rowel O. Atienza"],"categories":["cs.CV","cs.AI"],"headline":"New Method Tests How Vision-Language Models Process Information","summary":"Researchers developed FLIP, a technique to check if certain parts inside vision-language models handle tasks in a meaningful way or just cause random changes. By adjusting values right before the model makes a prediction, they observed how performance on tasks like object detection and counting changed. Their results suggest FLIP can validate how internal interventions affect the model's behavior, but it does not change or control the model itself. These findings are based on controlled tests and do not guarantee practical applications."},{"id":"2609.30992v1","title":"Practical Deterministic Linear-Time Modular Subset Sum","abstract":"We give a deterministic algorithm for exact modular subset sum that, for every modulus m, computes all reachable residues and one requested witness in O(m) time and O(m) auxiliary words. The input is a compact list of distinct residues with multiplicities, and the word-RAM supports constant-time modular arithmetic. The algorithm represents reachable residues as intervals along cycles of repeated addition, charging work on partial cycles to newly reached residues. Processing prime factors in increasing order keeps the cost of rebuilding and changing cycles linear. A classical theorem on subset sums of distinct invertible residues bounds the number of boundary lists by O(m^(3/4)); together they contain O(m) interval endpoints. Comparison sorting the short lists and radix sorting the long ones then takes O(m) total time. The algorithm is fast in practice, using arrays and interval lists rather than heavy data structures.","published":"2026-09-25T08:38:19Z","updated":"2026-09-25T08:38:19Z","authors":["Phuoc Dinh Le","Kha Le"],"categories":["cs.DS"],"headline":"New Fast Method Finds Modular Subset Sums Exactly","summary":"The researchers developed a step-by-step method to find all possible sums of numbers when working with modular arithmetic, doing so in an amount of time proportional to the size of the modulus. Their approach uses simple data structures like arrays and cleverly organizes calculations to keep the process efficient. While the method is mathematical and theoretical, the researchers report it works quickly in practice for the problem they studied."},{"id":"2609.30991v1","title":"Solving Graph Coloring Problems Using Feedback-Based Algorithm for Quantum Optimization","abstract":"The feedback-based algorithm for quantum optimization (FALQON), which is an extension of the quantum approximate optimization algorithm (QAOA), is a promising approach for finding approximate solutions to combinatorial optimization problems. The FALQON avoids the optimization of variational parameters by using information derived from iterative measurements. To solve a constrained combinatorial optimization problem, it is sometimes more efficient to use an XY mixer rather than the conventional X mixer. We applied the FALQON with the XY mixer to graph coloring and investigated the expectation value of the problem Hamiltonian, success probability of graph coloring, and fair sampling for several small-scale problems. The FALQON with the XY mixer exhibited better performance than that with the X mixer.","published":"2026-09-25T08:37:53Z","updated":"2026-09-25T08:37:53Z","authors":["Yuika Kawatomi","Yukina Tatsuta","Kazue Kudo"],"categories":["quant-ph"],"headline":"New Quantum Method Improves Graph Coloring Solutions","summary":"Researchers explored a quantum algorithm called FALQON, which helps find approximate solutions to complex problems without adjusting many parameters. They tested it on the graph coloring problem using a specific technique (XY mixer) and found it performed better than the traditional method (X mixer) in small test cases. Their findings show potential benefits of this approach, but results are limited to these initial simulations on small problems."},{"id":"2609.30990v1","title":"Teukolsky Master Equation for a Kerr-like metric with Mass Quadrupole Moment","abstract":"The Teukolsky Master Equation (TME) is a powerful technique used to study perturbations of a spacetime with gravitational, electromagnetic, and neutrino components. The well-known Kerr metric is a suitable candidate for this formalism because it is a Petrov type D metric. However, the Kerr metric represents the exterior spacetime of a perfect sphere and only black holes are described using this metric. For another astrophysical object, such as fast-spinning neutron stars, the Kerr metric is not valid due to the rotation induced deformations to the gravitational source. The main goal of the present work is to show a TME for a Kerr-like metric (KL) with a mass quadrupole moment parameter $q$. This new parameter will represent deformations from a perfect sphere; therefore, KL is a better candidate to study perturbations surrounding spinning stars.","published":"2026-09-25T08:37:04Z","updated":"2026-09-25T08:37:04Z","authors":["Pedro Gómez","Adrián Eduarte-Rojas","Francisco Frutos-Alfaro","Antonio Banichevich"],"categories":["gr-qc","math-ph"],"headline":"New Equation Improves Study of Spinning Star Spacetime Disturbances","summary":"Researchers extended a key equation, the Teukolsky Master Equation, to apply to a Kerr-like spacetime that includes a mass quadrupole moment, representing deformations from a perfect sphere. While the original Kerr metric suits black holes, it does not accurately model spinning neutron stars due to their shape distortions. The new approach aims to better study disturbances around these stars by incorporating their rotationally induced deformations. This work is theoretical and focuses on adapting the equation to a more general spacetime metric."},{"id":"2609.30989v1","title":"PICO: Projection-Informed Consistency Optimisation for 6DoF Surgical Tool Pose Estimation","abstract":"Purpose: Accurate 6 DoF pose estimation of surgical tools is critical for automa- tion, robotic proprioception, and safe interaction with the tissue operated on. Kinematics-based approaches suffer from accumulated errors due to the cable- driven nature of robotic arms, while vision-based methods often rely on external markers or trackers. Although more recent vision-based advances have been pro- posed, these two-stage pose estimation methods often lack real-time robustness due to accumulated errors and computational overhead. Methods: We propose a novel end-to-end trainable model, PICO. Our model employs a multi-task learning architecture to predict segmentation and depth maps, alongside regression of translation and rotation parameters. We define two proxy tasks that enforce geometric consistency in both 2D and 3D spaces, improving accuracy and robustness. For this, we propose a projection loss, and a point-to-point loss. Results: We evaluate our method on the SurgRIPE dataset, benchmarking its performance against state-of-the-art approaches using standard 6DoF pose esti- mation metrics. Our results demonstrate consistently strong performance across all four subsets, specifically in rotation, ranking second even under occlusion. It also demonstrates comparable translational performance, remaining competitive, especially in occluded cases. Conclusion: PICO demonstrates the effectiveness of multi-task learning and geometry-aware proxy tasks for robust and reliable surgical tool pose estimation, especially in occluded scenarios, highlighting potential for future applications.","published":"2026-09-25T08:35:04Z","updated":"2026-09-25T08:35:04Z","authors":["Lucy Fothergill","Pietro Valdastri","Dominic Jones","Duygu Sarikaya"],"categories":["cs.CV"],"headline":"New AI method improves surgical tool position tracking during surgery","summary":"Researchers developed PICO, an AI model that estimates the exact position and orientation of surgical tools in real time using images. Unlike older methods, PICO predicts tool shapes and distances, ensuring consistent 2D and 3D geometry without relying on external markers. Tested on a surgical dataset, the model showed strong accuracy, especially in challenging cases where tools are partly hidden. The study supports that combining multiple tasks and geometry checks can enhance tool tracking, though results are based on simulations and benchmark data only."},{"id":"2609.30988v1","title":"PhoenixSR: Generative Heterogeneous Distillation Unleashes Efficient Models for Real-World Super-Resolution","abstract":"Real-world image super-resolution (SR) requires recovering perceptually realistic high-resolution images from complex low-resolution observations while preserving faithful content. Diffusion-based SR benefits from strong generative priors but incurs substantial computational overhead, whereas feed-forward CNN and Transformer SR models are efficient yet often struggle to recover realistic high-frequency details. This motivates a natural question: can diffusion priors be transferred to existing diffusion-free SR networks without introducing diffusion components at inference time? To this end, we propose PhoenixSR, a generative heterogeneous distillation framework that transfers diffusion priors to independently designed feed-forward SR networks through score-based distribution matching. Rather than aligning heterogeneous features or imitating sampled diffusion outputs, PhoenixSR uses the pretrained diffusion model as distribution-level supervision, while paired SR supervision preserves reconstruction fidelity. To make distribution matching effective for fidelity-sensitive SR, we introduce Heterogeneous Distribution Adaptation, which adapts the target score to the SR domain, improves tracking of the evolving student distribution, and anchors training with paired supervision. We further employ Directional Reliability Weighting, a lightweight residual-consistency-based reweighting strategy that reduces unstable distributional guidance. All diffusion-related components are removed after training, leaving the original student architecture and inference cost unchanged. Experiments on three SR benchmarks and six feed-forward backbones, including SwinIR, HAT, Real-ESRGAN, and SeeMoRe, show consistent perceptual improvements with largely preserved reconstruction fidelity.","published":"2026-09-25T08:34:32Z","updated":"2026-09-25T08:34:32Z","authors":["Xin Di","Mingyu Shi","Yuanfei Bao","Long Peng","Yue Zhao","Jiaming Guo","Renjing Pei","Xueyang Fu","Yang Cao","Zheng-Jun Zha"],"categories":["cs.CV"],"headline":"New method improves image super-resolution without slowing it down","summary":"Researchers developed PhoenixSR, a way to teach fast image enhancement models using insights from slower, more detailed diffusion models. Their approach uses a special training method to transfer knowledge without adding complexity during actual image processing. Tests on several benchmarks showed their method improved the realism of enhanced images while keeping the quality faithful to the original content, all without increasing the model's runtime."},{"id":"2609.30987v1","title":"Self-Supervised Perceptually Interpretable Monocular Depth Estimation","abstract":"Self-supervised monocular depth estimation (MDE) enables depth prediction from monocular images without requiring ground-truth supervision, making it attractive for large-scale and real-world applications. Despite steady improvements in accuracy, most existing methods remain difficult to interpret, as depth is inferred from RGB representations that obscure the impact of individual perceptual image components. This lack of transparency limits systematic analysis of failure cases and reduces confidence in safety-critical settings. This paper presents a self-supervised framework for perceptually interpretable monocular depth estimation (PIMDE), designed to associate depth predictions with distinct perceptual components of the input image. Rather than operating directly on RGB inputs, the proposed method decomposes each image into a set of perceptual feature maps (PFMs), each encoding a specific visual cue. Distinct depth estimation branches process these PFMs independently to produce depth estimates (PIDEs), which are subsequently combined through an explicit fusion strategy. This formulation allows us to examine directly the contribution of each perceptual cue to the final depth prediction. Experiments conducted on the KITTI benchmark dataset demonstrate that PIMDE achieves performance comparable to established self-supervised MDE methods while providing additional insight into how different perceptual cues influence depth estimation. These results indicate that perceptual decomposition can support interpretability without sacrificing depth estimation accuracy.","published":"2026-09-25T08:34:26Z","updated":"2026-09-25T08:34:26Z","authors":["Zain Ul Abidin","George Dimas","Dimitris K. Iakovidis"],"categories":["cs.CV"],"headline":"New Method Makes Depth Estimates From Images Easier to Understand","summary":"Researchers developed a new way to estimate depth from single images without needing direct measurement data. Their method breaks down images into separate visual parts and uses these parts to predict depth, allowing them to see which aspects influence the results. Tests on standard data showed their approach works as well as existing methods while offering clearer insight into how image features affect depth prediction."},{"id":"2609.30986v1","title":"Evaluating Sycophancy in Chinese Large Language Models on Factual Questions Derived from Online Search Queries","abstract":"As large language models increasingly mediate information access, factually accurate and independent answers are critical. However, these models can exhibit sycophancy by aligning their responses with users' stated beliefs even when those beliefs are incorrect, potentially presenting misinformation as independently verified and reinforcing users' confidence in false claims. Prior work leaves unresolved whether introducing user beliefs causes correct responses to become incorrect or uncertain, or causes uncertain responses to become belief-aligned incorrect answers. It also remains unclear whether anti-sycophancy interventions preserve or restore factual accuracy or merely shift responses toward uncertainty. We analyze factual sycophancy in Chinese-language information seeking using yes/no fact-checking questions. Our analysis covers 364,941 responses from three frontier Chinese-based LLMs (DeepSeek, Qwen, and Doubao) to 12,165 factual questions derived from real-world Chinese search queries. We evaluate the models with and without reasoning across baseline, belief-conditioned, and anti-sycophancy prompting, tracing matched shifts among correct, incorrect, and uncertain responses. Under incorrect user beliefs, we distinguish belief-aligned errors from losses of factual confidence, in which initially correct answers become uncertain. Patterns vary across models and reasoning settings: reasoning is not a consistent safeguard, and anti-sycophancy instructions can reduce incorrect agreement while increasing uncertainty. In Chinese-language factual question answering, avoiding agreement with false beliefs is therefore not equivalent to preserving factual accuracy, highlighting the value of transition-level evaluation. Such behavior may undermine the reliability of LLM-mediated information access by reinforcing misinformation or weakening users' confidence in factually correct answers.","published":"2026-09-25T08:34:19Z","updated":"2026-09-25T08:34:19Z","authors":["Geng Liu","Feng Li","Mengxiao Zhu","Francesco Pierri"],"categories":["cs.CL"],"headline":"Study Examines How Chinese AI Models Handle False User Beliefs","summary":"Researchers studied whether Chinese large language models give accurate answers or wrongly agree with users' false beliefs when answering factual yes/no questions from real search queries. They tested three models with different prompts to see if models changed correct answers to uncertain or incorrect ones when faced with false beliefs, and whether warnings against agreeing with false beliefs helped. The study found that stopping models from agreeing with false beliefs sometimes just made answers more uncertain rather than more accurate. This suggests these AI tools might still weaken trust in correct information or unintentionally support misinformation."},{"id":"2609.30985v1","title":"Predicting Mutational Signature Exposures from H&E Whole Slide Images: A Pan-Cancer Feasibility Study","abstract":"Mutational signatures reveal cancer-driving processes with clinical relevance: MMR-deficient tumors respond better to immunotherapy, HRD tumors are sensitive to PARP inhibitors, and POLE-mutant tumors often have high mutation burdens influencing treatment response. Yet signature profiling remains limited by the cost, complexity, and turnaround time of whole-genome sequencing (WGS) or whole-exome sequencing. Histopathology is widely available and cost-effective, and H&E morphology has been linked to MSI, HRD, POLE-related processes, and driver mutations. Whether histology can recover the broader landscape of mutational signature exposures in a pan-cancer setting remains unknown. Here we introduce Hist2Sig, a deep learning framework predicting exposures to 30 COSMIC SBS signatures from H&E-stained slides. Trained on matched WGS and histology data from 7,063 TCGA patients across 29 cancer types, Hist2Sig was compared with a tumor type-only baseline to separate morphology-derived signal from tissue-of-origin priors. In internal cross-validation, it recovered relative signature compositions across most tumor types and achieved higher mean top-three overlap than the baseline in 19 of 29, with the largest gains in COAD, PRAD, GBM, and UCEC. It also recurrently recovered signatures of poorly characterized etiology, including SBS8, SBS12, SBS39, and SBS40a. In an external CPTAC cohort of 193 patients across five tumor types, Hist2Sig retained moderate concordance with observed compositions and outperformed the baseline in glioblastoma and pancreatic adenocarcinoma. These results support the feasibility of inferring mutational signature exposures from routine histology, while highlighting variation across tumor types. Hist2Sig provides a foundation for tumor-specific models that could complement sequencing in pre-sequencing triage or support decisions where WGS is unavailable.","published":"2026-09-25T08:32:03Z","updated":"2026-09-25T08:32:03Z","authors":["Flavio Sartori","Cesare Rollo","Isabella Caranzano","Tiziana Sanavia","Piero Fariselli","Corrado Pancotti"],"categories":["q-bio.QM"],"headline":"AI Predicts Cancer Mutation Patterns from Routine Tissue Slides","summary":"Researchers developed Hist2Sig, a deep learning tool that predicts detailed mutation patterns in cancers using common stained tissue slides, avoiding costly genetic sequencing. Trained on data from over 7,000 patients covering 29 cancer types, it showed promise in identifying mutation signatures across many tumors, outperforming simple models in some cases. Tests on a separate patient group confirmed moderate accuracy, though results varied by cancer type. The study suggests this method could help assess mutations when sequencing is unavailable, but it remains a feasibility study with variable performance."},{"id":"2609.30984v1","title":"THA: Weighted Finite-State Text Normalization and Inverse Text Normalization for Khmer","abstract":"Text-to-speech needs written text in spoken form, and speech recognition output needs the reverse. For Khmer, neither direction has a maintained open-source tool, and the script makes both harder: words are not separated by spaces, and number words occur inside ordinary words. We present Tha, a Khmer text normalization and inverse text normalization toolkit built from weighted finite-state transducers. It segments and classifies a whole line in one shortest-path search, and a second transducer rejects token boundaries inside a Khmer syllable. On Google's Khmer test suite, Tha agrees with the reference on all 274 cardinals up to one spelling variant, and on 2,906 real TTS prompts, 153 of the 158 sentences it rewrites are correct. Tha is open source under the Apache 2.0 license.","published":"2026-09-25T08:31:16Z","updated":"2026-09-25T08:31:16Z","authors":["Seanghay Yath"],"categories":["cs.CL"],"headline":"New Tool Normalizes Khmer Text for Speech Technologies","summary":"Researchers created Tha, an open-source software to convert written Khmer text into spoken form and back, addressing challenges like no spaces between words and complex number words. Tha uses advanced language processing methods to accurately segment and classify text lines. Tests showed Tha matches reference results for all tested numbers and correctly rewrites most sentences for text-to-speech prompts, supporting its usefulness for Khmer language technologies."},{"id":"2609.30983v1","title":"Tracing and Relearning Detection Evidence in Text-to-Speech Systems","abstract":"Recent audio deepfake detectors separate bona fide speech from synthetic speech, yet it remains unclear which stage of a text-to-speech system supplies the detection evidence. We address this with controlled resynthesis and detector adaptation in an F5-TTS-BigVGAN pipeline. Since vocoder reconstruction of a real mel can itself be separable from the source utterance, we fix the vocoder and trace the larger change in detector separation to acoustic generation. Adversarially fine-tuning the acoustic model, with no detector in its objective, raises EER against fixed detectors at comparable quality. However, adapting a detector only on the tuned model's VCTK outputs lowers its LibriSpeech EER from 19.42% to 7.46% and improves detection of unseen base F5-TTS outputs. These results suggest that acoustic-model updates can reduce the detection evidence available to fixed detectors, while detector adaptation keeps the updated outputs detectable in this pipeline.","published":"2026-09-25T08:29:14Z","updated":"2026-09-25T08:29:14Z","authors":["Eunji Shin","Kyudan Jung","Jihwan Kim","Minwoo Lee","Jaegul Choo"],"categories":["cs.SD","eess.AS"],"headline":"Studying What Helps Detect Fake Speech in AI Voice Systems","summary":"Researchers explored which part of AI systems that generate synthetic speech provides clues for detection. They examined a specific text-to-speech setup and found that changes in the acoustic model affect how easily fake speech can be identified. By fine-tuning the acoustic model and adapting detectors accordingly, they showed detection can be improved even against altered synthetic voices. Their findings come from controlled experiments with a particular system and highlight complexities in detecting AI-generated speech."},{"id":"2609.30982v1","title":"FARE: Forensic Acceptance Region Estimation for Catching Bait-and-Switch Image Generators","abstract":"Modern AI image generators are increasingly deployed as opaque APIs, where customers can query the deployed service, but cannot inspect model weights or architecture. This creates a practical challenge: a provider may pass governance certification with one generator and later silently switch to a cheaper and lower-quality one for deployment, compromising public trust or even safety in high-stakes domains. We study integrity auditing at deployment time and propose FARE (Forensic Acceptance Region Estimation). A certified generator is enrolled by training FARE on images sampled from that generator. After deployment, FARE can determine whether a generated image is consistent with the enrolled generator---using only that image. FARE's features are based on image generator-specific artifacts that have been proposed for forensic applications. FARE amplifies these features during training by finding hard samples that tighten the acceptance region and increase sensitivity to subtle changes in the certified generator. Across generator swaps, including substitutions with similar model versions and model variants, FARE is effective at detecting swaps, consistently outperforming existing baselines at strict operating points, and remains effective under the exact-model and decision-only attacks evaluated in this work.","published":"2026-09-25T08:28:21Z","updated":"2026-09-25T08:28:21Z","authors":["Kai Yao","Marc Juarez"],"categories":["cs.CV","cs.AI","cs.CR"],"headline":"New Method Detects Cheating in AI Image Generators","summary":"The researchers address a problem where companies might switch to lower-quality AI image generators after passing certification, risking trust and safety. They developed FARE, a tool trained on images from a certified generator to recognize if images still come from that approved system. FARE uses patterns unique to each generator and improves detection by focusing on challenging examples. Their tests show FARE reliably spots swaps between similar AI models, outperforming other methods, though the results are based on the scenarios they examined."},{"id":"2609.30981v1","title":"STORM-Bench: Evaluating Online Video QA under Evolving and Incomplete Evidence","abstract":"Reliable online video question answering requires tracking state transitions while selectively abstaining when visual evidence is insufficient. Existing benchmarks focus on static recognition or long-range retrieval, rarely evaluating these coupled capabilities under evolving and incomplete evidence. We present STORM-Bench, comprising 5,736 questions across 630 compact, change-dense episodes spanning five egocentric domains (STORM-Real) and two controlled simulation subsets (STORM-Sim) at 1 FPS. Questions are stratified by a proxy for accumulated change intensity (Low, Medium, High) and query-time answerability (Known, Uncertain). To measure reliability, we introduce STORM-BR, a harmonic metric over joint answer-status correctness that exposes abstention failures masked by aggregate accuracy, alongside STORM-BR-ATTR for uncertainty attribution. Across 14 video LLMs, online accuracy peaks at 60.3\\% (mean 51.7\\%), whereas STORM-BR ranges from 5.7\\% to 35.6\\% (mean 18.8\\%), driven by pervasive overconfidence on uncertain queries. STORM-Bench shows that task accuracy masks these gaps in epistemic reliability and state tracking. Benchmark and code are available at https://github.com/siruzhong/STORM-Bench.","published":"2026-09-25T08:27:29Z","updated":"2026-09-25T08:27:29Z","authors":["Siru Zhong","Shenghan Tan","Rihong Yan","Xiaohui Lv","Yuzheng Zhuang","Shuai Tao","Wulong Liu","Haohuan Fu","Yuxuan Liang"],"categories":["cs.CV"],"headline":"New test reveals limits in video AI's answer reliability","summary":"Researchers created STORM-Bench, a new challenge with thousands of questions about videos where scenes change and not all answers are clear. They tested 14 video AI models and found while models can answer some questions correctly, they often overconfidently answer when unsure, missing when to say \"I don't know.\" The study highlights that just measuring correctness hides important flaws in AI’s reliability and ability to track changes over time. These findings come from controlled tests and simulations provided by the researchers."},{"id":"2609.30980v1","title":"FeatMark: Feature-level Watermark Protection against Mimicry Attacks with Diffusion Models","abstract":"Text-to-image diffusion models enable data-efficient \"mimicry\" attacks, wherein adversaries fine-tune the model on a handful of public photos to synthesize convincing forgeries of a target individual. A common countermeasure is to embed imperceptible, low-energy watermarks, yet recent studies show these signatures are brittle: modest post-processing or lightweight adversarial perturbations readily suppress detection, exposing a fundamental tension between imperceptibility and robustness. We introduce FeatMark, a watermarking framework that shifts from pixel-level, energy-starved perturbations to inconspicuous semantic features: small, scene-consistent micro-features that remain natural to humans while providing a stronger, machine-verifiable provenance signal. FeatMark builds domain-specific feature banks that encode each watermark as a compact concept program, pairing open-vocabulary semantic cues with reliable edit regions and instruction templates. It then automatically selects features that are both feasible and executable and injects them through modular, mask-guided concept editing, yielding highly localized, scene-consistent micro-edits that are difficult to perceive. We conduct extensive experiments across VGGFace2, CelebA-HQ, and WikiArt, evaluating against 10 strong watermark removal/purification attacks (including regeneration-style purification) and several bespoke adaptive attacks tailored to FeatMark, to assess perceptual fidelity, watermark detection accuracy, and robustness. We further demonstrate FeatMark's extensibility to video mimicry attacks. The results show FeatMark remains virtually impervious, withstanding all evaluated attacks with negligible bit-accuracy and fidelity degradation.","published":"2026-09-25T08:26:49Z","updated":"2026-09-25T08:26:49Z","authors":["Haoyang Li","Ruoxi Sun","Qingqing Ye","Benjamin Zi Hao Zhao","Yaxin Xiao","Jason Xue","Haibo Hu"],"categories":["cs.CR","cs.CV"],"headline":"New Watermark Method Protects Against Fake Image Attacks","summary":"Researchers studied how text-to-image models can be tricked into creating fake images of people using a few public photos, which threatens image authenticity. They developed FeatMark, a new watermarking approach that adds subtle, natural-looking features instead of small pixel changes, making watermarks harder to remove. Their tests show FeatMark resists many advanced attempts to erase or hide watermarks while keeping images looking normal, though findings are based on controlled experiments and simulations."},{"id":"2609.30979v1","title":"CCRV-Bench: Constraint-Based Evaluation of Causal Reasoning in Vision-Language Models","abstract":"Vision-language models (VLMs) have demonstrated excellent performance in visual tasks, but their visual causal reasoning capabilities still lack reliable evaluation. Existing evaluations struggle to distinguish whether a model is performing causal reasoning based on visual evidence or relying on statistical correlations for shortcut learning, thereby potentially overestimating their actual capabilities. This paper proposes CCRV-Bench, a constraint-driven visual causal reasoning benchmark for single-image physical scenarios. We construct an orthogonal framework that evaluates four causal task dimensions: causal relation discovery, state prediction, causal diagnosis, and intervention. We further introduce entity symbolization, spatial grounding, the factual adversarial constraint, and minimalist output constraints to reduce shortcut cues while preserving the physical commonsense required by the task. Experiments across 15 multimodal models show that constraint sensitivity is task- and model-dependent: intervention has the largest average effective degradation among the four causal tasks, spatial grounding is the most damaging constraint on average, and the factual adversarial constraint improves DCR for all evaluated models. These results show that unconstrained performance does not determine constrained robustness and that a single aggregate score can obscure distinct failures in causal identification, spatial grounding, and constraint-compliant expression. CCRV-Bench provides a standardized framework for diagnosing image-grounded causal reasoning under controlled constraints. The code is available at https://github.com/0815linyuan/CCRV-Bench-Constraint-Based-Evaluation-of-Causal-Reasoning-in-Vision-Language-Models","published":"2026-09-25T08:26:06Z","updated":"2026-09-25T08:26:06Z","authors":["Linyuan Gao","Yuan Wu","Yi Chang"],"categories":["cs.CV"],"headline":"New Test Checks Vision-Language Models' True Causal Thinking","summary":"Researchers developed CCRV-Bench, a new way to test whether vision-language models truly understand cause and effect in images or just guess based on patterns. Their test looks at different types of causal thinking using controlled challenges to reduce shortcut guessing. Experiments on 15 models showed that performance varies and that passing some tests doesn't guarantee real understanding. The benchmark helps reveal specific weaknesses in how models reason about causes in pictures."},{"id":"2609.30978v1","title":"The maximum number of edges in minimal matching covered graphs","abstract":"A connected graph $G$ with at least two vertices is {\\em matching covered} if each of its edges lies in a perfect matching. A matching covered graph is {\\em minimal} if the removal of any edge results in a graph that is no longer matching covered. Lovász and Plummer [J. Combin. Theory, Ser. B 23 (1977) 127--138] proved by ear decompositions that every minimal matching covered bipartite graph $G$ different from $K_2$ has at most $(3|V(G)|-6)/2$ edges, and this bound is sharp for all $|V(G)|\\ge4$. In this paper, we prove that every minimal matching covered nonbipartite graph $G$ with at least 6 vertices has at most $5(|V(G)|-2)/2$ edges, and this bound is sharp for all $|V(G)|\\ge6$.","published":"2026-09-25T08:25:51Z","updated":"2026-09-25T08:25:51Z","authors":["Xiaoling He"],"categories":["math.CO"],"headline":"New edge limits found for special matching graphs","summary":"Researchers studied a type of graph called minimal matching covered graphs, where every edge is part of a perfect matching and removing any edge breaks this property. They built on earlier work that measured the maximum edges in bipartite graphs and established a new precise upper limit on edges for nonbipartite graphs with six or more vertices. Their findings are theoretical and confirm exact edge maximums for these graph types under the stated conditions."},{"id":"2609.30977v1","title":"Does Uniform Discrete Diffusion Need Time?","abstract":"Uniform discrete diffusion models (UDMs) commonly use explicit time conditioning, but we find that it can often be unnecessary in practice. In this paper, we first show that the population-optimal UDM predictor generally depends on time: time controls how much the model should trust the observed context. We then show that this dependence can become negligible in finite-data settings relevant to language. When a corrupted training sequence remains much closer to its original clean sequence than to competing training sequences, the empirical-optimal predictor is nearly insensitive to time over most of the diffusion trajectory, where the guarantee weakens toward the high-noise endpoint. Empirically, trained language UDMs exhibit limited time sensitivity over most of the trajectory, while time-agnostic predictors remain competitive with, and often outperform, time-conditioned models across datasets and training objectives. These results challenge the use of explicit time conditioning in UDMs: although the population optimum depends on time, explicitly conditioning on it may often be unnecessary in practice.","published":"2026-09-25T08:24:35Z","updated":"2026-09-25T08:24:35Z","authors":["Chunsan Hong","Chieh-Hsin Lai","Satoshi Hayakawa","Yuhta Takida","Jong Chul Ye","Yuki Mitsufuji"],"categories":["cs.LG","cs.AI","cs.CL"],"headline":"Time May Not Be Needed in Certain Language Diffusion Models","summary":"Researchers studied uniform discrete diffusion models (UDMs), which often use time as a key factor when making predictions. They found that while theoretically the best UDM prediction depends on time, in practical language-related settings with limited data, this time dependence is minimal. Their experiments showed that models ignoring explicit time information performed as well or better than those using it, suggesting time conditioning might often be unnecessary for these models in practice."},{"id":"2609.30976v1","title":"Geometric Dissipation Structures and the Koszul Form","abstract":"This paper investigates the canonical Koszul one-form at the interface of Jean Marie Souriau's Lie group thermodynamics and Paulette Libermann's symplectically complete foliations to describe geometric dissipation. On the Poincaré half-plane $\\mathbb{H}$ under $SL(2, \\mathbb{R})$, the equivariant momentum map level sets form a regular Libermann foliation framing the conservative boundaries. The Koszul-derived dissipative vector field acts strictly transverse to these leaves, driving a metriplectic flow that breaks Noether conservation laws and generates entropy. Quantum wise, we develop Berezin quantization on para-Kähler symmetric spaces, where the Koszul potential acts as a weight determining state density via the Fisher Souriau metric, and formalize Kostant's geometric quantization using the real dual polarizations of a bi-Lagrangian web.","published":"2026-09-25T08:21:29Z","updated":"2026-09-25T08:21:29Z","authors":["Prosper Rosaire Mama Assandje","Frédéric Barbaresco","Romain Nimpa Pefoukeu","Michel Bertrand Djiadeu Ngaha","Nozah Mba Frankel"],"categories":["math.DG"],"headline":"Researchers Explore Geometry Behind Energy Loss and Quantum States","summary":"The researchers study a special geometric object called the Koszul one-form to better understand how systems lose energy or dissipate. They analyze mathematical structures on a curved space known as the Poincaré half-plane, showing how energy-conserving behaviors can be separated from dissipative processes that increase entropy. They also extend their work to quantum settings, developing methods to describe state densities using advanced geometry, though their results remain theoretical and focus on abstract mathematical models rather than practical experiments."},{"id":"2609.30975v1","title":"A Comprehensive Study of Content Representations for Speech Synthesis","abstract":"Speech content representations are central to voice conversion, speech-to-speech translation, and multimodal language models, yet they are rarely compared under a common generative framework that directly measures what each representation contains. We address this by training a generative model conditioned solely on each representation and evaluating the generated audio along the content, speaker identity, and prosody axes. Across SSL features, supervised tokens, posteriorgrams, and neural audio codecs, we find two distinct regimes: representations that nearly reconstruct the original audio, and representations that effectively disentangle speaker identity. These results show that disentanglement depends not on supervision alone, but on the interaction between the training objective and the representation's information capacity: supervised representations only disentangle speaker identity when their capacity is sufficiently constrained.","published":"2026-09-25T08:20:56Z","updated":"2026-09-25T08:20:56Z","authors":["Diego Torres","Axel Roebel","Nicolas Obin"],"categories":["eess.AS","cs.LG","cs.SD"],"headline":"Researchers Compare Speech Content Representations for Voice Models","summary":"The researchers studied different ways to represent speech content, which is important for voice-related technologies like speech translation. They trained models using various speech features and evaluated how well each captured content, speaker identity, and speech rhythm. Their findings suggest that some representations closely recreate the original voice, while others better separate the speaker's identity from the content. They also found that successfully separating speaker identity depends on both how the model is trained and the complexity of the representation. The study is based on analysis within a shared modeling framework."},{"id":"2609.30974v1","title":"Coupled Usage-Sense Processes: Temporal and Attributable Lexical Semantic Change","abstract":"Lexical semantic change is usually summarized by a scalar distance between independently sampled period distributions. This measures how much a word changed, but does not reveal when it changed, which mechanisms and component movements carried the change, or which usages support the attribution. We introduce Coupled Usage--Sense Processes (CUSP), which derives these answers from a single marginal preserving temporal process. A hierarchical coupling relates contextual distributions through latent usage components, while Markov composition makes adjacent and longer span correspondences compatible. Displacement operators quantify change magnitude and timing, split variation exactly between movement of component centers and reorganization within components, and attribute it to transported component pairs. Word-local modes resolve distinct directions of change and their activity over time, while representative passages from attributed components ground the analysis in text. Under a Gaussian mixture specialization, we prove parametric recovery of the operators and squared distances. Synthetic experiments support the predicted rate. CUSP remains competitive on English and German DWUG and recovers controlled Janus profiles while maintaining compositionally coherent transport. A large corpus of US court opinions demonstrates transition, mode, and passage attribution in unlabeled natural text. CUSP thus makes magnitude, timing, mechanism, movement, modes, and textual evidence compatible views of one lexical history.","published":"2026-09-25T08:19:25Z","updated":"2026-09-25T08:19:25Z","authors":["Haruka Ezoe","Ryohei Hisano"],"categories":["cs.CL","stat.ML"],"headline":"New method tracks when and how word meanings change over time","summary":"Researchers developed Coupled Usage–Sense Processes (CUSP), a method to analyze how word meanings shift not just in size but also in timing and mechanism. CUSP links word uses across time to show when changes occur, what parts of meaning move, and which examples in text illustrate these shifts. Tested with simulations and real English and German text, including US court opinions, CUSP offers a detailed, connected view of how word meanings evolve. These findings come from mathematical modeling and require further practical exploration."},{"id":"2609.30973v1","title":"LipSSM: Structurally Lipschitz-Bounded Cascaded State-Space Model via Metric Transfer between Consecutive SSM Layers","abstract":"Lipschitz continuity is a fundamental principle in the design of certifiably robust deep neural networks (DNNs), wherein adjusting the Lipschitz constant, which quantifies network robustness, is of central theoretical importance. A standard approach to enforcing Lipschitz continuity requires each layer of a DNN to be Lipschitz continuous, thereby guaranteeing overall Lipschitz continuity. However, this layer-wise approach typically imposes overly conservative restrictions by producing a loose estimate of the overall Lipschitz constant, which limits the expressive capacity of the DNN and degrades empirical performance at a prescribed level of robustness. To overcome this loose estimation, the recently proposed LipKernel transfers information across layers to yield a much tighter overall Lipschitz bound than conventional layer-wise construction. In this paper, we extend this concept to cascaded state-space models (SSMs) to construct Lipschitz-continuous DNNs capable of modeling longer-term dependencies. The proposed architecture, named LipSSM, is theoretically justified and empirically evaluated.","published":"2026-09-25T08:19:02Z","updated":"2026-09-25T08:19:02Z","authors":["Natsuki Yoshino","Ren Uchida","Kazuki Matsumoto","Kohei Yatabe"],"categories":["cs.LG","eess.SY"],"headline":"New Method Improves Robustness Estimates in Deep Neural Networks","summary":"Researchers address a problem in designing robust deep neural networks, where existing methods impose strict layer-by-layer limits that reduce performance. They build on a recent idea that shares information between layers to better estimate overall network robustness. The team extends this idea to a type of neural network model called cascaded state-space models, creating a new architecture named LipSSM. Their paper provides theoretical justification and initial tests, but results are based on their specific evaluations without broader application claims."},{"id":"2609.30972v1","title":"Factorized axis convolutional gated recurrent unit with dynamic adaptive pooling for remaining useful life prediction of rolling bearings","abstract":"Convolutional neural networks (CNN) are widely used to predict the remaining useful life (RUL) of rolling bearings from time-frequency representations (TFRs) of vibration signals. However, during degradation, characteristic structures in TFRs align predominantly along the frequency or time axis, making it challenging for conventional CNN isotropic kernels to capture directional structure. Furthermore, global average pooling (GAP) averages across axes, potentially obscuring the locations and concentrations of salient activations. This study introduces a factorized-axis convolutional gated recurrent unit (GRU) that employs multiscale anisotropic convolution and dual-axis convolution block attention module to enhance directional features and highlight salient time-frequency regions. Dynamic adaptive pooling (DAP) adaptively aggregates the time-frequency-axis information from the extracted feature maps, whereas a GRU captures temporal dynamics in the latent representations and Monte Carlo dropout enables predictive uncertainty estimation. Experiments on two public bearing datasets demonstrate that the proposed model outperforms existing RUL prediction methods across operating conditions. Ablation experiments demonstrate that the factorized axis-wise design achieves lower mean errors than convolutional isotropic kernels. DAP yields clear improvements on one dataset while matching GAP on the other, highlighting the importance of anisotropic feature extraction and adaptive feature aggregation for TFR-based RUL prediction.","published":"2026-09-25T08:19:02Z","updated":"2026-09-25T08:19:02Z","authors":["Hanbyeol Park","Jungho Choo","Hyerim Bae"],"categories":["cs.AI"],"headline":"New AI method improves bearing life prediction using vibration data","summary":"Researchers developed a new AI model to better predict how long rolling bearings will last by analyzing vibration patterns over time and frequency. They improved existing techniques by focusing on directional features in the data and using adaptive methods to highlight important signals. Testing on public datasets showed their approach often predicts remaining life more accurately than previous methods, though some improvements vary depending on the dataset."},{"id":"2609.30971v1","title":"SciHorizon-eLab: An Agentic Protocol-to-Task Compiler for Scalable Benchmarking of Scientific Embodied Agents","abstract":"Embodied agents offer a promising route to automating scientific experimentation, yet their progress is constrained by the lack of reliable and systematic evaluation environments. Existing simulation-based laboratory benchmarks rely heavily on manual task engineering, making it challenging to systematically compile diverse scientific protocols into executable and verifiable embodied tasks at scale. To address this challenge, we introduce SciHorizon-eLab, an agentic protocol-to-task compiler that formulates scientific embodied task construction as a compilation problem. Given a natural-language protocol of scientific experiments, SciHorizon-eLab progressively compiles laboratory protocols into semantic-preserving embodied tasks through semantic grounding, executable task synthesis, and multi-stage simulation-based certification. The system generates semantically grounded environments, executable manipulation programs, and step-level success specifications, while enabling reproducible generation of expert demonstrations and execution traces. Using this pipeline, we further construct \\BenchName, a ready-to-use benchmark comprising 300 certified tasks across diverse laboratory operations. It supports HIL task execution, reproducible expert-demonstration generation, and ordered step-level evaluation. Across representative tasks, the strongest policy attains an average success rate of only 49.7%, with further evaluations revealing pronounced weaknesses in human and embodied agent coordination. We publicly release the code, benchmark data, and evaluation toolkit at https://github.com/SciHorizon-elab/SciHorizon-elab.","published":"2026-09-25T08:18:45Z","updated":"2026-09-25T08:18:45Z","authors":["Maokai Qin","Chuan Qin","Qi Zhang","Dianyu Liu","Zirui Liu","Hongting Niu","Yuanchun Zhou","Hengshu Zhu"],"categories":["cs.AI"],"headline":"New Tool Translates Science Protocols into Automated Lab Tasks","summary":"Researchers created SciHorizon-eLab, a system that converts written scientific experiment instructions into computer tasks for testing robot agents. This helps create scalable, reliable lab simulations without manual setup. They compiled 300 science tasks into a benchmark and found current robot policies succeeded less than half the time, showing challenges in automated lab work. The researchers provide their code and data publicly for further study."},{"id":"2609.30970v1","title":"Large-deviations theory for growing chemical reaction networks","abstract":"Growing biochemical systems are intrinsically noisy, with one important source of stochasticity arising from the discrete reaction events of the underlying biochemical network. Because growing systems do not generally admit stationary abundance distributions, it is unclear how to separate this intrinsic chemical noise from other sources of variability in population and single-cell data. Here we develop a large-deviation theory for exponentially growing stochastic chemical reaction networks by decomposing abundance into volume and composition. In these coordinates, balanced growth corresponds to a stable composition together with exponentially increasing volume. We show that composition fluctuations satisfy a large-deviation principle with speed equal to the volume of the growing system, and the corresponding quasipotential is selected by a solution of the contact Hamilton--Jacobi equation. We also derive a fluctuation theory for accumulated observables and show that their covariances decay with accumulated-volume, rather than with physical time. Applications to a minimal autocatalytic network and a coarse-grained cellular growth model demonstrate how the theory predicts composition quasipotentials, growth-rate fluctuations, and correlations between observables. The framework therefore provides a stochastic null model for single-cell heterogeneity generated by the intrinsic stochasticity of the underlying chemical reaction network.","published":"2026-09-25T08:17:27Z","updated":"2026-09-25T08:17:27Z","authors":["Praful Gagrani","Ignacio Madrid","Tetsuya J Kobayashi"],"categories":["q-bio.MN","cond-mat.stat-mech","physics.chem-ph"],"headline":"New Theory Explains Noise in Growing Biochemical Systems","summary":"Researchers studied how random fluctuations arise in growing biochemical systems due to the inherent randomness of chemical reactions. They developed a mathematical framework separating changes in system size from changes in composition, showing how fluctuations behave as the system grows. Using this approach, they predicted how variations occur in model chemical networks and cell growth, offering a way to understand variability in single-cell data caused solely by the randomness of chemical processes."},{"id":"2609.30969v1","title":"TACTIC: Understanding Tactile Encoders and Conditioning for Contact-rich Robot Manipulation Policies","abstract":"Tactile information is essential for contact-rich manipulation tasks in robotics. Vision-based tactile sensors make it particularly easy to design end-to-end manipulation policies with tactile sensing, as they enable the use of existing encoders from computer vision. However, this has led to a huge variety of architectures, training datasets, and evaluation protocols, making it difficult to determine which design choices best encode touch. In this work, we address this gap and present a comprehensive study of tactile encoders and fusion strategies across various contact-rich manipulation tasks in real-world experiments. To enable a controlled comparison, we train and evaluate all models under the same pipeline and experimental setup, comprising more than 2000 real-world rollouts. Our results go beyond other studies that only compare simulation performance, which does not necessarily translate to real-world settings, where large-scale evaluations are needed to obtain reliable statistics. Our key finding is that there is no universally optimal representation or fusion strategy for encoding visual-tactile. Instead, the best encoder backbone and fusion scheme depend strongly on the task.","published":"2026-09-25T08:17:23Z","updated":"2026-09-25T08:17:23Z","authors":["Seongjin Bien","Débora Oliveira Makowski","Carlo Kneissl","Reihaneh Mirjalili","Pankhuri Vanjani","Rudolf Lioutikov","Gitta Kutyniok","Florian Walter","Wolfram Burgard"],"categories":["cs.RO"],"headline":"Studying Touch Sensors for Better Robot Handling Skills","summary":"Researchers explored how different ways of processing touch sensor data affect robot performance in tasks that need careful contact. They trained and tested various models on over 2000 real-world trials to fairly compare approaches. Their findings show that no single method works best for all tasks; the choice depends on the specific job, highlighting the challenge of designing one-size-fits-all tactile systems for robots."},{"id":"2609.30968v1","title":"FAVoR: Measuring and Mitigating Author-Style Homogenization in Federated Personalized Generation","abstract":"Large language models are increasingly used as personalized writing assistants, but adapting a model across many authors can compromise individual writing style by pulling author-specific signals toward a shared register. Federated parameter-efficient fine-tuning (PEFT) offers a data-local setting for this multi-author adaptation problem: clients keep author text local while sharing compact adapter updates. However, we show that standard aggregation can preserve continuation utility while making different authors' generations less distinguishable in style space, a failure mode we define as author-style homogenization. We evaluate author-style retention with Angular Style Classification Encoder (ASCE)-based diagnostics on our main BlogText benchmark and ASCE-independent external authorship verification. Using this protocol, we find that common federated PEFT baselines can preserve semantic utility while averaging out author-specific signals. To address this homogenization, we instantiate FAVoR (Federated Authorial Voice Retention), an author-style residual mechanism for federated PEFT. FAVoR uses a shared-private adapter design: clients upload shared-adapter updates while retaining author-specific residual corrections locally. Across BlogText and external Mythos-Reddit validation, FAVoR improves author-style retention over standard and personalized federated PEFT baselines. These gains come with small continuation-utility trade-offs and are supported by component ablations, external verification, and cold-start transfer.","published":"2026-09-25T08:17:11Z","updated":"2026-09-25T08:17:11Z","authors":["Lu Han","Jingyao Zhang","Katy Ilonka Gero","Nguyen H. Tran"],"categories":["cs.CL"],"headline":"Preventing Loss of Personal Writing Styles in Shared AI Models","summary":"Researchers studied how large AI writing assistants, adapted to many authors, tend to blend individual writing styles into a generic one, a problem they call author-style homogenization. They tested this issue using a system where authors keep their text private but share small model updates, finding that standard methods preserve overall writing usefulness but lose distinct author voices. To fix this, the researchers developed FAVoR, a new technique that keeps shared updates while maintaining unique personal style elements locally, improving style retention with minor trade-offs."},{"id":"2609.30967v1","title":"MoMHa: Multi-Objective Optimization of LLM Harnesses over Accuracy, Safety, and Tokens","abstract":"Most work on improving large language models treats accuracy as the sole objective. We argue that the harness, the Python code surrounding the model that constructs prompts, routes calls, and parses outputs, is a first-class design surface whose quality is inherently multi-objective: an accurate harness that refuses no unsafe request, or that consumes an order of magnitude more tokens, is not a good harness. We present Meta-Harness, a system that casts harness design as search over three per-domain objectives (accuracy, behavioural safety, and token cost) solved by an agentic proposer (Claude Code) with full filesystem access to prior harness source, execution traces, and scoring artifacts. Our central finding is that a singlephase joint-reward proposer (MoMHa) outperforms every alternative, including a two-phase \"accuracy then tokens\" ablation, scalar-only feedback, and an accuracy-only baseline. We evaluate on seventeen domains: seven synthetic capability suites, seven real-world public benchmarks (HumanEval, MBPP, Spider, FEVER, MMLU-Pro, LawBench, NuminaMath), and three U-SafeBench-derived user-specific safety domains, using a 12-model fleet spanning four families. On the synthetic track MoMHa achieves a joint mean of 0.482 versus 0.198-0.422 for ten baselines, winning $7 / 10$ per-domain columns; on the real-world track it scores 0.461 versus 0.377 for the strongest baseline (DSPy), winning 5/7 columns, demonstrating that harness strategies transfer to unseen benchmarks without retraining on 8 of 12 target models. MoMHa attains the highest measured behavioral safety composite (U-SafeBench, 0.781) and uses 95 fewer tokens per example than the two-phase alternative. We will release all harness code, evaluation infrastructure, and crossmodel logs.","published":"2026-09-25T08:16:56Z","updated":"2026-09-25T08:16:56Z","authors":["Subhojyoti Mukherjee","Md Mehrab Tanjim"],"categories":["cs.AI"],"headline":"New method improves language model helpers on accuracy, safety, and cost","summary":"Researchers studied the code that manages large language models, called harnesses, which guide how models respond. They developed Meta-Harness, a system that searches for better harness designs by balancing three goals: accuracy, safety, and token usage. Tested across many benchmarks and models, their method called MoMHa outperformed other strategies, achieving better overall results and safer behavior while using fewer tokens. The findings are based on experiments and evaluations detailed in the paper, with all related code to be shared publicly."},{"id":"2609.30966v1","title":"Gradient Surgery for Physics-Informed Neural Networks","abstract":"Physics-Informed Neural Networks (PINNs) are trained by optimising a composite objective that combines data fitting with physics-based constraints, typically resulting in a highly imbalanced multi-task optimisation problem. Under these conditions, existing optimisation strategies are affected by conflicting task gradients, leading to slow convergence and unstable training, particularly for stiff and high-frequency partial differential equations. We analyse gradient conflicts throughout training of PINNs with standard optimiser and investigate Multi-Task Deep Learning (MTDL) optimisation methods. In our analysis across four benchmark problems we observed that PINN optimisation exhibits three distinct phases in which angle- and magnitude-based gradient conflicts alternate, with only one present at a time. Building on these observations, we propose PAM-GS, a physics-aware gradient surgery method that adaptively mitigates task interference during training according to the observed conflict types. Experiments on four representative PDE benchmarks demonstrate that PAM-GS combines competitive solution accuracy with consistently strong task-balanced performance, outperforming existing methods on most problems.","published":"2026-09-25T08:16:23Z","updated":"2026-09-25T08:16:23Z","authors":["Thomas Borsani","Giuseppe Di Fatta"],"categories":["cs.LG"],"headline":"Improving Training of Physics-Informed Neural Networks","summary":"Researchers studied why training physics-informed neural networks (PINNs), which solve equations using both data and physics rules, can be slow and unstable due to conflicts in learning goals. They analyzed these conflicts during training and found different types occurring in phases. To address this, they developed a method called PAM-GS that adapts to these conflicts to improve training balance and speed. Their tests on benchmark problems showed PAM-GS performed better than existing methods, though results are based on these specific simulations."},{"id":"2609.30965v1","title":"FRAM: Trajectory-Guided Visual Feature Selection for Compact Language-Conditioned Robot Manipulation","abstract":"Vision-Language-Action models achieve strong performance in robot manipulation, but often require large numbers of parameters. In this work, we propose the Future Representation Action Model (FRAM), a small policy that explicitly links the future end-effector trajectory to the current visual input. FRAM uses the image coordinates of the predicted trajectory as spatial pointers and reads local visual features related to the motion from the current image. This organizes the information for action generation into the reference position (Where), the visual state (What), and the future motion (Future). Trajectory labels are generated automatically from demonstrations and camera geometry, so no manual annotation is needed. With 138.7M parameters, including a frozen language encoder, FRAM reaches an average success rate of 92.2% over the four standard LIBERO suites, close to the 94.2% of $π_0$ with 3.3B parameters. Without extra training, it also reaches an average of 67.3% on LIBERO-Plus. Ablations confirm that both the future trajectory and the local visual features improve performance and robustness. On a real dual-arm UR5e, FRAM stacks cups using only wrist cameras, including choosing and switching between the left and right arms. These results show that selecting visual information based on future motion is an effective way to obtain both high performance and robustness in a small robot policy.","published":"2026-09-25T08:16:16Z","updated":"2026-09-25T08:16:16Z","authors":["Hiroshi Ito","Hyogo Hiruma","Yoshiki Kanai","Takahiro Yoshida","Akira Kanazawa","Hiroki Yamada"],"categories":["cs.RO"],"headline":"New Robot Model Uses Future Motion to Improve Task Success","summary":"Researchers developed a compact robot control system called FRAM that links future hand movements to current visual input. By focusing on where the robot will move next, it selects relevant visual features to guide actions without needing manual image labeling. Tests show FRAM performs nearly as well as much larger models on standard benchmarks and can stack cups using only wrist cameras, suggesting future motion helps small robot systems work effectively and reliably."},{"id":"2609.30964v1","title":"Dyonic edge modes in Abelian gauge theory","abstract":"We find new boundary conditions in four-dimensional Abelian gauge theory with general Chern--Simons boundary couplings. The boundary conditions allow dyonic edge modes and physical boundary symmetries. In particular, one of the new boundary conditions makes both electric and magnetic charges physical. We also study how our boundary conditions and charges transform under the $\\mathrm{SL}(2,\\mathbb Z)$ duality.","published":"2026-09-25T08:15:37Z","updated":"2026-09-25T08:15:37Z","authors":["Keito Shimizu","Sotaro Sugishita"],"categories":["hep-th"],"headline":"New boundary conditions reveal combined electric and magnetic edge modes","summary":"Researchers studied a type of mathematical physics theory called four-dimensional Abelian gauge theory and discovered new ways to set boundary conditions that allow for both electric and magnetic properties, known as dyonic edge modes, to exist at the boundaries. They also explored how these new boundary setups interact with symmetries and transform under a specific mathematical duality called SL(2,Z). The findings are based on theoretical analysis and provide a new perspective on boundary behavior in this field."},{"id":"2609.30963v1","title":"Where and When to Force: Routed Forcing for Streaming Avatars","abstract":"Audio-driven streaming avatar generation requires real-time synthesis of speech-synchronized videos with dynamic and diverse motion. Self Forcing uses Distribution Matching Distillation (DMD) to distill bidirectional video diffusion models into causal, few-step generators for real-time streaming. However, DMD minimizes a reverse KL divergence, which is inherently mode-seeking: it causes the student to discard high-dynamic modes and collapse onto static outputs, compressing both dynamics and diversity of generated videos. We find that this collapse is region-heterogeneous: person regions involving pose and gesture variations suffer the largest diversity loss, the audio-driven mouth region shows a small loss, and the background remains nearly stable. Based on this observation, we propose Routed Forcing, which routes the distillation objective by semantic region and noise stage to improve dynamics and diversity while preserving visual quality. Specifically, (1) Where to Force: Semantic-Region Routing applies Data-Forcing Distillation (DFD), which supervises the student with real videos, to the person region where diversity collapse is most severe, while retaining DMD for the mouth and background to preserve lip synchronization and scene stability. (2) When to Force: Noise-Stage Routing activates DFD at high noise stages, where real video serves as effective supervision to inject diverse and dynamic motion patterns. At low noise stages, DMD is used to refine details, avoiding blur and artifacts from spatial differences between real video and student-generated video. Experiments show that Routed Forcing improves dynamics by up to 45% and diversity by 7-25% over Self Forcing, while preserving video quality and lip synchronization.","published":"2026-09-25T08:15:11Z","updated":"2026-09-25T08:15:11Z","authors":["Zihan Su","Siwen Lu","Junhao Zhuang","Zeyue Xue","Haoyang Huang","Guanghao Li","Xiaofeng Tan","Chun Yuan","Nan Duan"],"categories":["cs.CV"],"headline":"New method improves lifelike motion in real-time talking avatars","summary":"Researchers studied how to make audio-driven avatars move more naturally and with diverse, dynamic motions in real-time videos. They found that previous methods tended to produce less varied movements, especially in body poses and gestures, while keeping lip-sync accurate. To address this, they developed Routed Forcing, which applies different training approaches to separate video regions and noise levels. Their experiments showed improved motion dynamics and diversity without losing video quality or lip synchronization. These findings are based on simulations and testing as described by the researchers."},{"id":"2609.30962v1","title":"IDM-Net: A Lightweight Illumination-Decoupled Modulation Network for Low-Light Image Enhancement","abstract":"Low-light image enhancement (LLIE) remains challenging for lightweight models because illumination restoration and color fidelity are difficult to optimize simultaneously in the RGB color space. Although recent color-decoupled methods separate luminance and chrominance representations, they primarily optimize luminance as an enhancement target, leaving its potential as an explicit guidance prior largely unexplored during feature reconstruction. To address this limitation, we propose IDM-Net, a lightweight Illumination-Decoupled Modulation Network for low-light image enhancement. IDM-Net adopts a dual-encoder architecture consisting of a structure encoder that extracts multi-scale appearance features from the RGB image and a lightweight illumination encoder that learns illumination priors from the decoupled luminance (Y) channel. To effectively exploit these priors, we introduce an Illumination-Guided Modulation (IGM) module that injects multi-scale illumination cues into the decoder through spatially adaptive affine modulation, enabling accurate brightness restoration while preserving natural color consistency. Furthermore, we design a lightweight Feature Refinement Block (FRB) to progressively suppress degradation artifacts and recover fine-grained image details during reconstruction. Extensive experiments on multiple standard low-light image enhancement benchmarks demonstrate that IDM-Net achieves competitive performance among lightweight LLIE methods while maintaining an excellent balance between restoration quality and computational efficiency.","published":"2026-09-25T08:14:57Z","updated":"2026-09-25T08:14:57Z","authors":["Cheng-Yen Hsiao","Jing-Ming Guo"],"categories":["cs.CV"],"headline":"New Lightweight Method Improves Dark Photo Brightness and Color","summary":"Researchers developed IDM-Net, a compact computer model that enhances low-light photos by separately analyzing brightness and color details. Their method uses two parts: one to capture the image's structure and another to focus on lighting information, which guides the brightening process. Tests show IDM-Net balances good image quality with low computer demands, but results are based on standard benchmarks and do not guarantee real-world performance."},{"id":"2609.30961v1","title":"Rough spectral asymptotics for commutators of general singular integrals","abstract":"Recently, Schatten class membership of commutators arising from Connes' quantised calculus has been characterised in a general framework, which covers multiple concrete situations of interest. In contrast to this, related results dealing with the asymptotic behaviour of the singular values of these commutators have been restricted to a relatively short list of specific examples only. In this work, we show that a version of the spectral asymptotic formula, involving comparable lower and upper limits in place of an exact limit, remains valid in great generality of singular integrals over metric measure spaces. A key proof ingredient of independent interest is a new asymptotic version of the Marcinkiewicz interpolation theorem. We also present a new approach to commutator upper bounds at the critical index; while not quite as general as more elaborate methods, it is general enough to reproduce the known upper bounds in Heisenberg and Carnot groups in a much simpler way.","published":"2026-09-25T08:13:04Z","updated":"2026-09-25T08:13:04Z","authors":["Tuomas Hytönen"],"categories":["math.FA"],"headline":"New insights into behavior of operators called commutators in math","summary":"Researchers studied the asymptotic behavior of singular values of certain mathematical operators known as commutators, which arise in a field related to quantum calculus. They extended previous results beyond a few specific examples to a broad class of singular integrals on metric spaces. Their work includes a new asymptotic version of an important mathematical tool, the Marcinkiewicz interpolation theorem, and introduces a simpler method to estimate operator bounds in certain groups. The results are theoretical and apply to abstract mathematical objects rather than practical applications."},{"id":"2609.30960v1","title":"Packet iSlip","abstract":"This paper examines input/output buffered crossbar switches under combined packet and cell data. Our switch architecture uses input buffering with Virtual Output Queues to avoid Head of Line Blocking. The switch fabric is a crossbar with no speedup. We use a modified iSlip [McKeown] crossbar scheduler geared towards packet data, called piSlip. Cell ports are largely unmodified from standard iSlip behavior. For packet output ports, we introduce changes to the grant pointer which minimizes output latency caused by packet reassembly. Our model uses several output states, including packet cut-through. From simulation results, we show that piSlip with virtual cut-through offers latency characteristics significantly better than unmodified iSlip with similar packet port interfaces. Simulation further shows that piSlip and iSlip have similar maximum and average buffering requirements.","published":"2026-09-25T08:13:03Z","updated":"2026-09-25T08:13:03Z","authors":["Marc Mosko"],"categories":["cs.NI"],"headline":"New Switch Design Improves Packet Data Latency in Simulations","summary":"The researchers studied a type of network switch that handles both packet and cell data. They modified an existing scheduling method (iSlip) to create piSlip, aiming to reduce delays when packets are reassembled at the output. Their computer simulations show that piSlip with a new 'virtual cut-through' approach reduces latency compared to the original method, while requiring similar buffer sizes. These findings are based on simulation models and have not been tested in real-world hardware."},{"id":"2609.30959v1","title":"VisTacAlign: Co-Training Dexterous Policies on Tactile Human and Robot Demonstrations","abstract":"Human demonstrations are a cheap source of data for dexterous manipulation, but co-training a robot policy on them requires closing the human--robot gap in every modality the policy consumes. We present VisTacAlign, a framework for co-training 3D-visual-tactile dexterous policies on human and robot demonstrations. Glove-tracked human hand motion is retargeted to a 17-DoF tactile robot hand with a one-time fingertip correction. The human hand is then erased from both stereo views and replaced by a posed robot-hand mesh painted with pixels from robot recordings, and a real-time stereo foundation model is re-run on the composite, so the human point clouds carry the same stereo errors and visibility as the robot ones. Finally, a capacitive tactile glove is aligned to the robot's fingertip sensors in its signal space, giving one interpretable per-finger force representation. A diffusion transformer consumes point-cloud, proprioceptive, and per-finger tactile tokens. On three real-world tasks requiring precise force -- Lego assembly, plucking strawberries of varying size, and activating and lifting a power drill -- adding aligned human demonstrations to existing robot data improves over robot-only policies, and ablations show that both tactile input and visual alignment are necessary. Project page: https://vis-tac-align.github.io","published":"2026-09-25T08:12:04Z","updated":"2026-09-25T08:12:04Z","authors":["Julien Poffet","Matthew Strong","Ankush Dhawan","Baiyu Shi","Shalika Neelaveni","Yujia Yuan","Zhenan Bao","Monroe Kennedy"],"categories":["cs.RO"],"headline":"Improving Robot Hand Control Using Aligned Human and Robot Touch Data","summary":"The researchers address the challenge of teaching robots precise hand movements by using demonstrations from both humans and robots. They developed VisTacAlign, a method that adjusts human hand data to match robot sensors and vision, allowing combined learning from both sources. Their approach improves robot performance on tasks requiring careful touch, like assembling Lego and picking strawberries, and they find that matching both touch and visual data between humans and robots is important for success."},{"id":"2609.30958v1","title":"Active-Space Quantum Simulation of N$_2$ Hydrogenation at a Ru Single-Atom Site on Ru(0001)","abstract":"Selective activation of dinitrogen (N$_2$) under mild conditions is difficult: N$\\equiv$N is one of the strongest bonds in chemistry, and most heterogeneous catalysts capable of breaking it require high temperature and pressure. Atomically dispersed transition-metal sites offer a computationally tractable route to studying the strongly correlated intermediates involved. We connect periodic DFT calculations with correlated active-space calculations on a finite, non-periodic surface fragment, demonstrating the workflow for the hydrogenation step RuH$_2$(N$_2$)* $\\rightarrow$ RuH(NNH)* at an isolated Ru$_1$ site on Ru(0001). A first-shell fragment is extracted from the periodically relaxed structure, an active space is selected with Active Atomic Valence Space (AVAS) around the Ru-N/N-H bond reorganization, and natural-orbital truncation reduces its cost while retaining the relevant correlated degrees of freedom. The reduced Hamiltonian is mapped to qubits with the Jordan-Wigner transformation and solved with the adaptive derivative-assembled pseudo-Trotter variational quantum eigensolver (ADAPT-VQE). Dynamic correlation beyond the active space is examined with strongly contracted NEVPT2 and DSRG-MRPT2. AVAS gives a 22-qubit active space, which natural-orbital truncation compresses to 16 qubits, reproducing the untruncated CASCI state energies to within 0.21 kcal/mol (reaction-energy error 0.2 kcal/mol). Statevector ADAPT-VQE in this reduced space converges for both reaction states under a fixed pool-gradient stopping criterion. NEVPT2 yields anomalously large, state-imbalanced corrections for the finite Ru fragment, while DSRG-MRPT2 retains an endothermic reaction energy at its default flow parameter, although its magnitude depends strongly on that parameter.","published":"2026-09-25T08:10:43Z","updated":"2026-09-25T08:10:43Z","authors":["Geet Gupta"],"categories":["physics.chem-ph","quant-ph"],"headline":"Quantum simulation studies breaking N2 molecule on ruthenium surface","summary":"Researchers studied how to break the very strong bond in nitrogen gas (N₂) under mild conditions using a single ruthenium atom on a surface. They combined standard calculations of the surface with advanced quantum simulations focused on the chemical bonds involved in adding hydrogen to nitrogen. Their method reduced the complexity of the problem to make it solvable with quantum algorithms, and they tested different ways to improve the accuracy of their results. The findings show a promising computational approach but are based on simulations of small surface fragments, so practical implications remain uncertain."},{"id":"2609.30957v1","title":"Towards Understanding Momentum Acceleration in River-Valley Loss Landscape","abstract":"The empirical success of pretraining large language models has inspired a deeper investigation into the underlying loss landscapes and the optimization dynamics. Recent empirical and theoretical study suggest that the training loss landscape often exhibits a \"river-valley\" structure, which features a low-loss manifold (river) flanked by sharp orthogonal directions with higher loss (mountains). In the long term, the optimization progress is determined primarily by the progress along the river. Within such a landscape, gradient descent with large learning rates can move faster along the river despite high apparent loss due to vertical oscillations, while a subsequent sharp decay in the learning rate suppresses these oscillations, revealing genuine optimization progress. This explains the recent success of warmup-stable-decay (WSD) learning rate scheduler which, unlike cosine scheduling, keeps stable high learning rate and decays before producing intermediate checkpoints. Building on this foundation, in this work we take a step further and study the role of momentum within such a loss landscape. We establish theoretical analysis that characterizes how momentum accelerates optimization by stabilizing large learning rates that can not be tolerated by vanilla GD without deviating significantly from the river. The enabled large learning rate in-turn gives greater speed along the river and makes faster essential progress in the long run. Another intriguing observation from theory is that for a river-valley landscape with very flat and slow-spinning river, the momentum itself does not contribute directly to acceleration in terms of the speed of tracking the river, while the main acceleration comes from the admissible larger learning rate.","published":"2026-09-25T08:09:59Z","updated":"2026-09-25T08:09:59Z","authors":["Miao Lu","Zeyu Bian","Kaiyue Wen","Beining Wu","Siyu Chen","Tianhao Wang","Zhiyuan Li"],"categories":["cs.LG","math.OC"],"headline":"How Momentum Helps Speed Up Learning in Complex Model Training","summary":"Researchers studied how momentum affects optimization when training large language models, focusing on a \"river-valley\" loss landscape. They provide a theoretical analysis showing momentum allows stable use of larger learning rates, which helps move faster along the low-loss \"river\" and improves long-term progress. However, in very flat landscapes, momentum itself does not directly speed up progress; instead, the acceleration mainly comes from using a larger learning rate that momentum enables."},{"id":"2609.30956v1","title":"Attacking Diophantus: Special Cases of Bag Containment","abstract":"Query containment is a fundamental decision problem in database theory: given two queries, determine whether, over all database instances, every answer produced by the first is also produced by the second. For conjunctive queries under set semantics, the problem is understood through the classical homomorphism-based characterisation. Under bag semantics, the interpretation underlying real relational databases, containment becomes a quantitative comparison of answer multiplicities. Despite decades of work, the decidability of bag containment for conjunctive queries remains open. This frontier is fragile: for slightly more expressive classes, bag containment is undecidable, with negative results relying on reductions from variants of Hilbert's 10th problem. This work develops a unified framework for bag containment of conjunctive queries that subsumes two previously studied decidable cases: projection-free and join-on-free containee queries. The framework yields decidability for a broader class, called join-uniform queries, while leaving the containing query arbitrary. This contrasts with techniques that impose restrictions on the containing query. The approach identifies tractable classes based on the internal unification structure of the query whose multiplicities must be bounded. Specifically, it reduces containment to a controlled Diophantine problem. Starting from the containee query, one builds a canonical model generated by all its possible unifications, over which multiplicities admit a finite arithmetic characterisation. Containment is proved equivalent to the non-existence of solutions of a corresponding Diophantine inequality system. Although these problems are undecidable in general, we show that the systems arising from join-uniform containment form a decidable subclass. Thus, the standard source of undecidability for bag containment becomes the core of the decision procedure.","published":"2026-09-25T08:07:30Z","updated":"2026-09-25T08:07:30Z","authors":["George Konstantinidis","Xinzhuo Li","Fabio Mogavero"],"categories":["cs.DB"],"headline":"New Method Helps Decide Complex Database Query Containments","summary":"Researchers study the problem of determining when one database query's results are always included in another's, considering how many times answers appear (bag semantics). They developed a new framework that extends known decidable cases to a broader group called join-uniform queries by linking the problem to solving certain math equations called Diophantine inequalities. Although such equations are generally hard to solve, the researchers identify a special subclass where these problems become decidable, advancing understanding of when query containment can be reliably determined."},{"id":"2609.30955v1","title":"Flow-TAG: Flow-based conditional latent transport for accurate spline approximation and data compression","abstract":"Robust curve fitting is essential in computer-aided design for transforming noisy, discrete data into accurate geometric models that ensure numerical stability across engineering workflows. B-spline models have become the industry standard for this task, offering a flexible and reliable framework characterized by local control and smooth shape representation. This paper presents flow-TAG--a data-driven framework based on a generative flow model with a 1D U-Net backbone capable of mapping the geometry of a curve to the optimal parametrization for cubic B-splines. By leveraging learned geometric patterns, flow-TAG exhibits superior parameterization performance, robustness to noise in the input data, and strong generalization capability to previously unseen 2D and 3D curves drawn from distinct data distributions. Flow-TAG yields fitted curves that achieve the lower root-mean-square error (55% lower on average) and Hausdorff distance (52% lower on average) relative to state-of-the-art data-driven methods. In addition, we investigate the practical applicability of our generative framework in the compression of ECG signals for wearable devices. The proposed compression setup provides a compression ratio of 13 with the signal distortion of around 5%, which is acceptable in the field.","published":"2026-09-25T08:06:00Z","updated":"2026-09-25T08:06:00Z","authors":["Roman Pavelkin","Luis A. Zavala-Mondragon","Fons van der Sommen"],"categories":["cs.CG"],"headline":"New method improves curve fitting and ECG data compression","summary":"Researchers developed flow-TAG, a method that uses machine learning to better fit curves from noisy data, important in design and engineering. Their approach improves accuracy compared to existing methods and works well on both 2D and 3D curves. They also tested flow-TAG for compressing ECG signals in wearable devices, achieving good compression while keeping distortion low. These findings are based on their data-driven model and simulations, showing promise but not clinical validation."},{"id":"2609.30954v1","title":"FTB Graph: Determining and Validating First-token Broadcasters and Language-Identity Head Circuits in Multilingual Language Models","abstract":"Large language models operating in multilingual contexts must resolve target response languages early in generation, yet the causal circuitry governing first-token language identity decisions remains poorly mapped. We present an end-to-end structural circuit analysis across six model architectures spanning four families: GPT-2, BLOOM-560M, Pythia-1B/2.8B, and Qwen2.5-1.5B Base/Instruct. Using Edge Attribution Patching (EAP) with FP16 active clamping, followed by exact activation patching verification with a 2,000-candidate-edge search ceiling, we extract directed acyclic graphs driving first-token language broadcasting. Across the standalone models, we observe deep or mid-to-deep broadcasting hubs, though the evidence is strongest for Pythia-2.8B and BLOOM-560M because GPT-2 and Pythia-1B leave few out-of-graph heads for comparison, while both Qwen2.5-1.5B variants invert the necessity check. Scaling from Pythia-1B to 2.8B expands node participation while maintaining a similar verified edge budget, producing sparser topology. The Qwen2.5-1.5B base and instruct circuits retain 84.7% Jaccard similarity, including the Layer 27 hub, indicating that first-token routing is largely established during pretraining and preserved by instruction tuning. Finally, EAP scores correlate weakly with exact patching deltas across most models, showing that linear gradient approximations can diverge from causal interventions in FP16 and motivating exact-patching verification for reliable circuit discovery.","published":"2026-09-25T08:05:03Z","updated":"2026-09-25T08:05:03Z","authors":["Arjun Pillai","Christian Hoang","Anjelo Laroza"],"categories":["cs.AI"],"headline":"Researchers Analyze Language Choice in Multilingual AI Models","summary":"The researchers studied how large multilingual AI language models decide which language to generate first in their responses. They examined six different models using advanced techniques to map the internal connections responsible for this language choice. Their findings suggest that specific parts of the models play key roles in this process, and that some early training effects persist after further tuning. However, the research is based on simulations and internal model analysis, and results vary between models, emphasizing the need for careful verification methods."},{"id":"2609.30953v1","title":"CoCoRerank: Towards Conventional Commit Message Generation by Component and Candidate Consistency Reranking","abstract":"Commit messages are essential for understanding software changes, yet automatic commit message generation typically treats a message as an unstructured text sequence. This limits its ability to support standardized development workflows, where commit messages are often expected to follow the Conventional Commits Specification (CCS) in the form type (scope): subject. In this paper, we study conventional commit message generation under the complete CCS format. We construct a new benchmark of 86,688 high-quality commits collected from open-source GitHub repositories, with each message normalized into type, scope, and subject through structural normalization and semantic quality filtering. Based on this benchmark, we propose a two-dimensional consistency-based reranking framework named CoCoRerank for LLM-based generation. CoCoRerank exploits horizontal consistency among the code change, type, scope, and subject, as well as vertical consensus across multiple generated candidates. Experiments with representative CMG baselines, LLM generators, reranking strategies, and ablation variants show that CoCoRerank improves both structural component prediction and subject generation quality. The results demonstrate that complete CCS supervision and multidimensional consistency modeling provide an effective foundation for accurate and standardized commit message generation. The artifact is publicly released at https://github.com/bluewhalebug/CoCoRerank.","published":"2026-09-25T08:04:37Z","updated":"2026-09-25T08:04:37Z","authors":["Shaopeng Jia","Yali Du","Ming Li"],"categories":["cs.SE"],"headline":"Improving Software Commit Messages with Consistency-Based AI Reranking","summary":"Researchers addressed the challenge of automatically creating clear and standardized commit messages that describe software changes. They built a large dataset of well-structured commit messages following a common format and developed a new method called CoCoRerank, which improves message quality by ensuring consistency between components and among multiple generated options. Their experiments suggest this approach helps generate more accurate and standardized messages, but findings are based on selected datasets and AI model evaluations detailed in the paper."},{"id":"2609.30952v1","title":"MVVBench: Benchmarking 4D Reasoning in Vision-Language Models","abstract":"Multi-view video understanding requires integrating spatial and temporal evidence across multiple, often non-overlapping camera streams: tracking entities as they transition between viewpoints, aligning events across time, and reasoning about latent 4D continuity rather than any single visible frame. We introduce MVVBench, a benchmark for multi-view video reasoning built from real world multi camera datasets. Questions are curated to be monocular-ambiguous along both the view and the temporal axis: each question is unanswerable from any single view in the designated input set, and the majority are further unanswerable from any single moment. Each question becomes uniquely solvable only by jointly reasoning across views and across time. MVVBench spans diverse dynamic scenes and probes six capabilities: implicit/explicit attribute identification, implicit/explicit relative distance, relative camera pose, and compositional counting, with human-authored QA and rigorous verification. Beyond benchmarking, we provide an extensive analysis of when and why current vision language models succeed or fail, characterizing errors due to temporal mis-localization, cross-view identity breaks, and brittle multi-hop reasoning. We then study inference-time elicitation strategies that unlock latent multi-view competence---task-specific chain-of-thought scaffolds and structured cross-view evidence aggregation---yielding substantial gains without retraining. Finally, we present preliminary evidence that reinforcement learning with verifiable rewards can elicit some latent multi-view competence in the base model, pointing to training-time approaches as a promising direction for future work. Together, MVVBench offers a rigorous evaluation of 4D multi-view reasoning and a foundation for future progress toward reliable embodied perception.","published":"2026-09-25T08:02:42Z","updated":"2026-09-25T08:02:42Z","authors":["Hyungjin Chung","Byeongjun Park","Joonseok Lee","Hojun Kim","Jaeho Choi","Byung-Hoon Kim"],"categories":["cs.CV","cs.AI"],"headline":"New Test Measures 4D Understanding in Multi-Camera Video Models","summary":"Researchers developed MVVBench, a new test to evaluate how well computer models understand videos from multiple cameras over time. The test includes questions that can only be answered by combining views from different cameras and moments, mimicking real-world challenges like tracking objects as they move. Their study shows current models often struggle with errors in timing, identity tracking, and complex reasoning, but certain techniques improved performance. The research highlights the difficulty of this problem and suggests new future training methods to help models better understand multi-view video data."},{"id":"2609.30951v1","title":"Bundled Contact Gradients: Stabilizing Differentiable Simulation for Deployable Dynamic Tasks","abstract":"Differentiable simulation provides analytic gradients of robot dynamics, enabling fast and sample-efficient first-order policy optimization. However, obtaining smooth and informative gradients through rigid-body contact typically requires softened contact models, often at the expense of physical fidelity and thereby limiting learned policies largely to simulation. This trade-off becomes particularly consequential for dynamic humanoid motions, where accurate contact dynamics are critical for transferring policies to the real world. Increasing contact stiffness in rigid-body simulation improves the fidelity of interactions, but also makes the dynamics increasingly sensitive to small state perturbations, producing high-variance gradients that can destabilize first-order policy learning. To address this, we propose \\emph{Bundled Contact Gradients (BCG)}, a contact-local randomized smoothing framework for differentiable policy learning. When stiff contact is detected, our method evaluates a local bundle of randomized perturbation rollouts around the stiff contact configuration and aggregates their gradient signal thereby reducing gradient variance. We demonstrate the effectiveness of our method by successfully training and transferring dynamic motions zero-shot onto a real-world Unitree G1 humanoid platform. Videos and supplementary information can be found at https://bundledcontactgradients.github.io/","published":"2026-09-25T08:02:29Z","updated":"2026-09-25T08:02:29Z","authors":["Dyuman Aditya","Jin Cheng","Clemens Schwarke","Quan Nguyen","Gaurav Sukhatme","Stelian Coros","Gabriele Fadini"],"categories":["cs.RO"],"headline":"New method improves robot motion learning with realistic contact dynamics","summary":"Researchers studied how robots learn to move dynamically, which requires precise contact with surfaces. They developed a technique called Bundled Contact Gradients that reduces unstable signals when modeling stiff contacts in simulations. This approach helped train policies that successfully transfer from simulation to a real humanoid robot, improving physical realism without losing learning stability. Their results come from simulations and real-world tests with a specific robot platform, as described in their abstract."},{"id":"2609.30950v1","title":"Low-Bit Recurrent States in Hybrid Language Models","abstract":"Hybrid language models maintain fixed-size recurrent states, but existing quantizers typically use eight bits or more. Quantization errors persist according to channel decay rates. We derive distortion weights from the observability Gramian and combine them with normalized state ranges for mixed-precision bit allocation, without calibration data, rotation, or training. We also quantize decay rates logarithmically. With per-token state quantization, a four-bit mean payload reduces excess negative log-likelihood by factors of 3.3--27.9 relative to the best of seven baselines across three hybrid models; metadata costs vary. At six bits, negative log-likelihood differs from the FP32-state baseline by less than 0.005 nats. Ablations separate gains from variable bit widths, decay weighting, and range normalization. With less frequent write-backs, gains diminish and depend on the model and budget.","published":"2026-09-25T08:01:56Z","updated":"2026-09-25T08:01:56Z","authors":["Hongren Chen","Jiayang He"],"categories":["cs.LG"],"headline":"Reducing Memory Size in Language Models Using Low-Bit States","summary":"Researchers studied how to reduce the size of internal memory states in language models by using fewer bits per value, which typically use eight bits or more. They developed a method to allocate bits more efficiently without extra training or calibration, leading to lower errors in the models' predictions. Their results, based on simulations with three hybrid models, show significant improvements in accuracy with four to six bits, though benefits depend on how often memory is updated and vary by model."},{"id":"2609.30949v1","title":"A unified approach to nonlinear Stein theorems","abstract":"We study regularity for the variable exponent quasilinear $\\mathfrak{p}\\left(x\\right)$-Laplace type equation on domains in Heisenberg groups and Euclidean spaces. We establish borderline continuity estimates for the appropriate first order derivatives of solutions. More precisely, we show that for any weak solution $u \\in \\mathbb{E}W^{1, \\mathfrak{p}(\\cdot)}\\left(Ω\\right)$ of \\begin{align*} \\operatorname{div}_{\\mathbb{E}} \\left( \\mathfrak{a}(x) \\lvert \\nabla_{\\mathbb{E}} u \\rvert^{\\mathfrak{p}\\left(x\\right)-2} \\nabla_{\\mathbb{E}} u \\right) = f \\qquad \\text{ in } Ω, \\end{align*} where $Ω\\subset \\mathbb{E}^{n}$, $\\mathfrak{a}$ is a uniformly positive bounded scalar function and the exponent function $\\mathfrak{p}$ is uniformly bounded away from $1$ and $\\infty,$ $\\nabla_{\\mathbb{E}}u$ is continuous in $Ω$ as soon as $f \\in L^{\\left( Q_{\\mathbb{E}}, 1\\right)}\\left(Ω\\right)$ and $\\mathfrak{a}, \\mathfrak{p}$ satisfies some conditions regarding the summability of their mean-oscillations. Here $\\mathbb{E}^{n}$ is either the Heisenberg group $\\mathbb{H}_{n}$ or the Euclidean space $\\mathbb{R}^{n}$ and $\\operatorname{div}_{\\mathbb{E}}$, $\\nabla_{\\mathbb{E}}$, $Q_{\\mathbb{E}}$ stands for the corresponding divergence, gradient and homogeneous dimension, respectively. We treat the elliptic and subelliptic cases in a unified manner using Euclidean techniques and our conditions on $\\mathfrak{a}$ and $\\mathfrak{p}$ are new and weaker than all the known sufficient conditions even in the Euclidean case. However, all the known sufficient conditions imply our conditions, achieving yet another unification.","published":"2026-09-25T08:01:07Z","updated":"2026-09-25T08:01:07Z","authors":["Arka Mallick","Swarnendu Sil"],"categories":["math.AP","math.FA"],"headline":"New unified method improves understanding of complex equations","summary":"Researchers studied a complicated type of equation that appears in special geometric spaces and regular Euclidean space. They developed new, less restrictive conditions under which certain derivatives of the solutions smoothly vary, improving upon earlier requirements. Their work unifies previous methods for both standard and more complex settings, using traditional Euclidean techniques, but it remains theoretical and focused on mathematical properties rather than practical applications."},{"id":"2609.30948v1","title":"PORL: Pretrained Offline Reinforcement Learning for the Job Shop Scheduling Problem","abstract":"The Job Shop Scheduling Problem (JSSP) is a fundamental combinatorial optimization problem in industrial optimization. This work introduces Pretrained Offline Reinforcement Learning (PORL), a hybrid approach that combines simulation-based online pretraining with offline fine-tuning on production-specific data. Reinforcement learning through online interaction enables exploration of general scheduling strategies, but typically relies on simulation environments and may suffer from a simulation-to-reality gap. In contrast, offline RL avoids direct interaction with the environment by learning from historical data, but its performance is strongly influenced by dataset quality and coverage. PORL combines the strengths of both paradigms by first learning a general scheduling policy through online interaction and subsequently adapting it offline to a target distribution. A KL-divergence-based policy constraint is introduced to limit deviations from the pretrained policy during fine-tuning. The approach is evaluated on JSSP instances with distribution shift and datasets generated from heuristic, noisy-expert, and random behavioral policies. The results show that PORL consistently achieves lower optimality gaps than standalone offline RL and the considered general scheduling baselines. Furthermore, its advantage over standalone offline RL increases as dataset quality decreases, indicating reduced sensitivity to the quality and coverage of the available offline data. The results suggest that offline adaptation of pretrained policies is a promising approach for industrial scheduling environments where direct online exploration is impractical.","published":"2026-09-25T08:00:23Z","updated":"2026-09-25T08:00:23Z","authors":["Mateo Toro Diz","Jonathan Hoss","Noah Klarmann"],"categories":["cs.LG","cs.AI"],"headline":"New Method Improves Industrial Job Scheduling Using AI and Data","summary":"Researchers studied a common industrial problem of scheduling jobs efficiently. They developed a method called PORL that first learns general scheduling strategies through simulated practice, then fine-tunes these using real historical data without needing direct trial-and-error in actual production. Their tests showed PORL outperforms other approaches, especially when available data is limited or lower quality. These findings are based on simulations and suggest potential for improving scheduling where direct real-world experimentation is difficult."},{"id":"2609.30947v1","title":"DAPEVO: Deep Adaptive Patch Frame-Event Visual Odometry","abstract":"Visual odometry is essential for autonomous navigation in GPS-denied environments, yet RGB-based methods remain vulnerable to motion blur, challenging illumination, and dropped frames. Event cameras complement conventional cameras with high temporal resolution and dynamic range, but their asynchronous measurements complicate reliable correspondence estimation. We present DAPEVO, a learned visual odometry system that estimates image and event correspondences independently at shared patch locations and fuses their correlation evidence before motion refinement. Each tracked patch maintains image and event descriptors, and a learned scalar gate combines modality-specific correlation embeddings for each patch--frame edge before a shared recurrent refinement and bundle-adjustment update. DAPEVO also supports event-only observations, enabling continued tracking when RGB frames are sparse or unavailable, while modality-aware keyframe culling preserves scarce frame constraints. On UZH-FPV, when retaining only one in six RGB frames, DAPEVO's mean absolute trajectory error (ATE) increases by only 36%, from 1.00 to 1.36m, whereas the ATE of DPVO and RAMP-VO rises by factors of $3.7\\times$ and $3.1\\times$, respectively. On TartanEvent, DAPEVO similarly remains below 1m ATE at 3Hz RGB input, while DPVO and RAMP-VO exceed 9m. Under degraded RGB input on TartanEvent, DAPEVO achieves an ATE of 0.60m, compared with more than 4m for both DPVO and RAMP-VO, while also outperforming event-only DEVO at 0.87m.","published":"2026-09-25T07:59:28Z","updated":"2026-09-25T07:59:28Z","authors":["Luca Gandolfi","Simone Nascivera","Roberto Pellerito","Rong Zou","Chiara Plizzari","Davide Scaramuzza"],"categories":["cs.CV"],"headline":"New Method Improves Robot Navigation Using Camera and Event Data","summary":"Researchers developed DAPEVO, a system combining regular images and special event camera data to better track movement when GPS is unavailable. Unlike existing methods, DAPEVO analyzes both types of data separately and then merges them, allowing it to keep accurate positions even with fewer or lower-quality images. Tests show DAPEVO reduces errors in trajectory estimation compared to other methods, especially when image data is limited or degraded, although results are based on specific datasets and simulations."},{"id":"2609.30946v1","title":"OneWorld: Learning Consistent Physics Across Actions in World Models","abstract":"Action-conditioned video world models aim to predict scene evolution under different actions, a capability that is essential for reliable planning, decision-making, and interaction in dynamic environments. However, futures generated independently from the same initial scene may each appear plausible while implying incompatible physical properties, such as friction or mass. This inconsistency can lead to contradictory predictions across interventions, making it difficult for the model to maintain a coherent understanding of the underlying world and limiting its reliability for planning and decision-making. To address these issues, we propose OneWorld, a shared-mechanism counterfactual generation framework that jointly models multiple action-conditioned futures under a common latent physical mechanism. A physical mechanism interpreter first infers a distribution over latent mechanisms from each action-outcome branch. These distributions are then aggregated into shared-world evidence, which captures whether the branches admit a common physical explanation while accounting for uncertainty in less informative branches. This evidence constrains flow training and guides sampling, encouraging consistency in the underlying physical mechanism while preserving the distinct outcomes induced by different actions. We further introduce a multi-intervention evaluation protocol in controlled environments, following the interaction settings of ACWM-Phys, to assess whether generated futures can be jointly explained by the same physical parameters, alongside standard measures of single-rollout prediction quality. Experiments in these environments show that OneWorld improves cross-intervention physical consistency while maintaining competitive single-rollout prediction quality.","published":"2026-09-25T07:59:11Z","updated":"2026-09-25T07:59:11Z","authors":["Ke He","Yichen Ding","Bin Yang"],"categories":["cs.CV","cs.AI"],"headline":"New Method Helps Models Predict Consistent Physics in Videos","summary":"Researchers developed OneWorld, a method to improve how models predict future video scenes after different actions. Unlike previous models that might predict believable but physically inconsistent outcomes, OneWorld uses a shared mechanism to keep physical properties consistent across different possible futures. Tests in controlled environments showed that OneWorld helps models maintain coherent physical explanations for varied actions, potentially making predictions more reliable for planning and decision-making, though results are based on simulations."},{"id":"2609.30945v1","title":"Coupled Meta-Adaptive Filtering for Active Noise Control Under Time-Varying Acoustic Paths","abstract":"Meta-adaptive filtering (Meta-AF) provides a data-driven alternative to hand-crafted adaptive filter updates by employing a learned optimizer throughout online adaptation. However, when Meta-AF is used for active noise control (ANC), time-varying acoustic paths remain a major challenge. In particular, the physics-informed optimizer features are constructed using a secondary path estimate and become mismatched when the physical path changes, leading to inaccurate filter updates and degraded noise reduction. To address this problem, this paper proposes a Coupled Meta-Adaptive Filtering Active Noise Control (CoMeta-AF-ANC) method, which applies meta-learning to jointly learn the control filter adaptation and acoustic path tracking within a closed-loop framework. The proposed Meta-Gated Joint Path Identifier (MG-JPI) simultaneously tracks the primary and secondary paths from the available ANC signals without auxiliary noise, while the updated secondary path estimate is fed back to reconstruct the Meta-AF controller features. A delayless dual-rate realization performs learned adaptation at the frame rate while generating the control signal at the sampling rate in the time domain. Evaluation using measured headrest acoustic paths shows that CoMeta-AF-ANC outperforms representative ANC algorithms in tracking time-varying acoustic paths, while maintaining higher stability across unseen head movement scenarios. It also generalizes well to real-world noises not encountered during training.","published":"2026-09-25T07:57:31Z","updated":"2026-09-25T07:57:31Z","authors":["Boxiang Wang","Zhengding Luo","Ziyi Yang","Dongyuan Shi","Xuexian Liu","Woon-Seng Gan"],"categories":["eess.AS"],"headline":"Improving Noise Control with Smarter, Adaptive Filters","summary":"Researchers studied active noise control, which reduces unwanted sounds but struggles when sound environments change over time. They developed a new method that uses meta-learning to adapt both the noise-filtering process and track changes in the sound paths together. Their tests using real-world data showed this approach better handles changing conditions and stays stable even with unexpected noises compared to existing methods. These results come from simulations and measurements, focusing on improving adaptive filter updates in noise control systems."},{"id":"2609.30944v1","title":"Representation-Aware Transport-Information Measure for Non-inclusive Discrete Supports","abstract":"Information-theoretic measures for comparing probability distributions are widely used across physics and other fields. When two discrete distributions have non-inclusive supports, however, the Kullback-Leibler (KL) divergence is in general not directly applicable, and various alternative divergences and distances have been introduced. These measures compare the resulting distributions themselves, but do not generally retain information about the representation transformations by which the discrete distributions are generated from underlying continuous ones. Here we introduce a representation-aware transport-information measure for discrete distributions with non-inclusive supports, formulated based on the standard KL divergence. We consider two continuous reference distributions, each transformed into a discrete representation through its own discretization scheme. Rather than comparing only the resulting discrete distributions or their continuous references, we additionally retain local information associated with the representation-change schemes. The resulting measure can therefore distinguish discrete representations that may have identical discrete probability landscapes but originate from different continuous references or discretization schemes. The construction is based on the transport-information cost of continuous-to-discrete representation in the framework of unavoidable canonical nonlinearity (UCN). UCN provides a non-arbitrary correspondence between the transport cost of discretization as an extrinsic geometric operation and the information-theoretic indistinguishability of nearby continuous distributions, thereby allowing a discrete representation to be associated with a local family of underlying continuous distributions on the statistical manifold.","published":"2026-09-25T07:56:38Z","updated":"2026-09-25T07:56:38Z","authors":["Koretaka Yuge"],"categories":["cond-mat.stat-mech"],"headline":"New Method Compares Complex Probability Distributions More Fully","summary":"Researchers developed a new way to compare certain types of probability distributions that come from different underlying continuous data but look similar in discrete form. Their method includes extra information about how continuous data is transformed into discrete categories, which existing measures often ignore. This approach, based on advanced mathematical concepts, helps distinguish cases where traditional comparisons fail, but their findings are theoretical and focus on the framework rather than practical applications."},{"id":"2609.30943v1","title":"LogicTree-RAG: Logic Tree-guided Retrieval-Augmented Generation for Long-form Patent Drafting","abstract":"Long-form technical text generation underpins knowledge-intensive workflows, yet remains challenging for large language models (LLMs) due to the need for globally consistent logical structuring and faithful technical reasoning beyond local coherence. Patent drafting is a canonical instance of this challenge, demanding holistic generation of a legally compliant and technically exhaustive document through sustained multi-expert collaboration. Existing approaches often focus on partial section generation or rely on manually crafted outlines, limiting scalable automation in realistic settings. In this work, we propose LogicTree-RAG, a logic tree-guided retrieval-augmented generation framework that induces a hierarchical logic tree as a global organizational backbone to organize and ground technical disclosures, without relying on expert-defined drafting priors. Each node in the logic tree represents a technical element and is constructed through evidence-guided recursive generation. A hybrid traversal mechanism then maps the logic tree into patent sections, enabling controllable and section-balanced generation. Extensive experiments show that LogicTree-RAG consistently improves content quality and language conformity over strong LLM-based baselines and achieves longer structured generation with high token efficiency, demonstrating the effectiveness of logic-centric generation for complex technical document drafting.","published":"2026-09-25T07:56:10Z","updated":"2026-09-25T07:56:10Z","authors":["Jiaqi Zhu","Naili Xing","Hexiang Pan","Haotian Gao","Jianwei Yin","Xiaokui Xiao","Beng Chin Ooi"],"categories":["cs.AI"],"headline":"New method improves automated drafting of long, complex patent documents","summary":"Researchers studied the challenge of using AI to write long, detailed technical texts like patents, which require clear logical structure and accurate reasoning. They developed LogicTree-RAG, a system that builds a hierarchical logic tree to organize technical content without needing expert outlines, guiding the AI to generate well-structured patent sections. Their experiments suggest this approach improves the quality and coherence of generated text compared to other AI methods, though the findings are based on controlled testing rather than real-world deployment."},{"id":"2609.30942v1","title":"Non-factor quantum dynamics and relativity","abstract":"We consider the effect of relativistic principles on the structure of the local evolutions of non-factor quantum systems such as those which naturally arise in quantum spacetimes and causal structures. Generalising a former result that had been given for factor systems, we show that any evolution which can be applied locally on a direct sum of factor systems and without producing superluminal signals must be linear, unitary, and respect the block decomposition of the system.","published":"2026-09-25T07:56:07Z","updated":"2026-09-25T07:56:07Z","authors":["Octave Mestoudjian","Matt Wilson"],"categories":["quant-ph"],"headline":"Researchers Study Local Changes in Complex Quantum Systems Under Relativity","summary":"The researchers explore how the rules of relativity affect local changes in complex quantum systems that are made up of smaller parts. They build on previous work done on simpler systems and find that any local change that doesn’t allow faster-than-light communication must follow specific mathematical rules: it must be linear, reversible (unitary), and respect the system's structure. Their findings apply theoretically to a type of quantum system relevant to models of spacetime."},{"id":"2609.30941v1","title":"Spackle: Completing Large View Single Image NVS with Adaptive Gaussians","abstract":"Single-image novel view synthesis (NVS) enables photorealistic rendering of un- observed viewpoints from a single input. Practical NVS systems require two key capabilities: robust reconstruction of occluded regions and high inference effi- ciency. While hybrid decoupled frameworks combining feedforward 3D Gaussian Splatting (3DGS) and diffusion models show promise for large-view-deviation NVS, they suffer from capacity competition: a fixed number of Gaussians forces resource shifts from visible to newly disoccluded areas, degrading original scene fidelity when the target view deviates significantly from the input. To address this, we propose Spackle, a lightweight residual learning framework that mit- igates capacity competition without sacrificing efficiency. Spackle operates in three stages: predicting base 3DGS attributes from given views, automatically identifying poorly reconstructed regions, and learning a residual 3DGS optimized exclusively for these areas. At inference, we combine the baseline and aug- mented Gaussians for NVS. We conduct comprehensive experiments and show that Spackle achieves state-of-the-art performance on large-view-deviation cases.","published":"2026-09-25T07:56:03Z","updated":"2026-09-25T07:56:03Z","authors":["Xuanzhi Liu","Yuhe Zhou","Xinyi Wu","Zhenyao Wu","Jinghao Chen","Ruize Han","Song Wang"],"categories":["cs.CV","cs.AI"],"headline":"Spackle improves 3D image views from a single photo","summary":"Researchers worked on improving how computers generate realistic images from new viewpoints using just one photo. They found that existing methods struggle to balance detail between visible and hidden parts when the new view changes a lot. Their solution, called Spackle, adds a smart adjustment step that focuses on fixing poorly recreated areas without slowing down the process. Experiments show Spackle performs better than current approaches, especially when viewing angles differ greatly from the original photo."},{"id":"2609.30940v1","title":"Financial Fragility in Societies of LLM Agents: Coordination Failures and Stabilizing Mechanisms","abstract":"Individually protective decisions can produce avoidable collective failures. As large language model (LLM) agents take on greater roles in financial decision-making, financial AI safety must therefore be considered not only at the level of individual agents, but also at the level of the systems they jointly create. We study this problem with FRAIL, a controlled experimental framework that places LLM agents in three dynamic financial environments---bank runs, debt rollover, and reward crowdfunding---where agents' decisions reshape the financial conditions faced by others. Across seven leading LLMs, we find widespread collective fragility even when no agent is instructed to destabilize the system: 77\\% of baseline bank-run episodes and 83\\% of debt-rollover episodes end in failure. We then compare three interaction mechanisms based on compensated commitments, centralized commitment agreements, and participant-led coalitions. All three improve aggregate outcomes, but no single mechanism performs best across all financial structures. Across mechanisms, successful stabilization shares a common temporal pattern: broad commitment forms early, before defensive behavior becomes self-reinforcing. Our findings show that individually capable agents do not automatically form safe financial systems, highlighting system-level evaluation and interaction design as central problems for financial AI safety. Code is available at https://anonymous.4open.science/r/FinFrail-CF26.","published":"2026-09-25T07:55:14Z","updated":"2026-09-25T07:55:14Z","authors":["Zhenhao Fu","Ruipeng Xu","Qibing Ren"],"categories":["cs.AI","q-fin.GN"],"headline":"AI Agents Struggle to Avoid Financial Crises Together","summary":"Researchers studied how AI language models make financial decisions when interacting in groups, focusing on situations like bank runs and debt crises. They found that even capable AI agents often fail collectively, causing financial system breakdowns without any agent aiming to do so. The team tested three ways to improve cooperation, all helping but none perfect. Their work suggests it's important to look at AI behavior not just individually, but as a whole system to make financial AI safer."},{"id":"2609.30939v1","title":"MACBT: A Multi-Agent Cognitive Behavioral Therapy Decision Support System with Longitudinal Memory","abstract":"Cognitive behavioral therapy (CBT) is an evidence-based first-line treatment for depression, yet its scale is constrained by the time clinicians spend on pre-session preparation, post-session documentation, and longitudinal cognitive-pathology tracking. We present a clinician-facing AI decision-support system that combines a multi-agent CBT framework (MACBT) with a CBT-specific longitudinal memory module (CD Memory). MACBT encodes the five-stage CBT workflow (assessment, Socratic questioning, cognitive restructuring, behavioral experiments, and treatment monitoring) into five collaborative agents. CD Memory tracks cognitive-distortion type, frequency, severity, and restructuring efficacy across sessions to generate pre-session pathology reports and intervention-priority recommendations. We construct a Chinese CBT dialogue corpus via dual-role large language model simulation and train a Qwen3-14B backbone with supervised fine-tuning and direct preference optimization. Evaluation with GPT-4 judges shows MACBT outperforms MeChat, SoulChat, PsyChat, and CPsyCounX in professionalism (2.62) and clinical authenticity (2.25). The full memory-augmented system further improves session quality by 12.6% and achieves a longitudinal mean of 2.29 on cross-session continuity, intervention progression, and personalization.","published":"2026-09-25T07:54:45Z","updated":"2026-09-25T07:54:45Z","authors":["De Jiang","Shuo Zhang","Weiwei Liao","Jianying Zhang","Chuanhui Yu","Hongen Liao","Kehong Yuan"],"categories":["cs.AI"],"headline":"AI System Supports Therapists in Managing Depression Treatment","summary":"The researchers developed an AI decision-support tool called MACBT to help therapists with cognitive behavioral therapy (CBT) for depression. MACBT uses multiple AI agents to follow the CBT process, and a special memory system tracks a patient's thinking patterns over time to guide therapy. They trained the system using simulated conversations and found it performed better than other AI tools in professional quality and clinical realism. The study's results are based on simulations and evaluations by AI judges, not real patient trials."},{"id":"2609.30938v1","title":"Landscape Limits of Quantum-Inspired Evolutionary Optimization across 256 continuous functions","abstract":"Quantum-inspired evolutionary optimization (QIEO) represents design variables as a set of qubits and searches a continuous, multi-dimensional landscape through rotation of the qubit's amplitude pair. Every generation rotates those amplitudes toward a single elite, which corresponds to that generation's best. The update is cheap, almost parameter-free, and well-suited for massive parallel implementation, which has encouraged its adoption in engineering, design, and planning applications. However, there are critical issues with this formulation, principally, the treatment of design variables as independent probability components which make it incapable of exploiting local curvature, anisotropy, or variable coupling. Despite this, QIEO is believed to hold promise, and has been used extensively to solve real-world problems, with significant qualitative and computational advantage over its classical counterpart, Genetic Algorithm (GA). A collection of 256 (actually 508; 256 unshifted + 252 shifted, 4 could not be shifted) continuous function are selected from the prior works, in such a way that they represent eleven landscape characteristics, namely continuity, differentiability, separability, scalability, modality, convexity, conditioning, symmetry, maximum dimensionality, dimension dependency, and the coupling pattern of the design variables. These functions are then solved by three QIEO variants, two GA encodings and Hansen's Covariance Matrix Adaptation Evolution Strategy (CMA-ES). The results are evaluated in terms of computational cost, solution precision, and specialization across landscape characteristics. They identify the conditions under which QIEO provides competitive performance, clarify where its independent-variable representation becomes limiting, and establish whether particular QIEO variants offer advantages for specific landscape characteristics.","published":"2026-09-25T07:53:40Z","updated":"2026-09-25T07:53:40Z","authors":["Rishi Govind","Ferdin Sagai Don Bosco","Kasturi Venkata Srikanth","Aman Mittal","Abhishek Singh","Aditya Singh","Abhishek Chopra"],"categories":["cs.NE","math.OC"],"headline":"Researchers test quantum-inspired optimization on diverse math problems","summary":"The researchers studied a special optimization method called quantum-inspired evolutionary optimization (QIEO), which uses a quantum-like approach to find good solutions in complex problem spaces. They tested QIEO on 256 continuous functions with different mathematical properties, comparing it to classical methods like genetic algorithms and another advanced technique. Their work identifies when QIEO works well and when its basic assumptions, like treating variables independently, cause it to struggle. The study clarifies how different versions of QIEO perform depending on the problem's characteristics."},{"id":"2609.30937v1","title":"Sub-Horizon and Quasi-Static Approximations in Interacting Dark Sector Models Involving Scalar Field Dark Energy","abstract":"We investigate the validity of the sub-horizon approximation (SHA) and quasi-static approximation (QSA) for linear matter perturbations in interacting scalar-field dark-energy models. We consider both quintessence and phantom scalar fields and derive their full relativistic perturbation evolution to compare with the corresponding SHA and QSA solutions. We find that, while the SHA reproduces the full evolution with percent-level accuracy for non-interacting and weakly interacting models on sufficiently sub-horizon scales, its accuracy progressively deteriorates with increasing interaction strength and towards larger scales. Moreover, unlike in $Λ$CDM and non-interacting scalar-field models, applying the SHA to interacting models does not fully recover the standard Newtonian fluid equations. The QSA generally provides a closer description of the full relativistic evolution than the SHA over the scales considered. By comparing the scalar-field model with $Λ$CDM using the full relativistic equations and, separately, using the SHA system, we evaluated whether the SHA is capable of capturing the full extent of the physical differences introduced by the alternative model. We find that the difference between these two comparisons becomes increasingly important for stronger interactions and smaller wavenumbers. Finally, we demonstrate that the resulting differences can reach several percent on scales accessible to large-volume hydrodynamical simulations, implying that the validity of the SHA cannot be assumed solely from the sub-horizon condition. Our results highlight the need to benchmark the approximation over the relevant interaction parameter and wavenumber ranges before employing interacting-dark-energy models in large-volume structure-formation simulations and precision comparisons with large-scale-structure observations.","published":"2026-09-25T07:53:24Z","updated":"2026-09-25T07:53:24Z","authors":["Priyanka Saha","Sharvari Nadkarni-Ghosh","Kaushik bhattacharya"],"categories":["gr-qc","astro-ph.CO"],"headline":"Testing Approximations in Dark Energy and Matter Interaction Models","summary":"Researchers studied how well common simplifications, called the sub-horizon and quasi-static approximations, work when modeling interactions between dark energy and matter in the universe. They found these simplifications become less accurate as interactions grow stronger or on larger scales. Their analysis shows the need to carefully check these approximations before using them in detailed cosmic structure simulations or observations, especially for models with significant dark energy-matter interaction. The study is theoretical and based on comparing different mathematical approaches."},{"id":"2609.30936v1","title":"Self-Play Search Distillation for Large Language Model Reasoning","abstract":"Improving reasoning abilities in Large Language Models (LLMs) requires high-quality data that exposes difficult decisions, competing alternatives, and their consequences. Data scarcity is driven by the low quality of synthetic data and the cost of human labeling. We introduce Self-Play Search Distillation (SPSD), a framework for generating superhuman synthetic data via self-play of MuZero-like networks trained on board games. SPSD uses executable environments to turn search into structured reasoning problems. At each state, the expert identifies a preferred decision, plausible alternatives, plausible opponent replies, and value estimates. By converting the self-play search records into superhuman chains-of-thought, we train LLMs with environment-grounded supervision. Although trained only on self-play search records, SPSD transfers to unseen mathematics. On Qwen3-4B-Base, it raises the mean over six mathematics benchmarks from 24.1 to 36.6 while increasing the held-out-game win rate from 15% to 45%. SPSD offers an annotation-efficient way to create high-quality synthetic data for improving LLM performance in reasoning tasks.","published":"2026-09-25T07:53:21Z","updated":"2026-09-25T07:53:21Z","authors":["Lorenzo Molfetta","Wai-Chung Kwan","Giacomo Frisoni","Luca Ragazzi","Gianluca Moro","Pavlos Vougiouklis","Jeff Z. Pan","Pasquale Minervini"],"categories":["cs.AI"],"headline":"New Method Helps AI Improve Reasoning Using Self-Play Data","summary":"Researchers developed Self-Play Search Distillation (SPSD), a technique that generates high-quality training data by having AI models play against themselves in game-like settings. This process creates detailed reasoning steps that help large language models learn better decision-making. Tested on math problems, SPSD improved performance significantly without needing costly human-labeled data, but the results are based on simulated self-play environments and transfer may vary."},{"id":"2609.30935v1","title":"Estimating and Orthogonalizing Unknown Pre-training Gradients for Continual Fine-tuning of Large Language Models","abstract":"Continual fine-tuning is essential for large language models (LLMs) to dynamically adapt to real-world environments, yet it inevitably suffers from catastrophic forgetting, particularly the performance degradation of previous tasks and LLMs' general-purpose knowledge. Although existing methods, such as orthogonal gradient projection, mitigate the forgetting across various fine-tuning tasks, they fundamentally fail to preserve pre-training LLMs' inherent general-purpose knowledge because the original data and gradients of off-the-shelf pre-training LLMs required by these methods are strictly unknown and highly diverse. To bridge this critical gap, we propose EoupCT, a novel framework designed to Estimate and Orthogonalize Unknown Pre-training gradients for Continual LLM fine-Tuning. Specifically, EoupCT estimates pre-training gradients by dynamically generating pseudo data that is most susceptible to forgetting for new tasks through a learnable soft prompt equipped with Gumbel-Softmax relaxation. Furthermore, we formulate a multi-objective optimization problem and introduce a first-order efficient Pareto optimizer that jointly optimizes LLM parameters and the soft prompt, rigorously enforcing orthogonality between new task updates and the estimated pre-training gradients. Extensive experiments across multiple LLMs demonstrate that EoupCT effectively preserves both task-specific proficiency and inherent general-purpose knowledge, successfully mitigating the catastrophic forgetting.","published":"2026-09-25T07:52:37Z","updated":"2026-09-25T07:52:37Z","authors":["Bing Wang","Changchun Li","Xin-Qiang Cai","Lin Yuanbo Wu","Ximing Li","Gang Niu","Masashi Sugiyama"],"categories":["cs.CL","cs.AI","cs.LG"],"headline":"New Method Helps AI Models Keep Old and New Knowledge","summary":"Researchers address the problem of large language models forgetting their original training when adapted to new tasks, known as catastrophic forgetting. They developed EoupCT, a system that estimates hidden original training signals and adjusts learning to protect past knowledge while learning new tasks. Their experiments on several models show EoupCT helps maintain both new task skills and original general abilities, but these findings rely on designed simulations rather than real-world deployment evidence."},{"id":"2609.30934v1","title":"ManiVid: Unified and Explainable Forensic Analysis of Manipulated Videos","abstract":"Rapid advances in AI-generated video (AIGV) have increased the risks posed by deceptive video manipulation. Unlike fully synthetic videos, manipulated videos retain most source content and alter only localized regions, making forensic analysis particularly challenging. Existing video forgery research faces two limitations in both data and methodology: (1) High-quality datasets and benchmarks tailored for manipulated videos remain scarce. (2) Multimodal large language models (MLLMs) extend forgery analysis beyond binary classification but struggle to use low-level forensic cues and provide precise pixel-level grounding. Specifically, we introduce ManiVid, a unified forensic analysis task covering forgery detection, artifact grounding, and anomaly explanation for manipulated videos. We construct ManiVid-38K, the first dataset to combine paired, open-vocabulary localized manipulations of general videos with authenticity labels, forgery masks, and anomaly explanations. It comprises about 19K manually verified real-fake video pairs, mostly at 1080P resolution, generated under 2 paradigms with 15 powerful generation models. We sample 1K pairs for ManiVidBench, balanced across six manipulation types and generation models for fair evaluation. We further propose ManiVidLens, a unified framework for explainable video forgery analysis. Its Forensic Evidence Router supplies shared low-level forensic evidence for multimodal reasoning and video segmentation. Its Prompt Distill Module converts grounding states into semantic and geometric prompts and distills spatial priors for mask decoding and full-video propagation. ManiVidLens achieves relative gains over the strongest comparison methods in artifact grounding (+21.1% mIoU; +21.3% J&F) and anomaly explanation (+131.3% ROUGE-L; +9.9% CSS). Its forgery detection remains comparable to dedicated classifiers (0.914 Acc; 0.913 F1).","published":"2026-09-25T07:51:50Z","updated":"2026-09-25T07:51:50Z","authors":["Hengrui Kang","Zhonghao Yan","Yuxuan Yang","Ruoyan Jing","Yuncheng Guo","Hao Chen","Kongming Liang","Zhanyu Ma","Conghui He","Weijia Li"],"categories":["cs.CV"],"headline":"New Method Analyzes and Explains Manipulated Videos with AI","summary":"Researchers address the challenge of detecting and explaining subtle changes in manipulated videos, which often keep most original content intact. They created ManiVid, a combined approach including a large dataset of real and fake videos with detailed labels, and ManiVidLens, a system that identifies video alterations and explains anomalies using shared forensic clues and advanced reasoning. Their results show improvements in pinpointing altered regions and explaining manipulations, while detection accuracy remains strong. Findings are based on data and simulations, focusing on improving forensic video analysis."},{"id":"2609.30933v1","title":"Discrete orthogonal polynomials related to Hahn difference operator","abstract":"This work studies general sequences of discrete orthogonal polynomials on linear lattices, including the so-called discrete Laguerre--Hahn orthogonal polynomials. We deduce the structure relations involving the orthogonal polynomials and their associated, as well as the explicit form of the fourth--order linear difference equation satisfied by these Laguerre--Hahn orthogonal polynomials. Furthermore, the particular cases corresponding to semiclassical and classical families of orthogonal polynomials, which lead to second--order difference equations, are also analyzed. Several examples illustrating the computation of the explicit coefficients of these difference equations for various families of orthogonal polynomials are also presented.","published":"2026-09-25T07:49:30Z","updated":"2026-09-25T07:49:30Z","authors":["J. F. Mañas-Mañas","J. J. Moreno-Balcázar","M. N. Rebocho","C. Rodríguez-Perales"],"categories":["math.CA"],"headline":"Researchers analyze special discrete orthogonal polynomials and their equations","summary":"The researchers studied certain sequences of discrete orthogonal polynomials, focusing on those related to the Hahn difference operator. They derived key relationships and the specific form of a fourth-order difference equation these polynomials satisfy, including simpler cases for classical polynomials with second-order equations. Their work includes examples showing how to calculate coefficients in these difference equations. The findings are theoretical and concern mathematical structures rather than practical applications."},{"id":"2609.30932v1","title":"A Uniform Bound on Optimal Strategy Length in Water Transport Problem","abstract":"We prove that every water transport problem on an $n$-vertex graph has an optimal strategy of length at most $n^{(2+o(1))n}$. More strongly, the convex hull of all strategy operators stabilizes within the same bound. We also give a five-vertex instance in which every optimal strategy repeats a nontrivial connected averaging set.","published":"2026-09-25T07:48:53Z","updated":"2026-09-25T07:48:53Z","authors":["Tianyi Tao","Bohan Yang"],"categories":["math.CO"],"headline":"Researchers Find Bound on Strategy Length in Water Transport Problem","summary":"The researchers studied a problem involving moving water through a network of points (vertices). They proved that the best possible method to solve this problem can be done within a certain length, which grows rapidly but predictably with the number of points involved. They also showed that the combination of all strategy operators becomes stable within the same limit. Additionally, they found a small example where the best strategy repeats a specific pattern. Their findings are theoretical and based on mathematical proofs."},{"id":"2609.30931v1","title":"Interpolating walk between discrete-time quantum walk and its intrinsic random walk on graph","abstract":"We consider an interpolation with the parameter $p\\in [0,1]$ between the discrete-time quantum ($p=0$) and its intrinsic random ($p=1$) walks on a finite connected graph. Its time evolution is defined by the convex combination of the Kraus (CPTP) maps of the quantum and random walks. The Kraus map of the random walk is represented by taking a kind of projection of that of the quantum walk. The time evolution of the interpolation is interpreted as the combination of the following two dynamics of $2$ walkers on the same graph correlated with each other: $2$ walkers move independently of each other (no-correlation) with probability $1-p$, while $2$ walkers always move into the same position (the maximal correlation) with probability $p$ at each time step. In this paper, we generalize the random walk to a new walk, namely, the correlated walk, which preserves the intrinsic property and relaxes the maximal correlation of the random walk. This correlation is determined by a partition of the arc sets in the underlying graph. We show that the eigenvalues of the interpolating walk between the correlated and quantum walks live in $\\{ z\\in \\mathbb{C}\\;|\\;1-p\\leq |z|\\leq 1 \\}$ and that the absorption state coincides with that of the correlated walk, which is characterized by the underlying graph structures.","published":"2026-09-25T07:47:35Z","updated":"2026-09-25T07:47:35Z","authors":["Yusuke Higuchi","Etsuo Segawa","Honoka Shiratori","Saori Yoshino"],"categories":["math-ph"],"headline":"Researchers blend quantum and random walks on graphs using a new method","summary":"The researchers study a way to smoothly transition between quantum and random walks on networks by mixing their behaviors with a parameter p. They introduce a new type of 'correlated walk' that allows different levels of connection between two walkers on the graph, rather than just fully independent or fully synchronized moves. Their analysis shows how the properties of this blended walk depend on the graph's structure, particularly focusing on the possible outcomes and steady states. These results come from mathematical analysis rather than experimental tests."},{"id":"2609.30930v1","title":"Tracking Dynamic Simplicial Complexes via Constrained State-Space Estimation","abstract":"Simplicial complexes (SCs) extend graphs to represent higher-order interactions, but couple their simplex levels through the inclusion property. While static SC inference is an emerging research direction, tracking time-varying SCs remains largely unexplored. A central challenge in tracking SCs is to account for the inclusion property. To this end, we build a nonlinear state-space model tailored to SCs. For prediction, we introduce a closure-aware Markov generation model whose conditional mean preserves simplicial inclusion and whose covariance captures edge-triangle dependencies. For correction, we encode inclusion constraints as nonlinear pseudo-measurements, allowing the constraints to inform both the state and covariance updates. We investigate progressively richer treatments of these measurements through a standard extended Kalman filter, an iterated extended Kalman filter, and a Laplace approximation. Experiments then demonstrate the benefits of exploiting the proposed dynamics and constraint information.","published":"2026-09-25T07:47:09Z","updated":"2026-09-25T07:47:09Z","authors":["Varun Sarathchandran","Bishwadeep Das","Elvin Isufi","Geert Leus"],"categories":["eess.SP"],"headline":"New Method Tracks Changing Complex Networks With Higher-Order Links","summary":"The researchers studied how to follow changes in complex network structures called simplicial complexes, which represent not just pairwise connections but groups of related elements. They created a mathematical model that respects the natural rules these structures must follow as they evolve over time. Using different versions of a Kalman filter, they showed in experiments that including these rules helps better track the network changes. Their findings are based on modeling and simulation, highlighting a new approach for studying dynamic higher-order interactions."},{"id":"2609.30929v1","title":"EPOC: Endpoint-Preserving Online Correction With Compressed Residual State for Multi-Horizon Time Series Forecasting","abstract":"Completed multi-horizon forecasts provide residual feedback for a fixed forecaster, but retaining full residual blocks increases auxiliary state. We propose Endpoint-Preserving Online Correction (EPOC) with a compressed residual state. It stores low-order discrete cosine transform (DCT) coefficients and the final value of the preceding residual block. Within each channel, the endpoint is shared across component-wise online ridge regressions that also use current-forecast coefficients. The fitted DCT correction is blended with the base forecast. We evaluate eight multivariate series with DLinear and PatchTST, three seeds, and two training variants, yielding 96 matched fixed-base conditions at a 24-step horizon. EPOC achieves mean condition-wise reductions in mean squared error (MSE) and mean absolute error (MAE) of 15.40% and 9.35% from the uncorrected base, respectively, with a median of 6,352 B in retained auxiliary arrays. It has lower paired MSE than the $δ$-Adapter, COSA, FAC, and OMPB in a majority of conditions and uses less state than each. Full ELF achieves the largest mean MSE reduction, 19.29%, but its median retained state is 474,048 B ($\\times$75 relative to EPOC). Equal-size summary controls favor the endpoint by 1.65--2.20% in paired MSE; a coefficient-reconstructed endpoint yields similar accuracy to the observed endpoint, highlighting its role as a shared input. Increasing the retained DCT component count from 4 to 8 adds 1.00 percentage point of MSE reduction for 5,728 B. On jointly trained bases, EPOC lowers MSE by 16.69--20.15% relative to globally blended TEFL-style adapters applied to the same base. The code and numerical records are available at https://github.com/keiotakmin/endpoint-preserving-residual-correction.","published":"2026-09-25T07:43:49Z","updated":"2026-09-25T07:43:49Z","authors":["Takumi Fujimoto","Hiroaki Nishi"],"categories":["cs.LG"],"headline":"New Method Improves Multi-Step Time Series Forecast Accuracy Efficiently","summary":"Researchers developed Endpoint-Preserving Online Correction (EPOC), a technique that improves forecasts over multiple future points by compressing and correcting forecast errors using a mathematical transform. Testing on various data sets showed EPOC reduced forecasting errors by about 15% while using far less extra memory than similar methods. The work is based on simulations and comparisons, showing EPOC strikes a good balance between accuracy and memory use in multi-step forecasting models."},{"id":"2609.30928v1","title":"UltraG-Bench: A Multi-task Benchmark for assessing Large Vision-Language Models on Pixel-level Evidence Grounding in Ultrasound","abstract":"Ultrasound is one of the most widely used medical imaging modalities, and recent large vision-language models(VLMs) have shown increasing capabilities in ultrasound image understanding. However, these models fail to provide pixel-level visual evidence aligned with their semantic predictions, and their fine-grained grounding capability in ultrasound remains largely unclear. We introduce UltraG-Bench, a large-scale multi-task benchmark for evaluating pixel-level evidence grounding in ultrasound. UltraG-Bench is built by annotating 40 public ultrasound segmentation datasets spanning 13 anatomical categories, and comprises three progressive tasks: instruction-guided segmentation, evidence-grounded VQA, and evidence-grounded report generation, with 331125, 666779, and 138832 annotations, respectively. Comprehensive evaluation of 14 state-of-the-art models reveals a substantial gap between semantic understanding and fine-grained pixel-level localization. We further propose UltraG-Agent, which combines the semantic reasoning capabilities of a VLM with the ultrasound-specific segmentation capability of UltraSAM3. Experiments show that UltraG-Agent substantially improves both semantic prediction and pixel-level visual grounding. Our dataset and code are available at https://github.com/zhuqh19/UltraG-Bench.","published":"2026-09-25T07:43:12Z","updated":"2026-09-25T07:43:12Z","authors":["Quanhao Zhu","Bo Xu","Rui Lin","Chenyuan Wang","Yu Shao","Boling Zhu","Jiuyan Sun","Liang Zhao","Hongfei Lin","Feng Xia"],"categories":["cs.CV","cs.AI"],"headline":"New Benchmark Tests Ultrasound Image Understanding by AI Models","summary":"Researchers created UltraG-Bench, a large set of ultrasound images with detailed labels, to test how well advanced vision-language AI models can pinpoint exact locations in images while explaining their predictions. They found current models struggle to align their understanding with precise image areas. To tackle this, the researchers developed UltraG-Agent, combining general reasoning with ultrasound-specific skills, which improved accuracy. These findings are based on evaluations of existing models and new methods using the benchmark they built."},{"id":"2609.30927v1","title":"Bounds for Unions of Several Parts in Balanced Graph Partitions","abstract":"Let $k\\ge3$ and $1\\le \\ell\\le k-1$. We study balanced $k$-partitions of a graph for which the union of any $\\ell$ parts induces few edges. We show that every graph $G$ with $n$ vertices and $m$ edges admits a balanced partition $V_1,\\ldots,V_k$ such that \\begin{equation*} \\max_{\\substack{A\\in\\binom{[k]}{\\ell}}}e_G\\left(\\bigcup_{i\\in A}V_i\\right)\\le\\frac{\\ell^2}{k^2}m+\\frac{\\ell^2(k-\\ell)}{k^2}(n-1)+\\frac{\\ell(k-\\ell)}{k(k-1)}\\sqrt{\\left(\\binom{k}{\\ell}-1\\right)m}. \\end{equation*} In the case $\\ell=2$, our result confirms a conjecture of Bollobás and Scott in a stronger form.","published":"2026-09-25T07:42:11Z","updated":"2026-09-25T07:42:11Z","authors":["Zhanping Yang"],"categories":["math.CO"],"headline":"New bounds found for dividing graphs into balanced parts","summary":"Researchers studied how to split a graph's vertices into balanced groups so that any combination of some groups has few connections between them. They proved a mathematical formula that limits the maximum number of edges in such unions, offering a precise upper bound. Their work includes confirming and strengthening a previous mathematical guess made by Bollobás and Scott when the union includes two parts. These findings are theoretical and concern properties of graph partitions."},{"id":"2609.30926v1","title":"Minimal m-subharmonic functions with nonmaximal Hessian measures","abstract":"For every $2\\le m\\le n$, we construct Hölder continuous functions on the closed unit ball that are minimal in the Cegrell class $\\F_m$, although their Hessian measures are not maximal in the $m$-subharmonic ordering. This answers a question posed by the second author. We also obtain a minimality criterion based on partial pluricomplex energy and an explicit family of Monge--Ampère examples on a product domain. For $m=1$, minimality and maximality are equivalent on a ball.","published":"2026-09-25T07:39:53Z","updated":"2026-09-25T07:39:53Z","authors":["Thai Duong Do","Van Thien Nguyen"],"categories":["math.CV"],"headline":"New examples challenge assumptions about special mathematical functions","summary":"Researchers constructed certain smooth functions within a specific mathematical class, called the Cegrell class, that are minimal yet do not have the largest possible Hessian measures, contrary to prior expectations. They also developed a way to determine minimality using an energy concept and provided explicit examples involving complex equations. Their findings answer a previously posed question but mostly apply within theoretical and mathematical contexts, especially for dimensions two and higher; for dimension one, minimality and maximality still coincide as before."},{"id":"2609.30925v1","title":"How to break the Miranda signature scheme over matrix Gabidulin codes","abstract":"The Miranda signature scheme relies on masking a matrix code which disposes of a masked underlying structure, and the knowledge of which allows for efficient error decoding. We consider a Gabidulin code which is expanded into a matrix code, that is only \\Fq-linear. An additional masking is then applied to it. The attack proposed here shares similarities with that of [Le26, https://arxiv.org/abs/2608.03328] on the EGMC encryption scheme, which also follows the paradigm described above. It consists in recovering the Fqm-linear structure of a masked matrix Gabidulin code by reducing to a MinRank instance to be solved over the extension field Fqm, but where the matrices have coefficients in Fq. Such an instance can be efficiently solved. However, unlike the previous attack, it is possible to reduce in polynomial time to such a MinRank instance in the case of the Miranda signature scheme. This results in a particularly efficient key recovery attack against the parameters proposed for Miranda. For example, for the proposed parameter set with m=79, the complexity drops from 146 bits to 46 bits in this attack.","published":"2026-09-25T07:39:27Z","updated":"2026-09-25T07:39:27Z","authors":["Adrien Vinçotte"],"categories":["cs.CR"],"headline":"Researchers find efficient way to break Miranda signature scheme","summary":"The researchers studied the Miranda signature scheme, which uses a special coded matrix to protect digital signatures. They developed an attack that uncovers hidden structures in the matrix by turning the problem into a simpler mathematical puzzle called MinRank. This approach can be solved efficiently, allowing them to recover secret keys faster than previously thought. Their findings suggest the security level for the proposed Miranda parameters is much weaker than expected, but these results are based on theoretical and mathematical analysis rather than physical testing."},{"id":"2609.30924v1","title":"Training-Free Pronunciation Transcription via Text-Constrained Acoustic Rescoring","abstract":"Accurate and efficient pronunciation transcription is essential for preparing text-to-speech training data at scale. Existing approaches have different limitations: grapheme-to-pronunciation (G2P) and speech-to-pronunciation (S2P) methods each capture only partial information, using only text or only speech, while speech-and-text-to-pronunciation (ST2P) methods use both but require costly pronunciation-annotated data. To address this problem, we propose a training-free ST2P pipeline that integrates both lexical and acoustic information at inference time. Lexical resources and G2P tools generate text-constrained candidates, and a left-to-right greedy search selects the best one using whole-sequence negative log-likelihoods from frozen pretrained S2P models. On three Japanese corpora, our method reduces Character Error Rate (CER) from 0.60--1.40\\% (text-only baseline) to 0.04--0.17\\% with reference transcripts, and 0.64--1.58\\% with ASR transcripts. It outperforms all baselines, including a trained ST2P model and commercial multimodal LLMs. Our greedy search method is 3--3.5$\\times$ faster than beam search at similar CER, and the cascade is 2$\\times$ faster than direct decoding ensuring the efficiency and accuracy. In Spanish, French, and preliminary English, it also surpasses four open multimodal LLMs and the best traditional methods.","published":"2026-09-25T07:38:49Z","updated":"2026-09-25T07:38:49Z","authors":["Hikaru Asano","Yotaro Kubo","So Kuroki"],"categories":["cs.CL","cs.SD","eess.AS"],"headline":"New method improves pronunciation transcription without extra training","summary":"Researchers developed a way to accurately convert speech and text into pronunciation transcriptions without needing costly annotated training data. Their method combines text and sound information using existing tools and a search technique to pick the best pronunciation candidate. Tests on Japanese, Spanish, French, and English showed it beats current methods, including some advanced AI models, while running faster. The results are based on simulations using existing data, focusing on lowering errors in transcription."},{"id":"2609.30923v1","title":"Improving Multi-Delay-ASL through specialized reconstruction","abstract":"Purpose: Although image reconstruction has received relatively little attention in ASL research to date, it has the potential to address several challenges in ASL. As well as speeding up measurements by increasing undersampling and reducing measurement artifacts, it can improve the signal-to-noise ratio (SNR) of a given data set and the reproducibility of examinations. Methods: This work focuses on extending a dedicated ASL reconstruction approach (ASL-TGV) (Spann et al. (2020)) to multi-delay data and applying it to a high-resolution pCASL test-retest dataset. To show not only improvement in the Perfusion Weighted Images (PWIs), Cerebral Blood Flow and Arterial Transit Time was estimated and the test-retest reliability was estimated using the within subject Coefficient of Variance (wsCV), the Intraclass Correlation Coefficient (ICC) and Root-Mean-Squared-Error (RMSE). Results: The Perfusion-Weighted-Images reconstructed from highly undersampled single-shot data using ASL-TGV are clearly improved compared to a fully-sampled reference from the same data even after denoising, especially for long PLDs and the outermost slices. SNR calculated in grey and white matter ROIs shows an improvement of 62\\% and 35\\% respectively. The CBF maps produced from the ASL-TGV images have an improved test-retest reliability. Conclusion: ASL-TGV, which is now implemented in the Berkeley Advanced Reconstruction Toolbox (BART), can be used with any type of ASL labeling or data acquisition and any existing image post-processing pipeline and can improve SNR for the PWIs and reproducibility of the CBF maps.","published":"2026-09-25T07:38:43Z","updated":"2026-09-25T07:38:43Z","authors":["Ingmar Sorgenfrei","Qinyang Shou","Ingrid Barth","Martin Uecker","Danny JJ Wang","Rudolf Stollberger"],"categories":["physics.med-ph"],"headline":"New image method improves brain blood flow scans","summary":"Researchers studied a special image reconstruction technique called ASL-TGV to enhance brain blood flow measurements from MRI scans. By applying this method to multi-delay data, they improved image quality and consistency compared to traditional approaches, especially in challenging conditions. Their findings suggest this technique can boost signal clarity and reliability, though results are based on test data and specific imaging setups."},{"id":"2609.30922v1","title":"JevSoup: System-One Routing for Training-Free LoRA Composition","abstract":"Building adaptable AI systems requires effective coordination of specialized capabilities across diverse tasks. Low-rank adaptation (LoRA) enables modular expertise, but existing routing approaches may require auxiliary data, additional training, or autoregressive decoding. We propose JevSoup, a training-free framework separating System One expert routing from System Two execution. Using only the input and expert descriptions, Jev selects two experts through structured probabilities. JevSoup retains the leading expert's update, projects the second onto the orthogonal complement of the first update's row space, and combines them with equal weights. Across 14 PorTAL tasks and three Qwen3 scales, JepSoup achieves absolute gains of up to 1.19\\% in task-macro and 1.21\\% in sample-micro accuracy over the strongest evaluated external baselines. Our code is available at https://github.com/Leowang980/JevSoup.","published":"2026-09-25T07:37:28Z","updated":"2026-09-25T07:37:28Z","authors":["Xiuying Wang","Jiahua Cheng","Shuotian Li","Yufan Cheng","Bowen Deng","Zhexuan Bai","Yichen Li"],"categories":["cs.AI"],"headline":"New Method Combines AI Experts Without Extra Training","summary":"Researchers developed JevSoup, a way to coordinate specialized AI components called experts without needing extra training or data. It selects and combines two experts based on input and their descriptions, aiming to improve AI task performance. Tests on multiple tasks showed modest accuracy improvements, but findings come from controlled experiments and focus on technical performance only."},{"id":"2609.30921v1","title":"Counterexamples to the inhomogeneous Duffin-Schaeffer Conjecture","abstract":"The Duffin-Schaeffer Conjecture, proven in 2020, gives a zero-full dichotomy for the approximation of irrational numbers with reduced fractions for arbitrary approximation functions. We construct approximation functions such that the inhomogeneous analogue of the Duffin-Schaeffer Conjecture fails. The counterexamples are obtained for all non-zero rationals and certain Liouville numbers.","published":"2026-09-25T07:36:36Z","updated":"2026-09-25T07:36:36Z","authors":["Manuel Hauke-Treuer","James Maynard","Andrew Pollington"],"categories":["math.NT"],"headline":"Researchers Find Exceptions to a Number Approximation Rule","summary":"The Duffin-Schaeffer Conjecture, proved in 2020, describes when irrational numbers can be well-approximated by fractions. These researchers studied a related version called the inhomogeneous Duffin-Schaeffer Conjecture and found specific cases where it does not hold. They created functions that serve as counterexamples for all non-zero rational numbers and certain special irrational numbers known as Liouville numbers. Their findings are based on mathematical construction and do not imply the original conjecture is false."},{"id":"2609.30920v1","title":"Large-Data Global Well-Posedness for the Defocusing Cubic Schrödinger Equation in 3D Convex Domains","abstract":"We establish the global well-posedness for the energy-subcritical defocusing cubic nonlinear Schrödinger equation (NLS) on the three-dimensional Friedlander model domain \\( Ω=\\{x\\in\\mathbb{R}:x\\geq0\\}\\times\\mathbb{R}^2_{y,z}, \\) subject to homogeneous Dirichlet boundary conditions, for arbitrarily large initial data in the coercive energy space $H_0^1(Ω)$. The phase-space geometry of this configuration features a strictly convex boundary that admits a non-empty glancing set, trapping high-frequency wave packets within a boundary layer via generalized whispering-gallery caustics. These concentration phenomena induce an intrinsic derivative loss in the sharp linear Strichartz estimates, establishing a major microlocal obstruction to closing the nonlinear Duhamel iteration directly at the energy level via classical perturbative frameworks. To bridge the regularity deficit between the conservation laws and the linear theory, we construct a boundary-adapted family of continuous--discrete Bourgain--Strichartz restriction spaces $X^{s,b}$ built from the spectral decomposition of the Dirichlet realization of the model Friedlander operator \\( Δ_g=\\partial_x^2+(1+x)\\partial_y^2+\\partial_z^2. \\) The normal variable is resolved through discrete Airy spectral modes, while the tangential variables are continuously Fourier analyzed. Within this functional framework, we implement a refined high--low frequency decomposition to formulate a perturbed nonlinear equation for the high-frequency remainder in the sub-energy space $H_0^s(Ω)$ for \\( \\frac{1}{2}<s<1. \\) The initial high-frequency datum satisfies a quantitative sub-energy decay estimate of the form \\( \\|P_{>λ}u_0\\|_{H_0^s(Ω)} \\label{eq:abs_decay} \\lesssim λ^{s-1}\\|u_0\\|_{H_0^1(Ω)}, \\) providing a small parameter to counteract the derivative loss.","published":"2026-09-25T07:36:00Z","updated":"2026-09-25T07:36:00Z","authors":["Len Meas"],"categories":["math.AP"],"headline":"Solving a 3D Wave Equation in a Complex Convex Space","summary":"Researchers studied a specific three-dimensional wave equation with complex boundary shapes that cause waves to trap and lose predictability. They developed a new mathematical framework adapted to these boundary effects to handle large and complex initial conditions. Their work shows a way to prove well-posedness — meaning solutions behave predictably over time — for this challenging problem, but their results are theoretical and rely on advanced mathematical constructions designed for this specific setup."},{"id":"2609.30919v1","title":"Upper and lower fluctuations in Shallit's law of leap years in Pierce expansions","abstract":"Shallit's law of leap years in Pierce expansions gives, for Lebesgue-almost every $x\\in[0,1]$, the upper and lower limits of a normalized difference between $Nx$ and the number of leap years up to the year $N$ determined by the Pierce expansion digits of $x$. In this paper, we show that, for every $x\\in[0,1]$, both limits are determined by the lower limit of a normalized logarithm of the product of the first $n$ digits of $x$. In particular, the upper and lower limits always have the same absolute value. As applications, we characterize the sets of points with prescribed upper and lower limits, and show that each of these sets, if non-empty, is dense in $[0,1]$ and its intersection with any non-empty open subset of $[0,1]$ has full Hausdorff dimension. We also compare these sets with the sets defined by the growth rate of the digits, and show that, for each finite parameter, their difference has full Hausdorff dimension.","published":"2026-09-25T07:34:01Z","updated":"2026-09-25T07:34:01Z","authors":["Min Woong Ahn"],"categories":["math.NT","math.CA"],"headline":"New insights on fluctuations in leap year counts from Pierce expansions","summary":"The researchers studied a mathematical law that relates leap years to a special number representation called Pierce expansions. They found a precise way to describe upper and lower limits of a difference involving leap years using properties of these expansions for every number between 0 and 1. Their results identify detailed sets of numbers with specific limit behaviors and show these sets are complex and densely spread, based on a mathematical measure called Hausdorff dimension. The findings come from theoretical analysis and clarify earlier work on this topic."},{"id":"2609.30918v1","title":"Robust to Which Model Change? A Unified Evaluation of Robust Counterfactual Explanations","abstract":"Robust counterfactual explanations promise recourse that still works after the model behind it changes. Whether they keep that promise depends on what the change is. A small perturbation of the parameters, retraining on new data, and a new architecture are different events, and each existing method is evaluated against the one it was built for. Reported robustness scores, therefore, answer different questions and cannot be compared. We propose a unified cross-family evaluation protocol that holds factual instances and generated counterfactuals fixed while testing every method against the same eight types of model change. The benchmark compares six robust methods and two standard baselines on four tabular datasets. It characterizes every changed classifier through its outputs and reports empirical robustness together with coverage, base validity, and proximity. We find that relative performance and failure modes vary across change families. Bounded parameter perturbations change 0.95\\% of test predictions on average, compared with 4.9\\% for bootstrap retraining. Methods with guarantees for these perturbations do not necessarily transfer to other changes. RobX transfers most consistently in our experiments, although greater stability can require larger interventions. We argue that robust CFE methods should be evaluated through a common protocol that specifies the model changes, measures their realized behavioral magnitude, and keeps generation performance separate from robustness.","published":"2026-09-25T07:29:42Z","updated":"2026-09-25T07:29:42Z","authors":["Marcin Kostrzewa","Maciej Zięba"],"categories":["cs.LG","cs.AI"],"headline":"Evaluating Stability of Model Explanations Across Different Changes","summary":"The researchers studied methods that explain AI decisions in ways that remain valid even when the underlying AI model changes. They found that existing methods are tested only against specific types of model changes, making comparisons unfair. To address this, they created a standardized way to test all methods against various model changes using the same data. Their results show that performance varies depending on the change type, and guarantees for one change do not always apply to others. They suggest a unified testing approach for clearer evaluation."},{"id":"2609.30917v1","title":"Convolutional non-parametric Gamma-Ray Signal and Background Separation","abstract":"In very-high-energy gamma-ray astronomy, signals must be separated from the residual background which arises from misidentified cosmic-ray protons. Building on previous work, we introduce an ensemble of variational autoencoders that aims to perform automatic signal-background separation with minimal assumptions that include separability of the spatial and energy distributions in both the signal and background, with no prior specification of the number of components or the point-like/diffuse nature of the signal itself. In addition, we do not assume knowledge of which region of the coordinate space is background-dominated or where the signal is supposed to be located in the field of view. We test the model on an analytic point-source mixture scenario, a realistic simulation of dark matter annihilation in the Galactic centre, and real observations of the Crab nebula and MSH 15-52 from the public H.E.S.S data release. The model proves capable of completely reconstructing the signal and background at a pixel by pixel level in all scenarios, whilst also denoising the inputs. Stable performance and reasonable error estimates are obtained even for low signal-to-background ratios.","published":"2026-09-25T07:29:23Z","updated":"2026-09-25T07:29:23Z","authors":["Scarlet Betterman","Emmanuel Moulin","Martin White"],"categories":["astro-ph.HE"],"headline":"New AI method separates gamma-ray signals from background noise","summary":"Researchers developed a new AI approach using variational autoencoders to distinguish true gamma-ray signals from background noise caused by cosmic rays. Their method does not rely on prior knowledge about where signals or background are located or how many signal components exist. Tests using simulated and real astronomical data showed the model can accurately reconstruct signals and reduce noise, even when signals are weak compared to background."},{"id":"2609.30916v1","title":"Emergence of Nanoscale Modulation in Liquid Crystals as a Result of Short-Range Order Parameter Condensation","abstract":"The discovery of the twist-bend nematic phase in liquid crystals composed of bent-core and dimeric molecules has revealed an unexpected mechanism for the spontaneous formation of nanoscale periodic structures in soft condensed matter. Unlike conventional liquid-crystalline phases, the twist-bend phase exhibits a nanoscale heliconical modulation despite the absence of molecular chirality. In this note, dedicated to the late R. Meyer, we discuss a Landau phenomenological interpretation of this phenomenon. The central idea is that a short-range orientational order parameter undergoes condensation, giving rise to a heliconical structure characterized by a finite wave vector. We emphasize that the conventional nematic order is already long-ranged, whereas the additional order parameter describes local orientational correlations hidden within the nematic state. The resulting phase transition bears a close analogy to the de Gennes theory of the nematic--smectic-A transition. Fluctuation effects are expected to drive the transition weakly first order. The theory also predicts a new Goldstone mode associated with the spontaneously broken continuous symmetry of the heliconical state","published":"2026-09-25T07:29:03Z","updated":"2026-09-25T07:29:03Z","authors":["E. I. Kats"],"categories":["cond-mat.soft","cond-mat.stat-mech"],"headline":"New Theory Explains Tiny Twisted Structures in Liquid Crystals","summary":"Researchers explored an unusual phase in liquid crystals where molecules form tiny twisted patterns without needing chiral (handed) molecules. They proposed a theory suggesting that a special kind of local molecular order appears, creating these nanoscale heliconical structures. Their model, based on established physics ideas, predicts certain behaviors during this phase change but mostly relies on theoretical analysis without direct experimental confirmation."},{"id":"2609.30915v1","title":"TS23/McDonald and FIES/NOT strike again. Two new warm Jupiter systems and outer companions in a third","abstract":"Warm Jupiters (gas giant planets with orbital periods of P$\\sim$8-200 d) are thought to be largely unaffected by the tidal interactions that can erase information about the dynamical histories of hot Jupiters. Their orbital eccentricities ($e$) and stellar obliquities are therefore expected to be preserved, making them valuable probes of giant-planet formation and migration. Our aim is to identify and characterise transiting warm Jupiters in order to expand the sample of well-characterised systems and enable further studies of giant-planet formation and migration. We analyse TESS transit photometry together with ground-based radial velocity measurements and perform a joint fit of the photometric and spectroscopic data from the TS23 and FIES spectrographs to determine the planetary and orbital parameters. We report on the discovery and characterisation of two new, transiting warm Jupiters; TOI-5120 b ($M_p = 1.23\\pm0.05$ M$_{\\rm Jup}$, $R_p = 0.969\\pm0.017$ R$_{\\rm Jup}$, $P = 24.91645\\pm0.00002$ d, $e = 0.10\\pm0.02$) and TOI-5699 b ($M_p = 0.25\\pm0.03$ M$_{\\rm Jup}$, $R_p = 0.996\\pm0.015$ R$_{\\rm Jup}$, $P = 22.01850\\pm0.00004$ d, $e = 0.30^{+0.09}_{-0.08}$). We also refine the parameters of the known warm Jupiter TOI-2158 b ($M_p = 0.87\\pm0.04$ M$_{\\rm Jup}$, $R_p = 0.899\\pm0.015$ R$_{\\rm Jup}$, $P = 8.60079^{+0.00002}_{-0.00001}$ d, $e = 0.028^{+0.013}_{-0.014}$). In addition, we detect an outer planetary companion, TOI-2158 c ($M_p \\sin i = 1.62\\pm0.13$ M$_{\\rm Jup}$, $P = 414.5^{+1.5}_{-1.6}$ d, $e = 0.54\\pm0.04$), and identify a long-term radial velocity trend that may indicate another planetary companion on a wide orbit. The diverse architectures of these systems, in particular the presence and absence of outer companions, highlight the range of dynamical pathways leading to warm Jupiters.","published":"2026-09-25T07:26:53Z","updated":"2026-09-25T07:26:53Z","authors":["E. Knudstrup","D. Gandolfi","W. D. Cochran","M. Endl","P. MacQueen","T. Pugh","M. S. Lundkvist","C. M. Persson","A. Stokholm","M. L. Winther","J. R. Larsen","J. L. Rørsted","S. H. Albrecht","K. A. Collins","C. Watkins","B. Edwards","D. R. Ciardi","C. A. Clark","A. S. Polanski","C. Ziegler","C. Briceño","A. W. Mann","M. Everett","S. Howell","B. Safonov","T. Gan","K. Barkaoui","P. Benni","R. P. Schwarz","K. Horne","R. Sefako","D. Muthukrishna"],"categories":["astro-ph.EP"],"headline":"Scientists Discover and Study Three 'Warm Jupiter' Exoplanet Systems","summary":"Researchers used space telescope data and ground-based measurements to find and analyze two new 'warm Jupiter' planets—large gas giants with orbits of about 20-25 days—and improved knowledge of a previously known one. They also found an additional outer companion planet in one system and signs of another possible distant planet. These findings offer clues about how such giant planets form and move in space, but the research mainly describes observations rather than explaining the exact processes involved."},{"id":"2609.30914v1","title":"Cross-Backend QIEO: Universal Runtime Portability across OpenMP5, CUDA, HIP, and Multi-Language Interfaces","abstract":"Quantum-inspired algorithms emulate quantum mechanical principles, such as, superposition, interference, and probabilistic amplitude evolution, on classical hardware by representing candidate solutions as qubit vectors and evolving them through rotation-gate operators. This approach offers higher optimization performance without physical qubits, and has been shown to achieve order-of-magnitude speedups (10--80$\\times$) over traditional solvers on combinatorial, high-dimensional NP-hard problems. A critical barrier to adoption, however, is the lack of a unified execution framework that delivers both algorithmic performance and hardware portability. We present \\textbf{Cross-Backend Quantum Inspired Evolutionary Optimizer (QIEO)}, the runtime core of BQP's BQPhy solver, which addresses this gap through a \\emph{single-source-of-truth} architecture. One C++ implementation of the QIEO algorithm is compiled once per hardware target and exposed to multiple high-level languages via thin binding layers. The framework dispatches to CPU (sequential), OpenMP~5 (multi-core), CUDA (NVIDIA), and HIP (AMD) backends at runtime, adapting kernels to each device's memory hierarchy and warp/wavefront execution model. The framework's real-world utility is validated through binding demonstrations that share the identical C++ runtime. BQPhy's Python library is demonstrated on a neural network hyperparameter optimisation achieving 88.60\\% test accuracy on MNIST. BQPhy's MATLAB's Toolkit is tested on wind farm layout optimisation attaining $365\\,399 \\pm 4\\,552$~MWh/yr, which is statistically indistinguishable from particle swarm optimisation and $+7.6\\%$ above genetic algorithms on a 32-variable constrained engineering problem. The Julia package tackles the Lotka--Volterra parameter estimation where BQPhy replaces native Julia solvers on the same residual, cutting mean SSE by $2.1\\times$.","published":"2026-09-25T07:25:20Z","updated":"2026-09-25T07:25:20Z","authors":["Aman Mittal","Ferdin Sagai Don Bosco","Kasturi Venkata Srikanth","Abhishek Singh","Aditya Singh","Abhishek Chopra"],"categories":["cs.DC","cs.CL","math.OC"],"headline":"New Software Runs Quantum-Inspired Algorithms on Many Hardware Types","summary":"Researchers developed Cross-Backend Quantum Inspired Evolutionary Optimizer (QIEO), a software framework that runs quantum-inspired algorithms efficiently across different hardware like CPUs, NVIDIA and AMD GPUs. Their approach uses one C++ core compiled for each target and connects to languages like Python, MATLAB, and Julia through simple interfaces. The team demonstrated this tool solving real problems, including neural network tuning, wind farm layout, and biological modeling, showing performance comparable or better than existing methods. These findings support software portability with maintained performance, based on tested practical examples."},{"id":"2609.30913v1","title":"Causeway: Restoring Task Accessibility for Instruction Switching in VLA Policies","abstract":"Vision-language-action (VLA) policies can execute many tasks from standard initial states, yet a new instruction may fail after another task has altered the robot's physical state. We study instruction switching, where a new task is issued during or after the execution of a different one. We observe that a target task that is reliably completed from its standard initial states can become inaccessible from states produced by a preceding task. We call such states task islands. We propose Causeway, a training-free inference-time intervention. Given the current state and a re-entry pose for the target task, Causeway back-propagates through the frozen decoding computation and applies a state-directed write within the action-stream representation. The VLA decodes the return motion itself, without parameter updates, a new action head, or external action generation. Across 71 cross-object pairs, three switch timings, and three VLA architectures on LIBERO-Goal, Causeway raises bare-switch success from 3-26% to 47-65% and increases the rate of reaching the handoff neighborhood by 42-72 percentage points across models. Additional experiments on LIBERO-Object and a real xArm platform show that the recovery extends beyond the main LIBERO-Goal setting, both in simulation and on a robot.","published":"2026-09-25T07:25:19Z","updated":"2026-09-25T07:25:19Z","authors":["Qingzi Wang","Kaixi Feng","Guangyao Shi","Xiyang Wu","Ang Li","Dinesh Manocha"],"categories":["cs.RO"],"headline":"New Method Helps Robots Switch Between Tasks More Successfully","summary":"Researchers studied how robots can struggle to start a new task if their physical state was changed by a previous one, creating what they call 'task islands.' They developed Causeway, a technique that, without retraining, adjusts the robot's internal plans based on its current position to help it successfully switch tasks. Their tests showed that Causeway significantly improves the robot's ability to start and complete new tasks after finishing others, both in simulations and on a real robot."},{"id":"2609.30912v1","title":"Music Source Separation via Stem Discovery","abstract":"Music source separation (MSS) methods aim to extract stems from music mixtures, which is important, for example, in karaoke, music remixing, and pedagogical applications. While earlier research on MSS systems has been dominated by models targeting narrow sets of general stems, there have recently been attempts to support broader source definitions. One of these methods uses audio queries to provide direct and descriptive control over the desired separation targets based on the sound itself. However, query-based separation remains cumbersome due to the need to provide audio examples with features matching the sources contained within mixtures. This paper proposes Music Source Separation via Stem Discovery (MuS3D), a query-based source separation framework that iteratively discovers active sources from the mixture. On correctly detected sources, our model matches manually queried baselines and surpasses state-of-the-art text-based models. Subjective evaluation indicates encoding artifacts as the limiting factor in current generative separation. The findings suggest that audio-based query representations offer an effective and automatable interface for source separation.","published":"2026-09-25T07:23:00Z","updated":"2026-09-25T07:23:00Z","authors":["V. Valtteri Kallinen","Eloi Moliner","Lauri Juvela","Vesa Välimäki"],"categories":["eess.AS","cs.SD"],"headline":"New Method Improves Music Source Separation with Audio Queries","summary":"Researchers studied how to separate different parts of music, like vocals or instruments, from mixed recordings, which helps in karaoke or remixing. They developed a new system called MuS3D that uses sound examples to find and separate these parts more effectively without needing exact audio samples. Their tests show MuS3D works as well as manual methods and better than text-based ones. However, audio quality issues still limit the system, and the research focuses mainly on this new approach rather than practical applications."},{"id":"2609.30911v1","title":"Infrared laser stimulated broadband white emission of transparent Cr:YAG ceramics obtained by solid state reaction sintering","abstract":"In the present work, the Light Induced White Emission (LIWE) was investigated as one of the sources of laser loss in transparent Cr:YAG ceramics. We have found that transparent Cr:YAG ceramics is capable of generating bright LIWE at excitation powers above a certain threshold. Intensity of LIWE strongly depends on the excitation power and ambient pressure. The host temperature estimated from Cr3+ luminescence was found to be below 600°C, while the most intense white emission was found between 50-400°C. The mechanism of the laser induced white emission was discussed in terms of Intervalence Charge Transfer (IVCT) in chromium mixed valence pair.","published":"2026-09-25T07:20:16Z","updated":"2026-09-25T07:20:16Z","authors":["M. Chaika","R. Tomala","W. Strek"],"categories":["cond-mat.mtrl-sci"],"headline":"Studying laser-induced white light in special transparent ceramics","summary":"Researchers investigated how transparent Cr:YAG ceramics produce bright white light when hit by an infrared laser above a certain power. They found the light’s brightness depends on the laser power and surrounding pressure, with the strongest light shown between 50-400°C. The study suggests this white light comes from interactions involving chromium ions in the material, but these findings are based on observations and a proposed explanation rather than confirmed practical uses."},{"id":"2609.30910v1","title":"Preparation Changes the Cost of Calibration for Quantum Control","abstract":"Error-correction records can reveal noise yet supply the information needed for control too slowly. In a surface code under Gaussian dephasing, encoded calibration learns a correlation that reverses the preferred control mainly through rare events. A product probe exposes it more often through the same checks, reducing ideal-record sample complexity from inverse-square to inverse-linear in phase variance. Counting both calibration and execution costs, we show that this gain can repay probe overhead and enable protected tasks within budgets that exclude encoded calibration.","published":"2026-09-25T07:18:40Z","updated":"2026-09-25T07:18:40Z","authors":["Xiu-Hao Deng"],"categories":["quant-ph"],"headline":"Improving Calibration Efficiency for Quantum Control Systems","summary":"Researchers studied how to better calibrate quantum control systems, which are affected by noise that is hard to track quickly. They found that using a simpler measurement method, called a product probe, can detect important noise patterns more efficiently than traditional error-correcting records. This approach reduces the amount of data needed for calibration from a steep to a more manageable rate. Their results suggest this method could make quantum tasks more feasible within practical resource limits, though findings are based on theoretical analysis."},{"id":"2609.30909v1","title":"Machine Unlearning for Large Language Models: Foundations, Advances, and Agentic Extensions","abstract":"Machine unlearning aims to remove target influence while preserving other capabilities. This survey compares methods, benchmarks, and evidence across large language models and systems using retrieval, memory, tools, and interacting agents. A five-layer framework connects removal requests, system boundaries, target locations, interventions, and supported claims. A seven-stage lifecycle and six evidence dimensions guide comparison. The review shows that target construction, retained data, and recovery tests affect reported outcomes. Evidence from model evaluations remains insufficient to establish removal across external state and subsequent updates, motivating evaluation that tracks dependencies and tests whether target influence returns.","published":"2026-09-25T07:17:20Z","updated":"2026-09-25T07:17:20Z","authors":["Xiaoyu Xu","Minxin Du","Li Bai","Junxu Liu","Yaxin Xiao","Kun Fang","Liu Yang","Huadi Zheng","Peizhao Hu","Qingqing Ye","Haibo Hu"],"categories":["cs.CR"],"headline":"Survey on Removing Specific Information from Large Language Models","summary":"Researchers reviewed different methods for removing targeted information from large language models while keeping their other abilities intact. They created a framework to compare approaches, considering how removal requests are made, where the data lives, and how success is tested. Their analysis shows that current tests and evaluations are not enough to confirm complete removal, especially when models update or interact with external data. They suggest more thorough evaluation to understand if the unwanted information truly disappears or can reappear later."},{"id":"2609.30908v1","title":"Summary-powered prediction under distribution shift","abstract":"Prediction models can perform poorly when the deployment population differs from the training population. Data from the target population would help, but individual-level target data may be inaccessible because of access restrictions or reporting conventions. We consider a multi-resolution setting in which individual-level data are available from a source population, while the target population is observed only through subgroup summaries. We propose SAGE, a one-step estimator that updates a source-trained predictor using a gradient estimated from these summaries. Motivated by diagnostics consistent with the random distribution shift model, we choose SAGE's step-size to account for both sampling and distributional uncertainty. Under this model, SAGE reduces mean asymptotic target excess risk relative to the source-trained predictor. We also show that, under the model, entropy-balancing weighted empirical risk minimization (EB), which reweights source observations to match the target summaries, is asymptotically equivalent to a full-step SAGE update. SAGE with the optimal step-size has asymptotic mean squared error no larger than that of EB. Across real-world datasets, SAGE generally improves on the source-trained predictor and one-step updates that ignore distribution shift, including when the random shift model only partially captures the observed shifts. Compared to EB, SAGE improves prediction more consistently across the sample size and shift settings studied.","published":"2026-09-25T07:14:18Z","updated":"2026-09-25T07:14:18Z","authors":["Ivy Zhang","Dominik Rothenhaeusler"],"categories":["stat.ME"],"headline":"New method improves predictions when data populations differ","summary":"Researchers studied how prediction models struggle when the group they are used on differs from the one they were trained with. They developed a method called SAGE that adjusts predictions using summary information from the new group instead of individual data. Their approach aims to reduce prediction errors by carefully updating the model, and tests on real datasets showed SAGE often improves predictions better than other methods, even when assumptions about data differences are only partly true."},{"id":"2609.30907v1","title":"Laser induced broad band white emission from transparent Cr4+:YAG ceramics: Origin of broadband emission","abstract":"Laser-induced white (light) emission was observed from transparent Cr:YAG ceramics irradiated with a focused continuous wave beam of light from an infrared laser diode. The laser-induced white emission is detected only on the surface of the sample and is not observed in volume. It is found that the intensity of the emission increases exponentially with the laser power density above the threshold. The impact of broadband emission on the power of the transmitted laser beam through the sample was measured. The disappearance of broadband emission due to displacement of the laser beam or an increase in ambient pressure leads to a decrease in the power of transmitted laser beam. Origins of the laser-induced white light emission along with its characteristic features are discussed in terms of multiphoton absorption, intervalence charge transfer and ionic space charge models.","published":"2026-09-25T07:13:59Z","updated":"2026-09-25T07:13:59Z","authors":["M. Chaika","W. Strek"],"categories":["physics.optics","cond-mat.mtrl-sci"],"headline":"Laser Creates Broad White Light on Cr:YAG Ceramic Surfaces","summary":"Researchers studied how shining an infrared laser on Cr:YAG ceramic materials causes bright white light to appear only on the surface. They observed that the brightness grows quickly as laser power increases beyond a certain point. By testing how this white light affects the laser passing through the material, the team linked the effect to complex processes involving multiple light absorptions and charge movements. Their findings come from laboratory observations and theoretical explanations discussed in the abstract."},{"id":"2609.30906v1","title":"ToolSearcher: Optimizing Tool Selection at Scale via Reinforcement Learning","abstract":"Large language models (LLMs) excel at natural language processing but struggle to interact with external environments. Tool learning provides a promising way to extend LLMs into actionable agents, where tool selection is a critical prerequisite for successful tool use. Existing work often assumes a small or predefined set of tools, leaving large-scale tool selection underexplored. Real-world repositories contain a vast and diverse array of tools, making it difficult for LLMs to effectively search, distinguish, and compose tools under context-length constraints. We identify large-scale tool selection as a new challenge for agentic reinforcement learning, highlighting that existing RL methods for knowledge-based question answering are inadequate for selecting tools while considering compatibility. To address this challenge, we propose ToolSearcher, a novel RL framework for effective multi-turn search and fine-grained optimization in large-scale tool selection. Specifically, we introduce category-constrained tool discrimination to improve the model's ability to distinguish functionally similar tools, event-level search modeling to explicitly optimize the discovery of target tools during multi-turn search, and trajectory-aligned credit allocation to provide fine-grained reward signals for different stages of the search-selection process. Extensive experiments on large-scale tool selection benchmarks demonstrate that ToolSearcher consistently outperforms a set of strong baselines in challenging settings involving iterative search and complex tool composition.","published":"2026-09-25T07:13:21Z","updated":"2026-09-25T07:13:21Z","authors":["Zhenlong Dai","Xujie Song","Zitong Wang","Tong Niu","Jian liu","Weiqiang Wang","Xiu Tang","Sai Wu","Chang Yao","Jingyuan Chen"],"categories":["cs.CL"],"headline":"New Method Helps AI Choose Tools More Effectively","summary":"Researchers studied how large language models (LLMs), which understand language well, find it hard to pick appropriate tools from many options. They developed ToolSearcher, a learning approach that helps LLMs search through and select the best tools step-by-step by improving their ability to tell similar tools apart and rewarding good choices along the way. Experiments show ToolSearcher works better than existing methods for complex tool selection, but these findings are based on simulation tests described by the researchers."},{"id":"2609.30905v1","title":"Network Analysis in Communication Research: Research Topics, Knowledge Organization, and Research Practices","abstract":"Communication network research explains access to information, patterns of participation, and the organization of public meaning through different relationships and observations. This integrative review connects research topics, knowledge organization, and research practices, using bibliographic analysis of a Web of Science candidate pool to guide selective reading. Three judgments emerge from the comparisons. First, the relevance of a connection depends on the task and the criterion of value: team members anticipate consulting colleagues whose expertise they recognize, while journalists distinguish monitoring sources, using their information, and citing them. Second, commonality at one level can coexist with differentiation at another: audiences share outlets while selecting different articles, and shared issue agendas accommodate different evaluations. Third, some differences remain unresolved: contrasting media-use influence findings cannot be explained simply by whether models include selection and content co-nomination. Reading the uses of homophily, transactive memory, sourcing, agenda-setting, and framing resources clarifies which expectations and observations support these judgments. Selected citation contexts also show how conceptual and measurement resources enter the same argument. The resulting agenda calls for comparisons of relationship types across group stages, source use across reporting tasks, and encountered content with recipients' interpretations. These purposive comparisons establish specific connections among literatures without estimating their prevalence or demonstrating a common causal mechanism.","published":"2026-09-25T07:13:10Z","updated":"2026-09-25T07:13:10Z","authors":["Pengjia Cui"],"categories":["cs.SI","stat.OT"],"headline":"Studying Communication Networks: How Relationships Shape Information Use","summary":"Researchers reviewed studies on communication networks to understand how people access information and share meanings. They found that the importance of connections varies depending on tasks, like teamwork or journalism, and that groups can share sources but interpret information differently. Some conflicting findings about media influence remain unclear. The researchers suggest more focused comparisons across different communication stages and tasks, but their work does not prove widespread patterns or causes—it's a conceptual overview linking existing studies."},{"id":"2609.30904v1","title":"QReason: Query-Focused Decoupled Chain-of-Thought for Efficient Passage Reranking","abstract":"Passage reranking plays a crucial role in information retrieval by refining the ordering of candidate passages to better reflect relevance. Existing listwise LLM rerankers with Chain-of-Thought (CoT) reasoning can handle complex queries effectively, but they suffer from substantial redundancy and high latency due to sliding-window strategies, which repeatedly generate highly similar CoTs. To address this, we propose QReason, a decoupled framework that separates query-focused reasoning from window-specific passage relevance assessment. Specifically, QReason introduces a dedicated rewriter that generates a ranking-oriented reasoning query once, capturing the query's core intent while avoiding redundant reasoning, and then reuses it across all windows with a non-reasoning reranker. The rewriter is trained via a two-stage process that first uses supervised fine-tuning with relevant-passage guidance through semantic evidence to produce deeply grounded, query-focused CoTs. It then applies reinforcement learning to align CoT generation with both the inference-time setting and the reranking objective, optimizing listwise metrics and passage-level discrimination to produce reusable reasoning chains for reranking. Experiments on the BRIGHT benchmark demonstrate that QReason significantly reduces redundant reasoning, achieves ranking performance comparable to or better than strong reasoning-based rerankers, and outperforms existing query rewriting models.","published":"2026-09-25T07:13:09Z","updated":"2026-09-25T07:13:09Z","authors":["Yang Zhang","Wenhan Liu","Qiannan Zhu","Mingming Li","Yuanfei Huang"],"categories":["cs.IR"],"headline":"New Method Improves Efficiency in Ranking Search Results","summary":"Researchers addressed the problem of slow and repetitive reasoning in systems that reorder search results to better match complex queries. They developed QReason, which first creates a single clear interpretation of the query's main intent, then applies this to evaluate passages efficiently without repeating reasoning. Testing showed QReason reduces unnecessary work while maintaining or improving how well results are ranked, based on experiments using a standard benchmark. The findings are based on simulations and comparisons within the study."},{"id":"2609.30903v1","title":"A Bonded-Particle Discrete-Element Model with Self-Gravity for the Collision and Reaccumulation of Rubble-Pile Asteroids: Formulation and Numerical Protocol","abstract":"I present a discrete-element approach developed to study the collisional and gravitational evolution of small self-gravitating granular aggregates. Each body is represented as a composite of cohesively bonded spherical grains, so that fragmentation and reaccumulation emerge from the microscopic competition between short-range dissipative contact forces, an irreversible bond-breaking criterion, and pairwise Newtonian self-gravity summed over every particle pair, including those internal to a single body. A numerical experiment is described: a random-sequential packing procedure that generates two bonded spherical aggregates, a controlled two-body encounter with tunable approach speed and impact parameter, and a systematic parameter sweep designed to map the conditions under which two bodies launched from rest under their mutual gravity alone either merge into a single remnant or fail to settle after a low-speed encounter. Of the 125 simulated collisions, 56.8\\% result in the formation of a single residue. The cohesive network remains intact for $Π_σ\\lesssim0.036$, is partially damaged with dispersion determined by the contact damping ratio, and breaks completely for $Π_σ\\approx3.59$","published":"2026-09-25T07:11:37Z","updated":"2026-09-25T07:11:37Z","authors":["Yohann Trivino"],"categories":["cond-mat.soft","astro-ph.EP"],"headline":"Modeling How Rubble-Like Asteroids Collide and Stick Together","summary":"Researchers developed a computer model that simulates small asteroids made of many bonded grains held together by gravity and sticky forces. They ran experiments where two such asteroid models collided at various speeds, observing whether they merged or broke apart. The study helps understand how these loosely held space rocks behave during impacts, but the findings are based on simulations of simplified conditions and require further testing."},{"id":"2609.30902v1","title":"Asymptotic profiles for solutions to the incompressible Navier-Stokes equations in the critical Fourier-Herz spaces","abstract":"We consider the large time asymptotic behavior of solutions to the initial value problem for the incompressible Navier-Stokes equations in the whole space $\\mathbb{R}^d$ ($d\\ge 2$). When the initial velocity belongs to $L^1(\\mathbb{R}^d)$, it is well-known that its spatial integral vanishes as a consequence of the divergence-free compatibility condition. Taking this property into account, Carpio (1996) and Fujigaki-Miyakawa (2001) derived higher-order asymptotic formulas for the strong solutions. In this paper, we investigate the large time behavior of solutions when the initial velocity belongs to the Fourier-Herz space $\\widehat{L}^1(\\mathbb{R}^d)$, which provides a broader framework than $L^1(\\mathbb{R}^d)$ and imposes an additional condition in the low-frequency region. In particular, in the two-dimensional case, we show that an asymptotic profile different from those obtained by Carpio and Fujigaki-Miyakawa arises.","published":"2026-09-25T07:11:29Z","updated":"2026-09-25T07:11:29Z","authors":["Ikki Fukuda","Ryosuke Nakasato"],"categories":["math.AP"],"headline":"Studying long-term patterns of fluid flow in advanced mathematical spaces","summary":"The researchers studied how solutions to the incompressible Navier-Stokes equations, which describe fluid flow, behave over a long time when starting conditions belong to a specific mathematical space called the Fourier-Herz space. This space is broader than previously considered spaces and includes extra conditions. They found that in two-dimensional cases, the long-term behavior of the fluid differs from earlier known results. Their findings are theoretical and focus on understanding these mathematical solutions rather than real-world experiments."},{"id":"2609.30901v1","title":"Thorium-229 as a Phonomagnetometer","abstract":"Thorium-229 possesses the only known low-energy nuclear transition suitable for spectroscopy with narrowband VUV lasers. While previous experiments have focused on application as a nuclear clock, this transition also offers a route to high-accuracy magnetometry. Experimental observations of magnetic fields generated by phonons carrying angular momentum remain inconclusive, highlighting the need for quantitative tests of the underlying physical mechanisms. In this article, we propose doping thorium-229 into a solid-state host to probe phonomagnetic fields in situ. We identify Na$_2$ThF$_6$, a chiral stoichiometric thorium compound, as a promising host material and evaluate its spectroscopic sensitivity to magnetic interactions under two excitation schemes: driving degenerate phonon modes with a circularly polarized laser, and applying a temperature gradient. Density functional theory simulations combined with quantitative estimates suggest that the temperature-gradient scheme yields a magnetic signal that appears too weak to be measurable, while a circularly polarized, high-intensity THz/VUV pump-probe driven phonomagnetic response may approach the shot-noise-limited detection threshold. While experimental challenges remain, the unique suitability of thorium-229 provides a testable pathway toward detecting phonomagnetic fields in a solid-state platform.","published":"2026-09-25T07:10:09Z","updated":"2026-09-25T07:10:09Z","authors":["Martin Pimon","Andreas Grüneis","Thorsten Schumm","Kjeld Beeks"],"categories":["cond-mat.mtrl-sci"],"headline":"Using Thorium-229 to Detect Magnetic Effects from Vibrations","summary":"The researchers explore how thorium-229, known for its rare nuclear properties, could help measure magnetic fields caused by vibrations in solids, called phonomagnetic fields. They suggest adding thorium-229 to a special crystal and use theoretical models to study two ways of detecting these magnetic signals. Their simulations indicate one method is likely too weak to detect, but another might barely reach detection limits. The findings outline a possible experimental approach but highlight that practical challenges remain."},{"id":"2609.30900v1","title":"Odd Cycle Transversal on $H$-free graphs","abstract":"\\textsc{Odd Cycle Transversal} is a classic $\\mathsf{NP}$-hard graph optimization problem asking for a minimum-weight set of vertices whose deletion makes the input graph bipartite, or equivalently, a maximum-weight induced bipartite subgraph. We show that \\textsc{Odd Cycle Transversal} is quasi-polynomial-time solvable on $kP_4$-free graphs, for every fixed $k \\in \\mathbb{N}$. In fact, we provide an $n^{O_k(\\log n)}$-time algorithm for the more general \\textsc{Max-Weight List $2$-Colorable Induced Subgraph}, where the notation $O_{k}(\\cdot)$ hides factors depending on $k$. Paired with known results from the literature, this allows us to obtain a complete complexity dichotomy for these two problems on $H$-free graphs into cases solvable in quasi-polynomial time and cases which are $\\mathsf{NP}$-hard, in particular resolving an open problem of Agrawal, Lima, Lokshtanov, Saurabh, and Sharma [SODA 2024]. Our algorithms are based on a new structural tool that may be of independent interest. We introduce the notion of $H$-amiable family and show that, for every fixed graph $H$ without isolated vertices and every fixed $k\\ge2$, every $kH$-free graph admits an $H$-amiable family of quasi-polynomial size that can be constructed in quasi-polynomial time. Besides yielding the aforementioned algorithms, this result gives, for every fixed connected graph $H$ and every fixed $k\\ge2$, a reduction from \\textsc{Max-Weight Independent Set} on $kH$-free graphs to the same problem on $H$-free graphs with $n^{O_{H,k}(\\log n)}$ overhead. In this setting, it improves the $n^{O_{H,k}(\\log^3 n)}$ overhead obtained by specializing the general reduction of Gartland and Lokshtanov [FOCS 2020].","published":"2026-09-25T07:08:47Z","updated":"2026-09-25T07:08:47Z","authors":["Esther Galby","Paloma T. de Lima","Andrea Munaro","Amir Nikabadi"],"categories":["cs.DS","cs.DM","math.CO"],"headline":"New algorithm helps solve odd cycle problems on special graphs","summary":"Researchers studied the Odd Cycle Transversal problem, which involves removing the fewest vertices to make a graph bipartite. They developed a quasi-polynomial time algorithm for graphs that avoid certain repeated structures, known as kP4-free graphs, extending to a related coloring problem. Using a new concept called an H-amiable family, they achieved a complexity classification for these problems on various graphs and improved known reductions in related graph optimization tasks. Their results are theoretical and based on mathematical structures rather than practical tests."},{"id":"2609.30899v1","title":"Room-temperature quantum-sensing molecular crystals grown in minutes","abstract":"Quantum sensing increasingly demands materials combining room-temperature optical spin addressability and coherent control with rapid, accessible fabrication. Yet semiconductor platforms require costly materials and specialized processing, while molecular systems often depend on bespoke synthesis or involved crystal growth. Here we apply microspacing in-air sublimation (MAS) to rapidly fabricate quantum-sensing materials within minutes. Microgram quantities of commercially available precursors crystallize on glass, in air and without vacuum or inert-gas handling. Monitored in situ, growth yields optically clear, stoichiometric single crystals of excellent quality, including plates and needles with intrinsic optical waveguiding. Donor-acceptor charge-transfer co-crystals pairing anthracene or biphenyl with tetracyanobenzene, together with localized-exciton pentacene:pentacenequinone (P:PQ), exhibit room-temperature optically detected magnetic resonance (ODMR). Charge-transfer ODMR resonances reach linewidths below $4\\mathrm{MHz}$, consistent with motional narrowing of mobile triplet excitons, while P:PQ supports coherent ensemble spin control with $T_2=0.56μ\\mathrm{s}$. Functionality across chemically and electronically distinct pairs establishes MAS as an accessible screen for molecular quantum-sensing materials, connecting molecular crystal engineering with low-cost, chemically tunable thin triplet-spin layers and waveguiding sensor geometries for quantum sensing and microscopy.","published":"2026-09-25T07:06:07Z","updated":"2026-09-25T07:06:07Z","authors":["Madhur Parashar","Guangzhao Chen","Emanuel Druga","Liang Z. Tan","Jeffrey Reimer","Ashok Ajoy"],"categories":["quant-ph"],"headline":"Fast, easy way to make room-temperature quantum sensing crystals","summary":"Researchers developed a quick method to grow high-quality molecular crystals for quantum sensing at room temperature using a simple air-based process. These crystals are made from common chemicals without special gases or vacuum, and they show promising properties for detecting magnetic signals through light. The study highlights this new method as a low-cost and accessible way to screen materials for quantum sensing, though the findings are based on controlled lab fabrication and optical measurements."},{"id":"2609.30898v1","title":"A Bregman inertial iteratively regularized extragradient method for bilevel variational inequality problems","abstract":"In this paper, we consider bilevel variational inequality problems, where the feasible set is the solution set of another variational inequality. We propose a Bregman inertial regularized extragradient method for solving these problems. By performing the inertial extrapolation in the dual space, the proposed method aligns the inertial step with the structure of the Bregman three-point identity. We also derive explicit non-asymptotic bounds for the generalized gap function. More precisely, using diminishing regularization, we establish convergence rates of $\\mathcal{O}(1/k^{1-b})$ for the optimality gap and $\\mathcal{O}(1/k^b)$ for the feasibility gap with $ 0 < b< 1$. For constant regularization parameter $η>0$, we establish an $\\mathcal{O}(1/k)$ rate for the optimality gap and an $\\mathcal{O}(1/k)+\\mathcal{O}(η)$ bound for the feasibility gap. Furthermore, numerical experiments on traffic networks and multi-portfolio Nash equilibrium problems demonstrate the practical effectiveness of the proposed framework.","published":"2026-09-25T07:04:42Z","updated":"2026-09-25T07:04:42Z","authors":["Kangming Chen","Atsushi Hori","Ellen H. Fukuda"],"categories":["math.OC"],"headline":"New Method to Solve Complex Mathematical Problems Improves Accuracy","summary":"Researchers developed a new mathematical method to solve bilevel variational inequality problems, where one problem's solution depends on another problem. They created an approach called the Bregman inertial regularized extragradient method, which improves how quickly solutions get closer to the correct answer. Their analysis shows explicit rates for how fast the method performs, and numerical tests on traffic and financial network problems demonstrate that the method can work well in practice. The results are based on mathematical theory and simulations."},{"id":"2609.30897v1","title":"From annotation to reasoning: Culture in language models","abstract":"How should we evaluate language models when more than one interpretation can be right? Cultural benchmarks often test factual knowledge, agreement with survey responses, or recognition of a predefined meaning. These tasks leave open whether a model can explain how a cultural reference works in a particular text, support a reading with evidence, or revise it after criticism. This is a question of interpretive depth, complementary to the breadth of cultural coverage. We argue that literary interpretation offers a useful setting for studying these capabilities. We focus on cultural referencing and reuse: how texts invoke, repeat, and transform earlier expressions across historical and linguistic contexts. Our central claim is that literary scholars can disagree about an interpretation while recognizing the quality of its support. We propose linking evidence-centered benchmarks, evaluation that preserves scholarly disagreement, and model-development experiments on literary data, contextual resources, and scholarly feedback. Danish literature provides a concrete starting point, with implications for other languages and domains. The aim is to develop alternative evaluation strategies that go beyond conventional benchmark metrics and guide model development toward cultural robustness in AI systems.","published":"2026-09-25T07:04:41Z","updated":"2026-09-25T07:04:41Z","authors":["Daniel Hershcovich","Alexander Conroy","Jens Bjerring-Hansen"],"categories":["cs.CL"],"headline":"New Ways to Evaluate Language Models on Cultural Understanding","summary":"Researchers explore how to better assess language models when multiple interpretations can be valid, especially regarding culture. They focus on literary texts to see if models can explain and support cultural references, not just recognize them. The study suggests involving literary scholars and using evidence-based methods to evaluate models, starting with Danish literature, aiming to improve AI's cultural understanding beyond standard tests. These findings are theoretical and propose new evaluation approaches rather than practical results."},{"id":"2609.30896v1","title":"Large language models underestimate and partly misrepresent cultural variation in everyday norms","abstract":"A key aspect of culture is a society's norms about everyday behavior. How accurately do large language models (LLMs) represent cultural differences in such norms? To answer this question we used the recent Global Study of Everyday Norms (GSEN), which collected ratings of 150 scenarios in 90 societies, as the human benchmark. We prompted GPT-5 to estimate each society's average rating for every scenario, and later repeated the benchmark in three other LLMs: GPT-5.4, Claude Opus 4.6, and Gemini 3.1 Pro. Compared to GSEN estimates, all four LLMs misrepresented cultural variation in two ways. First, they greatly underestimated its magnitude, estimating differences between societies to be, on average, less than half their measured size. Second, for many scenarios the LLMs poorly identified the pattern of variation, that is, which societies judged the behavior less acceptable and which societies judged it more acceptable. The pattern of variation was identified better for scenarios that elicit concerns about vulgarity, especially scenarios involving kissing and flirting. We also found that norms in more developed societies tended to be estimated somewhat more accurately, and that prompting in local survey languages rather than English produced only a modest improvement in accuracy. Local-language prompting also reduced, but did not remove, the underestimation of between-society differences. Cultural differences in everyday norms are only weakly and unevenly represented by LLMs.","published":"2026-09-25T07:04:13Z","updated":"2026-09-25T07:04:13Z","authors":["Kimmo Eriksson","Irina Vartanova","Pontus Strimling"],"categories":["cs.CY"],"headline":"Language Models Undervalue Cultural Differences in Everyday Norms","summary":"Researchers compared large language models' predictions to a global human study on social norms across 90 societies. They found that models like GPT-5 significantly underestimated how much norms vary between cultures and often got the specific patterns of these differences wrong. Some improvement appeared with scenarios about vulgarity and when using local languages, but overall, the models only weakly and unevenly capture cultural variation in everyday behavior."},{"id":"2609.30895v1","title":"Proof of the Kahn Saks Conjecture","abstract":"Let $\\mathbb{P}(x\\prec y)$ be the probability that $x$ precedes $y$ in a uniformly random linear extension of an $n$-element poset $P$, and define the balancing coefficient to be $δ(x,y)=\\min(\\mathbb{P}(x\\prec y),\\mathbb{P}(y\\prec x))$ with $δ(P)=\\max_{x,y}δ(x,y)$. We prove (Theorem 1) that sufficiently large width forces $δ(P)$ to be arbitrarily close to $1/2$, answering a long-standing conjecture of Kahn and Saks. In fact, we prove the stronger result (Theorem 2) that large width forces one of two configurations in our poset: either a nearly uniform order on $k$ vertices, or an almost fixed order on $t$ vertices with one further vertex inserted uniformly among the $t+1$ slots. We also show that, for fixed $k$, the first possibility must occur within any antichain $X$ of size $Ω(n^{2/3})$.","published":"2026-09-25T07:04:00Z","updated":"2026-09-25T07:04:00Z","authors":["Max Aires"],"categories":["math.CO","cs.DM"],"headline":"Researchers prove a long-standing math conjecture on order probabilities","summary":"The researchers addressed a problem about the likelihood that one element precedes another in a random ordering of a partially ordered set (poset), focusing on a measure called the balancing coefficient. They proved that if the poset is wide enough, this coefficient gets very close to one-half, confirming the Kahn Saks conjecture. Their work also identifies specific structural patterns that appear in such wide posets. These results come from theoretical mathematical proofs, not experiments or simulations."},{"id":"2609.30894v1","title":"Training Graph Foundation Models on The Web Graph","abstract":"We introduce Acacia, a graph foundation model, trained on the web graph. Acacia (i) supports arbitrary feature dimensionalities and semantics without additional training, (ii) supports a wide range of tasks, including node classification, link prediction, node clustering, and graph generation, without additional training, (iii) has in-context learning capabilities, and (iv) does not rely on pretrained LLMs. In particular, existing graph foundation models often require training additional classification heads or feature projectors to accommodate new graphs or new labels, whereas Acacia does not. Moreover, existing graph foundation models often gain their capabilities by being stitched together with pretrained LLMs, whereas Acacia is trained from scratch using only the Common Crawl web graph. This is also an important result because it provides evidence that graph models can acquire emergent capabilities from scratch like LLMs.","published":"2026-09-25T07:02:45Z","updated":"2026-09-25T07:02:45Z","authors":["Ryoma Sato"],"categories":["cs.AI","cs.LG"],"headline":"New graph model learns from the web without extra training","summary":"Researchers developed Acacia, a graph foundation model trained only on web data. Unlike other models, Acacia can handle different tasks like classifying or clustering nodes without extra training steps or relying on large language models. The researchers suggest this shows graph models can develop advanced abilities independently, but this conclusion is based on their training approach and initial results."},{"id":"2609.30893v1","title":"Two-coloring cubic graphs with small monochromatic components, but without singletons","abstract":"We combine two coloring aspects that work in opposite directions. One can 2-color the vertices of a cubic graph such that each monochromatic component is very small. One can also 2-color the vertices of a cubic graph such that each monochromatic component has degree at least 1. As an intended tool for solving a special case of Wegner's conjecture, Thomassen formulated a conjecture that combined the two previous properties. This led to the concept of a crumby coloring. However it turned out that there are cubic graphs without such coloring. Here we try to see what natural relaxations of the original concept might hold for each cubic graph. We show there exists a constant $c$ such that every cubic graph has a vertex 2-coloring such that every monochromatic component has at least 2 and at most $c$ vertices. We also prove an unbalanced version, which is the natural relaxation of the crumby coloring.","published":"2026-09-25T06:59:47Z","updated":"2026-09-25T06:59:47Z","authors":["János Barát","Zoltán L. Blázsik"],"categories":["math.CO"],"headline":"Researchers explore new vertex colorings in cubic graphs","summary":"The researchers studied how to color the vertices of cubic graphs (graphs where each vertex connects exactly to three others) using two colors, aiming to keep connected same-color groups small but with no isolated single vertices. They found that while perfect colorings meeting all previous criteria don't always exist, there is a way to color any cubic graph so that each same-color group has at least two and at most some constant number of vertices. These results are theoretical and help understand coloring constraints in cubic graphs."},{"id":"2609.30892v1","title":"Universal Formulas for de Jonquières Loci and Equivariant Contact Problems","abstract":"Given a family of genus $g$ curves $f:Ψ\\to S$ and a relative degree $d$ line bundle $L$ on $Ψ$ such that $f_*L$ is a vector bundle, the projectivization $\\mathbb{P}(f_*L)$ is stratified by relative de Jonquières loci, which parameterize sections of $L$ with prescribed vanishing orders. We give universal formulas for the classes of these loci in terms of the relative $O(1)$ on $\\mathbb{P}(f_*L)$ and tautological classes pulled back from the universal Picard stack $\\mathrm{Pic}_g^d$. When $f$ is the family of lines in $\\mathbb{P}^r$, we use the universal polynomials to compute $\\mathrm{GL}_{r+1}$-equivariant classes of loci of hypersurfaces and complete intersections admitting lines with prescribed contact orders.","published":"2026-09-25T06:56:21Z","updated":"2026-09-25T06:56:21Z","authors":["Smita Rajan"],"categories":["math.AG"],"headline":"Researchers Find Universal Formulas for Special Curve Sections","summary":"The researchers studied families of geometric objects called genus g curves and associated line bundles. They developed universal formulas that describe special subsets where sections of these bundles vanish in prescribed ways. Using these formulas, they also analyzed symmetrical classes of geometric shapes like hypersurfaces in projective space with specified contact properties. Their work provides theoretical results expressed through abstract mathematical objects and does not include experimental or direct practical applications."},{"id":"2609.30891v1","title":"Adaptive Pilot Selection for Unified Semantic Communication and Semantic Sensing in ISAC","abstract":"Semantic communication (SemCom) and integrated sensing and communication (ISAC) are promising technologies for future 6G wireless networks. Existing studies have applied semantic technology to either the communication module or the sensing module of ISAC. In this work, we propose SemISAC, which performs both SemCom and semantic sensing within a single dual-function waveform. SemISAC uses a joint semantic encoder that extracts task-specific information for both communication and sensing. We evaluate SemISAC in a vehicular scenario in which vehicles share pixel-wise segmentation of the road environment and, through sensing, classify surrounding objects and estimate their ranges. On the transmitter side, a deep learning encoder converts the input road-scene image into semantic symbols and places them on the data cells of an OFDM grid, while the remaining cells serve as pilots for channel state information estimation and sensing. The pilot configuration is adaptively optimized based on the channel conditions to balance communication and sensing requirements. At the receiver, a deep learning model reconstructs the segmentation from the received waveform, while the transmitting vehicle captures the reflected waveforms from surrounding objects and uses task-specific deep learning decoders for target recognition and range estimation. Simulation results show that SemISAC achieves a segmentation accuracy close to that of the dedicated SemCom module while outperforming both conventional and semantic baselines in target recognition and range estimation.","published":"2026-09-25T06:54:24Z","updated":"2026-09-25T06:54:24Z","authors":["Muhammad Abubakar Rashid","Muhammad Hannan Akram","Haejoon Jung","Syed Ali Hassan"],"categories":["eess.SP","cs.AI"],"headline":"Smart Wireless System Combines Communication and Sensing for Vehicles","summary":"Researchers developed SemISAC, a system that lets vehicles communicate and sense their surroundings using a single signal. It uses deep learning to turn road images into data for both sharing information and detecting objects nearby. Simulations show SemISAC performs well in accurately identifying road scenes and recognizing objects, balancing the needs of communication and sensing under different conditions."},{"id":"2609.30890v1","title":"From Segments to Trajectories: Evolving Affective Graphs with Evidence Retrieval for Continuous EEG Emotion Recognition","abstract":"Electroencephalography (EEG)-based emotion recognition is important for affective computing and human-computer interaction, yet most existing methods divide a long trial into short segments and assign each segment the label of its source trial. Although this strategy increases the number of training samples, it reduces an evolving emotional response to a segment-level, coarse-grained, and static prediction problem. In reality, emotion may continuously emerge, intensify, weaken, and fluctuate as a stimulus unfolds, motivating the prediction of a time-aligned affective trajectory from the complete EEG trial. This task requires coordinated modeling of how spatial neural organization evolves throughout the trial and how local emotional fluctuations interact with longer-term trends. In this work, we formally define and systematically investigate continuous EEG emotion recognition as whole-trial affective trajectory prediction. We propose EAGER, an Evolving Affective Graph framework with Evidence Retrieval for continuous EEG emotion recognition. EAGER comprises two complementary modules: Affective State-guided Topology Evolution models the evolving spatial organization of EEG activity, while Multi-scale Temporal Evidence Retrieval integrates short-term fluctuations with longer-range temporal trends for time-aligned prediction. Experiments on MAHNOB-HCI, SEED-VII, and REFED show consistent gains in trajectory-tracking metrics over representative methods, with competitive pointwise errors.","published":"2026-09-25T06:53:59Z","updated":"2026-09-25T06:53:59Z","authors":["Chi Yang","Jihong Wang","Chengxi Xie","Kai He","Huan Liu","Man Yao","Shile Qi","Yuzhe Zhang"],"categories":["cs.HC"],"headline":"New Method Tracks Emotions Continuously Using Brain Signals","summary":"Researchers studied how to better recognize emotions from EEG brain scans by looking at emotional changes over whole time periods, not just short segments. They developed a method called EAGER that models evolving brain activity and combines short- and long-term emotional patterns to predict emotional states continuously. Their tests suggest EAGER improves tracking of emotional trajectories compared to previous methods, though the findings are based on specific datasets and focus on modeling rather than real-world application."},{"id":"2609.30889v1","title":"PHASE: Compliance-Enabled Tactile Phase Retrieval for Few-Shot Insertion Learning","abstract":"Contact-rich assembly tasks such as peg-in-hole insertion remain difficult to learn from limited demonstrations. While retrieval-augmented imitation learning, which augments target demonstrations with relevant prior data, offers a promising direction, its applicability to contact-rich manipulation remains largely unexplored. Contact-rich insertion unfolds over multiple phases from search to insert, and retrieving phase-specific experience from prior data in principled ways remains an open question. Our key insight is that a compliant wrist enables the robot to sustain contact throughout execution, producing rich tactile and force signals that naturally reveal the phase structure of insertion and inform what should be retrieved. Based on this insight, we present PHASE (PHase-Aware Segmentation and REtrieval), a framework for compliance-enabled tactile phase retrieval that integrates multimodal contact-aware representation learning, variable-length phase segmentation from tactile signals, and phase-consistent retrieval for policy learning. We evaluate PHASE on real-world peg-in-hole insertion across five peg geometries, comparing against retrieval strategies drawn from state-of-the-art methods under a shared policy architecture. PHASE improves the overall success rate by 13 percentage points over the strongest non-phase-aware baseline, and improves performance under unseen initial positions by 30 percentage points. These results demonstrate that aligning retrieval with interaction-defined contact phases substantially improves robustness in few-shot insertion learning.","published":"2026-09-25T06:53:56Z","updated":"2026-09-25T06:53:56Z","authors":["Jeremy Siburian","Cristian C. Beltran-Hernandez","Tatsuya Matsushima","Yusuke Iwasawa","Masashi Hamaya","Mai Nishimura"],"categories":["cs.RO"],"headline":"New Method Helps Robots Learn Precise Insertion Tasks Faster","summary":"Researchers developed PHASE, a system that helps robots perform tasks like peg-in-hole insertion by recognizing different contact phases during the process. Using a flexible wrist and touch sensors, the robot gathers detailed feedback to identify these phases and find relevant prior experiences to improve learning from limited examples. Tests showed PHASE increased success rates significantly compared to other methods, especially when starting from new positions. The findings support that recognizing interaction phases can make robot learning more reliable, though this remains focused on peg insertion tasks."},{"id":"2609.30888v1","title":"Breaking the Infinite Barrier in the $\\frac{1}{3}$--$\\frac{2}{3}$ Conjecture","abstract":"The balance constant is a poset parameter measuring how well random linear extensions can be split according to their relative order on two elements. We give an $\\varepsilon$-improvement for the balance constant over the bound by Brightwell, Felsner and Trotter (1995), for some small $\\varepsilon>0$. This is the first general result towards the $\\frac{1}{3}$--$\\frac{2}{3}$ conjecture in over 30 years.","published":"2026-09-25T06:53:25Z","updated":"2026-09-25T06:53:25Z","authors":["Max Aires","Swee Hong Chan","Igor Pak","Greta Panova"],"categories":["math.CO","cs.DM"],"headline":"New Progress on a 30-Year-Old Math Conjecture About Ordering","summary":"Researchers studied a mathematical concept called the balance constant, which measures how two items can be ordered randomly. They made a small improvement to a previous mathematical bound related to this concept, originally set in 1995. This advancement is the first general progress in over 30 years toward a long-standing problem known as the one-third two-thirds conjecture. The findings come from theoretical work on partially ordered sets and rely on mathematical proofs rather than experiments."},{"id":"2609.30887v1","title":"From Tapping to Hopping: Augmenting Mobile GUI Agents with App-Native Deeplinks","abstract":"Mobile GUI agents complete tasks using GUI actions like taps and swipes. These actions are broadly applicable across applications, but reaching a navigation interface. A single deeplink call can replace a sequence of screen-by-screen GUI actions. We therefore introduce hybrid interaction, using deeplinks for direct navigation and GUI actions for other on-screen operations and fallback. To enable this, we discover candidate deeplinks through static analysis, validate them on real devices, and describe their observed landing screens. This process creates a verified and grounded deeplink catalog that pairs each working deeplink with a description of its landing screen. Using this catalog, we introduce GUI-Hopper, a improves task success in commercial applications on real devices, further demonstrating the benefits of hybrid interaction.","published":"2026-09-25T06:50:04Z","updated":"2026-09-25T06:50:04Z","authors":["Yuchen Sun","Chenglin Cai","Gongjie Zhang","Tianyu Xia","Quyu Kong","Panrong Tong","Zhengwen Zeng","Long Chen","Steven Hoi","Chongyang Zhang","Yue Wang"],"categories":["cs.AI"],"headline":"Researchers Enhance Mobile Agents with Direct App Links","summary":"Mobile GUI agents usually navigate apps through taps and swipes, which can be slow. Researchers propose combining these actions with 'deeplinks,' which jump directly to specific app screens, speeding up navigation. They created a verified catalog of deeplinks by analyzing apps and testing them on devices, then developed a system called GUI-Hopper that uses this hybrid approach. Their work shows potential improvements in completing tasks on real commercial apps, though details are based on their described experiments."},{"id":"2609.30886v1","title":"Conformal Prediction under Exponential-Tilt Joint Shift","abstract":"Conformal prediction can lose coverage when the data distribution changes after deployment. We study adaptation using labeled source data and unlabeled target inputs, allowing both the input distribution and its relationship with outcomes to change. We use Exponential Tilt Reweighting Alignment (ExTRA), introduced for classification by Maity et al. (2023), to estimate structured distribution shifts. We compare using its estimated weights in conformal calibration with additionally tilting the source predictive distribution. Shared learned predictors, estimated weights, calibration samples, and test observations isolate the effect of tilting. Existing theory gives both procedures target coverage with true weights and a common coverage bound with estimated weights. Identification calculations and an analysis of how scoring interacts with weight estimation error help explain why their performance can nevertheless differ. In a synthetic regression setting where the assumed models match the data-generating process and target inputs are informative about the shift, tilting reduces mean set length by about $30\\%$ relative to weighting alone, with both methods attaining coverage near nominal. Tilting can instead cause substantial coverage losses in synthetic classification and in regression when target inputs provide little information about the response shift. Real-data experiments also show no consistent benefit. Good coverage from weighted calibration alone does not ensure that adding predictive tilting will preserve coverage. Deciding when to apply this additional adjustment using only source labels and target inputs remains an open problem.","published":"2026-09-25T06:49:27Z","updated":"2026-09-25T06:49:27Z","authors":["Seungjin Choi"],"categories":["stat.ML","cs.LG","stat.ME"],"headline":"Researchers study ways to adapt predictive models to changing data","summary":"The study looks at how predictive methods called conformal prediction perform when data changes after a model is deployed. The researchers used a technique called Exponential Tilt Reweighting Alignment (ExTRA) to adjust for changes in data distribution and relationships. Their experiments with simulated and real data show that while some adjustments can reduce prediction sizes, they do not always improve reliability, and deciding when to use these adjustments remains uncertain based on available data."},{"id":"2609.30885v1","title":"Retraction-Based Gradient Projection Algorithms on Manifolds","abstract":"We introduce a framework for retraction-based convex optimization on Riemannian manifolds, which includes a notion of retraction-specific convex sets and retraction-based gradient projection algorithms. The standard theory of gradient projection algorithms generalizes easily to this framework. Within this framework, we establish convergence results for retraction-based gradient projection algorithms with various stepsize rules. As an application, we use our framework to study the weighted low-rank approximation. We also provide numerical validation of our convergence results on the image completion task.","published":"2026-09-25T06:47:31Z","updated":"2026-09-25T06:47:31Z","authors":["Conglong Xu","Hao Wu"],"categories":["math.OC","cs.LG"],"headline":"New Approach for Optimization on Curved Spaces","summary":"Researchers developed a framework for solving optimization problems on curved spaces, called Riemannian manifolds, using retraction-based methods. They extended existing gradient projection algorithms to work within this new framework and proved that these methods converge under certain conditions. The team applied their approach to a problem of finding weighted low-rank approximations and tested it with image completion tasks, supporting their theoretical results with numerical experiments."},{"id":"2609.30884v1","title":"CacheReforge: Bounded Recovery for Stale KV Caches under Evolving Adapters","abstract":"Large language models rely on KV caching to reduce repeated prefill computation in long context and interactive applications. As lightweight adapters evolve, cached states reflect earlier versions, so stale reuse distorts current model outputs, while complete affected suffix recomputation restores fidelity at substantial cost. We seek minimal recomputation that recovers current adapter behavior. Existing systems track token, context, or stable adapter identity, but neither represent caches from earlier adapter versions nor distinguish update propagation from the recomputation required for behavioral recovery. To address these gaps, we introduce CacheReforge, which represents stale KV caches as layerwise mixed-version objects. It combines per-layer adapter anchors, calibrated sensitivity, accumulated drift, and executable restart boundaries to select direct reuse, bounded recomputation, or complete affected-suffix recovery. We distinguish dependency depth from the functional recomputation horizon and use cumulative tail influence to characterize when bounded recovery preserves current-model behavior. We evaluate CacheReforge on Qwen2.5-1.5B and Qwen2.5-7B with continual LoRA updates, including 16K HotpotQA and 2WikiMQA workloads. CacheReforge reduces mean KL divergence by 92.4% relative to stale reuse, while recomputing only 5.44% of layers and reducing cache-maintenance time by 93.2% relative to fresh full prefill. These results show that version-aware recovery preserves model fidelity and most KV caching gains.","published":"2026-09-25T06:46:47Z","updated":"2026-09-25T06:46:47Z","authors":["Yuhang Cao","Yanzhou Mu","Chunrong Fang","Zhenyu Chen"],"categories":["cs.LG"],"headline":"Improving Large Language Model Cache Updates with Less Work","summary":"Large language models use memory caches to speed up repeated tasks, but when parts of the model change, old caches can cause errors. The researchers developed CacheReforge, a method that identifies which parts of the cache need updating and which can be reused safely. Tested on specific models and tasks, CacheReforge greatly reduces the work needed for updates while keeping outputs accurate. Their results are based on simulations with particular model setups and show promise for efficient cache management under changing conditions."},{"id":"2609.30883v1","title":"Warned alike, AI agents avoid the less-crowded road while people take it","abstract":"AI agents built on a few shared models increasingly act for many people. A shared forecast about others can align their choices and change how scarce capacity is allocated. We tested this feedback in a two-road congestion game. Adding one sentence warning that others might follow a routing tip made populations of 50 GPT agents crowd one road while avoiding the nearly empty alternative. Average travel time rose from 64 to 95 min, although any crowded-road agent could have saved 69 min by switching alone. The warning discouraged the very move it predicted. The pattern persisted for 100 rounds. Two other model families shifted the same way without locking onto one road. Twelve all-human groups (240 participants) stayed near balance under numerical reports or the tip and warning. In 24 mixed groups with a further 240 participants, imbalance grew with the share of agents in the registered analysis, while people increasingly took the road the agents avoided. Collective costs stayed below the allagent reference, but with 15 agents and 5 humans, agent seats averaged 80 min, compared with 44 min for human seats. Shared forecasts can thus sustain collective inefficiency among similar agents. A better group average can also hide an unequal burden. Evaluations of AI agents that share resources should test populations, treat messages as interventions and report who bears the costs.","published":"2026-09-25T06:44:10Z","updated":"2026-09-25T06:44:10Z","authors":["Takahiro Ezaki","Naoto Imura","Katsuhiro Nishinari"],"categories":["physics.soc-ph","cs.AI"],"headline":"AI agents overcrowd roads people tend to avoid, causing delays","summary":"Researchers studied how AI agents and people choose between two roads with different congestion levels. They found that when AI agents received a warning that others might follow a routing tip, they crowded the busier road, increasing average travel time. In contrast, human groups balanced their choices better, avoiding extreme crowding. Mixed groups showed AI agents stuck on crowded roads while people used less busy ones. This suggests shared AI warnings can unintentionally lead to inefficient group behavior and unequal burdens among users."},{"id":"2609.30882v1","title":"Effects of Transcript Compression on LLM-based Medical Misinformation Detection in Japanese YouTube Videos","abstract":"Large language models (LLMs) are increasingly used to assess long-form medical videos, but their effectiveness may depend on whether transcripts are provided in full or compressed through summarization, retrieval, or claim screening. This study examines how such transcript compression affects LLM-based veracity classification of Japanese medical YouTube videos. We compare four transcript input designs: full transcripts, LLM-generated summaries, RAPTOR-based retrievalaugmented generation (RAG), and Screening, which extracts candidate medical and health-related sentences. Using 74 long-form videos labeled as Real or Fake, we evaluate classification performance and analyze linguistic changes using J-LIWC, hedge expressions, and institutional or technical terms. The full-transcript Baseline achieved the best performance, whereas all compressed inputs increased false negatives, meaning that Fake videos were more likely to be misclassified as Real. Summary caused the largest performance drop, while Screening performed best among the compressed inputs but still omitted many medically relevant sentences. Linguistic analyses showed that these errors were not explained by a simple increase in certainty. Instead, Summary reduced affective, social, temporal, cognitive, and conversational cues, while Summary and RAG made institutional and technical terms more salient. These findings suggest that transcript compression can represent Fake videos as more coherent and authoritative inputs, thereby weakening cues needed for misinformation detection","published":"2026-09-25T06:42:44Z","updated":"2026-09-25T06:42:44Z","authors":["Yuya Wake","Sho Tsugawa","Toshiyuki Amagasa"],"categories":["cs.CL"],"headline":"How Transcript Shortening Affects Detecting False Medical Videos in Japanese","summary":"Researchers studied how shortening transcripts of Japanese medical YouTube videos influences the ability of large language models (LLMs) to distinguish real from fake content. They compared full transcripts with four types of compressed inputs and found that using full transcripts worked best, while all shortened versions increased mistakes, especially misclassifying fake videos as real. The study suggests that transcript compression can make false videos seem more credible by removing important linguistic cues needed for spotting misinformation."},{"id":"2609.30881v1","title":"Beating and coupling in pulsating stars: a unified Fourier description and observational diagnostics","abstract":"Long-term amplitude and phase variations in pulsating stars produce side peaks or multiplets in the Fourier spectrum of the light curves around the pulsation frequency and its harmonics. Such structures are often attributed to modulation, although linear beating and the non-linear coupling of close oscillations can also produce similar patterns. We develop a unified analytical Fourier description of these alternatives and identify observational relations that retain discriminatory power when their frequency patterns overlap. We represent the signal as a dominant, non-sinusoidal oscillation plus one or more secondary oscillations close to it, and analytically derive the frequencies, complex amplitudes, and phases resulting from linear superposition, quadratic and general non-linear coupling. We also illustrate the calculated phenomena using artificial light curves. We implement the resulting tests in a diagnostic tool \"freqdiag\", which compares modulation with primary--secondary quadratic coupling models anchored to either the higher- or lower-frequency side peak, using the same normalized complex amplitudes, and supplements the fit with hierarchical tests of side-peak amplitude progression. A sinusoidal secondary signal produces only one additional peak in the linear-beating case, whereas a non-sinusoidal secondary signal produces a harmonic sequence, whose separation from the primary harmonics increases with order. Quadratic coupling produces terms $nf_0+lf_B$, but their amplitudes and phases follow shifted-harmonic and combination relations. At general non-linear order, coupling and modulation occupy the same frequency grid. The formalism provides a practical framework for interpreting Blazhko-like RRc stars, HADS variables, and other pulsating stars in which modulation and coupling are difficult to separate observationally.","published":"2026-09-25T06:42:15Z","updated":"2026-09-25T06:42:15Z","authors":["József M. Benkő","Emese Plachy"],"categories":["astro-ph.SR","astro-ph.IM"],"headline":"New Method Helps Distinguish Patterns in Pulsating Star Signals","summary":"Researchers studied how brightness changes in pulsating stars create complex patterns in their frequency data. They developed a mathematical framework to describe three main causes: modulation, simple beating, and non-linear interaction of star oscillations. Their approach helps tell these effects apart even when patterns look similar, using a tool that analyzes and compares different models. This work aids interpreting data from certain variable stars where it’s tough to separate overlapping signal influences."},{"id":"2609.30880v1","title":"EXAONE Demand 1.0: A Time Series Foundation Model for Demand Forecasting","abstract":"Time series foundation models (TSFMs) are pretrained on series from diverse domains, where demand series make up only a small fraction. Demand data has properties that such corpora rarely contain: Short histories, frequent zeros, censoring by stock-outs, and exogenous events that the series does not record. To this end, we propose EXAONE Demand, built on 1) a demand-specific corpus and 2) a demand-aware adapter. For the corpus, we assemble 11.3M series and 48.4B observations from 73 sources, and a synthetic generator supplies the behaviour that open demand data under-represents. For the adapter, we attach low-rank branches to a frozen general-domain backbone, one for each of the four demand classes (smooth, intermittent, erratic, and lumpy), and a router that reads eight scale-free statistics of the input series decides how much each branch contributes. We build EXAONE Demand in two versions, one trained on real-world and synthetic demand together and one trained on the synthetic corpus alone. On 22 held-out datasets, both versions outperform 36 TSFMs, and real-world demand adds a gain over synthetic data alone.","published":"2026-09-25T06:37:39Z","updated":"2026-09-25T06:37:39Z","authors":["Seunghan Lee","Sangjun Han","Jun Seo","Junhyeok Kang","Jaehoon Lee","Tae Yoon Lim","Dongwan Kang","Hwanil Choi","Minjae Kim","Sungdong Yoo","Soonyoung Lee","Wonbin Ahn"],"categories":["cs.AI","cs.LG"],"headline":"New Model Improves Forecasting of Product Demand","summary":"The researchers identified that existing time series models often lack enough real demand data, which has unique challenges like short records and frequent zeros. They created EXAONE Demand, a model trained on a large dataset of real and simulated demand data and designed to handle different types of demand patterns. Their tests on 22 different datasets showed that EXAONE Demand outperformed 36 other models, and including real demand data improved results compared to using only synthetic data."},{"id":"2609.30879v1","title":"Joint-Sparse Transfer Learning for High-Dimensional Multi-Output Regression","abstract":"Multitask linear models can improve estimation and prediction by exploiting structure shared across responses, bridging taskwise fitting and complete pooling. In many applications, however, the objective is estimation in a data-limited target domain, while data-rich but heterogeneous source domains are available. Borrowing from these sources can improve efficiency but introduce bias. We develop a joint-sparse transfer-learning framework for high-dimensional multi-output regression that combines shared predictor structure across responses with source-target similarity. The framework yields two complementary estimators: a fused estimator that aggregates jointly fitted domain-specific coefficients and a target-based debiased estimator that adjusts for source-induced shifts. Our error bounds show how transfer increases the available information and sharing predictors across responses reduces selection costs. They also reveal a tradeoff: the fused estimator benefits from larger sources but may retain bias if source shifts point in similar directions, whereas debiasing trades this bias for additional estimation error governed by the smaller target sample. Comparison with a minimax lower bound identifies regimes where the bounds match up to logarithmic factors, where matching remains unresolved, and where projection onto a target-based convex set closes the gap. Simulations and an analysis of single-cell RNA and surface-protein profiles across cell types support the theory.","published":"2026-09-25T06:37:30Z","updated":"2026-09-25T06:37:30Z","authors":["Sunwoo Lim","Mladen Kolar"],"categories":["stat.ME"],"headline":"New method improves multi-output predictions using related data sources","summary":"Researchers developed a technique to better estimate relationships when predicting multiple outcomes, especially when the main data is limited but related datasets exist. Their method combines information shared across outcomes with careful adjustments to avoid bias from differences between datasets. They provide theoretical error limits and show through simulations and biological data that their approach can improve prediction accuracy, while acknowledging some tradeoffs depending on dataset sizes and differences."},{"id":"2609.30878v1","title":"TISD: On-Policy Self-Distillation with Trajectory Intervention","abstract":"On-policy self-distillation (OPSD) provides dense teacher targets, but evaluates them only along student-sampled rollouts. When the privileged teacher favors an alternative action at a visited prefix, OPSD can provide a target for the branch decision but cannot supervise the successor contexts induced by that action unless the student samples it. This creates a training-time data-collection bottleneck and suggests a different role for teacher-student disagreement: proposing a trajectory branch rather than identifying a sufficient local repair. Our diagnostic framework using controlled token interventions reveals that a teacher-preferred token at peak disagreement can improve student continuation success, while its local corrective value is limited. Motivated by this finding, we introduce a simple branch-regenerate-distill algorithm, Trajectory-Intervention Self-Distillation (TISD). TISD forces a teacher-selected branch action, returns suffix generation to the student, and distills the full trajectory under the privileged-context-conditioned teacher. Across the coding models, TISD improves average Avg@4 over SDPO by 1.2 percentage points. Across the science domains, it improves average Avg@128 by 0.8 points under an equal-step budget and by 0.3 points under an equal-time budget. These results support teacher-guided branching as a way to expose useful successor contexts for self-distillation.","published":"2026-09-25T06:37:02Z","updated":"2026-09-25T06:37:02Z","authors":["Taeckyung Lee","Rinat Amankos","Jeonghye Kim","Hyungjun Yoon","Woogyeol Jin","Sung-Ju Lee"],"categories":["cs.AI","cs.LG"],"headline":"New method improves AI learning by exploring alternate action paths","summary":"Researchers studied how AI models learn from a 'teacher' model when making decisions step-by-step. They found that traditional methods only train the AI on decisions it actually makes, missing valuable lessons from alternative choices. To address this, they developed a new approach called Trajectory-Intervention Self-Distillation (TISD), which forces the AI to follow teacher-suggested alternative steps and then learn from the resulting outcomes. Testing showed TISD slightly improved model performance in coding and science tasks, suggesting exploring different decision paths can aid AI training. These findings are based on algorithm simulations and controlled experiments."},{"id":"2609.30877v1","title":"Tight Stochastic Condition-Number Dependence in Nonconvex-Strongly-Concave Minimax Optimization","abstract":"We study whether the linear condition-number dependence in the stochastic complexity of SAPD+ is necessary for nonconvex-strongly-concave minimax optimization. For jointly $L$-smooth objectives with dual strong-concavity parameter $μ$, we prove a lower bound that matches the SAPD+ upper bound under the same Moreau-envelope stationarity criterion and the same primal-dual initialization gap. Specifically, when $σ\\ge\\varepsilon$, the worst-case complexity of zero-respecting algorithms is $Θ(κLGσ^2\\varepsilon^{-4})$ in the stated accuracy regime, where $κ=L/μ$, $G$ bounds the initial primal-dual gap, and $σ^2$ bounds the variance of a general unbiased first-order oracle. The lower bound is realized on a smooth problem class with a bounded dual box. Our construction routes each link of a nonconvex zero-chain through a dual gradient of magnitude proportional to $\\varepsilon/\\sqrtκ$, while an undiscovered primal coordinate prevents stationarity. It also yields the primal-gradient lower bound $Ω(LΔ(\\sqrtκ\\varepsilon^{-2}+κσ^2\\varepsilon^{-4}))$ after combination with the known deterministic bound, where $Δ$ bounds the initial primal function gap.","published":"2026-09-25T06:34:04Z","updated":"2026-09-25T06:34:04Z","authors":["Qihao Zhou"],"categories":["math.OC","cs.LG"],"headline":"Researchers establish tight limits on optimization complexity","summary":"This research investigates how difficult it is to solve a certain type of mathematical optimization problem involving both nonconvex and strongly concave functions. The researchers prove that the known complexity of an existing algorithm called SAPD+ cannot be improved in terms of a key parameter called the condition number. They use theoretical analysis to establish a matching lower bound on the effort needed by any similar algorithm under specific accuracy measures. Their results apply to smooth problems and reveal fundamental limits in this optimization context."},{"id":"2609.30876v1","title":"A vanishing-viscosity minimizing movement Fourier spectral method for scalar conservation laws","abstract":"We propose a vanishing-viscosity minimizing movement (VVMM) Fourier spectral method for scalar hyperbolic conservation laws under periodic boundary conditions. The method is a fully discrete space--time residual-minimizing scheme, Fourier spectral in space and polynomial in time, stabilized by vanishing viscosity and heat-kernel postprocessing. A basic analytical difficulty in Fourier discretizations of nonlinear conservation laws is the interaction between Fourier projection and the nonlinear flux; the resulting projection residual contains unresolved high-frequency components. VVMM addresses this issue by selecting, on each time slab, the numerical solution that minimizes {a penalized cut-off viscous conservation laws residual} and then applies heat-kernel postprocessing to damp unresolved modes. We prove a fully discrete $L^1$ error estimate, in any fixed space dimension, that separates the vanishing-viscosity error, the temporal interpolation error, and the spatial Fourier truncation error. {Under exact minimization of the penalized cut-off residual and suitable parameter coupling,} the estimate yields, for each fixed $γ>0$, an $L^1$ convergence rate $N^{-1/2+γ}$, with a constant depending on $γ$. Numerical experiments in one and two dimensions illustrate stable shock capturing without visible Gibbs oscillations and convergence behavior consistent with the theory.","published":"2026-09-25T06:33:27Z","updated":"2026-09-25T06:33:27Z","authors":["Lun Ji","Buyang Li","Fangyan Yao"],"categories":["math.NA"],"headline":"New method improves math simulations of waves and shocks","summary":"The researchers developed a new computational method to better simulate solutions of certain wave equations, called scalar conservation laws, especially when sharp changes or shocks occur. Their method uses a special mathematical approach involving Fourier techniques and adds a small smoothing effect that disappears over time to improve accuracy and stability. They proved mathematically that their method converges under certain conditions and demonstrated through computer tests that it can capture shocks smoothly without unwanted oscillations."},{"id":"2609.30875v1","title":"Quantum-circuit simulation of three-flavor neutrino oscillations: vacuum, matter, and CP diagnostics","abstract":"We simulate three-flavor neutrino oscillations on two-qubit quantum circuits. The flavor space is embedded as $\\ket{00}=ν_e$, $\\ket{01}=ν_μ$, $\\ket{10}=ν_τ$ with the $\\ket{11}$ channel kept inert, and the input mixing parameters (NuFIT~5.3, normal ordering) follow the global fit. We verify three physical settings against analytic or exact references: vacuum propagation, reproduced by a mass-eigenstate phase circuit to $\\sim 10^{-16}$; constant-density Mikheev--Smirnov--Wolfenstein (MSW) matter effects, showing resonant flavor conversion when the matter potential $A$ crosses $Δm^2_{31}$; and varying-density (supernova shock-shell) propagation by slicing plus second-order Suzuki--Trotter splitting, agreeing with exact evolution to $\\sim 10^{-6}$. We report a Suzuki--Trotter precision--resource calibration varying order, step number, and matrix/probability error. It delimits brute-force Trotterization on solar baselines: we quantify both standard alternatives, coherence averaging and adiabatic MSW. We also compute quantum-information diagnostics from the same model parameters, namely mode entanglement and the CP asymmetry. Finally, we extend the framework to open quantum systems via a Lindblad master equation with dephasing and absorptive channels. These results provide a reproducible link between low-energy oscillation observables and quantum-information measures.","published":"2026-09-25T06:32:42Z","updated":"2026-09-25T06:32:42Z","authors":["Daming Li"],"categories":["quant-ph"],"headline":"Simulating neutrino oscillations using quantum circuits","summary":"Researchers simulated how three types of neutrinos change flavor using two-qubit quantum circuits. They tested their simulations under different physical conditions, such as in a vacuum, constant matter, and varying matter densities, and found their results closely matched exact calculations. The study also explored quantum features like entanglement and incorporated effects like particle absorption, offering a link between neutrino behavior and quantum information. These findings are theoretical and based on simulated models rather than experimental data."},{"id":"2609.30874v1","title":"From Received Power to Certified Secret Keys: A General Method for Bridging Classical FSO Link Budgets and Decoy-State QKD","abstract":"Received optical power determines photon arrival flux, but photon arrival flux does not certify a quantum key. This paper gives a deployment-independent interface from a classical free-space optical (FSO) link budget to phase-randomized weak coherent pulse BB84, decoy-state estimation, and composable finite-key postprocessing. The channel interface separates geometric collection, atmospheric extinction, residual pointing, and detector efficiency, and states when common geometric and pointing expressions are approximations. The finite-key output follows the decoy-state analysis of Lim et al., with key- and test-basis counts estimated separately and an infinite-decoy asymptotic reference. The numerical study covers a five-level attenuation sweep, basis- and vacuum-probability sensitivities, a conditional Chernoff comparison, and integer-count Monte Carlo. For $3.33 \\times 10^8$ emitted pulses, the Hoeffding-based predicted one-kilobit-per-second distance limits are 2.727, 1.883, 1.472, 1.236, and 0.962 km from very clear to dense-fog parameterizations. These are conditional performance predictions, not experimental security certificates. Satellite, drone, high-altitude-platform, terrestrial, and train links enter through the same interface but require their own propagation, acquisition, and device inputs. The framework integrates established security tools into an engineering procedure without claiming a new security proof or a universal range.","published":"2026-09-25T06:32:18Z","updated":"2026-09-25T06:32:18Z","authors":["Hasan Abbas Al-Mohammed"],"categories":["quant-ph"],"headline":"Bridging Classical Optical Links and Quantum Key Security Methods","summary":"The researchers address how to connect traditional measurements of received light power in free-space optical (FSO) communication with the process needed to securely create quantum encryption keys. They developed a general method that links classical factors like atmospheric effects and detector efficiency with quantum key generation techniques, using existing mathematical analyses. Their results include simulations predicting secure communication distances under various conditions, but these are theoretical predictions and not experimental validations or new security guarantees."},{"id":"2609.30873v1","title":"A Second Torque Port for Series Elastic Actuators: Parallel-Integrated Design and Time-Scale Torque Allocation","abstract":"A series elastic actuator has a single torque port and pays for it twice: the geared motor must swing its own reflected inertia through the spring, so the amplitude it delivers collapses as $ω^{-2}$ in the command frequency $ω$ once it saturates, while commands below the transmission's breakaway friction never arrive at all. This letter opens a second torque port on the load side, placing a frameless direct-drive micro motor in parallel with a fixed-stiffness spring -- a parallel-integrated SEA, or Pi-SEA, whose delivered torque is read from spring deflection and micro current without a sensor -- and dividing the commanded torque between the two channels by time scale rather than by filter design. The micro torque loop is the fast subsystem, which makes the closed loop singularly perturbed and turns the separation the channels need into a bound to check rather than a crossover to tune; a leaky mid-ranging integrator returns the steady load to the spring; and the amplitude ceiling, read backwards, becomes a closed-form sizing rule that matches spring, geared motor and micro motor to the amplitudes and frequencies an application asks for. Against SEAs, the Pi-SEA widens the tracked band at small amplitudes and lowers the residual the joint imposes on its environment, each by an order of magnitude.","published":"2026-09-25T06:30:31Z","updated":"2026-09-25T06:30:31Z","authors":["Donghao Jia","Xiubo Xia","Junhang Liu","Zeyu Zhu","Xiaoyu Geng","Jian Sun"],"categories":["cs.RO","eess.SY"],"headline":"New Double-Torque Actuator Design Enhances Control and Responsiveness","summary":"The researchers studied a type of actuator that normally has one torque output, which limits performance at certain frequencies and forces. They introduced a second torque output using a small motor alongside a spring, creating a new parallel-integrated actuator called Pi-SEA. This design divides torque commands by timing rather than filters, improving control speed and stability. Their analysis, based on modeling and theory, suggests that Pi-SEA better tracks small movements and reduces force leftover in the system compared to traditional designs, but practical testing details are not provided."},{"id":"2609.30872v1","title":"The homology of $\\mathrm{Out}(F_8)$ and $\\mathrm{Aut}(F_7)$","abstract":"We compute the rational homology of $\\mathrm{Out}(F_8)$ and $\\mathrm{Aut}(F_7)$, and several related groups. We also show that the Eisenstein classes in $H_7(\\mathrm{Aut}(F_5))$ and $H_{11}(\\mathrm{Aut}(F_7))$ are zero.","published":"2026-09-25T06:27:52Z","updated":"2026-09-25T06:27:52Z","authors":["Maroš Grego","Tomáš Vítek","Thomas Willwacher"],"categories":["math.AT","math.QA"],"headline":"Researchers study algebraic structures related to symmetry groups","summary":"The researchers calculated certain mathematical properties called rational homology for two complex groups known as Out(F8) and Aut(F7), which relate to symmetries in abstract algebra. They also examined similar groups and demonstrated that specific elements called Eisenstein classes vanish in certain cases. The work improves understanding of these advanced algebraic objects, but the abstract provides limited information about broader implications or applications."},{"id":"2609.30871v1","title":"Systematic Evolution of Magnetic Anisotropy and Crystal-electric-field Ground States in SmTr2Ge2","abstract":"We investigated the crystal-electric-field (CEF) states in SmTr2Ge2 (Tr= Co-Cu and Ag) with the ThCr2Si2-type structure by magnetization and specific heat measurements. The magnetic anisotropy evolves systematically with Tr, changing from an easy ab plane for Tr = Co and Ni to an easy c axis for Tr = Cu and Ag. Combined analyses of the low-temperature susceptibility and magnetic entropy indicate a level crossing of the CEF ground state between Ni and Cu, together with a systematic evolution of the first excited CEF energy. The evolution of the magnetic properties correlates with changes in local structural parameters, providing an experimental basis for discussing how the ligand environment around the Sm ion is related to the 4f electronic state.","published":"2026-09-25T06:27:06Z","updated":"2026-09-25T06:27:06Z","authors":["Ryuji Higashinaka","Kokoro Higo","Tatsuma D. Matsuda","Yuji Aoki"],"categories":["cond-mat.str-el"],"headline":"Researchers Study Magnetic Changes in SmTr2Ge2 Materials","summary":"The researchers studied how the magnetic properties of SmTr2Ge2 materials change when different elements (Co, Ni, Cu, Ag) are used. They measured magnetization and heat capacity to find that the preferred magnetic direction shifts depending on the element. Their analysis suggests a fundamental change in the magnetic energy state occurs between Ni and Cu. They also linked these magnetic changes to small shifts in the material's structure, helping understand the relationship between atomic surroundings and magnetic behavior."},{"id":"2609.30870v1","title":"Counterexamples to the inhomogeneous Duffin-Schaeffer conjecture for a residual set of shifts","abstract":"Let $θ\\in\\mathbb Q\\setminus\\{0\\}$. We construct an approximating function $ψ:\\mathbb N\\to[0,\\frac12)$ for which \\[ \\sum_{q=1}^{\\infty}\\frac{\\varphi(q)}{q}ψ(q)=\\infty, \\] but the set of $x\\in[0,1]$ for which \\[ |qx-a-θ|<ψ(q),\\qquad \\gcd(a,q)=1, \\] holds for infinitely many $(a,q)\\in\\mathbb Z\\times\\mathbb N$ has Lebesgue measure zero. Thus the inhomogeneous analogue of the Duffin--Schaeffer conjecture fails for every nonzero rational shift. By a modification of the construction, we further show that the set of shifts for which the inhomogeneous Duffin--Schaeffer conjecture fails is residual in $\\mathbb R$.","published":"2026-09-25T06:23:10Z","updated":"2026-09-25T06:23:10Z","authors":["Yubin He","Lingmin Liao"],"categories":["math.NT"],"headline":"Researchers Find Exceptions to a Mathematical Approximation Conjecture","summary":"The researchers studied a mathematical problem about approximating numbers with fractions shifted by a rational amount. They constructed a specific example showing that a well-known prediction (the inhomogeneous Duffin-Schaeffer conjecture) does not always hold when the shift is a nonzero rational number. Their results come from theoretical constructions and indicate that for many shifts, the expected approximation property fails, contrasting with previous beliefs about this conjecture."},{"id":"2609.30869v1","title":"Observing Phase Space Evolution During Unimolecular Decay","abstract":"Direct time-resolved observation of large amplitude dynamics during ground state statistical unimolecular decay remains an experimental challenge. Here, an experimental Time-Resolved Coulomb Explosion Imaging (TR-CEI) study (J. Chem. Phys. 151, 174301, 2019) of the canonical unimolecular reaction NO$_2$ $\\rightarrow$ NO($^2Π, v=0, J)$ + O($^3P_j$) at threshold dissociation energies is modeled and analyzed. Quasi-classical trajectory (QCT) simulations on an accurate and validated ground state potential energy surface (PES), with energy and angular momentum constraints from known product state distributions, are used to characterize the dissociation dynamics. TR-CEI projects suddenly, via ultrafast Strong Field Ionization (SFI), the evolving neutral ground state wavepacket onto the repulsive doubly charged cationic state which then impulsively fragments into the NO$^{+}$ + O$^{+}$ ion pair. Within the Coulomb recoil approximation, which applies at longer NO--O (Jacobi) separations, the ion pair CEI-energy is inversely proportional to the scalar distance between the neutral ground state NO and O moieties at the moment of SFI, thus providing a direct measure of the evolving average NO--O distance during unimolecular decay. Results from QCT simulations agree favourably with the TR-CEI measurements. Phase space evolution, recurrence time distributions and Lyapunov exponent analysis are presented which support the conclusion that TR-CEI directly observes phase space evolution during unimolecular decay. For the present system, the time scale for the phase space distribution to achieve statisticality is estimated to be 2 to 3 ps.","published":"2026-09-25T06:20:17Z","updated":"2026-09-25T06:20:17Z","authors":["Abhirami Vijayakumar","Raidel Martin Barrios","Meenu Upadhyay","JingChun Wang","Andre Staudte","Albert Stolow","Markus Meuwly"],"categories":["physics.chem-ph"],"headline":"Tracking Atomic Movements During a Simple Molecular Breakup","summary":"The researchers studied how a nitrogen dioxide molecule breaks apart into nitrogen oxide and oxygen atoms at the most basic energy level. They used a special imaging technique to take snapshots of the molecule's changing shape over very short times, combined with computer simulations based on accurate energy models. Their findings suggest that this approach can directly observe how the molecule's atomic arrangement evolves during breakup and estimate that the system reaches a random distribution in about 2 to 3 trillionths of a second."},{"id":"2609.30868v1","title":"VLaRL: Augmenting Vision-Language-Action Models with Simulation-Trained Latent-Conditioned Residual RL","abstract":"Vision-language-action (VLA) models provide broad, instruction-conditioned manipulation behaviors, but their physical execution can remain imprecise during contact-rich interaction. Residual reinforcement learning (RL) can correct such errors while keeping the VLA frozen, but real-robot RL is costly and safety-critical. We propose VLA Latent-Conditioned RL (VLaRL), which enables residual RL for frozen VLAs to be trained in simulation and deployed on real robots without real-world RL or online adaptation. The key challenge is transferring the learned residual policy despite the visual gap between simulation and reality. Rather than requiring pixel-level visual correspondence, VLaRL uses the VLA's internal vision-language latent representation to condition residual control and as the sim-to-real transfer interface, and learns a lightweight mapper that transforms simulation-derived latents toward the real latent distribution. Across four contact-rich manipulation tasks and two VLA backbones, VLaRL improves real-world success in all task-backbone combinations, while controlled ablations demonstrate the importance of both latent conditioning and latent alignment for transferring simulation-trained residual control.","published":"2026-09-25T06:18:40Z","updated":"2026-09-25T06:18:40Z","authors":["Namiko Saito","Kinam Kim","Heecheol Kim","Katsushi Ikeuchi","Yasuyuki Matsushita"],"categories":["cs.RO"],"headline":"New Method Trains Robot Vision-Action Models Safely in Simulation","summary":"Researchers developed VLaRL, a technique that improves robot actions by combining pre-trained vision-language-action models with reinforcement learning trained entirely in simulation. This helps correct imprecise movements in complex tasks without costly or risky real-world training. Their method uses internal model representations to bridge differences between simulation and reality, improving success across multiple tasks. The study shows the importance of aligning internal model data for effective transfer, but results are based on simulation-trained models tested on real robots, not full real-world training."},{"id":"2609.30867v1","title":"Evidence-Grounded Auditing of Identification Assumptions in Climate-Policy Causal Evaluations","abstract":"Difference-in-differences (DID) studies are widely used to evaluate climate policy, but assessing the evidence supporting their identification assumptions remains challenging. We introduce ARGUS, a structured language-model pipeline that audits reported evidence against an eleven-dimension assumption-implication-evidence rubric and abstains when relevant evidence cannot be retrieved. We evaluate ARGUS using injected flaws, economics papers, and a small pilot with reconciled labels. On the 11-flaw benchmark, ARGUS detects 73% of planted flaws, compared with 18% for a keyword-based pipeline. Across 26 economics papers, ARGUS abstains on about 40% of paper-dimension assessments for lack of retrievable evidence. In a five-paper pilot with labels reconciled by two annotators, it assigns a higher risk level than the labels on 25 of the 33 assessments it completes. A rule fixed before the labels arrived removes most of this in-sample; weighted agreement stays low. ARGUS provides evidence-linked risk reports that localize potential weaknesses for expert review, without adjudicating causal claims. Code and data: https://github.com/yonghongzhang-io/ARGUS","published":"2026-09-25T06:18:01Z","updated":"2026-09-25T06:18:01Z","authors":["Yonghong Zhang","Yong Xie","Isabel M. Parra","Ricardo Correia"],"categories":["cs.CL"],"headline":"New Tool Helps Check Climate Policy Study Assumptions","summary":"Researchers developed ARGUS, a computer system that reviews how well climate policy studies support their key assumptions. ARGUS compares reported evidence to a detailed checklist and flags potential problems, detecting many hidden flaws better than simple keyword methods. However, it sometimes cannot find enough evidence to make a judgment and does not decide if a study's conclusions are correct. The team tested ARGUS on economics papers and a small expert-reviewed set, showing it can highlight areas needing expert attention. Results are based on testing and pilot studies."},{"id":"2609.30866v1","title":"Universal Drift Correction for Multidimensional Scanning Microscopy","abstract":"In scanning microscopy, drift causes the specimen to be sampled at positions displaced from the nominal probe positions. This displacement alters the spatial assignment of the recorded signals and biases quantitative measurements across two-dimensional imaging, channel-resolved spectroscopic mapping, and scan-position-resolved diffraction analysis. Here, we extend orthogonal-scan drift correction from 2D images to spectrum images and diffraction datasets. We demonstrate how to recover probe positions using either differently oriented multidimensional scans or structural reference images. The recovered positions are used either to resample the multidimensional data onto a regular grid or to assign each recorded signal to its corrected coordinate. Our method combines affine and non-rigid correction, requires no prior structural model, and is implemented as open-source, GPU-accelerated software that reduces processing times by two to three orders of magnitude, enabling routine and automated drift correction for quantitative multidimensional microscopy.","published":"2026-09-25T06:17:35Z","updated":"2026-09-25T06:17:35Z","authors":["Sangjoon Lee","William Millsaps","Dasol Yoon","Caitlyn Obrero","Guoliang Hu","Corrie Barnes","Cedric Lim","Andrew Barnum","Arthur R. C. McCray","Colin Ophus"],"categories":["cond-mat.mtrl-sci","cs.CV"],"headline":"New Method Corrects Drift in Multidimensional Microscopy Data","summary":"Scientists addressed the problem of drift, where microscope samples shift from expected positions, causing errors in imaging and measurements. They developed a method that recovers true probe positions from different types of scans without needing prior structural information. Their approach corrects data by adjusting coordinates, speeding up processing through open-source software. These findings support improved accuracy and efficiency in analyzing complex microscopy datasets, though the results focus on computational correction techniques rather than experimental validation."},{"id":"2609.30865v1","title":"Reliability-Regulated Trajectory Optimization for Progressive COLMAP-Free 3D Gaussian Splatting","abstract":"COLMAP-free 3D Gaussian Splatting (3DGS) bypasses computationally expensive structure-from-motion (SfM) pipelines, yet progressive camera pose tracking remains fundamentally vulnerable to error compounding---early pairwise tracking inaccuracies both corrupt subsequent frame initializations and remain permanently frozen in the scene representation. Rather than relying on heavyweight external neural priors or treating progressive tracking through isolated heuristic fixes, we propose a unified reliability-regulated trajectory optimization framework for progressive COLMAP-free 3DGS. At its core, our framework establishes an intrinsic, self-supervised bidirectional cycle-consistency mechanism that systematically regulates progressive camera trajectory estimation across two complementary temporal horizons: (1) Forward Motion Propagation, where the online reliability signal adaptively gates first-order kinematic warm-starts of rigid motion into upcoming pairwise registrations, supplying informed directional search priors while safely intercepting untrusted transitions; and (2) Retrospective Trajectory Correction, where the same reliability signal dynamically weights relative-pose consistency constraints within a sliding window of neighboring camera poses. By governing both prospective state initialization and retrospective trajectory consolidation through a unified reliability regulator, our self-contained framework resolves progressive drift without external priors or offline preprocessing. Extensive evaluations on Tanks and Temples and CO3D-V2 benchmarks show that our method substantially improves camera trajectory accuracy and novel-view rendering quality, outperforming existing unposed baselines. Code is available at https://github.com/Zijian1026/RRTO-CF3DGS.","published":"2026-09-25T06:16:42Z","updated":"2026-09-25T06:16:42Z","authors":["Zijian Wu","Jinliang Wang","Zidian Lin","Ying Song","Ziqian Lu","Hanjie Ma","Zhen Ye","Mingfeng Jiang"],"categories":["cs.CV"],"headline":"Improving 3D Scene Tracking Without Complex Preprocessing","summary":"Researchers addressed the problem of errors building up when tracking camera positions over time in 3D scene reconstruction without using traditional, costly methods. They developed a new system that carefully controls and corrects camera movement estimates using internal checks that work forward and backward in time, improving accuracy without needing external data or prior training. Their tests on standard benchmarks showed better camera tracking and 3D scene quality compared to previous methods that don't use known camera positions."},{"id":"2609.30864v1","title":"Persistent Negatives for Adversarial Black-Box On-Policy Distillation","abstract":"Black-box On-Policy Distillation (OPD) seeks to improve a student from its own generations when the teacher provides sampled responses but not token probabilities. Adversarial distillation offers one route: it learns a discriminator over prompt-matched teacher and student responses and uses its score as the policy reward. However, sampling discriminator negatives from the latest student at each step couples the learned reward to a negative distribution that changes after every policy update. We address this moving-target problem with persistent-negative adversarial distillation, a live-pool method that replaces a fraction of each discriminator batch with historical, prompt-matched teacher--student comparisons. Under matched discriminator compute, historical comparisons train the discriminator, while GRPO remains on-policy with fresh student responses. Our analysis identifies the Bayes-optimal reward as a teacher-to-negative log-density ratio and, under explicit assumptions, shows how persistent negatives anchor the discriminator and reduce reward-estimation MSE relative to fresh-negative training. Across two student families, three judges, and four judged-chat benchmarks, persistent-negative adversarial distillation consistently improves performance over current methods at matched discriminator compute. It also yields smoother fresh-policy discriminator trajectories, with fewer below-chance dips. These findings identify the discriminator's negative distribution as an important design axis in black-box on-policy distillation.","published":"2026-09-25T06:14:09Z","updated":"2026-09-25T06:14:09Z","authors":["Haixu Ma","Saad Lahrichi","Weiwei Li","Kevin Han","Weiqiang Wu","Peggy Yang","Dongzhuo Li","Ruiyi Li","Serena Li","Gedi Zhou","Mingze Gao","Abhishek Kumar","Xiangjun Fan","Lizhu Zhang"],"categories":["cs.CL","cs.AI"],"headline":"Improving AI learning using stable comparison examples","summary":"Researchers studied a technique called black-box on-policy distillation, where an AI student improves by comparing its answers to a teacher's sampled responses without detailed probabilities. They found that always comparing against the latest student responses makes the learning unstable. To fix this, they introduced a method using a mix of recent and past comparisons to train the system more reliably. Their experiments show this approach improves AI performance across several tests and creates smoother learning progress."},{"id":"2609.30863v1","title":"Developing a Roadmap to an AI-first Organization: A Case Study in Embedded Software Development","abstract":"The emergence of AI agents is expected to reshape software engineering by moving beyond AI as assistants towards systems capable of planning, executing, and evaluating development tasks with increasing autonomy. This transition is particularly significant for embedded software organizations, where strict requirements for quality, traceability, verification, and long-term maintainability often apply. This paper presents a case study of a large embedded systems company and its transition toward becoming an AI-first organization. Through a mixed method, we analyzed data collected from a semi-structured workshop with 40 participants, including scrum masters, architects, management, and product owners. The findings show that the participants expect agentic AI to affect team structure, required competencies, organizational strategies, and developers' roles within the organization. Based on these findings, the paper discusses implications for federated AI team formation, human-in-the-loop practices in such an organization, and the sustainable adoption of AI agents in embedded software engineering. We also present a concrete roadmap for the organization towards becoming an AI-first organization.","published":"2026-09-25T06:12:36Z","updated":"2026-09-25T06:12:36Z","authors":["Viktor Kjellberg","Srijita Basu","Simin Sun","Farnaz Fotrousi","Miroslaw Staron"],"categories":["cs.SE","cs.AI"],"headline":"How AI Could Change Embedded Software Development Teams","summary":"Researchers studied a large embedded software company shifting to prioritize AI agents that can plan and carry out development tasks more independently. They gathered insights from 40 team members about how AI might impact roles, skills, and team organization. The study presents a detailed plan for adopting AI in such organizations, highlighting challenges like maintaining quality and human oversight. These findings come from this specific case study and outline possible changes but do not test AI tools themselves in practice."},{"id":"2609.30862v1","title":"Evaluation of portability and performance of an OpenMP5 offloaded Quantum-Inspired Evolutionary Optimization Across the GPU Ecosystem","abstract":"Quantum-inspired evolutionary optimization (QIEO) is a new class of population-based metaheuristic optimization algorithms which represents design variables as a set of qubits and searches a continuous, multi-dimensional landscape through rotation of the qubit's amplitude pair. Every generation rotates those amplitudes toward a single elite, which corresponds to that generation's best. The per-generation cost scales as $O(N_p N_g)$ for $N_p$ chromosomes and $N_g$ genes (decision variables). Production use of such solvers is rarely confined to a single machine class. Prototypes are run on laboratory servers- before moving to rented cloud workstations for more involved campaigns. The largest problems are reserved for leadership-class accelerators. This paper asks whether a \\emph{single} OpenMP~5 source of QIEO, offloaded with \\texttt{\\#pragma omp target}, is a viable production path in each of those settings. We report three independent, campaigns of the 0/1 knapsack problem against a same-source multi-core Intel CPU baseline. The study comprises approximately 3,000 runs spanning varying chromosome and gene counts, evaluated using both chromosome-level and gene-level offload strategies on the NVIDIA Tesla V100 SXM2, NVIDIA A100 80GB, and AMD Instinct MI300X GPUs. Deployment-specific nuances such as Volta's constant-memory cliffs, Ampere's L2 persistence and \\texttt{cp.async}, CDNA~3's Infinity Cache and XCD occupancy are addressed to ensure high performance of these platforms. Results reveal gene-parallel offload achieved geometric-mean speedups of 90$\\times$, 136$\\times$, and 155$\\times$ over a single CPU core on the V100, A100, and MI300X, respectively, and 12$\\times$, 17$\\times$, and 16.6$\\times$ over 72 host threads. Furthermore DetermineElite, the $O(N_p)$ selection of the generation-best chromosome, is found to be better suited to the host than to the device.","published":"2026-09-25T06:12:10Z","updated":"2026-09-25T06:12:10Z","authors":["Kasturi Venkata Srikanth","Ashish Singh","Ferdin Sagai Don Bosco","Aman Mittal","Abhishek Singh","Aditya Singh","Abhishek Chopra"],"categories":["cs.PF","math.OC"],"headline":"Testing a Quantum-Inspired Algorithm Across Different GPUs","summary":"Researchers studied how well a new optimization method, inspired by quantum computing, runs on different graphics processing units (GPUs) using a single programming approach called OpenMP 5. They tested the method on various GPUs and compared speeds to a standard multi-core CPU, finding large speed improvements on GPUs. They also found that certain parts of the algorithm work better on the regular CPU instead of the GPU. The study helps understand how this approach performs across different hardware but is limited to specific simulated problem tests."},{"id":"2609.30861v1","title":"SkillEvoReg: Regularizing Agent Skill Evolution Against Overfitting","abstract":"Language-model agents increasingly improve by converting execution experience into reusable external skills. Yet repeated skill updates form a learning process of their own: locally useful edits can accumulate into redundant or task-specific instructions, while new updates can disrupt behavior that previously worked. We study this problem as skill-evolution overfitting and introduce SkillEvoReg, a general regularization framework for skill evolution inspired by anti-overfitting techniques in neural-network training. SkillEvoReg combines training-time skill dropout, which perturbs update generation, and complexity-aware local regularization, which controls unnecessary structural growth, with causal counterexample validation (CCV), which provides targeted behavioral validation of candidate-specific regressions. We instantiate the framework across heterogeneous skill-evolution systems while retaining each system's native skill evolver and task evaluator. Across SkillOpt, SkillEvolBench, and ContinualSkillBench, SkillEvoReg consistently controls skill-state growth while preserving competitive downstream capability, improves several transfer and later-stage evolution outcomes, and identifies update-level regressions that structural metrics alone cannot reveal. These results suggest that explicit regularization is a useful complement to increasingly capable skill updaters.","published":"2026-09-25T06:11:26Z","updated":"2026-09-25T06:11:26Z","authors":["Guanyu Nie","Fangzhou Zhu","Shixiong Kai","Xiongwei Han","Tao Zhong","Mingxuan Yuan"],"categories":["cs.AI"],"headline":"New Method Helps AI Skills Improve Without Losing Past Successes","summary":"The researchers address a problem where AI agents that learn new skills risk overfitting—getting too specialized or forgetting what worked before. They introduce SkillEvoReg, a technique that helps manage skill updates by adding checks and controls during learning to avoid redundant or harmful changes. Testing SkillEvoReg on different AI skill systems showed it effectively limits unnecessary skill growth and maintains overall performance, highlighting the value of regularization when evolving agent skills. The findings rely on simulations and evaluations within existing AI skill frameworks."},{"id":"2609.30860v1","title":"Classification of Generalised Triorthogonal Codes through Length 54","abstract":"Magic state distillation is a widely considered primitive in fault-tolerant quantum computation for the preparation of high-fidelity non-Clifford resources. The most commonly considered class of such protocols are generalised triorthogonal codes; these distil $n$ noisy input $\\mathrm{T}$ states into purified third-level diagonal magic states. Extensive prior work has searched this space of protocols, often using heuristic methods that do not guarantee optimality. Existing classification work is limited to protocols distilling $k$ output $\\mathrm{T}$ states with $n+k\\leq 38$ (Nezami and Haah, Phys. Rev. A 106, 012437 (2022)). In this work, we significantly expand this classification to all protocols with lengths $n\\leq 54$. We restrict to distance $d\\geq 3$ to keep the classification to a sensible size, and because efficient searches are well understood at distance $2$ (Singh et al., arXiv:2606.28518). Moreover, our results are complementary to synthillation (Campbell and Howard, Phys. Rev. A 95, 022316 (2017)), which can distil third-level states from $\\mathrm{T}$ states, but only at distance $2$. Under optimality in terms of input count, space footprint, and distance for a given output, we find $74$ optimal generalised triorthogonal protocols in our range, $65$ of which are new to the literature. To achieve our classification, we extend the classification of unital triorthogonal spaces of Nezami and Haah from length $38$ to $54$ using a directional derivative method. Using these as the stabiliser spaces, we add logical rows that satisfy the triorthogonality constraints to create full protocols.","published":"2026-09-25T06:10:26Z","updated":"2026-09-25T06:10:26Z","authors":["Adam Wills","Shubham P. Jain","Shraddha Singh"],"categories":["quant-ph"],"headline":"Researchers classify advanced quantum error-correcting codes up to length 54","summary":"The researchers studied a type of quantum code used in fault-tolerant quantum computing to improve the quality of special quantum states needed for computation. They expanded previous classifications of these codes from length 38 to 54, focusing on codes that can detect certain errors more effectively. They identified 74 optimal code protocols within this range, 65 of which were previously unknown. Their results are based on theoretical analysis and mathematical methods, providing a more complete map of these codes but without experimental testing."},{"id":"2609.30859v1","title":"AC Power Flow Contingency Analysis Using a Single Deep Neural Network","abstract":"Contingency analysis using the AC power flow (AC-PF) model is a critical tool for accurate grid security assessment, but its computational burden increases with the number of operating scenarios and outage configurations to evaluate. Recent ML-based approaches typically require outage-specific training data, leading to offline training costs that scale with the number of contingencies. This work proposes a framework that reuses a single ML model trained solely on basecase AC-PF data to estimate post-contingency operating states under arbitrary single-line outages. The proposed approach formulates post-contingency state prediction as a fixed-point iteration. If the ML model is a deep neural network (DNN), we derive sufficient conditions that guarantee convergence and develop semidefinite programming (SDP) formulations to certify these conditions for a given DNN. Numerical tests on the IEEE 118-bus system demonstrate that the proposed SDP formulations are tight, that the certified conditions hold for all tested contingencies, and that the resulting method produces accurate post-contingency state estimates within only a few iterations.","published":"2026-09-25T06:08:45Z","updated":"2026-09-25T06:08:45Z","authors":["Md Obaidur Rahman","Junjie Qin","Vassilis Kekatos"],"categories":["eess.SY","cs.LG"],"headline":"New AI Method Predicts Power Grid Problems Efficiently","summary":"Researchers addressed the challenge of analyzing power grid security, which usually requires heavy computations for many outage scenarios. They created a single deep neural network model trained on normal grid data to predict how the grid behaves after different line outages. Their approach uses a mathematical process ensuring the model's predictions converge reliably. Tests on a standard power system showed this method accurately estimates post-outage states quickly, potentially reducing the need for multiple trained models."},{"id":"2609.30858v1","title":"On the Cauchy problem for the Tzitzéica equation: Soliton resolution conjecture and asymptotic analysis","abstract":"We study the Cauchy problem for the Tzitzéica equation, which is an important integrable model arising in affine differential geometry and characterizing proper affine spheres. Recently, Huang, Wang and Zhu (Math. Ann. \\textbf{395}, 18 (2026)) reported the long-time asymptotic result for the Tzitzéica equation for the case of the purely continuous spectrum. Inspired by their work, we conduct in-depth research on the soliton resolution conjecture and asymptotic analysis of the Tzitzéica equation. Compared with the previous work, by developing the $\\bar{\\partial}$-nonlinear steepest descent method, we reveal that the solution of the Tzitzéica equation exhibits three different asymptotic regions depending on $ξ:=x/t$. For the region $ξ\\in(-\\infty,-1]\\cup[1,\\infty)$, where there are no stationary phase points, we rigorously prove that the solution of the Tzitzéica equation decays algebraically to zero with error $O(t^{-1})$. In the region $ξ\\in(-1,1)$, the phase function $θ(z)$ has two stationary phase points. The corresponding asymptotic approximations can be characterized with an $N$-soliton solution as well as an interaction term between soliton solutions and the dispersion term with a residual error of order $O(t^{-3/4})$. In the transition regions $\\{(x,t):1-\\varepsilon<|ξ|<1\\}$, the corresponding asymptotic approximation can be characterized by the $N$-soliton solution together with the interaction terms between the soliton solution and the dispersion term, with a residual error of order $O(t^{-3/4})$. Our results provide a rigorous analysis of the long-time asymptotic behavior of solutions of the Tzitzéica equation in different regions and confirm the soliton resolution conjecture for the Tzitzéica equation.","published":"2026-09-25T06:07:43Z","updated":"2026-09-25T06:07:43Z","authors":["Shou-Fu Tian","Jia-Fu Tong"],"categories":["math.AP","math-ph","nlin.SI"],"headline":"Researchers analyze long-term behavior of Tzitzéica equation solutions","summary":"The researchers studied the Tzitzéica equation, a mathematical model important in affine geometry, focusing on how solutions evolve over time. Using advanced mathematical techniques, they identified three regions where the equation's solutions behave differently, proving that solutions decay or can be described by soliton waves plus small error terms. Their work supports the soliton resolution conjecture, showing a detailed understanding of the equation's long-time behavior based on rigorous analysis of theoretical results."},{"id":"2609.30857v1","title":"Rare-earth Bilayer Triangular Lattice as a Platform for Tunable Ground States and Field-induced Magnetic Phases","abstract":"Rare-earth bilayer triangular lattice (TL) antiferromagnets provide a versatile platform for realizing diverse magnetic states by combining geometric frustration, interlayer coupling, and strong single-ion anisotropy. Here, we report a family of R2O2Se (R = Sm, Eu, Tb-Lu; ROSe) single crystals featuring bilayer equilateral TLs. Magnetic susceptibility and specific-heat measurements reveal predominantly antiferromagnetic (AFM) interactions and diverse magnetic ground states across the series, including successive AFM transitions in TbOSe and single AFM transitions in SmOSe, DyOSe, HoOSe, and YbOSe. Notably, DyOSe and HoOSe exhibit pronounced 1/2 magnetization plateau-like features for fields along the c-axis, revealing field-induced magnetic states. In HoOSe, complementary thermodynamic and magnetization measurements resolve three critical fields and a rich field-temperature phase diagram containing five distinct magnetic phases. These results establish rare-earth bilayer TLs as a chemically tunable materials platform for accessing novel quantum spin states through the interplay of lattice geometry, single-ion anisotropy, and competing magnetic interactions.","published":"2026-09-25T06:03:05Z","updated":"2026-09-25T06:03:05Z","authors":["Jianqiao Wang","Fangli Li","Yangyang Yu","Quan Xiao","Zhibin Qiu","Liusuo Wu","Shu Guo"],"categories":["cond-mat.mtrl-sci"],"headline":"Rare-earth crystal layers reveal complex magnetic behaviors","summary":"Researchers studied crystals made of rare-earth elements arranged in two-layer triangular patterns. They found various magnetic states due to interactions between atoms and the crystal structure, including changes triggered by applying magnetic fields. The study shows that these materials can be chemically adjusted to explore new magnetic states, but the findings are based on measurements and observations of specific crystal samples, not broader practical applications."},{"id":"2609.30856v1","title":"Learning Chance-Constrained MDPs with Bellman Distributional Certificates","abstract":"Safe reinforcement learning (RL) commonly enforces expected-cost constraints, but such expectation safety may fail to control the probability of rare high-cost trajectories. Chance-constrained MDPs (CCMDPs) impose a stronger probability-level requirement, but are widely viewed as harder because the chance constraint is nonconvex and depends on the full trajectory rather than a Bellman-linear expectation. In this paper, we reveal that this computational difficulty does not necessarily imply a higher statistical price. For tabular discounted CCMDPs with fixed bounded successor support and access to a certified planning oracle, we establish a model-based upper bound, with a matching lower bound up to logarithmic terms. Technically, our key idea is the \\emph{Bellman distributional certificate}, which constructs a Bellman recursion for constraint violation probabilities before policy selection. The certificate can be reused across candidate policies; combined with shared row-wise reverse-KL confidence sets, it gives a policy-uniform trajectory-KL transfer without a union bound over policies or time--budget Bellman tables. For stochastic policies, we give a model-free variance-reduced policy-gradient algorithm with a finite-sample expected KKT-residual guarantee and independent validation of every accepted policy. Numerical experiments on synthetic CCMDPs and an IEEE 14-bus energy storage control benchmark illustrate the safety and mechanism behavior of the proposed algorithms.","published":"2026-09-25T06:03:03Z","updated":"2026-09-25T06:03:03Z","authors":["Chenbei Lu","Hongyu Yi"],"categories":["cs.LG"],"headline":"New Method Improves Safety in Reinforcement Learning with Rare Risks","summary":"Researchers studied a type of decision-making model focused on controlling the chance of rare but costly events, called chance-constrained Markov decision processes (CCMDPs). They developed a new technique called Bellman distributional certificates that helps better estimate and manage the probability of unsafe outcomes in learning algorithms. Their approach offers theoretical guarantees, works with existing planning tools, and was tested in simulations and an energy storage control problem to show improved safety handling. The findings are mainly theoretical and demonstrated in controlled experiments."},{"id":"2609.30855v1","title":"MDSkin-Net: Multi-Task Skin Lesion Analysis Driven by Pattern Analysis Priors and Spatial Alignment Regularization","abstract":"Reliable skin lesion segmentation and classification are central to dermoscopic computer-aided diagnosis. Existing multi-task frameworks couple the two tasks architecturally without clinical knowledge, while knowledge-injecting approaches rely on the macroscopic ABCD rule, which was not designed for dermoscopy. Dermoscopic diagnosis is grounded in Pattern Analysis, a microscopic framework structured around dermoscopic features. We propose MDSkin-Net, which incorporates cue-level Pattern Analysis priors into a hybrid CNN-Transformer architecture. At its core is a Pattern Analysis-Guided Attention Module (PAGAM) comprising three priors motivated by distinct dermoscopic cues: an improved Efficient Channel Attention (iECA), a Multi-Scale Spatial Attention (MSSA), and a Biased Asymmetry Attention (BAA). We further introduce a multi-scale spatial alignment regularization (MSAR) that uses the segmentation ground-truth mask as hierarchical soft supervision, confining the classification head to lesion-localized evidence and coupling both task pathways through a shared spatial prior. Trained exclusively on the ISIC 2017 training split without external dermoscopy data, the MDSkin-Net ensemble transfers robustly under zero-shot evaluation, reaching a Dice Similarity Coefficient (DSC) of 92.38% and a melanoma AUC of 97.84%on PH2, and a DSC of 88.92% on the ISIC 2018 Task 1 test set. On the in-domain ISIC 2017 benchmark, the ensemble attains a mean Area Under the Curve (AUC) of 91.60% across the two classification tasks (melanoma and seborrheic keratosis vs. rest), and a DSC of 84.72% for segmentation. Classification remains competitive with baselines; in-domain segmentation trails single-task specialists, yet the proposed priors and alignment regularization yield representations that generalize consistently across cohorts of different scales.","published":"2026-09-25T06:02:34Z","updated":"2026-09-25T06:02:34Z","authors":["Yijian Li","Saad Bedros","Paul Bigliardi","Mei Bigliardi Qi","Vassilios Morellas","Nikolaos Papanikolopoulos"],"categories":["cs.CV"],"headline":"New AI model improves skin lesion detection using detailed pattern knowledge","summary":"Researchers developed MDSkin-Net, an AI system that analyzes skin lesions by integrating detailed pattern analysis typically used by dermatologists. It combines different attention methods within a neural network to better identify lesion features and aligns segmentation and classification tasks to focus on relevant areas. Tested on standard datasets, MDSkin-Net showed strong performance in identifying melanoma and segmenting lesions, especially when applied to new unseen data. The results support the value of incorporating clinical pattern knowledge but note that segmentation still lags behind specialized single-task models."},{"id":"2609.30854v1","title":"The KV Cache Is the New Memory Wall","abstract":"Autoregressive LLM inference at long context is bounded by memory bandwidth, not arithmetic throughput, and the binding resource shifts from model weights to the Key-Value (KV) cache as sequence length grows. For Llama-3-70B in BF16, the 140 GB weight footprint exceeds the 80 GB HBM of a single accelerator, and one 128k-token sequence adds 42 GB of KV cache. Techniques that compress, evict, page, share, or offload KV state have proliferated, but reported gains use inconsistent workloads, hardware, and quality metrics, preventing cross-paper comparison. This SoK paper unifies the field analytically, with a protocol that strictly separates derived and reported claims. We derive closed-form arithmetic intensity as a decaying function of context length, parameterized by hardware topology for NVIDIA H100, NVIDIA B200, and AMD MI300X, including per-die bandwidth partitioning and the crossover lengths where KV traffic overtakes weight traffic. We classify the literature into five domains, quantization, token eviction, KV paging, prefix caching, and heterogeneous tiering, evaluating one method per domain at 128k context under a single protocol. The central finding is a three-regime structure: below a hardware-specific crossover, weight traffic dominates and KV compression yields negligible speedup; beyond it, KV traffic dominates and each domain trades quality for bandwidth savings approaching the roofline bound. Paging and prefix sharing are lossless but address capacity, not bandwidth. Quantization and eviction cut bandwidth directly, with degradation that accelerates below 4-bit precision and turns discontinuous for eviction on position-sensitive tasks. Tiering converts the bandwidth wall into an interconnect problem bounded by PCIe or NVLink rather than HBM. We close with design rules for selecting a compression domain given hardware, context length, and quality budget.","published":"2026-09-25T06:01:27Z","updated":"2026-09-25T06:01:27Z","authors":["Tejinder Singh"],"categories":["cs.DC","cs.LG","cs.PF"],"headline":"Researchers Study Memory Limits in Large AI Models' Long Contexts","summary":"The researchers investigate how memory bandwidth, not computing speed, limits large language models when processing very long text sequences. They analyze how the Key-Value (KV) cache grows with context length and becomes the main memory bottleneck, surpassing model weights in demand. By creating a unified framework, they classify existing methods to reduce KV memory use and evaluate their effects on speed and quality, finding trade-offs depend on hardware and compression techniques. Their results come from analytical modeling, not direct experiments with all methods."},{"id":"2609.30853v1","title":"Faster Linear Programming with $\\sqrt{\\mathrm{rank}}$ Linear System Solves","abstract":"Lee and Sidford [LS19] showed that the linear program $\\min\\{c^\\top x\\ :\\ A^\\top x=b,\\ l\\leq x\\leq u\\}$ over $x\\in\\mathbb{R}^m$, where $A\\in\\mathbb{R}^{m\\times n}$, can be solved to accuracy $ε$ using $O(\\sqrt n\\log^{13}m\\cdot\\log(mU/ε))$ solves of linear systems in $A^\\top\\mathbf DA$ for positive diagonal matrices $\\mathbf D$, where $U$ bounds the magnitudes of the input and of the initial point. Theirs is the first such bound governed by $\\mathrm{rank}(A)=n$ rather than by the number of constraints $m$. Song [Son19] mentioned that reducing the $\\log^{13}m$ factor is an interesting future direction. Given an interior point and a finite certified magnitude bound $U$, we give an algorithm that uses $O(\\sqrt n\\log^{4}m\\cdot\\log(mU/ε))$ solves of such linear systems, improving the Lee--Sidford bound by a factor of $\\log^{9}m$. We conjecture that $O(\\sqrt n\\log(mU/ε))$ such linear-system solves suffice.","published":"2026-09-25T05:59:48Z","updated":"2026-09-25T05:59:48Z","authors":["Zhao Song"],"categories":["math.OC","cs.DS"],"headline":"Improving Speed for Solving Certain Linear Programs","summary":"Researchers studied a method to solve linear programming problems, which involve finding the best outcome given constraints. Previous work showed these problems could be solved using a number of linear system solutions related to the problem's size and complexity. The researchers improved this method by significantly reducing one factor that slowed the process, making the algorithm faster in theory. Their results are based on mathematical analysis and assume starting within certain conditions; actual practical performance or applications are not addressed in detail."},{"id":"2609.30852v1","title":"Energetically Driven Structure Matching for Autonomous Total X-ray Scattering Experiments","abstract":"The emergence of autonomous laboratories motivates rapid conversion of experimental data into reliable atomistic models on time-scales compatible with closed-loop optimization. Here we develop an energetically driven structure matching framework for analysis during ongoing total X-ray scattering experiments. Using data from gold nanoparticles, we match against idealized spherical, octahedral, decahedral, and icosahedral geometries, their machine-learned interatomic potential (MLIP)-relaxed structures, and molecular dynamics (MD) ensembles. Idealized models are fast to generate but can misassign morphology and systematically underestimate size by neglecting surface relaxation, strain, and thermal disorder. MLIP relaxation markedly improves both, while MD ensemble averaging agrees best with experiment. We therefore introduce a hierarchical workflow combining rapid idealized screening with targeted MLIP and MD refinement of top candidates, delivering improved structural feedback without interrupting autonomous operation. This framework provides a route towards autonomous campaigns in which the target structure itself can be updated in response to the evolving energy landscape of structures compatible with the experimental data.","published":"2026-09-25T05:58:37Z","updated":"2026-09-25T05:58:37Z","authors":["Emil J. P. Frost","Martin A. Karlsen","Jonas H. Jensen","Rodrigo Moreno","Jørn Lambertsen","Jonathan Quinson","Andy S. Anker","Alexander Bagger","Tejs Vegge"],"categories":["cond-mat.mtrl-sci","cond-mat.mes-hall","physics.comp-ph"],"headline":"New Method Matches Atomic Structures During X-ray Experiments","summary":"Researchers developed a method to quickly interpret data from X-ray scattering experiments to identify atomic arrangements of materials like gold nanoparticles. They compare simple ideal shapes, improved shapes refined by machine-learned potentials, and detailed simulations to find the best match. Their approach combines fast initial guesses with more accurate refinements, helping experiments adjust in real time without stopping. The study shows simulations agree best with experiments, but the method is based on modeling and has not been tested beyond these cases."},{"id":"2609.30851v1","title":"Non-vanishing of the structure-sheaf cohomology of affine henselian schemes and henselian affinoids","abstract":"Let $(A,I)$ be a henselian pair such that $A$ is a domain of characteristic zero and $I\\ne0$. We prove that any flat affine henselian scheme of finite presentation over $(A,I)$ has nonzero first structure-sheaf cohomology whenever one of its fibers over a point in $V(I)$ has positive dimension. This in particular gives a negative answer to the question by Devadas in the mixed-characteristic case. As an application, we also prove that the first structure-sheaf cohomology does not vanish for henselian affinoids of positive dimensions.","published":"2026-09-25T05:57:48Z","updated":"2026-09-25T05:57:48Z","authors":["Yoshinori Gongyo","Fumiharu Kato","Takayuki Uchiba"],"categories":["math.AG","math.AC"],"headline":"Researchers find key cohomology property in special algebraic structures","summary":"The researchers studied henselian pairs, which are algebraic structures with specific properties, in characteristic zero. They proved that certain geometric objects built from these pairs have nonzero first structure-sheaf cohomology if one of their fibers has positive dimension. This result answers a previously open question negatively in a mixed-characteristic setting. Their findings also apply to henselian affinoids, showing the same cohomology does not vanish in positive dimensions. The conclusions are based on theoretical proofs within algebraic geometry."},{"id":"2609.30850v1","title":"Strangeness Production at GlueX","abstract":"The GlueX experiment at Jefferson Lab is well suited for exclusive studies of strangeness production, combining a linearly polarized photon beam at $8.0 < E_γ< 8.8$ GeV with a large-acceptance detector that reconstructs weakly decaying hyperons. We outline a research program built on three aspects: (i) spectroscopy of strange mesons, where many excited $K^*$ states remain unestablished and crypto-exotic strange hybrids must be identified from combined spectrum and decay information, anchored by the $K^*(892)$ as a standard candle; (ii) spin dynamics, where the first precise measurements of the complete spin-density matrix elements in the photoproduced $K^*(892)Λ$ system beyond the Schilling formalism constrain the $t$-channel strange production mechanism; and (iii) hadronization, probed through the newly published $p\\bar{p}$, $Λ\\barΛ$, and $p\\barΛ$ photoproduction cross sections and emerging $Λ\\barΛ$ spin-correlation measurements. We place these efforts in the context of the broader GlueX strangeness program and give an outlook toward the high-statistics GlueX-II era.","published":"2026-09-25T05:55:42Z","updated":"2026-09-25T05:55:42Z","authors":["Hao Li"],"categories":["hep-ex","nucl-ex"],"headline":"Studying Strange Particles with GlueX at Jefferson Lab","summary":"Researchers at Jefferson Lab's GlueX experiment investigate how particles containing strange quarks are produced and behave. They use a polarized photon beam and a detector that tracks certain decaying particles to study strange mesons, spin behavior in particle production, and how particles form. Their program aims to identify new strange particle states and understand production mechanisms, based on recent measurements. The findings are part of ongoing work and future studies with improved data are planned."},{"id":"2609.30849v1","title":"Enhancing Assessment of Self-Consistency in LLM Explanations using Perturbation Strength","abstract":"Prior work has examined the self-consistency of LLM-generated explanations using surface-level perturbation methods. However, the strength of these perturbations is not explicitly measured and controlled. In this work, we propose an LLM-as-a-judge approach to measure perturbation strength in a unified manner across input and CoT perturbations. We then evaluate the self-consistency in explanations generated from various LLMs under controlled strength conditions, ensuring a fair comparison across perturbation types. Experiments show that our proposed LLM-based perturbation strength measure outperforms other embedding- and probability-based approaches and that input perturbations generally affect LLMs more strongly than CoT perturbations. Our work suggests that judgments about a model's self-consistency is fair only within the same perturbation type.","published":"2026-09-25T05:53:13Z","updated":"2026-09-25T05:53:13Z","authors":["Phuong Q. Le","Kemal Kurniawan","Jey Han Lau"],"categories":["cs.CL"],"headline":"Measuring Consistency in AI Explanations Using Controlled Changes","summary":"Researchers studied how consistent large language models (LLMs) are when explaining their answers. They introduced a new way to measure the strength of changes (perturbations) made to inputs or reasoning steps, allowing fair comparison of model consistency. Their experiments found this new measure works better than others and that input changes affect model explanations more than reasoning changes. They caution that judging model consistency is only fair when comparing the same type of changes."},{"id":"2609.30848v1","title":"When Is Molecular-Dynamics-Predicted Ionic Conductivity Reliable in Solid Electrolytes?","abstract":"Molecular dynamics is widely used to predict ionic conductivity in solid electrolytes, but the reliability of these predictions is often difficult to assess. Our analysis identifies finite trajectory length, limited cell size, and insufficient sampling as intrinsic limitations of finite atomistic ion-transport calculations. Cubic Li7La3Zr2O12 is used as a representative solid electrolyte to quantify their consequences. These limitations can remain hidden behind apparently linear mean-squared displacements and well-behaved Arrhenius relations, leading to inaccurate diffusivities, activation energies, and extrapolated ionic conductivities. Such inaccuracies can misrank candidate solid electrolytes and consequently misdirect computational screening and experimental validation. Here, we establish the local mean-squared-displacement exponent, α(t), as a quantitative reliability criterion that links dynamical convergence to errors in diffusivity and Nernst-Einstein ionic conductivity. The criterion further determines the minimum trajectory length required to achieve a prescribed accuracy as a function of temperature. Independent replicas reduce statistical uncertainty, while selective single-axis expansion mitigates finite-size effects. By establishing when simulated ionic conductivity is quantitatively trustworthy, this approach enables more reliable materials ranking and more efficient use of computational and experimental resources, thereby accelerating the development of high-performance solid electrolytes for solid-state batteries.","published":"2026-09-25T05:53:08Z","updated":"2026-09-25T05:53:08Z","authors":["Yiwei You","Shaofei Chen","Zhifeng Wu","Pushun Lu","Eric Jianfeng Cheng","Songyan Chen","Shunqing Wu"],"categories":["cond-mat.mtrl-sci"],"headline":"Ensuring Reliable Ionic Conductivity Predictions in Solid Electrolytes","summary":"Researchers studied why computer simulations predicting ionic conductivity in solid electrolytes can sometimes be inaccurate. They found that limited simulation time, small sample size, and insufficient sampling can hide errors, leading to wrong estimates of how well ions move. Using a specific solid electrolyte as an example, they developed a method to judge when simulation results are reliable and how long simulations should run for accuracy. This helps scientists better compare materials and focus resources when developing batteries."},{"id":"2609.30847v1","title":"Decision-Gated Surrogate-Assisted Stochastic Optimization with Independent High-Fidelity Certification for Photovoltaic Hosting-Capacity Planning","abstract":"High photovoltaic (PV) penetration planning requires repeated nonlinear PV and power-flow evaluations under uncertain irradiance and demand, particularly when two competing objectives, penetration and total electrical loss, are optimized simultaneously subject to reliability constraints. This paper develops a decision-focused, surrogate-assisted stochastic multiobjective framework in which a calibrated PV model coupled with a distribution-network simulator provides the high-fidelity reference, while matched data-only and physics-informed surrogates accelerate optimization and are evaluated empirically for decision reliability across the loss-penetration tradeoff. Although the two surrogates exhibit nearly identical held-out errors, optimizer-realistic screening yields markedly different operational validity rates of 44.4% and 93.4%, respectively, motivating selection of the physics-informed model on decision performance rather than test accuracy. Surrogate inference cuts the cost of a 60-scenario design evaluation from 286.0 to 0.46 ms, an approximately 620 time speedup. A decision-gated enrichment strategy further targets Pareto-critical, constraint boundary, and surrogate-disagreement regions, reducing prediction error by 39.1%. Independent 500 scenario certification verifies all 80 screened candidates and identifies 46 non-dominated designs; the selected compromise attains 70.74% expected penetration with 297.9 kW expected total electrical loss. Cross-feeder transfer eliminates baseline voltage violations on the 85- and 69-bus systems, but produces overvoltage and a 62.7% increase in feeder-network loss on the 33-bus system, demonstrating that certified penetration is topology-specific and demands decision-level auditing, independent certification, and feeder-specific validation.","published":"2026-09-25T05:51:37Z","updated":"2026-09-25T05:51:37Z","authors":["Ali Abubakar","Samuel Essamuah Assabil","Akhtar Hussain","Van-Hai Bui"],"categories":["eess.SY"],"headline":"Improving Solar Power Planning with Faster, Smarter Computer Models","summary":"Researchers addressed the challenge of planning solar panel integration in electrical networks, which requires complex and repeated calculations under uncertain conditions. They developed a new approach using simpler, faster computer models (surrogates) to speed up optimization while still relying on detailed simulations for accuracy. Their findings show that models incorporating physical knowledge perform better for decision-making than those based on data alone. They also highlight that certification of solar capacity must be specific to each network, as different setups affect outcomes differently, emphasizing the need for careful, independent validation."},{"id":"2609.30846v1","title":"I-Parakeet: Integer-Only Conformer ASR on Mobile NPU","abstract":"In this paper, we propose I-Parakeet, an integer-only implementation of NVIDIA's Parakeet-CTC (0.6B parameters) that runs on a smartphone NPU without any floating-point operator or CPU fallback. Modern Conformer ASR models are hard to deploy on edge devices because of their size, and quantized models still fall back to floating point for numerically sensitive operations. This prevents them from fully exploiting integer accelerators such as mobile NPUs. To achieve this, our contributions are threefold. First, we derive an integer formulation of the relative-positional self-attention at the core of the Conformer. We fuse its two score branches with different quantization scales and the relative shift into integer-only operations. Second, we introduce a minimax-optimized Swish approximation that minimizes the maximum error of the Swish output. Third, a layer-wise range analysis of activations yields two targeted remedies: an INT16 grid for the BatchNorm output and percentile calibration for the heavy-tailed pre-encoder activations. I-Parakeet achieves 4.97% WER on LibriSpeech test-other, running on a Qualcomm NPU at a real-time factor of 0.048, 7.5x faster than a CPU baseline.","published":"2026-09-25T05:47:40Z","updated":"2026-09-25T05:47:40Z","authors":["Taichi Nishimura"],"categories":["cs.CL","cs.SD","eess.AS"],"headline":"Integer-Only Speech Recognition Runs Efficiently on Mobile NPUs","summary":"Researchers developed I-Parakeet, a version of a speech recognition model that uses only integer calculations to run on smartphone neural processors without relying on slower floating-point operations. They redesigned key parts of the model to work fully with integer math and optimized certain functions to maintain accuracy. Their tests showed it works well for recognizing speech while running much faster than a standard CPU, though these results focus on specific technical improvements and performance measures."},{"id":"2609.30845v1","title":"$U(1)$ Gauge-Equivariant Representation Learning of Bloch States","abstract":"Representation learning, or featurization, of mean-field Bloch states provides an important route for incorporating electronic information into machine-learning models of first-principles condensed-phase systems. A challenge in representing Bloch states is the gauge redundancy of each state. In this work, we propose to address this issue by explicitly incorporating gauge equivariance in the machine-learning framework. We develop a $U(1)$-equivariant autoencoder that compresses the plane-wave-basis Bloch states obtained from density functional theory into a low-dimensional latent representation. We achieve an average overlap of 0.970 between original wavefunctions and the reconstructions from a 64-dimensional latent space using a dataset of two-dimensional insulators. We further show that $U(1)$ equivariance endows the latent space with physically meaningful structures, including smoothness and topological information. Finally, we present a proof-of-principle application in which the latent representations are used to predict the off-diagonal elements of the $GW$ self energy matrix, and demonstrate the importance of enforcing gauge equivariance in the prediction.","published":"2026-09-25T05:43:31Z","updated":"2026-09-25T05:43:31Z","authors":["Chengyan Zhang","Xian Xu","Bowen Hou","Diana Y. Qiu"],"categories":["physics.comp-ph"],"headline":"Improving electronic state models with gauge-aware machine learning","summary":"Researchers addressed the challenge of redundant information in representing electronic states of materials by incorporating a special symmetry called gauge equivariance in a machine learning model. They created a $U(1)$-equivariant autoencoder that compresses complex wavefunction data into a simpler form while retaining accuracy. Their approach, tested on simulated two-dimensional materials, showed the model's internal structure reflects important physical properties, and using this method improved predictions related to electronic interactions. The findings are based on computational experiments and show the potential of symmetry-aware learning in material science."},{"id":"2609.30844v1","title":"Sharp Basic Velocity Estimates for the Plane Steady Navier--Stokes Equations through Vorticity","abstract":"We consider some new velocity estimates for general steady Navier-Stokes solutions in plane domains. Namely, we obtain some optimal estimates of the difference between mean values of the velocity over two concentric circles in terms of the $L^2$-norm of the vorticity in the annulus between the circles. The sharpness of the estimates is demonstrated by some unexpected counterexamples.","published":"2026-09-25T05:39:53Z","updated":"2026-09-25T05:39:53Z","authors":["Mikhail Korobkov","Zhixiang Zhuang"],"categories":["math.AP"],"headline":"Researchers Improve Velocity Estimates for Fluid Flow Equations","summary":"The researchers studied equations that describe steady fluid flow in flat areas and found new, precise ways to estimate the difference in flow speed between two rings. They related this difference to a measure of the fluid's rotation in the area between the rings. Their work includes examples showing these estimates are as accurate as possible. The results are theoretical and focus on improving understanding of fluid behavior described by the Navier-Stokes equations."},{"id":"2609.30843v1","title":"A Quantization-Constrained Parameter Method for Studying $α$ and Cluster Decay","abstract":"We propose a Quantization-Constrained Parameter Method (QCPM) for determining the Woods-Saxon (WS) potential depth $V_0$ and diffuseness $a$ in studies of $α$ and cluster decay half-lives. In this approach, the parameters are derived analytically by imposing the Bohr--Sommerfeld (BS) quantization condition, thereby eliminating the need for parameter fitting. The calculated diffuseness values exhibit a strong dependence on nuclear shell structure. The reliability of the QCPM-derived WS potential is validated through optical model analyses of elastic $α$ scattering data. The resulting $α$ and cluster decay half-lives show good agreement with experimental values. Our results highlight the significant role of daughter-nucleus deformation in improving the consistency between theoretical predictions and experimental data. In addition, we introduce a modified Woods--Saxon (mWS) potential that effectively approximates the surface behavior of folding potentials incorporating the nuclear medium effect. This modification leads to closer agreement with the measured half-lives. Free of adjustable parameters such as $V_0$ and $a$, the QCPM enhances the predictive power of phenomenological potentials for $α$ and cluster decay studies.","published":"2026-09-25T05:38:29Z","updated":"2026-09-25T05:38:29Z","authors":["Nguyen Thi Huyen Nga","Le Hoang Chien","Nguyen Tri Toan Phuc","Chau Van Tao"],"categories":["nucl-th"],"headline":"New method improves predictions for alpha and cluster decay rates","summary":"Researchers developed a Quantization-Constrained Parameter Method (QCPM) to determine parameters of a nuclear model without fitting to data, using a quantization condition. They applied this method to calculate alpha and cluster decay half-lives, finding results that matched experimental data well. Their approach also highlights the importance of nuclear shape changes and introduces a modified potential to better capture nuclear effects, enhancing theoretical predictions without relying on adjustable parameters."},{"id":"2609.30842v1","title":"Impedance Cloning: Learning Equilibrium Point Parameters for Contact-Rich Manipulation","abstract":"Contact-rich manipulation requires robots to regulate force against surfaces whose geometry deviates unpredictably from training conditions. Trajectory-based imitation learning, which reproduces observable outputs, breaks down under such shifts. We propose Impedance Cloning, which instead imitates the biomechanical priors that generate motion -- the stiffness and equilibrium point -- and thereby passively absorbs contact uncertainty. Because these parameters encode intent rather than outcome, they generalize across surface geometries where trajectory reproduction does not. We extract them from bilateral teleoperation demonstrations via a particle filter without force/torque sensors and evaluate the framework on two CRANE-X7 manipulators. In a wiping task with joint-space actions, the trajectory-based baseline loses contact below -6 cm, whereas the proposed method maintains a consistent 4-5 N contact force above -6 cm, with a gradual decrease below; with Cartesian-space actions, its force-height slope over 0 to +8 cm is 0.13 +/- 0.03 N/cm, versus 0.34-0.83 N/cm for fixed-impedance baselines. In a pick-and-place task with 10 diverse cups (100 trials), the proposed method succeeds in 84 trials, outperforming the fixed-impedance baseline (74/100) and performing comparably to a variable impedance control baseline (82/100) with one demonstration instead of ten. In a grasping task, the representation reduces torque tracking error with both ILBiT and Mamba backbones, confirming its generality across architectures.","published":"2026-09-25T05:36:42Z","updated":"2026-09-25T05:36:42Z","authors":["Hayato Takahashi","Ryoga Oishi","Yuki Kasuga","Toshiaki Tsuji"],"categories":["cs.RO"],"headline":"New Robot Learning Method Improves Contact Handling with Surfaces","summary":"Researchers studied how robots handle tasks involving touching and applying force to unpredictable surfaces. They developed 'Impedance Cloning,' a way for robots to learn underlying motion intentions like stiffness rather than copying exact movements. Tested on robot arms doing wiping and picking tasks, their method maintained better and more consistent contact forces compared to traditional approaches. The researchers also showed it worked well across different control systems. These results come from simulations and controlled experiments, showing promise but not tested in all real-world scenarios."},{"id":"2609.30841v1","title":"Why Jailbreaks Succeed in Diffusion Language Models: An Energy Landscape Analysis","abstract":"Existing attacks and defenses for diffusion-based large language models (dLLMs) target specific vulnerabilities but lack a shared framework explaining why attacks succeed. We propose one by interpreting safety alignment as shaping the denoising energy landscape: a well-aligned model routes harmful queries toward safe outputs through an energy barrier that separates the two regions. Current jailbreak attacks reduce to two strategies for circumventing this barrier: obscuring the query's safety disposition at initialisation, or intervening mid-trajectory to force the denoising path across the energy barrier. From this perspective and the result that masked diffusion models minimise kinetic energy during denoising, we derive three complementary, training-free detection signals: a step-0 ratio that reads the initial safety disposition from the logit distribution before generation begins, and two trajectory-velocity signals that track kinetic energy in complementary subspaces of the logit space. An attack must either reveal its intent at initialisation or expend kinetic energy to cross the barrier in at least one monitored subspace, so the three signals cover each other's blind spots in the energy budget by construction. Evaluation across three dense dLLMs (LLaDA-8B, LLaDA-1.5, Dream-7B) and a sparse mixture-of-experts dLLM (LLaDA-MoE-7B) confirms this complementarity. In stress tests of known attacks, every configuration that evades detection also fails to produce harmful content, suggesting that the detection and barrier-crossing thresholds are hard to separate.","published":"2026-09-25T05:35:36Z","updated":"2026-09-25T05:35:36Z","authors":["Thong Bach","Dung Nguyen","Thao Minh Le","Truyen Tran"],"categories":["cs.AI"],"headline":"Researchers Analyze Why Jailbreaks Bypass Safety in Language Models","summary":"The researchers studied why certain attacks, called jailbreaks, succeed in tricking advanced language models designed to avoid harmful responses. They propose a new way to understand this by imagining a model’s safety system as an 'energy barrier' that guides unsafe inputs to safe outputs. Their study suggests jailbreaks either hide their intent early or push through this barrier later, allowing the team to develop signals that detect these attacks. Tests on several models showed the detection method is effective, but findings are based on theoretical analysis and simulations."},{"id":"2609.30840v1","title":"Aligning One-Step Generative Models with Reward-Weighted Transport Distillation","abstract":"One-step generators enable high-quality visual generation with a single network evaluation, but their post-training is difficult: general implicit generators provide neither tractable likelihoods nor denoising trajectories, and many rewards are non-differentiable. We introduce Reward-Weighted Transport Distillation (RWTD), a post-training method that requires only generated samples and scalar reward evaluations. Rather than aligning solely to the conventional reward-tilted reference distribution, RWTD constructs an adaptive target that mixes separately tilted current and reference distributions. The current component incorporates improvements discovered during training, while the reference component anchors the target to the pretrained generator. RWTD realizes this target through feature-space optimal transport and fixed-point regression. Theoretical analysis shows that the fixed-point distributions of RWTD interpolate between off-policy reward tilting of the reference and on-policy tilting of the current model, providing a principled approach to balancing reward adaptation with retention of prior knowledge. Empirically, RWTD substantially improves the GenEval score of the one-step SANA Sprint 1.6B backbone from 0.73 to 0.80, while separate preference alignment experiments demonstrate strong cross-reward generalization that yields balanced improvements and preservation of compositional capabilities.","published":"2026-09-25T05:35:32Z","updated":"2026-09-25T05:35:32Z","authors":["Austin Wang","Ziheng Cheng","Lexing Ying"],"categories":["cs.LG","cs.CV"],"headline":"New method improves one-step AI image generators using reward feedback","summary":"Researchers developed a technique called Reward-Weighted Transport Distillation (RWTD) to improve AI models that create images in one step. RWTD uses feedback scores from generated images to adjust the model without needing complex calculations. The method balances learning from new rewards while keeping previous knowledge. Tests showed RWTD improved image quality scores and maintained the model's ability to combine features. Findings are based on theoretical analysis and experiments with pre-existing AI image generators."},{"id":"2609.30839v1","title":"Attention-Based Adaptive Policies for Simultaneous Speech-to-Text Translation","abstract":"Simultaneous speech-to-text translation (Simul-S2TT) consists of generating partial translations while the incoming audio frames are processed by the system. However, the streaming nature of this setup creates the challenge of deciding the best moment to perform an accurate translation while minimizing the delay. To address this challenge, we utilize the cross-attention mechanism of the encoder-decoder architecture to find the right alignment between the input speech frames and the target text tokens. In this paper, we propose the Recent Frame Attention Policy (RFAP) and the Dual-Condition Attention Policy (DCAP) that allow offline trained speech-to-text translation models to be used in streaming scenarios without requiring additional training. Results on three different language translation pairs over the CVSS-C corpus show that the RFAP is able to surpass other policies with gains of up to 4.0 BLEU while reducing the translation delay by almost 1 second. Moreover, the DCAP is able to preserve a high translation quality when the latency is very low.","published":"2026-09-25T05:34:46Z","updated":"2026-09-25T05:34:46Z","authors":["Filip Tăşădan","Ema Tomanová","Ondrej Lopuch","Paweł Bilko","Anders Søgaard"],"categories":["cs.LG"],"headline":"New Methods Improve Live Speech-to-Text Translation Timing and Accuracy","summary":"Researchers studied how to better translate spoken language into text as the speech happens, balancing accuracy with delays. They developed two new policies that use parts of existing translation models to decide when to translate audio input, without needing extra training. Tests on multiple language pairs showed these methods can provide more accurate translations with less delay, though the findings come from controlled experiments using existing datasets."},{"id":"2609.30838v1","title":"Peer-Grounded Counterfactual Path Planning for Chronic Health Management","abstract":"Effective behavioral intervention in chronic disease management requires not a single prescription but a sequence of incremental steps, each grounded in what real, similar individuals have demonstrably achieved. Counterfactual explanation offers a natural computational route to such guidance, answering what change in behavior would have produced a better outcome. But existing methods return a target state without a route to it, guarantee no monotone health improvement along the way, and draw no evidence from peer behavior -- asking a patient to close a wide gap in one move, which is precisely the recommendation structure least likely to be attempted. We propose POROS (Peer-Grounded Optimal Routes Over States), a domain-agnostic framework rooted in Bandura's self-efficacy theory and Festinger's social comparison theory that constructs a Behavioral Progression Graph -- a directed acyclic graph over observed patient states in which every edge requires both peer-grounded behavioral proximity and strict health outcome improvement. Every edge is therefore a behavioral change that individuals in the cohort have demonstrated is achievable within a single period. Minimum-cost paths through this graph decompose otherwise inactionable behavioral gaps into incremental, peer-grounded steps. We evaluate POROS on two independent longitudinal cohorts of patients with diabetes. For patients below the 70% clinical threshold for time in range (TIR, blood glucose within 70-180 mg/dL), it reduces the mean gain required per step from 26.3 percentage points (pp) to 5.5 pp on one cohort and from 31.1 pp to 5.7 pp on the other, decomposing large behavioral jumps into the incremental steps that self-efficacy requires. Across both cohorts, 97-98% of multi-hop paths cross patient boundaries, embedding social comparison by construction.","published":"2026-09-25T05:33:08Z","updated":"2026-09-25T05:33:08Z","authors":["Saman Khamesian","Hassan Ghasemzadeh"],"categories":["cs.LG"],"headline":"New method suggests small, achievable health steps using peer data","summary":"Managing chronic diseases often needs gradual behavior changes, but existing advice can be unrealistic. Researchers created POROS, a tool that maps patient behaviors into small, achievable steps based on what similar patients have done. Tested on diabetes patients, POROS breaks large health goals into easier parts, potentially improving motivation by showing achievable progress and comparing to peers' experiences. These findings come from analyzing patient data and do not guarantee outcomes, but suggest a way to guide better health management."},{"id":"2609.30837v1","title":"MOPD-Router: Rethinking Teacher Routing in Multi-Teacher On-Policy Distillation","abstract":"Multi-teacher on-policy distillation (MOPD) integrates specialized capabilities into a single student, but existing practice typically hard-routes each prompt to a domain-matched teacher for the entire rollout. This dependence on prompt-level domain labels restricts using unlabeled training mixtures and leaves complementary signals from other teachers unused. We introduce MOPD-Router, a framework that routes supervision over the full teacher pool at each token, without domain labels or training a separate routing model. Its plug-in interface supports different metrics for selecting and weighting teacher-specific OPD signals. Within this interface, we propose ExpertAlign, which scores each teacher by whether its correction to the student at the current token expresses the specialization that teacher acquired during post-training, and compare it against two reference metrics built on teacher confidence (Entropy) and teacher-student discrepancy (Novelty). Experiments on unlabeled and domain-labeled training mixtures under strong-to-weak and same-size distillation scenarios show that ExpertAlign achieves the strongest overall performance in all four settings. On unlabeled data, it improves the overall score by 5.88 (+12.3%) points over Mean aggregation; on domain-labeled data, it outperforms standard MOPD by 3.95 (+7.8%) points without using available domain labels. These results demonstrate token-level routing can exploit cross-domain complementary supervision, and reduce exclusive reliance on prompt-level domain assignment. Code is available at: https://github.com/TURLEing/MOPD-Router.","published":"2026-09-25T05:32:03Z","updated":"2026-09-25T05:32:03Z","authors":["Tianze Xu","Yanzhao Zheng","Zhentao Zhang","Yuanqiang Yu","Chao Ma","Jihuai Zhu","Lelun Wu","Lyumanshan Ye","Pengfei Liu","Baohua Dong","Hangcheng Zhu","Ruohui Huang","Gang Yu"],"categories":["cs.LG","cs.AI"],"headline":"New method improves multi-teacher machine learning supervision","summary":"Researchers studied a way to combine knowledge from multiple specialized teacher models into one student model. Instead of assigning each task to a single matching teacher based on domain labels, their method, MOPD-Router, chooses supervision at a finer level for each word in the output, without needing domain labels or extra training. Experiments show their approach, especially with the ExpertAlign scoring method, performs better across tasks using both labeled and unlabeled data. The research highlights benefits of flexible, token-level teacher guidance but is based on specific controlled scenarios."},{"id":"2609.30836v1","title":"PTC-Decoder: Towards Intelligent SLMs on Offline Resource-Constrained Edge Devices","abstract":"Deploying small language models (SLMs) on offline, resource-constrained edge devices such as remote sensing satellites presents a fundamental challenge: their limited reasoning capacity hinders reliable execution of multi-step agent tasks requiring complex tool orchestration. Existing plan-solve paradigms rely on prompt-based enforcement, which our experiments show SLMs almost entirely disregard: weak models fail to invoke the plan. We propose PTC-Decoder (Plan-Tool Constrained Decoder), a training-free, plug-and-play decoder framework that combines (1) a Plan-to-Act paradigm, which elevates planning to an atomic tool and forces its invocation at the first inference step, and (2) TC-Decoder, a deterministic finite automaton that imposes token-level hard constraints on tool names while preserving freedom over parameter generation, thereby retaining SLM reasoning capability. Evaluated on 200 real remote-sensing satellite tasks across 7 SLMs, PTC-Decoder yields a statistically significant mean overall score gain of +1.21 (p<0.01), 95% CI [+1.13, +1.29]), with consistent improvements across models and other datasets. An ablation study that removes TC-Decoder causes substantial performance degradation across all quality metrics without reducing computational cost, confirming TC-Decoder as the primary driver. PTC-Decoder thus offers a lightweight yet effective solution for improving step-level reliability, with final-answer accuracy remaining an open challenge. In essence, we enforce plan adherence by constraining the permissible output vocabulary during inference, without requiring retraining.","published":"2026-09-25T05:30:35Z","updated":"2026-09-25T05:30:35Z","authors":["Minghui Yu","Ke Mu","Gang Wu"],"categories":["cs.AI"],"headline":"New Method Helps Small Language Models Follow Plans on Limited Devices","summary":"Researchers studied how small language models struggle to carry out multi-step tasks on offline devices like satellites due to limited reasoning. They developed PTC-Decoder, a tool that forces the model to start with a plan and restricts certain outputs to improve task following, without retraining the model. Tested on remote sensing tasks, this method improved performance but the researchers note that overall final-answer accuracy still needs improvement."},{"id":"2609.30835v1","title":"Near-infrared spectra of chemically peculiar stars and applications to kilonova spectra","abstract":"Kilonova spectra provide information on rapid neutron-capture nucleosynthesis in neutron star mergers. However, it remains challenging to identify individual elements in kilonova spectra, primarily due to incomplete atomic data for heavy elements at near-infrared wavelengths. Recent work demonstrated that spectra of chemically peculiar stars offer excellent laboratory for studying kilonova spectra. In this paper, to expand our understanding of strong transitions of heavy elements at near-infrared wavelengths, we present high-resolution near-infrared spectra of 14 chemically peculiar stars covering a range of effective temperatures, surface gravities, and abundance patterns, obtained with Subaru/IRD. The Sr II, Ce III, and Gd III lines, previously suggested to be among the strongest transitions, are detected in the spectra of most of our targets, confirming their intrinsic importance. In addition, Eu II and Nd III lines are detected in several stars. By comparing observed equivalent widths and abundance ratios, we show that under a highly heavy-element-enriched abundance pattern, Eu II and Nd III lines may contribute to kilonova spectral features comparably to Sr II. Importantly, we find no unidentified lines systematically stronger than these transitions despite the larger stellar sample. These results further support the primary importance of the Sr II and Ce III transitions and provide new insights into the potential contribution of Eu II and Nd III.","published":"2026-09-25T05:27:05Z","updated":"2026-09-25T05:27:05Z","authors":["Nanae Domoto","Masaomi Tanaka","Wako Aoki","Salma Rahmouni","Tomoyuki Kudo","Hiroki Harakawa","Teruyuki Hirano","Takayuki Kotani","Takashi Kurokawa","Masayuki Kuzuhara","Jun Nishikawa","Takuma Serizawa","Motohide Tamura","Akitoshi Ueda"],"categories":["astro-ph.SR","astro-ph.GA","astro-ph.HE","physics.atom-ph"],"headline":"Studying Star Spectra to Understand Kilonova Light Signatures","summary":"Researchers examined detailed near-infrared light spectra from 14 unusual stars to better understand the signals seen in kilonovae—explosions from merging neutron stars. They focused on specific heavy elements thought to shape kilonova light but found no new unidentified strong signals beyond those elements. Their work supports the idea that particular elements like strontium and cerium play key roles in kilonova spectra, while europium and neodymium might also contribute. These findings rely on comparing star observations and help improve interpretation of kilonova measurements."},{"id":"2609.30834v1","title":"KBSS-InCLOSE II: First Detailed Insights on the Inner CGM of Low-Mass $z\\sim2.3$ Galaxies","abstract":"We present results from an extension to the Keck Baryonic Structure Survey (KBSS) that focuses on the Inner Circumgalactic Medium (CGM) of QSO Line Of Sight Emitting galaxies at z$\\sim$2.3 (InCLOSE). We analyze two low-mass galaxies $\\log{(\\rm{M_*} / \\rm{M_\\odot})} \\leq 9$ that are within small a projected distance of a QSO $D_{\\rm tran} \\leq 50~{\\rm kpc}$ ($D_{\\rm tran} /R_{\\rm vir} \\leq 0.75$). One galaxy is detected as a bright Lyman- Emitter with Keck/KCWI (confirmed with follow-up Keck/MOSFIRE spectra), and the other as a serendipitous line emitter with Keck/MOSFIRE. Both galaxies have nebular and morphological properties consistent with those of typical low-mass z$\\sim$2.3 star-forming galaxies. Analysis of their CGM absorption as seen with Keck/HIRES spectra of the background QSOs shows no detections of low-ionization metal absorption (low-ions; e.g., Si II), ubiquitous detections of intermediate- (e.g., C IV) and high-ions (O VI), kinematically complex absorption ($\\geq$7 components per galaxy halo) spread over $|Δv|\\pm 150~{\\rm km~s^{-1}}$, and no unambiguously unbound gas is detected in either galaxy halo. We analyze the thermal properties of a subset of CGM components, finding that the majority (6/10) have temperatures consistent with heating from the metagalactic UV background ($\\log{(T/\\rm{K})_{\\mathrm{med}}}=4.0$) while the remainder possess short-lived, intermediate temperature gas that would require additional heating or rapid replenishment ($4.6 \\leq \\log{(T/\\rm{K})} \\leq 5.1$), the internal energy of these absorbers are dominated by thermal broadening, their internal (turbulent) velocities are all subsonic, while their motions through the halo are likely supersonic. These results hint that the z2.3 CGM changes with stellar mass in terms of kinematic complexity and unbound gas fraction, while thermal properties remain similar.","published":"2026-09-25T05:25:24Z","updated":"2026-09-25T05:25:24Z","authors":["Evan Haze Nunez-Cravin","Charles C. Steidel","Gwen C. Rudie","Evan N. Kirby","Charis M. Hall","Nikolaus Z. Prusinski","Zhuyun Zhuang","Yuanze Ding","Alice E. Shapley","Ryan F. Trainor"],"categories":["astro-ph.GA"],"headline":"Studying Gas Around Small Galaxies 11 Billion Years Ago","summary":"Researchers studied the gas surrounding two small galaxies seen when the universe was about 3 billion years old. Using telescope data, they found complex gas motions and mostly warm gas influenced by background radiation, with no clear signs of gas escaping the galaxies. Their work suggests that while the motion of gas varies with galaxy size, its temperature characteristics remain similar in this early universe period."},{"id":"2609.30833v1","title":"Fast Plans, Faithful Actions: Closing the Planning-Execution Gap in Hierarchical Vision-Language-Action Models","abstract":"Hierarchical vision-language-action (VLA) systems consist of a high-level vision-language planner and a low-level action expert that generates continuous actions. This hierarchical design has practical value only if the planner can generate plans fast enough to meet real-time control requirements, and the resulting plans actually contribute to the generation of action. We study one such system, a waypoint hierarchy pipeline adapted from $π_{0.5}$, and find that neither requirement is satisfied. This baseline relies on token-level autoregressive decoding (Token-AR) to generate a waypoint plan, requiring 57 very expensive vision-language model (VLM) forward passes. However, we find that erasing the waypoint endpoints has little effect on task success. Two findings reveal the misalignment of planner-executor: the planner generates outputs at an excessively fine granularity, and the executor underuses plans as a control condition. We address the latency issue with waypoint-aligned block-autoregressive decoding (Block-AR), and plan underuse issue with normalized goal modulation (NGM), a layer-wise goal path constrained by phase gating and anti-shortcut training so that the waypoint influences action generation maintaining other signals. Our method reduces the maximum number of VLM forward passes from 57 to 8 on LIBERO, including one prefix prefill, and achieves an $8.7\\times$ reduction in planning latency on a Rokae dual-arm robot. With normalized goal modulation and anti-shortcut training, Block-AR's success rate on LIBERO-Long increases from 91.0% to 96.2%, while its average success rate across the four suites increases from 95.85% to 98.45%. On three bimanual tasks with this robot, success rates remain comparable across methods.","published":"2026-09-25T05:22:23Z","updated":"2026-09-25T05:22:23Z","authors":["Chuanliang Xie","Boyu Ma","Gen Li","Yizhou Liu","Houwang Chen","Xinyu Zhou","Jianfei Yang"],"categories":["cs.RO"],"headline":"Improving Speed and Accuracy in Robot Vision-Language Planning","summary":"Researchers studied hierarchical systems where a high-level planner guides low-level robot actions using vision and language. They found current planners are too slow and generate plans that the robots barely use. To fix this, they developed faster plan generation and a method to better integrate plans into robot actions. Their approach greatly reduced planning time and improved task success rates in simulations and robotic tests, though the improvements varied by task and setting."},{"id":"2609.30832v1","title":"Subject-Invariant Cross-Modal Decoding of Perceived Speech from Brain Recordings","abstract":"Perceived speech decoding based on non-invasive brain-computer interface (BCI) signals has been extensively studied in recent years. Research in this field primarily faces two challenges: extracting neural representations with rich spatiotemporal information and achieving cross-subject generalization. Although separate studies have proposed methods to cope with these issues, a unified approach that simultaneously tackles both challenges remains lacking. To fill this gap, we propose the Subject-Invariant Cross-Modal Perceived Speech Decoding (SICMD) method, which integrates functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG). We conduct comprehensive analyses of the fusion method, fusion position, encoder architecture, and model inputs. Our results demonstrate that the proposed method improves Top-1, Top-10, and Rankacc by more than 10.6%, 10.1%, and 1.7%, respectively, compared to baseline methods in cross-subject perceived speech decoding tasks, while reducing training costs by 88.8% and 60.5% compared to multi-subject and intra-subject decoding settings. Further visualization experiments also confirm the effectiveness of our approach.","published":"2026-09-25T05:22:13Z","updated":"2026-09-25T05:22:13Z","authors":["Aoke Zhang","Jing Chen"],"categories":["cs.SD","cs.AI","eess.AS"],"headline":"New Method Decodes Heard Speech from Brain Signals Across People","summary":"Researchers studied how to decode speech that people hear using brain signals without invasive methods. They combined two brain imaging techniques to better capture brain activity and created a method that works across different individuals. Their approach improved decoding accuracy and reduced training time in tests using brain data, but the findings come from controlled experiments using these specific technologies and may not yet apply broadly."},{"id":"2609.30831v1","title":"CDBG: Causally Motivated Dual-Invariance Learning against Topological and Predictive Shifts in EEG Workload Recognition","abstract":"Generalizing Electroencephalography (EEG)-based mental workload recognition to unseen subjects remains a formidable challenge due to severe inter-subject variability. While functional brain graphs effectively model distributed cognitive dynamics, their inherent subject-specificity induces two coupled distribution shifts: a class-conditional topological shift in the underlying functional connectivity, and a predictive mechanism shift in the learned representation-to-label mapping. Motivated by the subject-induced distribution shifts, we propose CDBG, a Causally motivated Dual-invariance learning framework for Brain Graphs. CDBG disentangles and mitigates these shifts via a two-stage rationale learning pipeline. First, it employs stochastic edge masking to extract sparse, workload-predictive graph rationales, regularized by workload-conditional Laplacian spectral alignment to enforce topological invariance across subjects. Second, it applies subject-wise Invariant Risk Minimization (IRM) to the graph representations, ensuring environment-wise risk stationarity. Extensive experiments on a self-built air traffic controller EEG cognitive workload dataset and multiple public datasets under a strict leave-one-subject-out protocol demonstrate that CDBG significantly outperforms state-of-the-art cross-subject and graph-based baselines, improving the Macro-F1 score by up to 4.23%, while simultaneously providing neurophysiologically interpretable functional rationales.","published":"2026-09-25T05:22:06Z","updated":"2026-09-25T05:22:06Z","authors":["Yuzhe Zhang","Wenmin Zhou","Chengxi Xie","Kai He","Jihong Wang","Huan Liu","Man Yao","Daoqiang Zhang"],"categories":["cs.HC"],"headline":"New method improves brain workload recognition across different people","summary":"Researchers addressed a challenge in using brain signals to recognize mental workload because brain activity patterns vary a lot between individuals. They developed a new approach, called CDBG, that identifies brain connections related to workload and makes the method stable across different people. Their tests on brain data, including from air traffic controllers, showed CDBG performed better than previous methods and offered meaningful insights into brain activity. These findings are based on experiments and not yet confirmed in real-world use."},{"id":"2609.30830v1","title":"AGATE: Provenance-Based Runtime Defense Against Compositional Attacks on LLM Agents","abstract":"LLM agents can produce harmful effects through sequences of ordinary operations. Judging such actions requires establishing both the authority that permits them and the origin of the data they carry. We present AGATE, an authorization and data-provenance gate at instrumented agent-harness boundaries. Operator declarations and host approval events ground authorization; delegated actions are constrained by grants that bind to exact parameters, expire, and permit a limited number of uses. Source registration connects observed inputs to subsequent transfers, while an effect ledger tracks repeated requests. Deterministic checks make decisions without an LLM in the decision path and retain their grounds with execution evidence for forensic replay. Adapters integrate three production harnesses -- DeepSeek Harness, OpenCode, and OpenClaw -- without modifying host code, translating each host's native observation and veto points into a single shared gate interface; the judgment core is identical in all three, and only enforcement depth differs. Our evaluation combines 153 exercised attack-chain records with deployment, utility, and reconstruction experiments. The deployment observations expose how tool declarations and data checks govern business actions, including a bypass through parameter rewriting. Six of eleven benign file-processing scenarios contain denial events, revealing the utility cost of content-based provenance policies. Across 252 runs on 63 sanitized scenarios, replay agrees with live graph projections for all 63 scenarios on each of two platforms. These results establish the feasibility of provenance-based runtime judgment and identify content transformation, legitimate reuse, and observation coverage as concrete limits.","published":"2026-09-25T05:18:02Z","updated":"2026-09-25T05:18:02Z","authors":["Xiaorui Zhang","Zhuoran Cheng","Kailin Liu","Zhaoxi Sun","Shiyu Fan","Tongyu Yuan","Bin Yuan","Weizhong Qiang","Deqing Zou"],"categories":["cs.CR","cs.SE"],"headline":"New System Checks Data Origins to Block Harmful AI Actions","summary":"Researchers developed AGATE, a tool that tracks where data comes from and what permissions AI agents have before allowing them to act. It uses strict rules and logs actions to prevent misuse without relying on AI to make judgment calls. Tests with different setups showed AGATE can successfully oversee AI operations, but challenges remain with data changes, legitimate data reuse, and fully monitoring all actions. The findings show the potential and limits of using data history for real-time AI control."},{"id":"2609.30829v1","title":"Geometric Optimization Parameterized by Piercing Complexity","abstract":"Packing and covering problems for geometric regions have been studied under many notions of complexity, including VC-dimension, union complexity, shallow-cell complexity, and fatness. Although these restrictions often yield constant-factor approximation algorithms, they do not by themselves generally lead to PTASs. A recurring feature of known hardness constructions is that one region may be \\emph{pierced} by many others: a region $B$ pierces $A$ when $A\\setminus B$ is disconnected. We study geometric instances through the \\emph{piercing degree}. Since piercing is symmetric for Jordan regions, this is the maximum degree of the corresponding piercing graph. Our main result is that, for every fixed piercing degree, the standard local-search algorithms give PTASs for the unweighted \\emph{Discrete Independent Set} and \\emph{Set Cover} problems. The proof constructs a sublinear balanced separator for an appropriate locality graph and applies it adaptively throughout the recursive local-search analysis. This guarantee depends only on the piercing degree; in particular, it places no bound on the number of components created by an individual piercing pair. We also prove a polynomial shallow-trace bound depending only on the piercing degree. As consequences, for every fixed piercing degree, weighted Set Cover admits a deterministic $C_r$-approximation and weighted Discrete Independent Set admits a deterministic $O(r+1)$-approximation. These results extend the known guarantees for non-piercing families and apply, for example, to axis-parallel rectangles when every rectangle is pierced by only a bounded number of other rectangles.","published":"2026-09-25T05:15:56Z","updated":"2026-09-25T05:15:56Z","authors":["Aritra Banik","Rajiv Raman","Saurabh Ray"],"categories":["cs.CG","cs.DS"],"headline":"New approach improves solutions for complex geometric packing problems","summary":"Researchers studied geometric packing and covering problems, focusing on a concept called piercing degree, which measures how regions overlap in a way that splits others into separate parts. They showed that when this piercing degree is limited, existing local-search algorithms can provide very close approximations for two key problems: selecting independent regions and covering with minimum sets. Their proofs rely on mathematical structures called separators and apply even without limiting how many pieces each overlap creates. These findings extend previous results and offer new ways to handle shapes like overlapping rectangles. The results are theoretical and based on fixed piercing complexity."},{"id":"2609.30828v1","title":"HIRE: History-Conditioned Interaction Reasoning and High-Rate Execution for Visually Aliased Precision Manipulation","abstract":"Precision manipulation with contact-critical interactions is often history-dependent: visually similar observations can correspond to different latent interaction states and therefore require different actions, while small execution errors can alter task outcomes. Policies relying on the current visual observation alone cannot resolve such ambiguity; force-aware and memory-augmented methods enrich physical or temporal context, while reactive high-rate policies improve local contact response, yet long-horizon temporal reasoning and precision execution remain largely decoupled in existing methods, limiting reliable progression in visually aliased precision manipulation. To bridge this gap, we introduce History-Conditioned Interaction Reasoning and Execution (HIRE), a cross-rate framework comprising a history-conditioned Interaction-State Reasoner (ISR) and a high-rate Interaction-Manifold Executor (IME). ISR encodes ordered wrench history with a temporal wrench encoder and Force Perceiver as persistent physical evidence for state-consistent action generation, while IME structures contact-critical motion into intrinsic progress and transverse correction for precise execution; their cross-rate loop allows the resulting physical traces to inform subsequent reasoning. In real-robot experiments across surface, insertion, and rotational interactions, HIRE achieves at least 90% completion across all evaluated task stages while improving interaction-state disambiguation, execution precision, and generalization. More broadly, HIRE provides a unified reasoning--execution perspective on precision manipulation under history-dependent partial observability, where physical interaction both realizes task intent and reveals latent-state evidence for future decisions. Code will be released upon publication.","published":"2026-09-25T05:13:23Z","updated":"2026-09-25T05:13:23Z","authors":["Rongji Li","Wenhao He","Cewu Lu","Xingyu Chen","Xu-Yao Zhang"],"categories":["cs.RO"],"headline":"New method improves robot precision in hard-to-see tasks","summary":"Robots often struggle with tasks where similar views hide different states and tiny errors cause failures. Researchers developed HIRE, a system that uses past force information and fast precise movements together to better understand and act in these tricky conditions. Tested on real robots, HIRE improved task success rates and accuracy by combining long-term reasoning with precise execution, showing promise for more reliable manipulation where sensing is limited."},{"id":"2609.30827v1","title":"On the Compositions of Interstellar Objects","abstract":"As physical samples of distant planetary systems, the compositions of interstellar objects (ISOs) should correlate with the properties of their diverse parent stars. We combine a chemical partition model with a sample of stellar elemental abundances from APOGEE DR17 to predict the abundances of commonly-observed cometary volatiles in ISOs. We find that the compositions of ISOs vary significantly between different stars. In particular, we predict a strong correlation between an ISO's ammonia abundance and its parent star's metallicity: high-metallicity stars will create ISOs rich in ammonia. This suggests the production rates of NH$_3$'s photolytic daughter products are an ideal observational tracer for the origins of ISOs. We infer that 2I/Borisov formed around a star of near-solar metallicity ($-0.4\\lesssim[\\mathrm{Fe}/\\mathrm{H}]\\lesssim0.3$), and corroborate the velocity- and isotope-based inferences that 3I/ATLAS formed around a lower-metallicity star ($-0.8\\lesssim[\\mathrm{Fe}/\\mathrm{H}]\\lesssim0.0$). The production rates of 2I and 3I imply both only contain hypervolatiles such as CO and N$_2$ that are trapped in other less-volatile ices, similar to typical Solar System comets. Despite this, we argue that true hypervolatile-ice-rich ISOs may exist; they would exhibit high CO and N$_2$ production rate ratios, possibly similar to those of C/2016 R2 (PanSTARRS). Our predictions provide context for future ISO discoveries, opening a path to link the properties of their origin systems to observable production rates.","published":"2026-09-25T05:11:48Z","updated":"2026-09-25T05:11:48Z","authors":["Matthew J. Hopkins","Michele T. Bannister","Bertram Bitsch","Chris Lintott","John C. Forbes"],"categories":["astro-ph.EP","astro-ph.GA"],"headline":"Studying Interstellar Objects' Makeup to Trace Their Star Origins","summary":"Researchers used a chemical model and data on star compositions to predict the chemical makeup of interstellar objects (ISOs) that come from other star systems. They found that the chemicals in ISOs, like ammonia, change depending on the type of star they come from, and this might help identify where they formed. By applying their model, they estimated the star origins of two known ISOs and suggested how future observations could better link ISOs' chemistry to their home stars. Their findings are based on models and comparisons, not direct sampling."},{"id":"2609.30826v1","title":"SUCRe: Selective Uncertainty-Aware Contrastive Representation for Graph Transfer Learning","abstract":"Graph transfer learning (GTL) provides a promising paradigm for adapting knowledge from source graphs with sufficient labels to label-scarce target graphs. However, existing approaches often assume that transferred knowledge is uniformly reliable, ignoring the different transferability of samples caused by structural and distribution shifts across graphs. This limitation leads to negative transfer and unnecessary computational overhead. In this work, we propose SUCRe, a selective uncertainty-aware contrastive representation method for GTL. The key idea is to selectively adapt and transfer graph knowledge according to its estimated reliability. Specifically, we introduce structure-aware entropy-based matching discrepancy, which jointly models feature uncertainty and structural coherence to ensure accurate feature adaptation between graphs. Moreover, we develop a domain-aware semi-hard negative sampling strategy that constructs informative contrastive sets by filtering unreliable cross-domain relationships, reducing computational redundancy while enhancing representation discrimination. Extensive experiments on graph transfer benchmarks demonstrate that SUCRe achieves competitive performance with improved efficiency.","published":"2026-09-25T05:11:25Z","updated":"2026-09-25T05:11:25Z","authors":["Mingcan Wang","Junchang Xin","Zhongming Yao","Bing Tian Dai","Kaifu Long","Zhiqiong Wang"],"categories":["cs.GR"],"headline":"New Method Improves Knowledge Transfer Between Different Graphs","summary":"Researchers studied how to better transfer information from well-labeled graphs to related graphs with fewer labels, a process called graph transfer learning. They found that current methods assume all transferred information is equally reliable, which can cause errors and waste computation. To address this, they developed SUCRe, which selectively transfers knowledge based on how trustworthy it is, using techniques that measure uncertainty and structure differences. Their experiments suggest SUCRe performs well and is more efficient, though results are based on benchmark tests."},{"id":"2609.30825v1","title":"Stable Recovery and Benign Overparameterized Landscapes for Phase Retrieval from Coded Diffraction Patterns","abstract":"Coded diffraction patterns (CDPs) provide a structured and physically relevant model for phase retrieval, but the dependence among Fourier measurements generated by a common mask makes sharp stability analysis challenging. For a fixed unit-norm signal $\\boldsymbol{x}_\\star \\in \\mathbb{C}^n$, let $\\boldsymbol{X}_\\star=\\boldsymbol{x}_\\star\\boldsymbol{x}_\\star^*$, and let $\\mathcal A$ be the lifted CDP measurement operator. We prove that, with $L=O(\\log n)$ random masks, the following uniform lower isometry holds with high probability: $\\|\\boldsymbol{X}-\\boldsymbol{X}_\\star\\|_F \\lesssim \\log^2(2n) \\frac{\\|\\mathcal A(\\boldsymbol{X}-\\boldsymbol{X}_\\star)\\|_2}{\\sqrt{nL}}, \\boldsymbol{X}\\succeq\\boldsymbol{0},$ from which we derive two consequences. First, for $\\boldsymbol{y}=\\mathcal A(\\boldsymbol{X}_\\star)+\\boldsymbol{e}$, PhaseLift-type convex programs achieve the Gaussian-type stable recovery bound $\\|\\widehat{\\boldsymbol{X}}-\\boldsymbol{X}_\\star\\|_F \\lesssim \\frac{\\log^2(2n)}{\\sqrt{nL}}\\|\\boldsymbol{e}\\|_2.$ Second, in the noiseless case, the nonconvex factorized loss has a benign landscape when the factor width satisfies $r = O(\\log^5(2n))$: every second-order critical point $\\boldsymbol{V}\\in\\mathbb C^{n\\times r}$ satisfies $\\boldsymbol{V}\\boldsymbol{V}^*=\\boldsymbol{X}_\\star$. The key ingredient is a uniform operator-norm bound over row subsets of the dependent CDP measurement matrix, which permits the removal of a controlled set of adaptively selected rows while preserving tangent injectivity.","published":"2026-09-25T05:10:32Z","updated":"2026-09-25T05:10:32Z","authors":["Jian-Feng Cai","Zhibo Jin","Tong Wu","Ruizhe Xia"],"categories":["cs.IT"],"headline":"New Insights into Reliable Phase Retrieval Using Coded Diffraction Patterns","summary":"The researchers addressed the challenge of recovering signals from coded diffraction patterns, a method important in imaging, which involves complicated dependencies among measurements. They proved mathematical properties that ensure stable recovery of the original signal using both convex and certain nonconvex optimization methods, under specific conditions related to the number of masks used. Their findings are theoretical and rely on high-probability guarantees, focusing on the mathematical landscape that makes recovery feasible and stable in idealized settings without noise or with controlled noise."},{"id":"2609.30824v1","title":"Crypto-bound identity-verified capability tokens for coordinating distributed AI agents: A proposal","abstract":"The prospect of fully autonomous transactional agents did not appear on the horizon until the advent of high capability language models. With such models, the operational benefits of adaptive task orchestration and independent (but constrained) decision making are tantalizing for enterprises and individuals alike. However, each such agent carries with it a serious attack surface in the form of prompt injection which can compromise any soft \"guard rails\" that may have been placed in context. The consequences of these attacks include credential ex-filtration which, if left unmitigated, renders the whole category of such agents unusable due to breach of trust. Furthermore, expecting a growth of autonomous agents, a security framework for them would require a form of decentralization to scale. Drawing on the proposed OAuth Agent Authorization Profile and the W3C DID and VC standards, we propose a framework based on the principles of capability based security with decentralized agent identity whereby agents access services based on tokens that are cryptographically bound to the agent's and issuer's identities and specify their scope. Services can validate that the delegation chain only involves scope attenuation before acting on any given token. We show that such a layer that lives outside the language model's context window in a secure module can enable agents to act within enforceable security boundaries.","published":"2026-09-25T05:07:28Z","updated":"2026-09-25T05:07:28Z","authors":["Srikumar Subramanian","Shubhashis Sengupta"],"categories":["cs.CR"],"headline":"New Crypto Tokens Aim to Secure Autonomous AI Agents","summary":"Researchers propose a security system for autonomous AI agents using cryptographic tokens tied to each agent's verified identity. These tokens limit what the agents can do and prevent attacks like stealing credentials through manipulated instructions. Their approach builds on existing security standards and aims to create a scalable, decentralized way to ensure AI agents operate safely within strict boundaries. The results are theoretical, outlining a framework rather than practical implementation."},{"id":"2609.30823v1","title":"Mishiba's Conjecture on the Coaction of $\\infty$-adic Multiple Zeta Values","abstract":"We study multiple zeta values in positive characteristic. We construct, using $\\infty$-adic multiple zeta values, a coaction of the $\\infty$-adic multiple zeta value algebra modulo $ζ_A(q-1)$ on the $\\infty$-adic multiple zeta value algebra itself, and prove that it agrees with Mishiba's coaction constructed via special values of Carlitz multiple polylogarithms. We also determine the coaction on the $\\star$-inverse values and show that the antipode on the $\\infty$-adic multiple zeta value algebra modulo $ζ_A(q-1)$ is given by the $\\star$-inverse operation. In particular, these results prove Mishiba's conjecture concerning the coaction and the antipode for $\\infty$-adic multiple zeta values.","published":"2026-09-25T05:05:48Z","updated":"2026-09-25T05:05:48Z","authors":["Hung-Chun Tsui"],"categories":["math.NT"],"headline":"Researchers Solve Mishiba's Conjecture on Special Mathematical Values","summary":"The researchers investigated special number patterns called multiple zeta values in a particular mathematical setting known as positive characteristic. They built and analyzed an operation called a coaction on these values and confirmed that it matches an earlier construction by Mishiba. Additionally, they studied related inverse operations and demonstrated specific properties, thereby proving Mishiba's conjecture about these mathematical structures. Their findings are theoretical and involve advanced algebraic concepts related to multiple zeta values."},{"id":"2609.30822v1","title":"Adaptive Interaction Graphs for Particle Simulation","abstract":"Learned particle simulators based on graph neural networks achieve strong one-step accuracy, but errors compound over long horizons. An underexplored variable is the interaction graph: existing methods fix its topology via k-nearest neighbors or a static radius rule, regardless of local model confidence. We propose making this graph adaptive: a per-particle variance head, trained jointly with the acceleration head under a heteroscedastic Gaussian NLL loss, drives a trajectory in which high-uncertainty particles receive an expanded neighborhood. This is done at little extra inference cost by using the previous step's uncertainty estimate. A key discovery is that the variance head learns a meaningful notion of uncertainty: high-variance particles concentrate near complex regions, such as splash zones or free surfaces. When this signal drives graph topology, the resulting AdaptGNS simulator achieves a strict Pareto improvement on WaterDrop and a modest gain on Sand. Given the model's stronger performance on WaterDrop, we hypothesize that adaptive graphs are most useful when complexity is concentrated in space. Our code can be found at https://github.com/aidenzhou8/AdaptGNS.","published":"2026-09-25T05:02:48Z","updated":"2026-09-25T05:02:48Z","authors":["Aiden Zhou"],"categories":["cs.LG","cs.CE"],"headline":"Improving Particle Simulations with Adaptive Interaction Graphs","summary":"Researchers studied particle simulations that use graphs to represent interactions between particles. They found that fixing these graphs in advance can cause errors to build up over time. To address this, they developed a method that adjusts the graph based on uncertainty estimates, expanding connections for particles in complex areas. Their approach, AdaptGNS, showed better performance in simulating water droplets and some improvement with sand, suggesting adaptive graphs help most when complexity is localized. These findings are based on simulations and model results reported by the researchers."},{"id":"2609.30821v1","title":"Bounding extreme events in dynamical systems over an infinite time horizon","abstract":"This work addresses the computation of extreme values of observables along trajectories of dynamical systems over an \\ks{infinite time horizon}. The setting is intrinsically global: trajectories may exhibit large transient excursions, approach their largest values only asymptotically as time tends to infinity, or escape to infinity in finite time. The proposed framework treats these \\mk{ scenarios} in a unified way, so that the resulting extreme values reflect both the transient and the asymptotic behavior of the dynamics. The main difficulty is that neither the time interval nor the state space is assumed to be compact. We address this by combining a radial compactification of the state space with a compactified clock variable and discounted occupation measures. This yields a description of trajectory ensembles on compact sets via occupation measures and the associated Liouville's equation. Extreme values are then characterized through a family of reachability problems for level sets of the observable, which can be solved by a bisection procedure. On the theoretical side, we prove a superposition principle for an infinite-horizon Liouville's equation with a time-distributed terminal measure. This result shows that the measure formulation represents ensembles of stopped admissible trajectories, including trajectories \\ks{that never stop}, and establishes equivalence between the trajectory-level and measure-level descriptions.","published":"2026-09-25T04:58:29Z","updated":"2026-09-25T04:58:29Z","authors":["Karolína Sehnalová","Milan Korda"],"categories":["math.OC","math.DS"],"headline":"New method estimates extreme behavior in complex systems over infinite time","summary":"The researchers focused on finding the most extreme values that can occur in evolving systems observed over an unlimited time period. They developed a mathematical framework to handle situations where system states can change dramatically, approach extremes slowly, or even run off to infinity. Their approach uses advanced techniques to represent system behavior compactly and solves related problems to identify these extremes, supported by theoretical proofs linking different mathematical descriptions of system trajectories."},{"id":"2609.30820v1","title":"Quantizing Looped Transformers: Feedback Exposure and Calibration Blindness","abstract":"Looped transformers reuse weights across recurrence steps, making low-bit quantization especially attractive. We identify two distinct failure modes of standard post-training quantization. On Huginn-3.5B, per-channel INT4 fails primarily at the non-residual loop-entry adapter, while quantizing the residual core is much less damaging. We call this feedback exposure: a quantized layer perturbs the recurrent state without an identity path, and the resulting error is fed back at later steps. Controlled experiments on linear filters and Mamba state-space models show that feedback exposure also occurs outside transformers. Grouped INT4 reveals a separate failure, calibration blindness: our one-step GPTQ baseline builds its Hessian from step-0 activations, leaving input directions used later in the recurrence nearly unweighted. Across nine checkpoints from seven looped architectures, one-step GPTQ is worse than round-to-nearest (RTN) on the primary task metric for five checkpoints. Accumulating the GPTQ Hessian across recurrence steps outperforms both one-step GPTQ and RTN on all nine checkpoints and recovers bf16-level accuracy on Huginn. These results separate two questions for PTQ on looped models: where quantization error enters the recurrence, and which states calibration sees.","published":"2026-09-25T04:57:33Z","updated":"2026-09-25T04:57:33Z","authors":["Nux Li"],"categories":["cs.LG","cs.CL"],"headline":"Improving Quantization Accuracy in Looped Transformer Models","summary":"Researchers studied how low-bit quantization affects looped transformers, models that reuse weights repeatedly. They found two main problems: 'feedback exposure,' where quantization errors accumulate without correction in certain layers, and 'calibration blindness,' where standard calibration methods miss important state changes over time. By adjusting calibration to consider multiple steps, they improved accuracy across various models. Their findings come from controlled experiments and simulations, highlighting challenges in quantizing looped transformers and suggesting more effective calibration methods."},{"id":"2609.30819v1","title":"Learning Provable Neural Network Observer for Uncertain Dynamical Systems","abstract":"In many safety-critical applications, control of uncertain dynamical systems relies on observers that estimate states and external disturbances. Neural network observers can improve estimation accuracy, but certifying their Lyapunov stability via Linear Matrix Inequality (LMI) constraints leads to large-scale semidefinite programs (SDPs) that are difficult to solve for large networks. To overcome this scalability bottleneck, we propose a novel two-stage training framework for provably stable neural network observers. Our approach decouples the optimization into a point-guided Lyapunov pre-training phase, which rapidly achieves high estimation accuracy and local stability over sampled states, followed by an LMI fine-tuning phase that efficiently satisfies a strict global Lyapunov stability certificate. We provide formal theoretical guarantees for local stability radii and probabilistic coverage over a prescribed compact error-state domain under specified regularity and sampling assumptions. Experiments on nonlinear control benchmarks and X-29 aircraft ablations show that our LMI-certified neural network observers train significantly faster than direct LMI-based methods and generalize robustly across diverse systems, achieving improved tracking accuracy over a range of observer baselines. The code is available at https://github.com/Berry-Myon/LearningNeuralNetworkObserver.","published":"2026-09-25T04:56:18Z","updated":"2026-09-25T04:56:18Z","authors":["Zhangyi Wang","Jiaxu Liu","Chen Song","Chao Xu","Shengze Cai"],"categories":["cs.LG","math.OC"],"headline":"New Method Improves Safety of AI State Estimators in Control Systems","summary":"Researchers studied how to make neural network based observers reliably estimate states and disturbances in uncertain systems, which is important for safety. They developed a two-step training method that first learns accurate local estimates quickly, then fine-tunes to ensure global stability using mathematical guarantees. Their approach is faster to train and performs better than previous methods in tests on control problems and aircraft models, though results are based on theoretical guarantees and simulations."},{"id":"2609.30818v1","title":"Evaluation Is All You Need for Multi-Modal Autonomous Driving","abstract":"Multi-modal planning is promising for autonomous driving by representing multiple plausible behaviors in ambiguous and long-tail scenarios. Existing methods mainly focus on improving trajectory multi-modality, enhancing trajectory representations, or reshaping the candidate distribution. Nevertheless, we identify a pronounced generation-evaluation asymmetry in multi-modal planning: despite strong oracle performance, existing planners often fail to reliably select the best available candidate, leaving substantial planning potential unrealized. To address this challenge, we propose iDriveVLA, a multi-modal planning framework that improves the candidate trajectory space while enabling more reliable and context-aware trajectory evaluation. Specifically, iDriveVLA introduces a unified trajectory evaluator comprising a Safety-aware Scorer for quality and risk estimation, together with a VLM-guided Modulator for scene-adaptive criterion weighting. We further develop an oracle-aligned progressive training strategy consisting of candidate imitation pretraining, candidate space refinement, and semantic ranking alignment. On the public NAVSIM v1 leaderboard, iDriveVLA achieves a new state-of-the-art performance of 94.95 PDMS, surpassing the human-expert reference.","published":"2026-09-25T04:56:15Z","updated":"2026-09-25T04:56:15Z","authors":["Zeyu He","Shiqi Liu","Ke Chen","Yun Yan","Jinzi Wu","Dianqiao Lei","Sirui Wang","ShuRui Peng","Tao Chen","Zhuo Huang","Yu Wu","Yadong Shao","Zhichao Li","Ke Sun","Yang Guan","Keqiang Li","Shengbo Eben Li"],"categories":["cs.RO","cs.AI"],"headline":"New Method Improves Decision-Making for Self-Driving Cars","summary":"Researchers studied how autonomous driving systems choose the best path from many options, finding current methods struggle to pick the safest and most suitable routes despite generating good candidates. They developed iDriveVLA, a system that better evaluates and adapts to different driving scenes by scoring safety and context, and uses a careful training approach. Tests on a public benchmark showed iDriveVLA performs better than previous models and even expert human drivers in selecting the best driving paths. This work is based on simulations and evaluations on existing datasets."},{"id":"2609.30817v1","title":"Quadratic bounds for uncompletable words and matrix mortality","abstract":"Every finite nonempty incomplete uniquely decipherable code with maximum word length $k$ has an uncompletable word of length at most $4k^2-3k$. The bound is independent of the number of codewords and their total length. Deleting a complete codeword cycle gives a finite path-counting identity; Kraft equality then supplies a short word of deficient compressed mass. Cyclic averaging and padding turn it into an uncompletable word. Conditional expectation makes the construction polynomial-time and also decides completeness. First-return words extend the bound to mortal families of nonnegative integer $n\\times n$ matrices with joint spectral radius at most one, provided every strongly connected component has a vertex meeting every cycle. Such a family has a zero product of length at most $4n^2-3n$. A binary partial deterministic family with $2k-1$ states has shortest zero product of length $k^2+k-1$, establishing the optimal quadratic order. The bounds and the explicit-code algorithm, including its polynomial work bound, are proved in Lean.","published":"2026-09-25T04:54:22Z","updated":"2026-09-25T04:54:22Z","authors":["Rahul Chandelkar","Samrath Chadha"],"categories":["cs.FL","math.CO"],"headline":"Researchers find quadratic length limits in special coding and matrix problems","summary":"The researchers studied certain code systems where some sequences can't be completed and special sets of matrices that multiply to zero. They proved these difficult-to-complete sequences or zero product matrix sequences are no longer than roughly a constant times the square of a key parameter. They developed a polynomial-time method to find these sequences and confirmed their results mathematically using the Lean proof system. Their work helps understand limits on sequence lengths in these theoretical problems, but practical applications are not discussed."},{"id":"2609.30816v1","title":"Pathways of early evolution from the perspectives of a riboreplisome -- the ultimate RNA machine of life","abstract":"Translation of the genetic code into proteins is universal across life, with ribosomes serving as the ancient molecular machines enabling this information transformation. We propose that early replication and protein biosynthesis were inseparable functions performed by a single ancient RNA molecule: the riboreplisome. This hypothesis addresses fundamental challenges facing RNA world, protein world, and RNA-protein co-evolution theories -- particularly the problem of non-Darwinian starting points and the interdependence of replication and translation machineries. We present a 12-step evolutionary pathway from simple RNA replicase to proto-ribosome, supported by quantitative modelling demonstrating that only two steps require non-Darwinian transitions while the remaining ten proceed through standard selection. The single-molecule riboreplisome architecture provides an approximately 6-fold evolutionary advantage over multi-component \"RNA soup\" scenarios, principally by maintaining genetic linkage and retaining catalytic products close to the replicating genome (resisting diffusion loss); structural resistance to parasitic \"cheater\" molecules follows as a corollary. The riboreplisome framework explains the origins of ribosomal, transfer, and messenger RNA as derivatives of a single ancestral molecule, while providing a driving force for cellular compartmentalisation. Although the original riboreplisome is likely lost to deep time, molecular remnants may persist in extant biological RNA. Identification or reconstruction of riboreplisome-like molecules could advance both origin-of-life research and synthetic biology applications.","published":"2026-09-25T04:53:20Z","updated":"2026-09-25T04:53:20Z","authors":["Alice Cleynen","Archa H. Fox","Nikolay E. Shirokikh"],"categories":["q-bio.PE"],"headline":"Ancient RNA Machine May Explain Life's Early Evolution Steps","summary":"Researchers propose that an ancient RNA molecule called the riboreplisome combined early genetic replication and protein production, solving key puzzles about how life began. They developed a 12-step evolutionary model supported by computer simulations, showing most steps followed natural selection, with few needing other explanations. This single-molecule system could have helped protect genetic material and set the stage for cells. While this idea is theoretical, it offers a new way to understand early life and could inform future research."},{"id":"2609.30815v1","title":"Engineering Ferromagnetism in Wide Bandgap w-AlN for Spintronic Applications: Insights from DFT Calculations","abstract":"Spintronics drives the search for functional materials, pursuing room-temperature dilute magnetic semiconductors to realize practical, energy-efficient devices. This paper presents a theoretical investigation into the magnetic properties of w-AlN doped with Cr, Ru, and Rh atoms, using spin-polarized density functional theory calculations with supercell models. Calculations of point-defect formation energies as a function of the Fermi level predict that Cr4+, Ru4+, and Rh3+ are the most favorable charge states substituting Al in w-AlN. With these preferred charge states, Cr-doped AlN is stable in the ferromagnetic state, preferable for spintronic devices, across a wide concentration range (1.85 to 16.67% of Al). Conversely, Ru- and Rh-doped AlN are unstable in the ferromagnetic state relative to the antiferromagnetic state. Further investigation into the electronic density of states reveals a fascinating evolution: for Cr concentrations below 5.56%, the Fermi level resides within the band gap directly above the valence band maximum, keeping the system insulating. It transitions to a high-spin half-metallic state between 7.40% and 12.96%, and finally transforms into a normal metal at 16.67% Cr. This behavior is absent in the DOS of the antiferromagnetic model of Ru- and Rh-doped systems, where insulating and metallic behaviors instead appear non-sequentially with varying concentrations.","published":"2026-09-25T04:52:11Z","updated":"2026-09-25T04:52:11Z","authors":["Chinnappan Ravi"],"categories":["cond-mat.mtrl-sci"],"headline":"Theoretical Study of Magnetism in Aluminum Nitride for Spintronics","summary":"Researchers used computer simulations to study how adding certain metal atoms (chromium, ruthenium, rhodium) affects the magnetic properties of aluminum nitride, a wide bandgap material. They found that chromium doping can create a stable ferromagnetic state, potentially useful for spintronic devices, while ruthenium and rhodium do not favor such magnetism. The study reveals how the material’s electronic behavior changes with chromium concentration, but these insights are based on theoretical models and calculations rather than experiments."},{"id":"2609.30814v1","title":"SeA-RVINS: Semantic-Aware Tightly Coupled RTK-Visual-Inertial System with Correlation-Preserving Robust Estimation for Urban Navigation","abstract":"Reliable absolute pose estimation in urban environments is undermined by outlier measurements and incorrect temporal associations that can persist in tightly coupled estimators. Global Navigation Satellite System (GNSS) observations provide globally referenced measurements but are prone to multipath effects. Visual-inertial sensing supplies local motion constraints, but false visual associations can corrupt the estimator. We present SeA-RVINS, a fixed-lag factor-graph Real-Time Kinematic (RTK) visual-inertial system for robust urban pose estimation. A semantic-aware learned stereo frontend rejects unreliable tracks before persistent landmarks enter the graph. For double-differenced GNSS measurements, SeA-RVINS applies Dynamic Covariance Scaling through configurable batch, scalar, and latent-pivot robust formulations while retaining the shared-pivot correlation structure. We propose a hybrid ambiguity-continuation strategy that shares one ambiguity state over short arcs with verified continuity and softly links successive arcs through random-walk factors. On an approximately 20-km route from the public TEX-CUP dataset, including about 50\\% deep-urban driving, the latent-pivot configuration achieves 100\\% availability and a 1.6-m maximum horizontal error, with 96.16\\% and 99.90\\% of epochs below 1.0 and 1.5 m, respectively. The implementation is released as open-source software","published":"2026-09-25T04:49:11Z","updated":"2026-09-25T04:49:11Z","authors":["Wang Hu","Bo Wu"],"categories":["cs.RO","eess.SY"],"headline":"New System Improves Urban Positioning Using GPS and Visual Data","summary":"Researchers developed SeA-RVINS, a system that combines satellite navigation and visual-inertial data to better estimate location in cities. The method uses smart software to ignore unreliable visual information and improved math techniques to handle noisy GPS data affected by buildings. Tests on a 20-kilometer urban route showed the system was reliable and usually accurate within 1.5 meters. These results come from researchers' simulations and experiments, and the software is available for public use."},{"id":"2609.30813v1","title":"A Benchmark and Diagnostic Study of Epistemic Admission in Shared Agent Memory","abstract":"Evaluating claim admission in shared agent memory is challenging because repeated claims may be mistaken for independent evidence. An agent may copy or paraphrase a retrieved belief, while admitting a false claim exposes subsequent agents to it. To study this problem, we introduce the Correlated Promotion Benchmark (CPB), which evaluates whether candidate claims should be admitted to shared memory.CPB-Static constructs a frozen test split from publicly annotated sources with fixed gold actions. CPB-Live runs multi-agent teams over a shared store, records all writes and retrievals, and tracks source lineage defined by each scenario. A separate consumer answers from the store alone. We evaluate eight admission policies across four agent families. Our results show that policies which deduplicate sources reject many true claims alongside false ones, whereas policies preserving answer coverage admit nearly as many false claims as unrestricted sharing. Gating on declared source type reduces false adoption to 0.06--0.09, compared with 0.22--0.47 for other answering policies. Once an uncontested false belief enters memory, the consumer asserts it in 0.97--0.99 of probes across all families. No non-oracle policy consistently rejects false claims across verbatim copies, paraphrases, and paraphrases declared authoritative. These findings reveal the limitations of admission policies without access to source lineage.","published":"2026-09-25T04:48:31Z","updated":"2026-09-25T04:48:31Z","authors":["Xiaoyang Li","Yiqi Wang","Chencheng Zhu","KE XU","Wencheng Yang","Zequn Sun","Pingan Song","Yiqun Duan","Taotao Cai"],"categories":["cs.AI"],"headline":"Studying How Agents Decide Which Claims to Remember","summary":"Researchers investigated how groups of artificial agents decide which claims to add to their shared memory, a tricky problem because repeated claims might not be independent evidence. They created a benchmark to test different rules agents use to accept claims and tracked how false claims spread. Their study found that current policies either reject many true claims or allow many false ones, and once a false claim gets in, it is nearly always repeated. The study highlights limits in current methods without detailed source tracking."},{"id":"2609.30812v1","title":"Dynamic Service Recommendation with Congestion-Dependent Joining: Near-Optimal and Constant-Factor Approximation Algorithms","abstract":"Modern service platforms often provide customers with real-time congestion information, such as anticipated waiting times, before they decide whether to use a service. This creates an intertemporal tradeoff in service recommendation: directing a customer to a service may generate immediate value, but the resulting congestion can make that service less attractive to future customers. We study this tradeoff through a finite-horizon stochastic optimization problem in which a platform dynamically recommends among multiple services, each represented by a queue, with customers' joining probabilities decreasing with congestion. We first formulate the problem as a Markov decision process and show that optimal decisions generally depend on the joint congestion state, giving rise to a prohibitively large state space. We develop two complementary approximation algorithms that overcome this curse of dimensionality. First, we provide a quasi-polynomial-time approximation scheme that uses truncation and rounding of the model primitives to compress the joint congestion state, while carefully accounting for how these approximations affect the stochastic evolution of the system to establish near-optimality. Second, we develop a polynomial-time LP-guided algorithm that replaces the joint-state representation with service-level marginal information and achieves a (1-1/e) approximation guarantee under substantially more general service and joining dynamics. Finally, we show that monotone joining behavior, whereby customers become less likely to join as a service becomes more congested, marks a fundamental tractability boundary: without monotonicity, the problem is NP-hard to approximate within any constant factor. This establishes monotonicity as a key behavioral property that enables tractable dynamic service recommendation despite the complexity introduced by congestion-dependent joining.","published":"2026-09-25T04:48:18Z","updated":"2026-09-25T04:48:18Z","authors":["Yi-Chun Akchen","Sena Asli Bozkurt","Chen-An Lin"],"categories":["math.OC"],"headline":"Improving Service Recommendations Considering Wait Times and Congestion","summary":"The researchers address how online platforms can recommend services to users when wait times and congestion affect user decisions. They model this as a complex decision problem and propose two algorithms to approximate the best recommendations efficiently, even when many services and customers are involved. Their findings highlight that assuming users join less when services are busier is crucial for making the problem manageable. These results come from mathematical models and simulations, not real-world testing."},{"id":"2609.30811v1","title":"Counterfactual Online Conformal Prediction Under Adaptive Logging","abstract":"Online conformal prediction can fail when predictions shape actions and actions determine which outcomes enter calibration. Standard adaptive methods may retain marginal coverage while systematically miscovering the counterfactual outcomes of rarely selected actions. This paper formalizes the failure through counterfactual coverage and introduces Propensity-Weighted Online Conformal Prediction, an inverse-propensity-weighted recursion that debiases calibration. A doubly robust variant further reduces nuisance bias to the product of outcome-model and propensity errors. Under positivity, the resulting coverage rate matches an information-theoretic lower bound up to logarithmic factors. Experiments on synthetic decision tasks, open bandit data, and financial rebalancing show that PW-OCP and DR-OCP improve counterfactual coverage and downstream regret without sacrificing prediction-set sharpness.","published":"2026-09-25T04:43:37Z","updated":"2026-09-25T04:43:37Z","authors":["Xinyu Qiao","Yichen Lin","Kaihong Ji","Xue Wang","Tao Yao"],"categories":["cs.LG"],"headline":"Improving prediction accuracy when actions affect data collected","summary":"Researchers studied how predicting outcomes online can go wrong when the predictions influence which actions are taken, affecting the data collected for calibration. They developed new methods that adjust for this bias by weighting data inversely to the likelihood of actions, improving the accuracy of predictions about rarely chosen actions. Their tests with simulations and real data suggest these methods better cover possible outcomes and reduce errors, though the results rely on certain technical conditions and focus on theoretical guarantees."},{"id":"2609.30810v1","title":"The conductance filtration: sink-relative persistence and exact edge sensitivity for heat transport in particle-filled composites","abstract":"We formulate the thermal conductance $g_e$ of a particle pair as three branch types -- contact, near-contact and bond -- within a single construction, and we introduce the filtration that takes this $g_e$ as its axis (the conductance filtration) together with persistence taken relative to the sink boundary. We then show that the dissipation share is exactly the logarithmic sensitivity of the effective conductance, and that a finite change of a single edge is predicted exactly in closed form. The allocation of the dissipation among branch types, and the idealisation ratio of each mechanism (interface, matrix and particle), are thereby obtained as diagnostic quantities on one and the same graph representation.","published":"2026-09-25T04:40:52Z","updated":"2026-09-25T04:40:52Z","authors":["Daisuke Yasufuku"],"categories":["cond-mat.mtrl-sci","cond-mat.dis-nn","cs.CE"],"headline":"New method analyzes heat flow in particle-filled materials","summary":"The researchers developed a way to categorize heat conductance between particles into three types and created a method called the conductance filtration to study this. They linked how energy is lost (dissipated) to the sensitivity of overall conductance and found exact formulas to predict effects of changing individual connections. Their approach provides detailed diagnostics on heat transfer behavior in composite materials, but the findings are theoretical and based on mathematical analysis rather than experiments."},{"id":"2609.30809v1","title":"Deep-Learning Solvers and Surrogates for Infinity and p-Laplace Problems","abstract":"We investigate the use of neural network solvers for infinity and $p$-Laplace problems, which are fundamental in nonlinear analysis and have practical applications. Our approach employs Physics-Informed Neural Networks (PINNs) and Deep Operator Networks (DeepONets) to address computational challenges associated with large $p$ values, ranging from $2$ to $1000$, on various 2D and 3D domains. Our method offers advantages over traditional physics-based solvers, especially in three dimensions where mesh-based solvers become very costly for these problems. We also establish conditional convergence results for PINN approximations of both problems and a universal approximation result for DeepONet on the parametric $p$-Poisson problem. We demonstrate the effectiveness of these neural network solvers through numerical experiments and compare their performance with conventional methods.","published":"2026-09-25T04:39:09Z","updated":"2026-09-25T04:39:09Z","authors":["Tak Shing Au Yeung","Ka Chun Cheung","Hannah Potgieter","Steven J. Ruuth","Simon See"],"categories":["math.NA","cs.LG"],"headline":"Using Neural Networks to Solve Complex Mathematical Problems","summary":"Researchers explored neural network techniques to solve difficult mathematical problems called infinity and p-Laplace problems, which have many practical uses. They used special neural networks to handle large parameter values in 2D and 3D spaces, showing their method can be more efficient than traditional approaches, especially in 3D where standard methods are costly. The researchers also provided mathematical evidence supporting the reliability of their neural network solutions and tested their approach through computer experiments compared to conventional methods."},{"id":"2609.30808v1","title":"Optimal Sample Exponents for Direct Discrete-Gaussian SVP Search on Haar Random Lattices","abstract":"We determine the optimal sample exponent for direct discrete Gaussian SVP search on Haar random unimodular lattices, counting zero outputs. The output is one sampled vector, optionally divided by the greatest common divisor of its lattice coordinates. For every fixed approximation factor $1\\leqγ<\\sqrt e$, the exponent is $γ^2/(2e)-\\logγ$ in natural logarithmic units; it is zero for $γ\\geq\\sqrt e$. For exact SVP this gives $0.2653689\\ldots$ in base two. The converse permits arbitrary positive widths chosen from the lattice and all previous outputs, and a fixed width attains every exponent above the threshold. Aggarwal, Dadush, Regev, and Stephens-Davidowitz already give a width within a factor two of optimal for exact SVP on each lattice. We determine the explicit Haar typical rate and show that primitive reduction preserves it. An explicit finite dimensional converse controls all widths simultaneously, including recovery from long multiples; together with finite attainment bounds, it yields query guarantees in both directions. The proof uses random lattice moments and a pointwise Gaussian bound optimized over the width. We account for sampling error and separate sample requirements from generation costs in comparisons with the random lattice search of Pouly and Shen and later algorithms.","published":"2026-09-25T04:36:06Z","updated":"2026-09-25T04:36:06Z","authors":["Masahiro Kaminaga"],"categories":["math.PR"],"headline":"Researchers find best sampling rates for a lattice search method","summary":"The researchers studied how to efficiently find short vectors in complex lattice structures using a specific random sampling method. They determined the optimal number of samples needed for different approximation levels and showed how these rates apply typically for random lattices. Their results are theoretical, based on mathematical analysis and simulations, clarifying limits and performance of this approach compared to others in lattice-based problems."},{"id":"2609.30807v1","title":"Front-based construction of quantum droplets, bubbles, and hole states in a Bose mixture","abstract":"We study nonlinear holes, bubbles, kink fronts, and quantum-droplet-like states in an elongated Bose mixture governed by an extended Gross-Pitaevskii equation with attractive cubic mean-field interactions and repulsive quartic Lee-Huang-Yang contribution. We classify the constant-amplitude backgrounds: the Bogoliubov-de Gennes spectrum shows that the upper branch is modulationally stable, whereas the lower branch is unstable to long-wavelength perturbations. Grand-potential and pressure determine the vacuum-finite-density coexistence point and the unique background supporting stationary kink and antikink fronts. Symmetric holes on the stable background have a density deficit and integral width that diverge near coexistence as the hole separates into two nonlinear fronts. A multiplicative kink-antikink construction connects bell-shaped states to broad flat-top droplets and approaches the stationary quantum-droplet family as the front separation increases. An antikink-kink sum with equal signs generates bubble-like density depletions without a phase jump, whereas their difference produces dark-hole-like profiles with a $π$ phase difference between the asymptotic backgrounds. Bubble depth increases with front separation, distinguishing gray-like and dark-like profiles. Real-time simulations show that deep bubbles persist under weak random perturbations, demonstrating finite-time robustness without establishing spectral stability. In the asymmetric two-component model, localized relative-density modes can induce spinor-driven breakup on a modulationally stable background, while robust holes persist near the kink limit. Collisions of kink-antikink droplet-like states display phase-dependent merging, asymmetric particle transfer, and effective reflection. These results connect nonlinear fronts with bubbles, phase-jump holes, and self-bound droplets in beyond-mean-field Bose mixtures.","published":"2026-09-25T04:35:22Z","updated":"2026-09-25T04:35:22Z","authors":["Sherzod R. Otajonov","Fatkhulla Kh. Abdullaev"],"categories":["cond-mat.quant-gas","nlin.PS"],"headline":"Researchers explore waves and droplets in special quantum gas mixtures","summary":"The researchers studied various wave patterns and droplet-like states in mixtures of ultra-cold gases called Bose mixtures, using advanced mathematical models. They identified stable and unstable backgrounds and described how different wave structures form and interact, including bubbles and holes with or without phase changes. Real-time simulations showed some droplet and bubble states are robust to small disturbances, but this does not prove full stability. Their findings theoretically connect complex wave fronts and quantum droplets in these gases, based on mathematical and simulation analysis."},{"id":"2609.30806v1","title":"An inverse scattering problem for the nonlinear Schrödinger equation under partial harmonic confinement","abstract":"We study an inverse problem for a nonlinear Schrödinger equation with a harmonic potential in some spatial directions and free propagation in the remaining directions. The nonlinearity has an unknown power and a real, spatially dependent coefficient. For small incoming data in a weighted energy space, we construct the wave operator and show that it determines both the power and the coefficient when the latter depends either on the confined variables or on the free variables. The proof uses the first nonlinear term of the wave operator and concentrated product data. We explain why dependence on the free variables requires a stronger lower bound on the power. We also discuss the outgoing wave operator, the small-data scattering operator, and coefficients depending on all spatial variables.","published":"2026-09-25T04:34:45Z","updated":"2026-09-25T04:34:45Z","authors":["Pranav Kumar","Li Li"],"categories":["math.AP"],"headline":"Researchers explore identifying unknowns in a complex wave equation","summary":"The researchers studied a mathematical problem related to a nonlinear wave equation that behaves differently in certain spatial directions. They focused on figuring out unknown factors in the equation, such as a power and a spatially varying coefficient, using small initial inputs. The team showed how these unknowns can be determined under specific conditions, relying on detailed analysis of the wave behavior. Their findings are theoretical and based on mathematical proofs rather than experimental data."},{"id":"2609.30805v1","title":"XPhysICS: Cross-Physical-Domain Threat Grounding for Industrial Control Systems Security","abstract":"Industrial control system (ICS) threats documented for one plant can express cyber-physical effects relevant to another, but semantic similarity alone does not establish whether those effects are structurally admissible or evaluable on a target. We present XPhysICS, a provenance-aware, target-conditioned method that separates analyst-guided source abstraction from deterministic grounding into target-specific validation slices. Given a fixed source abstraction, vocabulary and schema, and machine-validated target contract, XPhysICS evaluates candidate mappings using five eligibility criteria: role compatibility, implemented type compatibility, stage coherence, slice viability, and rule-surface applicability. Grounding acceptance, slice adequacy, dynamic realizability, consumer applicability, and consumer outcome remain distinct evidence layers. We evaluate 83 structured source-threat abstractions across water treatment, water distribution, hydro/water-energy, and chemical-process targets. Controlled target-side studies of SWaT-to-water-treatment and WADI-to-water-distribution groundings produce clean, nominal-confounded, and near-threshold consumer outcomes; nine Hydro/GRFICS cases extend bounded validation-slice execution. We also evaluate bounded predictive, state-aware, and phase-aware consumer lanes, the unmodified upstream GeCo implementation, and a paper-derived reproduction of a physics-guided search method over three frozen groundings. Results show that cross-domain ICS threat reuse requires traceable source semantics, explicit target-conditioned grounding criteria, and careful separation of subsequent target-side evidence.","published":"2026-09-25T04:33:17Z","updated":"2026-09-25T04:33:17Z","authors":["Sangshin Park","Jainta Paul","Lawrence Ponce","Md Raihan Ahmed","Mu Zhang","Luis Garcia"],"categories":["cs.CR","cs.AI"],"headline":"New method improves safety checks for industrial cyber-physical threats","summary":"The researchers address the challenge of applying known industrial control system (ICS) threat information from one plant to another, recognizing that similar threats may behave differently depending on the system. They developed XPhysICS, a method that carefully matches and tests these threats against specific target systems using strict criteria. Their results suggest that safely reusing threat information across different industrial domains requires clear tracking of threat origins, target-specific checks, and separating various validation steps. These findings come from controlled studies and theoretical evaluations across several industrial scenarios."},{"id":"2609.30804v1","title":"Fast factorization in diagram monoids","abstract":"We give explicit algorithms that factor elements of the standard diagram monoids into their usual generators. These algorithms generalize sorting from permutations to partial matchings and set partitions. In every case the worst-case complexity is $n^2$, which is optimal for algorithms that explicitly list the factors. We also determine the average complexity of our algorithms.","published":"2026-09-25T04:31:27Z","updated":"2026-09-25T04:31:27Z","authors":["Matthias Fresacher","Willow Stewart","Daniel Tubbenhauer"],"categories":["math.RA","cs.DS","math.CO"],"headline":"New efficient algorithms for breaking down diagram monoid elements","summary":"Researchers developed clear algorithms to break down elements in diagram monoids into basic building blocks, extending the idea of sorting from simple rearrangements to more complex structures like partial matchings and set partitions. They showed these algorithms run in the best possible time for listing all factors, with the worst case taking time proportional to the square of the input size. They also analyzed how these algorithms perform on average. The work is theoretical and focuses on algorithm design and complexity."},{"id":"2609.30803v1","title":"Einstein four-manifolds of positive sectional curvature and ADM mass","abstract":"We prove that closed Einstein four-manifolds $(M,g)$ with positive sectional curvature and Euler characteristic and signature satisfying $2χ(M)-3|τ(M)|\\le4$ must be homothetically isometric to the round $S^4$ or to $\\mathbb{CP}^2$ with the Fubini-Study metric. The proof relies on an upper bound on the ADM mass of conformal blow-ups of $(M,g)$, combined with a lower one from \\cite{GM}.","published":"2026-09-25T04:27:45Z","updated":"2026-09-25T04:27:45Z","authors":["Matthew Gursky","Andrea Malchiodi"],"categories":["math.DG"],"headline":"Understanding Special Curved 4D Spaces Using Mathematical Tools","summary":"The researchers studied special four-dimensional shapes called Einstein four-manifolds that have positive curvature and meet certain technical conditions. They showed that such shapes must be very similar to two well-known curved spaces: a round 4-dimensional sphere or a complex projective plane with a standard geometry. Their proof used advanced mathematical tools including bounds on a quantity called the ADM mass, which measures certain geometric properties after altering the shape in a specific way. The findings are theoretical and rely on mathematical inequalities."},{"id":"2609.30802v1","title":"Understanding the Role of Prompt Template in Knowledge Distillation for Safety Alignment","abstract":"Prior research has demonstrated that the choice of prompt template during Supervised Fine-Tuning (SFT) significantly impacts the robustness of safety alignment afterwards. However, the influence of template selection during Knowledge Distillation (KD) from teacher to student remains largely unexplored. Thus, we fill this gap by analyzing how different template configurations influence the pre-existing safety alignment of the student. We observe a significant degradation of safety alignment present in the aligned base instruct-tuned model. Specifically, we find that utilizing chat templates renders the model more compliant with harmful queries compared to a non-chat template. These findings are consistent across three models: LLaMA, Gemma and Qwen model families and are evaluated across multiple safety benchmarks. We further show that using a non-chat template during distillation better preserves the base student's internal representations, while chat template distillation induces a larger representational shift. Code: https://github.com/anjilab/role-of-prompt-template-in-kd","published":"2026-09-25T04:26:49Z","updated":"2026-09-25T04:26:49Z","authors":["Anjila Budathoki","Manish Dhakal","Benjamin M. Ampel","Yi Ding"],"categories":["cs.CL"],"headline":"How Prompt Templates Affect AI Safety During Knowledge Distillation","summary":"Researchers studied how different prompt templates impact the safety of AI models when knowledge is transferred from a teacher model to a student model. They found that using 'chat' style templates during this transfer makes models more likely to respond harmfully, while 'non-chat' templates better preserve safety features. This pattern appeared across several AI model types and safety tests. The findings come from analyzing model behavior and internal structures, highlighting an area needing care in AI training methods."},{"id":"2609.30801v1","title":"Hyponormal Toeplitz operators with operator valued normal symbols","abstract":"In this article, we first give a characterization of hyponormal Toeplitz operators with operator valued normal symbols. Latter, we examine when the class of hyponormal Toeplitz operators with operator valued symbols are analytic or normal.","published":"2026-09-25T04:24:45Z","updated":"2026-09-25T04:24:45Z","authors":["Neeru Bala","Pratima Pandey"],"categories":["math.FA"],"headline":"Understanding Special Mathematical Operators Called Toeplitz Operators","summary":"Researchers studied a type of mathematical object known as hyponormal Toeplitz operators that involve operator valued normal symbols, which are specific functions used in advanced mathematics. They provided a description of these operators and investigated conditions under which these objects behave in particular ways called analytic or normal. The findings are theoretical and focus on characterizing these operators rather than practical applications."},{"id":"2609.30800v1","title":"Movable-Element STARS-Assisted NOMA Communications","abstract":"This paper proposes a movable-element (ME) enabled simultaneously transmitting and reflecting surface (ME-STARS)-assisted non-orthogonal multiple access (NOMA) communication system, where the positions of STARS elements can be flexibly reconfigured to introduce additional spatial degrees-of-freedom. Our objective is to maximize the sum achievable rate by jointly optimizing the active beamforming at the base station (BS) as well as the passive beamforming and MEs positions at the STARS. To tackle the resultant highly-coupled non-convex optimization problem, an alternating optimization-based algorithm is proposed, in which the original problem is decomposed into three subproblems. For the BS and ME-STARS beamforming subproblems, the successive convex approximation method is employed. For the MEs positions optimization, penalty terms are incorporated into the objective function to handle the constraints, thereby transforming the original problem into an unconstrained formulation. The gradient descent algorithm is subsequently applied to iteratively approach the sub-optimal solution. Numerical results demonstrate that: i) ME-STARS-assisted NOMA communications achieve superior performance compared to the NOMA communications assisted by STARS with fixed-position elements; and ii) the performance gain brought by the MEs is significant even with restricted moving region size.","published":"2026-09-25T04:23:51Z","updated":"2026-09-25T04:23:51Z","authors":["Jingjing Zhao","Yiheng Zhang","Kaiquan Cai","Xidong Mu","Yuanwei Liu"],"categories":["eess.SP"],"headline":"New Smart Surface Improves Wireless Communication Speed","summary":"Researchers studied a new type of smart surface with movable parts to help improve wireless communication systems. They developed a method to adjust both the base station's signals and the positions of the surface's elements to increase data rates. Their computer simulations show this approach performs better than surfaces with fixed parts, even when the movable parts can only shift in a small area."},{"id":"2609.30799v1","title":"Stability in recovering the order of a time-fractional evolution equation","abstract":"We consider the recovery of the fractional order in a time-fractional evolution equation with a source of the form $p(t)f$, where the temporal factor $p$ is also unknown but belongs to a bounded admissible class and has a prescribed nonzero value at $t=0$. From one scalar observation we prove a uniform small-time expansion and obtain a Hölder estimate for the fractional order. More precisely, if a fixed derivative order $γ$ lies below every admissible fractional order, then the stability exponent is $α_+/(2α_+-γ)$; for the undifferentiated observation this gives the exponent $1/2$. We also show that differentiating the data to an order larger than all admissible fractional orders is supercritical: the corresponding derivative is unbounded near $t=0$ whenever the two fractional orders are different. A uniform estimate for the two-parameter Mittag--Leffler function, including the diagonal case, is proved in the appendix.","published":"2026-09-25T04:23:22Z","updated":"2026-09-25T04:23:22Z","authors":["Ravshan Ashurov","Masahiro Yamamoto"],"categories":["math.AP"],"headline":"Researchers study how to identify fractional order in complex time equations","summary":"The researchers investigate how to determine an unknown fractional order in a specific time-based evolution equation where another unknown time factor is also involved. They prove a mathematical estimate that shows how stably the fractional order can be recovered from a single measurement taken early in time. Their findings include limits on differentiating the data and provide a detailed mathematical function estimate, but the results are theoretical and based on specific mathematical conditions."},{"id":"2609.30798v1","title":"Evaluating Real-Time Voice Agents: From Component Quality to Grounded Outcomes","abstract":"Real-time voice agents have moved from research prototypes to production deployments, yet the literature describing them is fragmented across three communities that rarely cite one another: speech foundation modelling, turn-taking psycholinguistics, and agentic evaluation. Architecture papers report latency, turn-taking papers report prediction accuracy, and agentic benchmarks report task success, so no single number describes whether a deployed agent is actually good. We address that gap with three evidence-based claims, each traceable to a corpus of 38 primary sources organised into an application-centric taxonomy of six categories. First, architecture choice is a deployment constraint rather than a settled verdict: a 2026 enterprise tutorial reports that no fully self-hostable end-to-end system yet meets production constraints, while a chunked cascade independently reaches state-of-the-art duplex behaviour, showing duplex behaviour is separable from duplex architecture. Second, evaluation has shifted decisively from component quality toward grounded outcomes, with recent benchmarks verifying backend state rather than trusting what the agent claims to have done. Third, the dyadic assumption in most models and benchmarks is breaking down: multiparty turn-taking and multi-speaker reasoning benchmarks show that deciding when not to speak, and reasoning about who may be told what, are first-class capabilities two-participant framings cannot measure. For each source we state the problem it targets, its mechanism, and its reported evidence, alongside the search strategy, inclusion criteria, and a verification step that caught a misattributed arXiv identifier in circulation. We propose TRG (Timing-Recovery-Grounded), a reporting standard characterising an agent by timing, post-disruption recovery, and state-verified outcome together, with a conditional fourth axis for multiparty deployments.","published":"2026-09-25T04:22:20Z","updated":"2026-09-25T04:22:20Z","authors":["Shivam Negi","Arpit Rawat","Rashi Jain"],"categories":["cs.AI"],"headline":"New Standard Proposed for Evaluating Real-Time Voice Agents","summary":"Researchers reviewed 38 studies on real-time voice agents, which currently have fragmented evaluation methods focusing on different aspects like speed, turn-taking, or task success. They found no single measure fully captures an agent’s quality and highlight challenges such as handling conversations with multiple speakers. The researchers suggest a new reporting standard called TRG that accounts for timing, recovery from disruptions, and verified task outcomes to better assess voice agents, based on evidence from diverse sources. Their findings mainly organize existing research rather than test new systems."},{"id":"2609.30797v1","title":"HasMem: Hard-Origin Adaptively Softened Memory for Long-Term LLM Agents","abstract":"Text-based memory and context compression support reuse of past interactions. Resizing continuous memory changes the input to a frozen LLM, coupling capacity allocation with readout. We propose Hard-Origin Adaptively Softened Memory (HasMem). Frozen hard-prompt embeddings provide a verifiable initial state. A controller adjusts memory widths, a Writer re-encodes resized entries, and Reader and Global provide readout adaptation and cross-turn state. On all $535$ questions in a reconstruction probe derived from the Multi-Session Chat (MSC) development split, the main configuration achieves lexical F1 of $95.3$ ($+4.4$ percentage points) at $93.6\\%$ of the hard reference's framed memory positions. With approximately matched per-question target body budgets, six configurations at mean per-entry retention around $0.83$--$0.91$ exceed rule-based re-encoding by $8.0$--$23.6$ exact-match (EM) percentage points. With fixed model parameters and rule target width ratio $0.75$, Global's EM gain passes a user-level exact paired test with Bonferroni correction over eight comparisons. On all $500$ LongMemEval-S questions, local lexical F1 rises from the hard reference's $3.4$ to $8.9$, and answer negative log-likelihood (NLL) falls from $12.257$ to $5.274$. F1 gains accompany lower EM on both evaluations.","published":"2026-09-25T04:22:05Z","updated":"2026-09-25T04:22:05Z","authors":["Zihong He","Junxiao Shen","Chen Liang","Hai-Ning Liang"],"categories":["cs.AI"],"headline":"New Memory Method Improves Long-Term AI Chat Accuracy","summary":"Researchers developed a method called HasMem to improve how AI language models remember and reuse past conversations. By adjusting memory size and re-encoding stored information, their approach improved performance on tests measuring understanding and recall. Their results, based on computer experiments, show better accuracy on reconstruction and long-term memory tasks compared to prior techniques, though some metrics varied."},{"id":"2609.30796v1","title":"ConsultMind:Towards Automated Diagnostic Consultation via Uncertainty-Aware Reasoning","abstract":"Diagnostic consultation is an online sequential decision-making process in which clinicians gather evidence through patient interaction until a diagnosis is sufficiently supported. Automating this process requires adaptive inquiry and interpretable decisions. Bayesian networks offer a natural foundation by updating diagnostic posteriors as evidence accumulates, but their use in open-ended consultation raises two challenges: linking diagnostic hypotheses to potential inquiries and translating evolving posteriors into consultation decisions. We introduce AutoDisym, an automated pipeline that integrates diagnostic knowledge with heterogeneous diagnosis-labeled clinical narratives to construct a Disorder--Symptom Bayesian Network (DSBN). Building on the DSBN, we propose ConsultMind, an uncertainty-aware framework that updates disorder posteriors after each response and uses posterior uncertainty to guide inquiry and diagnosis. We evaluate both methods across psychiatry, respiratory medicine, fever clinics, and three public datasets. The results show that AutoDisym can automatically construct high-quality DSBNs and that ConsultMind consistently improves diagnostic performance and explanation soundness. For example, AutoDisym achieves macro-averaged F1 scores of 81.37 for canonical symptoms and 72.19 for manifestations using GPT-5.6-Sol. ConsultMind improves Top-1 and Top-3 diagnostic accuracy by up to 22.15 and 37.89 percentage points, respectively. Physician evaluation further shows that ConsultMind improves the quality of ranking explanations, differential diagnoses, and diagnosis rationales across LLMs of different scales. This work offers a promising approach to automatic diagnostic consultation.","published":"2026-09-25T04:20:54Z","updated":"2026-09-25T04:20:54Z","authors":["Xiao Sun","Yuming Yang","Yun Chen","Jiang Zhong","Junnan Zhu","Xinyi Jiang","Haoyang Zeng","Ruirui Chen","Yining Wang","Xinyu Zhou","Rong Tang","Kaiwen Wei"],"categories":["cs.AI"],"headline":"New AI system aims to improve medical diagnostic consultations","summary":"Researchers developed ConsultMind, an AI framework designed to mimic how doctors gather patient information and update diagnoses as new evidence appears. They built a tool called AutoDisym to automatically create a structured model linking disorders and symptoms from clinical texts. Tested across several medical fields, these methods improved diagnostic accuracy and the quality of explanations in computer simulations and expert reviews. However, the findings are based on automated models and evaluations rather than live clinical trials."},{"id":"2609.30795v1","title":"Motion Style Slider: Endpoint-Supervised Continuous Style Control for Human Motion Diffusion","abstract":"Existing human motion diffusion methods provide strong motion generation quality, and recent style transfer models can inject target style cues, but fine-grained continuous control of style intensity remains underexplored. In production, style intensity is subjective across artists and directors, so the practical requirement is not a universal absolute unit, but a reliable monotonic control axis. We propose Motion Style Slider, a motion-to-motion style transfer framework for endpoint-supervised continuous control. Given a content motion and a style motion, we construct a style direction in a learned motion-style embedding space and condition diffusion generation with a scalar intensity. The training objective combines diffusion denoising with latent intensity regularization to encourage smooth and monotonic style scaling without requiring intermediate-intensity ground-truth motions. Our framework is compatible with pretrained motion diffusion backbones and supports heterogeneous style datasets, including the multi-actor style motion dataset. To test out-of-range usability, we additionally introduce a small real-capture over-reaction extension and evaluate large-intensity behavior against these unseen targets. Experiments measure controllability, interpolation/extrapolation behavior, content preservation, and motion realism, with ablations on direction construction and loss design.","published":"2026-09-25T04:20:11Z","updated":"2026-09-25T04:20:11Z","authors":["Chen-Chieh Liao","Yichen Peng","Yiyi Cai","Yûi Ono","Hiroki Hanaoka","Erwin Wu","Hideki Koike","Shuichi Kurabayashi"],"categories":["cs.CV"],"headline":"New Tool Lets Creators Smoothly Adjust Style in Human Motion Videos","summary":"Researchers developed Motion Style Slider, a method to smoothly control the intensity of style changes in generated human motion videos. Their approach learns a style direction from example motions and adjusts style strength with a simple slider, without needing many intermediate examples for training. They tested it on diverse motion datasets and showed it can control style continuously while keeping motion realistic and preserving original movement content. The results come from simulated experiments and focus on technical performance rather than real-world use."},{"id":"2609.30794v1","title":"Uniqueness in determining the temporal source factor and the fractional order with an unknown initial value","abstract":"We consider inverse problems for a time-fractional diffusion equation with a second-order elliptic operator and the homogeneous Dirichlet boundary condition in a bounded domain of $\\mathbb R^N$. The initial value is unknown and is not assumed to be zero. The main result concerns the simultaneous determination of the fractional order and the time-dependent source factor from one weighted spatial observation. For irrational orders in $(0,1)$ and bounded source factors satisfying an exponential integrability condition, we prove that a nonzero observation given for all positive times uniquely determines both unknowns. We also describe the information about the initial value determined by this observation. A counterexample shows that uniqueness of the source factor may fail at the classical order $α=1$, even when the observation is nonzero. In the second part, we establish two uniqueness results for the fractional order when the elliptic operators, initial values, and source terms in the compared equations may be different. The first result is proved by the Laplace transform, whereas the second uses short-time observations and the asymptotic behavior of the Mittag--Leffler functions.","published":"2026-09-25T04:20:03Z","updated":"2026-09-25T04:20:03Z","authors":["Ravshan Ashurov","Masahiro Yamamoto"],"categories":["math.AP"],"headline":"Researchers identify unique time-fractional factors from limited data","summary":"The researchers studied a complex equation modeling how substances spread over time with unknown starting conditions. They showed that for certain non-integer fractional orders between 0 and 1, one specific time-based spatial measurement can uniquely reveal both the fractional order and the changing source factor. Their findings also describe what can be learned about the unknown initial state. However, they found this uniqueness may fail for the classical case when the fractional order is exactly 1. Additional uniqueness results were proven using mathematical transforms and short-term observations."},{"id":"2609.30793v1","title":"Long-term X-ray variability of the NLS1 Ark 564","abstract":"We present a systematic spectral study of the X-ray emission from the Narrow- Line Seyfert 1 (NLS1) galaxy Ark 564 using Swift/XRT observations spanning sixteen years. The source exhibits significant variability in both the soft and hard X-ray bands, along with a strong positive correlation between them, indicating a closely coupled emission mechanism within the accretion flow-corona system. A simple absorbed power-law model is statistically insufficient to describe the spectra, confirming the presence of an additional soft component. To investigate the origin of the soft excess, we performed multi-model spectral fitting. Incor- porating a warm Comptonization component yields a statistically improved fit, supporting the scenario in which the soft excess arises from Comptonization in a warm, optically thick corona. We also applied a relativistic reflection model to representative observations; however, given the moderate spectral resolution and limited high-energy coverage of Swift/XRT, the current data do not allow us to robustly determine whether reflection provides a superior or physically preferred description. Our results indicate that warm Comptonization offers a consistent explanation for the observed soft excess in Ark 564 within the limitations of the available data, while deeper broadband observations are required to defini- tively distinguish between competing models and comprehensively characterise the long-term X-ray spectral variability of the source","published":"2026-09-25T04:19:59Z","updated":"2026-09-25T04:19:59Z","authors":["Neha P R","Nikita S Arun","Savithri H. Ezhikode","Ranjeev Misra","Bari Maqbool","Jeena K","Joe Jacob"],"categories":["astro-ph.HE"],"headline":"Studying Long-Term X-Ray Changes in Galaxy Ark 564","summary":"Researchers examined 16 years of X-ray data from the galaxy Ark 564 to understand its changing emission. They found that simple models couldn't fully explain the X-ray features, and adding a warm, dense cloud around the black hole helped improve the fits. The study supports the idea that soft X-ray signals come from this 'warm corona,' but the data aren't detailed enough to rule out other explanations. More advanced observations are needed to confirm these findings."},{"id":"2609.30792v1","title":"BS Cassiopeiae: A Contact Binary with Starspot-Driven Variability and Evidence for Additional Components","abstract":"We present a comprehensive photometric and spectroscopic analysis of the contact eclipsing binary BS Cas, based on 22-yr multi-site light curves and 2-yr high-resolution spectra. The long-term photometry shows pronounced year-to-year changes in the light-curve shape, including eclipse-depth reversals between the primary and secondary minima and variations in the O'Connell effect. The first double-lined radial-velocity (RV) curves of BS Cas are derived from the spectra using broadening-function (BF) profiles. Sharp peaks absent from the 2018 BFs appear in the 2020 BFs, suggesting an additional component along the same line of sight as BS Cas. Simultaneous modeling of the BVR light curves and the RVs shows that BS Cas is a contact binary in which the more massive component is cooler than its less massive companion, with a mass ratio of 2.515 and a fill-out factor of 32.3%. Multi-epoch modeling suggests that the long-term photometric variations arise from evolving starspots and changes in the third-light contribution. The eclipse-timing variation diagram confirms a secular orbital period decrease at a rate of $-2.305 \\times 10^{-7}~\\mathrm{days\\,yr^{-1}}$ consistent with angular momentum loss and mass transfer, along with an 18.376-yr modulation interpreted as a light-travel-time effect from a third body. The TESS timings reveal anti-correlated variations between the primary and secondary minima that appear to result from differential rotation of high-latitude starspots. These results suggest that BS Cas is a physically complex contact binary in which long-term photometric and eclipse-timing variations reflect multiple processes, including magnetic activity and the influence of a third body.","published":"2026-09-25T04:18:35Z","updated":"2026-09-25T04:18:35Z","authors":["Min-Ji Jeong","Chun-Hwey Kim","Kyeongsoo Hong","Mi-Hwa Song","Hye-Young Kim","Jang-Ho Park","Cheongho Han","Joh-Na Yoon"],"categories":["astro-ph.SR"],"headline":"Study reveals complex behaviors in binary star system BS Cassiopeiae","summary":"Researchers analyzed 22 years of light observations and 2 years of spectral data from BS Cassiopeiae, a pair of stars orbiting each other very closely. They found changes in brightness patterns caused by spots on the stars and identified signs of a third object affecting the system. The study suggests that variations in light and timing are due to magnetic activity, mass transfer between stars, and gravitational effects from an additional companion. These conclusions are based on observed data modeling and indicate a complex stellar system."},{"id":"2609.30791v1","title":"Topological Superconducting Phases in a Strained Altermagnet-Superconductor Heterostructure","abstract":"We investigate topological superconductivity in a heterostructure consisting of a two-dimensional $s$-wave superconductor and a $d$-wave altermagnet with Rashba spin-orbit coupling. In particular, we study the effects of strain and hopping anisotropy on the topological superconducting phases. By calculating the Chern number, we obtain topological phase diagrams as functions of the chemical potential and the strength of the effective magnetic field induced by the magnetic proximity effect. We find that strain-induced lattice distortion allows topological superconducting phases to emerge over a broad parameter region. Our findings suggest that strain that lowers symmetry can serve as a route to realizing topological superconductivity in altermagnet-superconductor heterostructures.","published":"2026-09-25T04:15:02Z","updated":"2026-09-25T04:15:02Z","authors":["Keita Yoshizawa","Ryo Okugawa","Takami Tohyama"],"categories":["cond-mat.supr-con","cond-mat.mes-hall"],"headline":"Strain helps create special superconducting states in layered materials","summary":"Researchers studied a layered material made of a superconductor and a special kind of magnet, focusing on how stretching (strain) affects its properties. They used theoretical calculations to map how different physical parameters influence the emergence of unique superconducting states linked to topological features. Their results show that applying strain can broaden the conditions where these special superconducting phases appear, suggesting strain as a useful tool to achieve such states in these materials."},{"id":"2609.30790v1","title":"Towards Universal Representation-Based Process Control","abstract":"Many temporal process learning and monitoring pipelines operate in local windows, making window-level decisions unavoidable in practice. In such settings, classical statistical tests can be applied to individual windows, but they typically evaluate predefined parametric hypotheses-such as unit-root or moment-based conditions-thereby limiting flexibility when reference behavior is defined empirically from task- or domain-specific data. In this work, we view window-level monitoring as a process control problem and reformulate it as reference-based hypothesis testing, where the null hypothesis is specified by an empirical reference distribution rather than a fixed parametric model. We operationalize this perspective through a representation-based, nonparametric framework that combines pretrained time series encoders, kernel density estimation, and conformal calibration, yielding finite-sample valid inference in learned representation space. Classical notions such as stationarity and cyclostationarity arise as natural instantiations of empirical reference sets within this framework. Through experiments, we demonstrate sensitivity to window-level distributional deviations while maintaining well-calibrated inference under stable reference regimes, highlighting the applicability of the proposed approach to a broad class of time series process control and monitoring tasks.","published":"2026-09-25T04:14:55Z","updated":"2026-09-25T04:14:55Z","authors":["Jinmyeong Choi","Taesup Kim","Artur Dubrawski"],"categories":["cs.LG"],"headline":"New Method Improves Monitoring of Time-Based Data Windows","summary":"Researchers studied how to better monitor changes in data streams that are divided into time windows. They developed a flexible approach that compares observed data against a learned reference, rather than relying on rigid predefined models. Their method uses advanced data representations and statistical techniques to detect deviations in data patterns while ensuring reliable decisions, based on experiments with simulated scenarios."},{"id":"2609.30789v1","title":"Interpretable-by-Design Descriptor Portfolios Match a 2048-Dimensional Foundation Embedding on Low-Data Molecular Assays","abstract":"In low-data structure-activity prediction, the choice of molecular representation can matter more than the choice of predictor, and tabular foundation models sharpen that effect. We ask whether a portfolio of compact, semantically named descriptor blocks can reach the accuracy of a 2048-dimensional CheMeleon embedding while staying auditable at the feature level, meaning that every input dimension carries a model name and a recorded training provenance. Starting from a fixed 11-dimensional physicochemical base, we greedily concatenate provenance-screened blocks using the labelled context alone. Across nine ADME/Tox assays and 50 evaluation cells, scored on common-coverage subsets restricted to the molecules that every representation covers, the portfolio reaches a mean test AUC of 0.762, against 0.764 for CheMeleon and 0.756 for Mordred. The pooled gap to CheMeleon is +0.003 AUC (task-bootstrap 95% CI [-0.020, +0.030]), which satisfies our predeclared pooled parity gate but not the per-assay gate. At 25 context labels the headline rule again satisfies the pooled gate; at 10 labels it does not. We also report four predeclared candidate-selection rules that we falsified. Post-freeze checks over ten seeds and three previously unseen assays support pooled competitiveness for compact, auditable representations; a same-width random-bundle control does not establish that greedy membership itself adds accuracy. Assay-level differences remain unresolved.","published":"2026-09-25T04:13:17Z","updated":"2026-09-25T04:13:17Z","authors":["Yiqi Yao","Miquel Duran-Frigola"],"categories":["cs.LG"],"headline":"Researchers compare molecular data summaries for drug testing accuracy","summary":"The researchers explored whether simple, understandable groups of molecular features can predict drug-related outcomes as well as complex 2048-number summaries. They combined small sets of chemically meaningful descriptors and tested these against a large standard dataset of molecular tests. Their findings suggest that these compact, interpretable feature sets achieve nearly the same accuracy on average, but performance varies by specific test. The study highlights benefits of easier-to-understand models but also notes unresolved differences across individual tests and some limitations in confirming why certain features improve accuracy."},{"id":"2609.30788v1","title":"Observation of the electronic Pomeranchuk effect in generalized Wigner crystals of twisted MoS$_2$","abstract":"Moire superlattices in transition metal dichalcogenides provide a highly tunable platform for exploring strongly correlated electronic phases, such as generalized Wigner crystals. While these crystalline states typically melt with increasing thermal fluctuations, an electronic analogue of the Pomeranchuk effect can stabilize the localized solid phase at elevated temperature through isospin entropy. Here, we report the observation of an electronic Pomeranchuk effect at the fractional filling factors of $ν= 1/3$ and $ν= 1/4$ in AB-stacked twisted bilayer MoS$_2$ with twist angles of 4.1$^\\circ$ and 3.9$^\\circ$, respectively. At ultra-low temperature, the system exhibits a highly conducting, itinerant behavior at these fractional fillings, characteristic of a compressible Fermi-liquid ground state. Upon heating, the system exhibits a counterintuitive increase in the longitudinal resistivity, signaling an isospin-entropy-driven transition into a localized generalized Wigner crystal state. Our findings highlight the unique capacity of flat bands in twisted bilayer MoS$_2$ for stabilizing highly degenerate magnetic configurations, offering new insights into the thermodynamic phase diagrams of low-dimensional correlated systems.","published":"2026-09-25T04:13:09Z","updated":"2026-09-25T04:13:09Z","authors":["Xinjie Fang","Linfeng Wu","Naitian Liu","Le Zhang","Wenqiang Zhou","Jing Ding","Zhangyuan Chen","Hanxiao Xiang","Yue Xiao","Kenji Watanabe","Takashi Taniguchi","Shuigang Xu"],"categories":["cond-mat.mes-hall","cond-mat.mtrl-sci","cond-mat.str-el"],"headline":"Researchers observe unusual solid state in twisted bilayer MoS2","summary":"The researchers studied a special layered material called twisted bilayer MoS2 and found that heating it can cause electrons to form an unusual solid-like arrangement, called a generalized Wigner crystal. Normally, heating melts solids, but here, increased temperature stabilizes this state due to an effect linked to electron properties called isospin entropy. Their observations offer new understanding of how electrons behave in flat, layered materials at very low temperatures and certain electron densities."},{"id":"2609.30787v1","title":"FSSDataBase: a reconstructable dataset of simulated frequency-selective surface structures and scattering responses","abstract":"Data-driven design of frequency-selective surfaces (FSSs) requires reusable datasets that link structural geometry to electromagnetic response under documented simulation conditions. Here we present FSSDataBase, an openly available collection of 5,000 procedurally generated single- and multilayer FSS unit cells simulated using Ansys HFSS. The dataset covers 10-20 GHz with transverse-electric and transverse-magnetic relfection and transmission responses, including magnitude and phase, for recorded incidence angles of $0^{\\circ}$ and $30^{\\circ}$. Each record links binary structural masks and JSON-based reconstruction metadata to raw response samples and derived magnitude labels. Structural descriptors, response-distance measures and task-oriented cost functions characterize geometric and electromagnetic variation and support screening for polarization stability, angular-selective bandwidth and matched-amplitude, 180° phase-separated structure pairs. Accompanying code supports model reconstruction, data processing and configuration-driven generation of additional samples. Beyond conditional forward modelling and response-based structure retrieval, FSSDataBase has the potential to serve as a reference dataset for deep-learning-based FSS inverse design, supporting fair comparisons across network architectures under shared data splits,input conditions and evaluation protocols.","published":"2026-09-25T04:10:44Z","updated":"2026-09-25T04:10:44Z","authors":["Xinke Kuang","Shiyun Ma","Yuanyuan Wang","Jiang Wu"],"categories":["physics.app-ph"],"headline":"New Open Dataset Links Surface Designs to Electromagnetic Responses","summary":"Researchers created FSSDataBase, a public dataset of 5,000 simulated frequency-selective surface structures, showing how design relates to electromagnetic behavior between 10-20 GHz. The data includes detailed simulation results for different angles and wave types, along with code to recreate and extend the dataset. The team presents this as a tool to help develop and compare machine learning models that design such surfaces, based on consistent data and testing methods. The findings are based on computer simulations documented in the dataset."},{"id":"2609.30786v1","title":"Long-time validity of the Korteweg-de Vries approximation for the Boussinesq equation","abstract":"We consider the Korteweg-de Vries (KdV) approximation for the good Boussinesq equation. Building upon the low-regularity local-in-time justification established by Hong and Yang \\cite{HY2024}, we prove that this approximation remains valid on logarithmically long time intervals. The key ingredient is a persistence of regularity argument, which allows the local approximation to be iterated using the rescaled conservation law for the Boussinesq equation.","published":"2026-09-25T04:07:20Z","updated":"2026-09-25T04:07:20Z","authors":["Younghun Hong","Junyeong Jang"],"categories":["math.AP"],"headline":"Researchers confirm long-term approximation for a wave equation","summary":"The researchers studied the Korteweg-de Vries (KdV) approximation, a simplified model for waves described by the Boussinesq equation. They extended previous work by showing this approximation remains accurate for much longer time periods than before. Their results rely on mathematical arguments involving conservation laws, but these findings are theoretical and focus on understanding the equations rather than practical applications."},{"id":"2609.30785v1","title":"Job Class Thermal Intent Aware Liquid Cooling Allocation for AI Data Centers","abstract":"GPU-dense AI data centers need to run on liquid cooling as air simply cannot shed the heat at these power densities. Yet the cooling loops themselves are blind to what workloads are about to run; they crank up flow only after a sensor catches a temperature climb, which can take 30 to 50 seconds. We built Job-Class Thermal Intent (JCTI) to close that window. The scheduler already knows a job is coming and what class it belongs to; JCTI feeds that information straight to the cooling controller so it can stage coolant before the heat shows up. We pulled the thermal signatures for each job class out of MLPerf GPU power traces and tuned arrival patterns against Alibaba cluster data. Over 120 paired Monte Carlo trials the numbers come out to 56.4% fewer thermal violations and 60.2% less cumulative overshoot than a straight PI loop. As AI data centers evolving towards gigawatt grid loads with highly fluctuating power swings, thermally-aware scheduling reduces sudden demand and improves load prediction in grid side. Cooling and scheduling have been running as two separate systems for years despite each one knowing something the other needs, JCTI wires them together.","published":"2026-09-25T04:06:10Z","updated":"2026-09-25T04:06:10Z","authors":["Krishna Chaitanya Sunkara"],"categories":["cs.DC","cs.PF","eess.SY"],"headline":"New Cooling System Predicts AI Workload Heat to Save Energy","summary":"Researchers developed a system called Job-Class Thermal Intent (JCTI) that helps liquid cooling in AI data centers prepare for heat before it happens. By using information about upcoming computer tasks, JCTI adjusts cooling earlier than traditional methods that wait for temperature rises. Testing with simulated data showed JCTI reduced cooling problems by over half compared to standard controls, suggesting it could help manage power demand and improve energy efficiency in large AI facilities. These results are based on simulations combining real workload patterns and power traces."},{"id":"2609.30784v1","title":"Symbiotic Architecture for Post-Hoc Audio Extension of Frozen Language Models","abstract":"This paper proposes an architecture for equipping large language models (LLMs) with audio-understanding capabilities without fine-tuning their weights. The proposed symbiotic architecture employs an injector module that writes audio-conditioned vectors directly into the target LLM's short-term memory, i.e., the key-value (KV) cache, enabling the LLM to behave as an audio language model (ALM). The architectural advantages are twofold. First, it improves the scalability of ALMs: because the proposed method bypasses the LLM during audio injection, the injection cost is governed by the injector width rather than the backbone width, and can therefore scale more slowly than the cost of full-backbone prefilling. Second, since the training scheme does not update the LLM weights, the original capabilities of the LLM are preserved without the risk of degradation from fine-tuning. The effectiveness of the proposed method is evaluated on both audio-understanding tasks (automatic speech recognition, audio question answering, and acoustic scene classification) and text-only tasks. We confirm that, while activating fewer parameters during audio prefilling, our architecture outperforms the conventional method with a frozen LLM and approaches the performance of a fine-tuned ALM, all while preserving the backbone LLM's original text-only task performance by construction.","published":"2026-09-25T04:06:01Z","updated":"2026-09-25T04:06:01Z","authors":["Yotaro Kubo","Qi Sun","Yujin Tang"],"categories":["cs.SD","cs.CL","eess.AS"],"headline":"Adding Audio Understanding to Language Models Without Changing Them","summary":"Researchers developed a method to give large language models the ability to understand audio without changing their internal settings. They created a separate module that feeds audio information into the model's short-term memory, letting it process audio like language. This approach keeps the original model's text skills intact, is more efficient to scale, and performs well on several audio tasks, nearly matching models that have been fully adjusted for audio understanding."},{"id":"2609.30783v1","title":"Skip the Talk, Re-Focus on Vision: Latent Reasoning for Reasoning Segmentation in Multimodal Large Language Models","abstract":"Reasoning segmentation aims to interpret implicit textual queries and enable fine-grained visual perception, which is critical for applications such as human-computer interaction and embodied agents. Existing methods typically generate explicit Chain-of-Thought (CoT) by multimodal large language models (MLLMs) before localizing the target. Although intuitive, such explicit verbal reasoning introduces substantial attention interference: redundant textual tokens disrupt attention during perception-token generation and also increase the effective distance between visual tokens. To address this issue, we propose LIRSeg, which fully replaces explicit CoT with a compact set of learnable latent tokens for reasoning segmentation. LIRSeg is trained in two stages: spatial alignment grounds the latent tokens in object-relevant visual evidence, and GRPO further optimizes them with segmentation rewards. To make these compact latent tokens more informative, we introduce three complementary mechanisms from an information perspective: extreme-advantage sampling for selecting informative training signals, decoupled exploration-stability updates for learning complementary representations, and latent diversity amplification for preventing representational collapse. Extensive experiments on benchmarks demonstrate that LIRSeg consistently improves both segmentation accuracy and reasoning efficiency. Compared with the VisionReasoner baseline, LIRSeg achieves absolute gIoU improvements of 4.9% on ReasonSeg, 7.1% on MUSE, and 4.7% on MMR, while achieving a approximately 16x reduction in reasoning tokens. Code is available in supplementary materials.","published":"2026-09-25T04:04:55Z","updated":"2026-09-25T04:04:55Z","authors":["Tianhang Guo","Yulin He","Wei Chen","Wenjuan Zhou","Yuhang Li","Xinbiao Gan"],"categories":["cs.CV","cs.AI"],"headline":"New method improves visual understanding by skipping explicit verbal reasoning","summary":"Researchers studied how multimodal large language models interpret visual scenes linked to textual queries. They found that traditional approaches using explicit verbal reasoning create distractions and reduce efficiency. To fix this, the researchers developed a new technique called LIRSeg that uses compact latent tokens for reasoning instead of explicit text, improving accuracy and speed in visual segmentation tasks. Their experiments showed better performance on multiple benchmarks, though these findings are based on modeling and need further practical validation."},{"id":"2609.30782v1","title":"Pulsed Accretion onto Eccentric Binaries in Highly Misaligned Circumbinary Disks","abstract":"We present three-dimensional smoothed particle hydrodynamics simulations of highly misaligned circumbinary disks (CBDs) around moderately eccentric equal-mass binaries ($e_\\mathrm{b}=0.5$). We show that the binary accretion is modulated on the binary orbital period and exhibits two pulses near periastron. The dominant pulse peaks before periastron for an initial binary-disk misalignment of $60^\\circ$, shifts to after periastron at $90^\\circ$, and occurs at an even later post-periastron phase at $120^\\circ$. We further show that the two pulses are accompanied by a time-dependent response of the circumstellar disks (CSDs) and by different distributions of accreting material within the cavity and around the CSDs. The qualitative pre- versus post-periastron distinction is also present in individual binary orbits despite variations in pulse amplitude. Our results motivate future tests of whether pulse timing is related to binary-disk orientation.","published":"2026-09-25T04:04:43Z","updated":"2026-09-25T04:04:43Z","authors":["Ruiqi Yang","Jeremy L. Smallwood","Shang-Fei Liu","Alessia Franchini"],"categories":["astro-ph.EP"],"headline":"Simulating Gas Flow onto Tilted Two-Star Systems","summary":"Researchers used 3D computer simulations to study gas flow onto pairs of stars orbiting each other with a tilted disk of material around them. They found that the gas accretion happens in two bursts during each orbit, with the timing depending on how misaligned the disk is. The study reveals complex interactions between the stars and surrounding gas, suggesting the orbit and disk orientation affect when and how gas falls onto the stars. These findings come from simulations and need further testing in real systems."},{"id":"2609.30781v1","title":"Missingness-Aware Conformal Prediction Under Cross-Hospital Distribution Shift","abstract":"Clinical measurements are recorded for some patients but not others, at rates that differ across hospitals, and marginal conformal coverage does not ensure coverage within groups defined by missingness. We propose a missingness-aware conformal calibration procedure for mortality prediction under cross-hospital distribution shift. It selects a measurement on an independent sample, groups patients by whether that measurement is recorded, and applies Mondrian calibration within each group, so no calibration outcome is reused. We evaluate the procedure across hospitals in eICU and across care units within one MIMIC-IV hospital, using three predictors. Relative to pooled calibration, it reduces the average worst-group coverage gap on its selected groups in all six settings, with a median reduction of 1.9 percentage points; paired site-bootstrap intervals exclude zero in five. These gains do not extend uniformly. Calibration by predicted risk achieves smaller gaps on a broader panel of missingness groups, and when eICU hospitals are evaluated separately, the gain shrinks for all three predictors and reverses in sign for one. We explain this discrepancy with a hospital-level decomposition. Pooling reweights hospitals through a covariance between group shares and coverage errors, and lets errors of opposite sign cancel: weighting explains the reversal, and cancellation accounts for most of the attenuation for the other two predictors. Constructed population distributions show that pooled and within-hospital evaluations can rank calibration methods oppositely even without sampling noise. Pooled improvement alone therefore cannot establish better coverage within hospitals, even when the calibration groups are fixed.","published":"2026-09-25T04:04:14Z","updated":"2026-09-25T04:04:14Z","authors":["Liang You","Dongwen Ou","Hengyu Shi","Siyuan Dai"],"categories":["cs.LG","stat.AP"],"headline":"New method improves prediction accuracy across hospitals with missing data","summary":"Researchers studied problems predicting patient deaths when medical records vary across hospitals. They developed a technique that adjusts predictions by grouping patients based on missing measurements and calibrating separately within each group. Tested on real hospital datasets, their method reduced errors in some situations but didn’t always help, especially when looking at individual hospitals. The researchers explain that combining data from many hospitals can hide these differences, so overall improvements don’t guarantee better predictions inside every hospital."},{"id":"2609.30780v1","title":"Sampling Safe Futures: Multimodal Trajectory Planning for Personalized Safety in Anthropomorphic AI","abstract":"Anthropomorphic artificial intelligence systems increasingly remember personal details, display empathy, and are engaged with as social counterparts, creating forms of risk that emerge from the evolution of the user-system relationship over time. Existing safeguards largely operate at the level of individual conversational turns and cannot determine whether a sequence of seemingly acceptable interactions is cumulatively moving a particular user toward harm. This paper introduces personalized trajectory-level safety, a framework that treats relational safety as a sequential decision problem over a latent escalation state inferred from the user's messages and influenced by the system's responses. At each turn, a screening step first discards any response strategy that does not preserve at least one safe continuation of the interaction under every plausible model of the user. Among the remaining strategies, we formulate action selection as multimodal trajectory sampling, and use a Generative Flow Network to generate diverse future evolutions in proportion to their plausibility, safety, and utility. The system then selects the strategy that preserves the largest fraction of safe and useful continuations. We evaluate the framework in simulation, calibrated on statistics reported for real human-chatbot interactions, and using response strategies derived from public benchmarks. Results show that trajectory-aware decision making substantially reduces the frequency of harmful states while keeping helpful interaction. This work reframes safety for anthropomorphic AI from response-level filtering to personalized control over the future evolution of human-AI relationships. The source code is available at https://github.com/benedettapicano/ANTHROPOMORPHIC_SAFETY_TRAJ.","published":"2026-09-25T04:04:02Z","updated":"2026-09-25T04:04:02Z","authors":["Benedetta Picano","Dusit Niyato"],"categories":["cs.HC"],"headline":"New AI method aims to keep long chats safe and helpful","summary":"Researchers address how AI systems that interact socially can accidentally cause harm over time by building up risky conversations. They developed a method that looks ahead at possible future chats, choosing responses that keep interactions safe and useful based on a model of the user's state. Tested in simulations reflecting real human-chatbot data, their approach reduced harmful situations while maintaining helpfulness. The work shifts AI safety from judging just single replies to managing the overall relationship over time."},{"id":"2609.30779v1","title":"The weak type $(1,1)$ estimate of Dirichlet Means for unbounded noncommutative Vilenkin systems","abstract":"Let $\\mathcal{R}$ be the hyperfinite $\\mathrm{II}_1$ factor. Considering the partial sum operators $(\\mathcal{S}_n)_{n\\geq 1}$ of the noncommutative Vilenkin-Fourier series associated with an arbitrary admissible Vilenkin group, we prove that there exists a universal constant $c>0$ such that \\begin{equation*} \\sup_{n\\geq1}\\|\\mathcal{S}_n(f)\\|_{L_{1,\\infty}(\\mathcal{R})} \\leq c\\|f\\|_{L_1(\\mathcal{R})},\\quad f \\in L_1(\\mathcal{R}), \\end{equation*} and, for every $1<p<\\infty$, $$\\sup_{n\\geq1}\\|\\mathcal{S}_n(f)\\|_{L_p(\\mathcal{R})} \\leq c\\frac{p}{p-1}\\|f\\|_{L_p(\\mathcal{R})},\\quad f \\in L_p(\\mathcal{R}).$$ Besides the transference technique, the main novel ingredient is a modified version of noncommutative Calderón-Zygmund decomposition established in \\cite{CCP2022}. Consequently, we resolve the problem of weak type $(1,1)$ estimate communicated to the authors by Fedor Sukochev, and substantially improve the strong type (p,p) estimates obtained in \\cite{DFdePS2001} by achieving the optimal order $\\frac{p}{p-1}$.","published":"2026-09-25T04:02:03Z","updated":"2026-09-25T04:02:03Z","authors":["Guixiang Hong","Tiantian Zhao"],"categories":["math.FA"],"headline":"New mathematical bounds found for certain noncommutative operators","summary":"Researchers studied special mathematical operators called partial sum operators related to a type of series in an advanced algebraic structure known as the hyperfinite II1 factor. They proved precise estimates showing how these operators behave on specific function spaces, improving previous results and solving an open problem about weak type (1,1) estimates. Their work relies on advanced methods including a modified decomposition technique. The findings are theoretical and pertain to abstract mathematical objects without immediate practical applications."},{"id":"2609.30778v1","title":"An $n^{8/5+o(1)}$-Time $Ω(λ^3)$-Approximation for Longest Common Subsequence","abstract":"Let $λ$ denote the ratio of the length of a longest common subsequence of two length-$n$ strings to $n$. Rubinstein, Seddighin, Song and Sun [RSSS19] gave an $Ω(λ^3)$-approximation for LCS running in $\\widetilde O(n^{39/20})$ time, where $39/20=1.95$. Song [Son19] mentioned that improving the $n^{1.95}$ running time is an interesting open question. We give an algorithm that computes an $Ω(λ^3)$-approximation of the longest common subsequence in $n^{8/5+o(1)}$ time. This improves the exponent $1.95$ to $1.6+o(1)$.","published":"2026-09-25T04:01:48Z","updated":"2026-09-25T04:01:48Z","authors":["Zhao Song"],"categories":["cs.DS"],"headline":"Faster Algorithm for Approximating Longest Common Subsequence","summary":"Researchers improved the speed of an algorithm that estimates the longest common subsequence (LCS) between two strings. Previous work provided a certain quality approximation running in about n^1.95 time; the new method achieves a similar quality but runs faster, in roughly n^1.6 time. This advancement addresses an open problem, but the results focus on improving theoretical running time and approximation factors rather than exact solutions."},{"id":"2609.30777v1","title":"Moving Horizon Estimation for Quadrotors: An $\\mathcal{L}_1$ Adaptive Optimizer Approach","abstract":"Moving Horizon Estimation (MHE) is a state estimation method based on finite-horizon optimization that can offer higher accuracy at the cost of increased computation compared to Kalman filter-based approaches. We present a linear smoothing MHE formulation as a dense Quadratic Program (QP), and a solver consisting of a continuous-time Newton's method augmented with the $\\mathcal{L}_1$ Adaptive Optimizer ($\\mathcal{L}_1$-AO). While MHE is inherently time-varying, conventional approaches treat it as a sequence of independent, time-invariant problems and employ iterative solvers at each time step, which can be both inaccurate and computationally burdensome. In contrast, time-varying solvers track the optimal solution with fewer iterations by exploiting the temporal evolution of the problem, thereby reducing the computational load. In this research, we enhance both the performance and efficiency of MHE through a time-varying solver with an $\\mathcal{L}_1$-AO augmentation that compensates for the prediction inaccuracy, which is common in practice due to noisy sensors and the lack of prior knowledge of the system. Simulation results on a quadrotor platform show that the $\\mathcal{L}_1$-AO-augmented approach solves the MHE optimization problem more efficiently than the baseline time-invariant solver and achieves higher estimation accuracy under challenging conditions, compared with both the Extended Kalman Filter and the standard MHE.","published":"2026-09-25T04:01:22Z","updated":"2026-09-25T04:01:22Z","authors":["Thinh Nguyen","Minkyung Kim","Sandeep Banik","Jinrae Kim","Naira Hovakimyan"],"categories":["cs.RO","eess.SY"],"headline":"Improving Drone State Estimation with a New Optimization Method","summary":"The researchers studied Moving Horizon Estimation (MHE), a technique that estimates a system's state more accurately than common methods but requires more computation. They developed a new solver using a time-varying approach combined with an adaptive optimizer to make MHE more efficient and accurate, especially in noisy and uncertain conditions. Simulations on a quadrotor drone showed their method outperformed traditional solvers and the Extended Kalman Filter, but these findings are based on simulations and not tested in real-world experiments."},{"id":"2609.30776v1","title":"A hydrogen-informed Rice-Beltz model for crack-tip dislocation emission under mixed-mode loading","abstract":"The fundamental competition between crack-tip cleavage and dislocation emission dictates the ductile-to-brittle transition in crystalline solids. In hydrogen-charged environments, this delicate balance is disrupted, yet existing models often struggle to self-consistently capture the atomic-scale thermomechanical modulation of the emission barrier. Here, we propose the hydrogen-informed Rice-Beltz framework to quantify the critical stress intensity factors (SIFs) required for initial dislocation nucleation under mixed-mode (I+II) loading. By integrating the elastic interaction of the dilatational field of interstitial hydrogen and the applied stress fields against the incipient dislocation core, the model recovers the non-monotonic relation between the most probable SIF and the hydrogen concentration. Unlike the classical pure mode-I (or mode-II) scenario, the minimization of the local strain energy density is employed to generate closed-form nominal driving forces, which are subsequently embedded into a transition-state-theory framework. This atomistically-informed approach yields the most probable SIF for the first dislocation emission event as a function of the loading rate, slip angle, and hydrogen concentration, providing a rigorous, parameter-free boundary condition for the onset of hydrogen-modulated crack-tip plasticity.","published":"2026-09-25T03:59:40Z","updated":"2026-09-25T03:59:40Z","authors":["Kai Zhao"],"categories":["cond-mat.mtrl-sci","physics.app-ph"],"headline":"New model predicts crack behavior in metals exposed to hydrogen","summary":"Researchers developed a new theoretical model to understand how tiny cracks in crystalline solids grow or stop when hydrogen is present. The model calculates the stress needed for dislocations—small shifts in the crystal structure—to start near crack tips under combined loading conditions. This approach links atomic-scale effects of hydrogen to crack behavior, helping explain the complex relationship between hydrogen levels and crack propagation. The findings are based on simulations and theory, offering a detailed method without experimental data yet."},{"id":"2609.30775v1","title":"Parnassus for the CLD Detector: A Generative Machine-Learning Surrogate for Detector Simulation and Reconstruction at the FCC-ee","abstract":"Detector simulation and event reconstruction will be computationally expensive for future $e^+e^-$ collider programs and are currently critical bottlenecks for accurate feasibility studies. To address this challenge, we build a Parnassus model for the CLD detector concept. Parnassus is a framework for automatically tuning a surrogate model, in our case, a conditional flow matching neural network, to emulate a full detector simulation and reconstruction. We train on $e^+e^-\\to Z\\to q\\bar q$ events at $\\sqrt{s}=91.2$ GeV processed through a Geant4 simulation of the CLD detector concept and the Pandora particle-flow reconstruction and reproduce single-particle kinematics, particle-IDs, and impact-parameter distributions. We also examine jet- and event-level features, inclusively and split by flavor, and find excellent fidelity, significantly better than the parameterized program Delphes. Furthermore, we train a transformer-based flavor tagger on the reconstructed particle-flow constituents and show that the surrogate preserves the $b/c/s/q$ discrimination of the full CLD reconstruction. The Parnassus CLD model achieves a generation cost of about 1.2 ms (40 ms) per event on a single GPU (CPU), over three (two) orders of magnitude faster than full simulation and reconstruction. Our model is publicly available for feasibility and design studies.","published":"2026-09-25T03:57:55Z","updated":"2026-09-25T03:57:55Z","authors":["Umar Sohail Qureshi","Benjamin Nachman","Caterina Vernieri"],"categories":["physics.ins-det","hep-ex","hep-ph","physics.data-an"],"headline":"Faster Collider Detector Simulations Using Machine Learning","summary":"Researchers developed a machine-learning approach called Parnassus to speed up simulations of particle detectors for future electron-positron colliders. They trained their model on detailed simulated data of particle collisions, finding it reproduces detector outputs closely and much faster than traditional methods. While this study is based on simulated data and specific detector models, the researchers show their approach maintains accuracy in key physics features and is publicly available for further study."},{"id":"2609.30774v1","title":"Dialogue-Based Streaming Audio-Visual Target Speaker Extraction with Predictive Dialogue Information","abstract":"In face-to-face, real-time communication, a talking agent must track the target speaker through natural pauses, turn-taking, and backchannels, often amid background cross-talk. Most target speaker extraction (TSE) studies, however, rely on simulated mixtures with full or sparse overlap and ignore the turn-taking of real conversations. We therefore introduce, to our knowledge, the first benchmark for online audio-visual TSE (AV-TSE), built from intact dyadic interactions with independent third-party interference. Observing that anticipating upcoming activity from semantic, acoustic, and facial cues benefits online AV-TSE, we propose an LLM-based target-speaker voice activity projection (TS-VAP) module. Unlike conventional VAP with separated speaker channels, it forecasts the future activity of the target and conversational partner directly from the overlapping mixture, drawing on the linguistic and conversational knowledge of a speech-LLM, and uses this prediction to guide a low-latency separator. We further combine this predictive context with historical and synchronous speaker context. Experiments show that TS-VAP consistently improves streaming extraction across multiple AV-TSE backbones, and that further combining historical, synchronous, and predictive context yields nearly 1 dB gain on real AV conversations. Project page: https://jjjjiaozi.github.io/TS-VAP/.","published":"2026-09-25T03:56:44Z","updated":"2026-09-25T03:56:44Z","authors":["Shuhan Zhang","Wenxuan Wu","Haizhou Li"],"categories":["cs.SD","eess.AS"],"headline":"New Method Enhances Real-Time Speaker Isolation in Conversations","summary":"Researchers studied how to better isolate a target speaker's voice during real-time, face-to-face conversations, which involve natural pauses and overlapping speech. They created a new benchmark using actual two-person interactions with background noise and developed a system that predicts future speaking activity using language and facial cues. Their approach improved the ability to separate speakers in real conversations, though results are based on experiments with their specific models and data."},{"id":"2609.30773v1","title":"Learning Natural Conversational Behavior in Tandem Speech-to-Speech Models with Randomized Guidance","abstract":"Tandem speech-to-speech architectures couple a responsive speech frontend with an asynchronous text backend. In KAME, a large language model (LLM) serves as the backend, supplying candidate responses as guidance to the speech frontend while the user is still speaking. Ordinary conversation recordings capture the eventual response but not the guidance the backend would supply during the user's utterance. Generating the missing guidance with a simulator LLM adds substantial data-preparation overhead when training on real conversations. We propose randomized intermediate guidance, which derives guidance directly from the conversation corpus rather than simulating backend LLM behavior. During training, target responses provide informative guidance, while randomly sampled responses provide potentially irrelevant updates during the utterance. This combination aims to teach the frontend to use backend information selectively. On synthetic dialogues, KAME trained with this recipe achieves response quality comparable to that of the LLM-generated and similarity-based baselines. Training on 3.8k hours of real conversations improves smooth turn-taking and audio-judge naturalness over synthetic-data KAME while retaining a response-quality advantage over Moshi. These results show that randomized guidance offers a practical route to combining the response-quality benefits of tandem models with natural conversational behavior learned from real speech.","published":"2026-09-25T03:55:18Z","updated":"2026-09-25T03:55:18Z","authors":["Manato Yaguchi","Yotaro Kubo","Hikaru Asano","So Kuroki"],"categories":["cs.CL","eess.AS"],"headline":"New Method Helps Speech Models Learn Natural Conversation Timing","summary":"Researchers studied tandem speech-to-speech systems where a speech part reacts to user speech while a text-based backend plans replies. They introduced \"randomized intermediate guidance,\" a way to train the speech part using actual conversation data without complex simulations. This method mixes helpful and random signals during training so the system learns when to rely on backend info. Tests on synthetic and real conversations showed it improves natural turn-taking and response quality, suggesting a practical approach to better talk-based AI."},{"id":"2609.30772v1","title":"Efficient Spin-Orbit Coupling in Algebraic Diagrammatic Construction Theory: A State-Interaction Approach","abstract":"Spin-orbit coupling and electron correlation play important roles in a broad range of chemical and physical phenomena. In this work, we systematically assess the performance of single-reference algebraic diagrammatic construction combined with state interaction (SI-ADC) for describing electronic structure in the presence of spin-orbit coupling. We find that second-order SI-ADC methods (SI-ADC(2)) achieve zero-field-splitting accuracies comparable to those of higher-level and more computationally demanding approaches, including variational four-component ADC and two-component ADC methods based on spin-orbit effective Hamiltonians. We further demonstrate the broad applicability of SI-ADC by evaluating a diverse range of spin-orbit-coupled properties, including electron affinities, ionization potentials, neutral excitation energies, core-ionization energies, and magnetic g-tensors, for systems ranging from small molecules to transition-metal complexes and f-block compounds. Across these applications, SI-ADC(2) methods provide consistently accurate results, likely benefiting in part from favorable error cancellation between electron-correlation and spin-orbit effects. In contrast, SI-ADC(3) methods are generally less accurate, particularly for systems containing transition-metal and heavy elements. Overall, these results establish SI-ADC as an efficient and broadly applicable framework for incorporating spin-orbit coupling into ADC calculations of electronic and spectroscopic properties.","published":"2026-09-25T03:52:54Z","updated":"2026-09-25T03:52:54Z","authors":["Nicholas Yiching Chiang","Rajat Majumder","Terrence L. Stahl","Ning-Yuan Chen","Alexander Yu. Sokolov"],"categories":["physics.chem-ph"],"headline":"New Method Improves Study of Spin-Orbit Effects in Molecules","summary":"Researchers tested a method called state-interaction algebraic diagrammatic construction (SI-ADC) to better understand how electrons behave when spin and motion interact, important in chemistry and physics. They found that the second-order SI-ADC method gives accurate results similar to more complex techniques but with less computational effort. The study shows this method works well for various properties and types of molecules, especially with lighter elements, but higher-order versions may be less reliable for heavier elements. The findings are based on theoretical calculations using this approach."},{"id":"2609.30771v1","title":"Pinch's conjecture on $a$-convexity","abstract":"For $a$ in $\\mathbb{R}$, a subset $V$ contained in $\\mathbb{R}^{n}$ is said to be $a$-convex if $x, y \\in V \\Longrightarrow a x + (1-a) y \\in V$. According to Pinch [Math. Proc. Cambridge Philos. Soc., 1985], the $a$-convex hull of $V$ is the intersection of all of the $a$-convex subsets of $\\mathbb{R}^{n}$ that contain $V$, and Pinch also defines $D(a)$ as the $a$-convex hull of $\\{ 0, 1 \\}$ in $\\mathbb{R}^{1}$. Pinch conjectured that if $a$ is a totally real algebraic integer and $D(a)$ has no limit points, then every algebraic conjugate of $a$ other than $a$ is in $(0, 1)$. We succeed in proving this conjecture, which seems to have remained open.","published":"2026-09-25T03:51:28Z","updated":"2026-09-25T03:51:28Z","authors":["John M. Campbell"],"categories":["math.NT","math.CO"],"headline":"Mathematicians prove Pinch's conjecture on special number sets","summary":"Researchers studied a problem about special sets called 'a-convex' sets, which are defined using a number 'a' and points in space. They focused on a particular case involving the number 'a' and a set formed from 0 and 1, related to a question posed by Pinch in 1985. Pinch conjectured a specific property of these sets when 'a' is a certain type of number and the set lacks limit points. The researchers claim to have successfully proved this mathematical conjecture, which had remained unsolved."},{"id":"2609.30770v1","title":"NavGen: Visual Generative Models as a Scalable Data Engine for Embodied 3D Navigation","abstract":"General-purpose robot models increasingly rely on large and diverse datasets. For embodied 3D navigation, however, existing data sources face a fundamental trade-off: simulated data can be generated at scale but often suffer from the visual sim-to-real gap, whereas real-world flight data provide realistic observations but are costly to collect. This paper studies another direction: the use of high-fidelity visual generative models as scalable data engines for embodied 3D navigation. We introduce NavGen, a text-to-video data generation pipeline that produces diverse vision-language navigation (VLN) episodes across indoor and outdoor scenes. We also propose a style-diversification method that scales up long-tail data that are difficult and costly to collect. The resulting dataset contains approximately 400K navigation episodes. We evaluate our dataset against existing UAV navigation datasets across multiple metrics, and find that the model trained on our data generally improves with scale, outperforming those trained on existing datasets. To validate real-world transferability, we deploy the trained model in world-action-model paradigm to real-world flying experiments. The final model achieves a 75\\% success rate across different navigation tasks and environments.","published":"2026-09-25T03:49:44Z","updated":"2026-09-25T03:49:44Z","authors":["Xijie Huang","Yongyang Wan","Chengbin Dong","Zimo Ding","Mo Zhu","Yijin Wang","Zhiyang Liu","Fei Gao","Yuze Wu","Xin Zhou"],"categories":["cs.RO","cs.AI"],"headline":"Using AI-Generated Videos to Train Robots for 3D Navigation","summary":"Researchers explored using advanced AI to create realistic video data for training robots to navigate 3D spaces. They developed NavGen, a system that generates large, diverse sets of navigation scenarios indoors and outdoors, producing about 400,000 episodes. Their AI-trained models performed better than those trained on traditional data, with tests showing a 75% success rate in real-world flying tasks. These findings suggest AI-generated data could help overcome limitations of current simulation or real-world datasets, though results are based on specific navigation tasks and tested environments."},{"id":"2609.30769v1","title":"Query-Conditioned Prototype Adaptation for Cross-Domain Few-Shot Learning: Single-Query Inference, Controlled Comparisons, and Failure Modes","abstract":"Cross-domain few-shot learning requires adapting a classifier to a new visual domain from very few labelled examples without target-time parameter updates. We isolate one question: under a fixed global representation, what does joint query-support adaptation contribute to prototype construction? The Within-Instance Prototypical Transformer (WIPT) implements single-query test-time prototype adaptation by jointly transforming one unlabelled query and the labelled support embeddings, then forming query-specific class means. Using a shared frozen ViT-S/16 encoder, miniImageNet source training, and CUB, EuroSAT and ISIC targets, we replicate the key comparisons across five independent training seeds. In 1-shot evaluation, WIPT improves frozen ProtoNet in every run on CUB (+0.21 percentage points) and EuroSAT (+2.07), but decreases ISIC (-0.22). In 5-shot evaluation, ProtoNet remains strongest overall, while WIPT consistently improves a capacity-matched support-only Transformer on ISIC (+0.99). Joint processing of up to five queries yields no reliable accuracy gain; in a head-only 5-shot benchmark, g = 5 reduces analytical attention-token pairs by 73% and peak allocated memory by 29% relative to g = 1, although latency is non-monotonic. Across all target/shot conditions, WIPT changes uncertain ProtoNet decisions far more than confident ones, and rescue/break decomposition accounts for the observed gains and losses. Source-shift and scorer controls further show that the benefit is not universal. Overall, WIPT provides a streaming-compatible form of test-time prototype adaptation that can improve difficult low-shot cross-domain decisions without target-time optimization.","published":"2026-09-25T03:49:42Z","updated":"2026-09-25T03:49:42Z","authors":["Rushab Rasik Karania","Tomas Maul"],"categories":["cs.CV","cs.LG"],"headline":"Researchers Explore Adapting Visual Classifiers with Few Examples","summary":"The researchers studied how to improve image recognition when only a few labeled examples from a new domain are available, without updating the model during testing. They developed a method that adapts prototypes (representative examples) by jointly processing a new image and known examples to create query-specific class averages. Their tests showed some improvement on certain datasets but not all, indicating the method helps in some cases but is not universally better. The approach works without changing model parameters at test time."},{"id":"2609.30768v1","title":"Does Thinking Help Fairness? Reasoning Tokens Resolve Some Biases but Create More","abstract":"Thinking in reasoning language models (RLMs) has been subject to debate on whether it resolves or amplifies bias. Prior works have shown competing conclusions in both directions. Using a within-model thinking-vs.-non-thinking ablation across QwQ-32B, DeepSeek-R1-Distill-Qwen-32B, and Qwen3-32B on three high-stakes decision tasks (Adult, COMPAS, Credit), we show that thinking has an asymmetric dual effect on counterfactual fairness: it both resolves counterfactual flips produced by the non-thinking baseline and creates new flips at near-saturating model confidence. In all nine (model, dataset) combinations, the created flips outnumber the resolved flips by roughly 5 times. To explain the effect, we treat the thinking trace itself as a measurable site of fairness change and study it through two dynamic instruments: 1) We propose Counterfactual Depth Probability Gap (CDPG) to track bias evolution along thinking depth, and observe that bias propagates and amplifies with thinking. 2) We also formulate the Bias Transition Matrix (BTM) to show how predictions of counterfactual pairs change from non-thinking to thinking, and find that the asymmetric dual effect originates in the pair-state joint transition.","published":"2026-09-25T03:46:53Z","updated":"2026-09-25T03:46:53Z","authors":["Deng Pan","Joe Germino","Yihong Ma","Elizabeth Daly","Nuno Moniz","Ting Hua","Nitesh Chawla"],"categories":["cs.AI"],"headline":"Thinking in AI models reduces some biases but creates more","summary":"Researchers studied whether reasoning steps in language models affect fairness in decision tasks like credit and criminal risk assessment. They found that while 'thinking' fixes some unfair predictions, it also causes more new biased outcomes with high confidence. The study analyzed this dual effect across multiple models and datasets, showing bias tends to grow through the reasoning process. These findings come from simulations comparing model behavior with and without reasoning steps, highlighting complex impacts on fairness."},{"id":"2609.30767v1","title":"Lipschitz stability in the simultaneous recovery of wave speed and initial source","abstract":"We consider a inverse problem for a hyperbolic equation. The objective is to simultaneous recover both the spatially varying coefficient, whose square root represents the wave speed, and the initial displacement from a single pair of Dirichlet-Neumann Cauchy data. By introducing a suitable transformation and differentiating the resulting equation in time, we derive a coupled system for the differences of the unknown coefficient and initial displacement. Combining weighted energy estimates with Carleman estimates, we establish a Lipschitz stability estimate for both unknown quantities in appropriate Sobolev norms. The proof relies on a priori regularity assumptions, non-degeneracy conditions on the initial data, and structural inequalities controlling the interaction between the coefficient and initial-displacement differences. Our result extends coefficient-only stability results to a simultaneous coefficient--initial-displacement identification problem.","published":"2026-09-25T03:46:46Z","updated":"2026-09-25T03:46:46Z","authors":["Shiqi Ma","Qi Zhai"],"categories":["math.AP"],"headline":"Researchers improve methods to recover wave speed and initial source","summary":"The researchers studied a mathematical problem involving waves, aiming to find both the wave speed and the initial disturbance from limited measurements. They developed a method using advanced estimates to show a stable way of recovering these two unknowns simultaneously. Their findings extend previous work that focused only on the wave speed, but the results depend on specific assumptions about the data and mathematical conditions, and are theoretical rather than experimental."},{"id":"2609.30766v1","title":"WeEnv: The Environment for Agentic Reinforcement Learning at WeChat","abstract":"Agentic reinforcement learning (RL) differs from conventional RL in that every task executes inside a complex environment, e.g., a virtual machine or a container. We find that agentic RL pays a heavy environment tax: a large share of the iteration time goes to the environment rather than to learning. The root cause is the lack of a full-lifecycle solution to environment management. We present WeEnv, which manages environments across packaging, initialization, and provisioning. WeEnv packages components as independently published layer groups and composes them at initialization, so that updating a component republishes one small group rather than every artifact containing it. To speed up environment initialization, WeEnv launches environments instantly and fetches contents on demand. During task execution, WeEnv provisions CPU and memory elastically, adjusting each environment's quota from its observed usage to fit the varying demands. WeEnv reduces the initialization by 5.6-14.2x over E2B, Docker, and AgentENV, cutting its share of the iteration time from up to 53.4% to 9.1%. WeEnv is deployed for agentic RL at WeChat.","published":"2026-09-25T03:43:44Z","updated":"2026-09-25T03:43:44Z","authors":["Yang Yu","Jing Lei","Shaoxun Zeng","Xinyu Gao","Jindi Shi","Ci Lei","Junjie Zhang"],"categories":["cs.DC"],"headline":"New System Speeds Up Complex AI Learning Environments","summary":"Researchers developed WeEnv to tackle slowdowns in agentic reinforcement learning, where tasks run inside complex environments like virtual machines. They created a system to better manage these environments by breaking them into smaller parts, starting them faster, and adjusting computing resources as needed. Their tests show WeEnv cuts environment setup time by 5 to 14 times compared to other tools, reducing delays that once took over half the process time. The findings are based on deploying WeEnv at WeChat and focus on improving environment management, not the learning itself."},{"id":"2609.30765v1","title":"Insurance Reserve Intelligence Platform","abstract":"Insurance reserve estimation is a fundamental actuarial task supporting premium pricing, solvency assessment, financial reporting, capital planning, and risk management. Classical reserve methods based on Thiele's differential equation provide a rigorous and interpretable foundation for life insurance valuation, but repeated reserve calculations become computationally expensive in sensitivity analysis, optimization, and large-scale scenario evaluation. This paper presents an Insurance Reserve Intelligence Platform for term-life reserve modelling that combines a classical Thiele-equation solver with a Physics-Informed Neural Network (PINN) enhanced by Knowledge-Informed Neural Network (KINN) losses. The framework includes synthetic policy generation, risk-adjusted premium calculation, classical reserve trajectory generation, reserve-ratio dataset construction, configurable neural training, validation diagnostics, sensitivity and elasticity analysis, prototype optimization workflows, and interest-rate scenario testing. A key refinement is the use of premium ratio and the explicit separation of pricing-time and scenario-time interest-rate semantics. The final model uses seven features: elapsed time, issue age, pricing interest rate, scenario interest rate, premium ratio, sum assured, and mortality intensity. It predicts a standardized reserve ratio instead of raw reserve values, improving numerical stability across policies with different sums assured. The model achieved an R2 of 0.9887, MAE of 785.48, and RMSE of 1212.76 on the test set. On 200 policies, PINN/KINN inference was approximately 119.53 times faster than the classical solver. Results show strong predictive accuracy, physics consistency, and boundary performance, while highlighting remaining limitations in monotonicity and out-of-distribution generalization.","published":"2026-09-25T03:42:37Z","updated":"2026-09-25T03:42:37Z","authors":["Anugya A","Saket Mohanty","Abhilash Timmapur","Somya Rai"],"categories":["cs.AI"],"headline":"New AI Platform Speeds Up Life Insurance Reserve Calculations","summary":"Researchers developed a hybrid platform combining classical mathematical methods with neural networks to estimate life insurance reserves faster and accurately. Their system predicts a standardized reserve ratio using seven key features, improving stability and achieving strong accuracy. The AI approach performed about 120 times faster than traditional methods on test cases. However, the researchers note challenges remain in ensuring consistent behavior and handling data unlike what was seen during training."},{"id":"2609.30764v1","title":"A Sample of Long-Period Binaries with Dormant Compact-Object Candidates from LAMOST DR13 MRS and Gaia DR3","abstract":"Identifying dormant compact objects in non-interacting binaries is crucial for constraining the hidden population of stellar remnants. We present a systematic search for long-period compact-object candidates by combining Gaia DR3 astrometry with multi-epoch LAMOST DR13 medium-resolution spectroscopy. Starting from sources with elevated RUWE values, we required at least five LAMOST spectroscopic epochs and peak-to-peak RV variations exceeding 20 km/s. After RV extraction, orbital fitting, and subsequent screening, we obtained Bayesian Keplerian orbital solutions for 19 long-period single-lined binary candidates, with periods spanning approximately 30-1500 days. Multi-wavelength spectral energy distribution fitting was used to constrain the visible stars and evaluate the expected luminosities of unseen companions under a main-sequence assumption. Two systems are identified as the highest-priority compact-object candidates in our sample, each with a minimum companion mass exceeding 1 solar mass. Their spectra and optical-to-infrared photometry show no detectable secondary-light contribution, although such a contribution would be expected if the unseen companions were normal main-sequence stars. These systems therefore represent the strongest compact-object candidates in our sample, although further observations are required to distinguish massive white dwarfs or neutron stars from faint stellar companions or unresolved hierarchical systems. We also identify five intermediate-priority and 12 lower-priority systems for follow-up spectroscopy, ultraviolet imaging, and future astrometric constraints.","published":"2026-09-25T03:40:20Z","updated":"2026-09-25T03:40:20Z","authors":["Junxing Zhou","Wei-Min Gu"],"categories":["astro-ph.SR","astro-ph.GA"],"headline":"Scientists Find Possible Hidden Dead Stars in Binary Systems","summary":"Researchers searched for pairs of stars where one is a hidden dead star, like a black hole or neutron star, by combining star movement data from Gaia with repeated light measurements from LAMOST. They studied 19 binary star systems with long orbits and found two strong candidates where the invisible companion is likely a compact object, as its light is missing but its mass is too large to be a normal star. Although promising, more observations are needed to confirm what these unseen companions are."},{"id":"2609.30763v1","title":"HCOE: Hyperbolic Clinical Ontology Embeddings from Biomedical Language Models","abstract":"Biomedical language models (LMs) encode textual semantics but do not explicitly preserve medical code hierarchies. We present Hyperbolic Clinical Ontology Embeddings (HCOE) for hierarchy-aware clinical concept representation. HCOE maps frozen BioBERT embeddings into a Poincare ball, combining parent-side and child-side ontology-guided contrastive learning with coarse-to-fine ontology-path aggregation. It uses International Classification of Diseases (ICD) codes organized by Clinical Classifications Software (CCS) and Anatomical Therapeutic Chemical (ATC) medication hierarchies. Evaluations show that HCOE performs best on ICD/ATC clinical relation prediction and CCS-to-PheCode hierarchy transfer. On the MIMIC-IV dataset, HCOE also achieves the best performance on mortality prediction, readmission prediction, medication recommendation, and rare drug prediction.","published":"2026-09-25T03:39:25Z","updated":"2026-09-25T03:39:25Z","authors":["Yixuan Li","Weihao Li","Ziyang Song"],"categories":["cs.AI","cs.LG"],"headline":"New Method Improves Medical Code Understanding in AI Models","summary":"The researchers developed Hyperbolic Clinical Ontology Embeddings (HCOE) to help AI better understand the hierarchical relationships in medical codes. They transformed existing medical language model data into a special geometric space that respects code hierarchies, improving predictions related to clinical relations and patient outcomes. Their tests on medical data showed HCOE outperformed other methods in tasks like predicting mortality and recommending medications, but these findings are based on specific datasets and model setups."},{"id":"2609.30762v1","title":"A Generalized Stein Lemma for Quantum Channels","abstract":"We prove a generalized quantum Stein lemma for parallel discrimination of an arbitrary finite-dimensional channel against families of alternative channels. The alternatives are compact and convex, closed under tensor products, and contain a faithful replacer. At every fixed type-I error tolerance, the optimal type-II error exponent equals the regularized Umegaki channel relative entropy minimized over the alternatives, and both defining limits exist. Inputs may be entangled across channel uses and with a reference system. The main technical result constructs exactly trace-preserving approximations with exponentially small diamond error, dominated in completely positive order by free channels at every rate above the common exponent. The proof combines uniform auxiliary map approximation, iterative reduction of the domination rate, and tensor amplification. It yields asymptotic equipartition with trace-preserving smoothing at every subexponentially vanishing error, an exponential strong converse, and stability under vanishing diamond-norm perturbations. Under additional permutation invariance and closure under insertion of a fixed faithful input state and discarding of its output, we also prove an explicit finite-block completion bound. This quantitative construction uses weighted discarding, local operator corrections, and comparison with auxiliary extensions. Applications include state-preserving, coherence-restricted, covariant, entanglement-breaking, and positive-partial-transpose channels. For isometric targets against entanglement-breaking or positive-partial-transpose alternatives, we obtain exact finite-block testing and smoothing formulas.","published":"2026-09-25T03:39:14Z","updated":"2026-09-25T03:39:14Z","authors":["Minbo Gao","Zhengfeng Ji","Chenghua Liu"],"categories":["quant-ph"],"headline":"New Quantum Tool Measures How Well Channels Can Be Differentiated","summary":"Researchers developed a general mathematical result to distinguish a specific quantum process from many alternatives, even when complex correlations are involved. They proved how to optimally tell them apart with minimal error, using certain advanced quantum entropy measures. The findings are theoretical, based on abstract mathematical approximations and limit processes, and support various types of quantum channels, but rely on strong assumptions about channel properties and error tolerances."},{"id":"2609.30761v1","title":"Timo: $\\textbf{T}$aming Mult$\\textbf{i}$modal Diffusion Transformer for Human $\\textbf{Mo}$tion Generation","abstract":"Most existing human motion generation (HMG) methods use cross-attention modules to inject text semantics, but ignore the importance of bidirectional modeling between motion and text tokens, which limits text comprehension. A straightforward idea is introducing multimodal diffusion transformers (MMDiT), which have shown effective joint text--visual modeling in vision generation, into HMG. However, we find that articulated motion is temporally coherent but weakly correlated across joints, in which directly applying an MMDiT with flow matching produces poorly coordinated and jerky motion. In this work, we propose Timo, a novel kinematics-aware MMDiT framework tailored for HMG. Timo combines fully shared multimodal attention for bidirectional text--motion modeling with flow matching, geometric and rotational-kinematics supervision that compares actual rotations and their changes over time, and a two-stage curriculum progressing from broad motion learning to detailed caption alignment. Further, we construct a benchmark of $40{,}025$ held-out clips from six public datasets spanning diverse actions, assessing six complementary dimensions under a common evaluator and scoring protocol. Our model substantially outperforms state-of-the-art methods in both quantitative and qualitative evaluations. Remarkably, Timo surpasses Kimodo on five of six dimensions, achieving a $40.8$% relative improvement in the average benchmark score. Project page: https://kyfafyd.wang/projects/timo. Demo page: https://timo.kyfafyd.wang.","published":"2026-09-25T03:39:00Z","updated":"2026-09-25T03:39:00Z","authors":["Zhao Wang","Jiangtao Hu","Jack Yu","Tao Yu"],"categories":["cs.CV"],"headline":"New AI model improves text-based human motion generation","summary":"Researchers developed Timo, a new method that better connects text descriptions with human motion data for generating movements. Unlike previous methods, Timo uses a special transformer model that understands how body parts move over time and improves coordination using kinematics principles. The team tested Timo on a large benchmark with diverse actions and found it outperformed existing models, suggesting it better creates realistic and well-aligned motions from text instructions. Their findings are based on experiments combining multiple datasets and evaluation metrics."},{"id":"2609.30760v1","title":"An Upper Bound for the Mean Speed of Transition Fronts and Unbounded Front Widths for Fisher KPP Equations in Almost Periodic Media","abstract":"In this paper, we investigate transition fronts and spreading solutions of Fisher--KPP equations in one-dimensional almost periodic media, both on the real line and on the lattice. Let $λ_1$ be the supremum of the spectrum of the linearized operator acting on $L^2(\\R)$ or $\\ell^2(\\Z)$, respectively, and let \\(L(λ_1)\\) denote the spatial Lyapunov exponent at the spectral parameter \\(λ_1\\). We prove that, if $L(λ_1)>0$, the global mean speed of any transition front is at most $λ_1/L(λ_1)$. Moreover, if a solution of the Cauchy problem spreads faster than this bound, its transition width is unbounded along a sequence of times. This occurs for a class of initial data with slowly decaying exponential tails. This contrasts with periodic media, where pulsating fronts exist at every speed above the minimal speed. As applications, we consider almost Mathieu coefficients and a continuous quasiperiodic coefficient with two frequencies. Together with the results of Nadin--Rossi and Liang--Wang--Zhou--Zhou \\citep{LWZZ24}, our results provide an almost complete picture of the admissible speeds of generalized transition fronts, while leaving the critical cases open.","published":"2026-09-25T03:37:53Z","updated":"2026-09-25T03:37:53Z","authors":["Xing Liang","Linfeng Xu","Qi Zhou","Tao Zhou"],"categories":["math.AP"],"headline":"Limits on Speed and Width of Waves in Complex Media","summary":"Researchers studied how waves called transition fronts spread in one-dimensional spaces with almost repeating patterns. They found a maximum average speed for these fronts based on mathematical properties of the system. If a wave moves faster than this speed, its width becomes very large at certain times. These results highlight differences from perfectly repeating environments and help clarify possible wave speeds, though some special cases remain unresolved."},{"id":"2609.30759v1","title":"Design and Characterization of a Variable-Length Continuum Mechanism with Force Locking","abstract":"The utility of flexible continuum mechanisms for dexterous navigation is often impaired by their low stiffness, making them ineffective at manipulation in high-force scenarios. To address this challenge, we propose a novel continuum mechanism that achieves both flexible and rigid behavior by antagonistic extension and contraction of a rod-driven continuum helical structure. The helical design combines variable-length capacity with force locking for workspace and stiffness enhancement. In this article, we present the detailed design of the proposed mechanism and characterize its performance through experiments that quantify bending and stiffness. The results demonstrate 180 degree bending range of motion with an average distal positioning error of <10%. Further tests demonstrate that force locking directly improves axial stiffness and thus indirectly increases bending stiffness anisotropically, with maximum bending stiffness along load paths with a large axial component. Tensioning the driving rods provides additional stiffness tunability in the force-locked state, where increasing rod tension proportionally increases bending stiffness with a dimensionless gain of 0.56.","published":"2026-09-25T03:37:30Z","updated":"2026-09-25T03:37:30Z","authors":["Katelyn King","Veronica Fish","Allison M. Okamura"],"categories":["cs.RO"],"headline":"New Flexible Robot Arm Design Stiffens on Demand","summary":"Researchers created a flexible robot mechanism that can switch between bendable and stiff states using a special helical rod system. Their experiments showed it bends up to 180 degrees with small positioning errors and that locking the rods increases stiffness, especially when pushed along certain directions. They also found they can control how stiff it gets by adjusting rod tension. These findings come from controlled tests and help understand trade-offs between flexibility and strength in such designs."},{"id":"2609.30758v1","title":"LLPR: Location-aware learning and physics-based reconstruction for raindrop removal from a single image","abstract":"Raindrops can cause occlusion and distortion in the background scenes due to their adherence to windows or camera lenses. Existing raindrop removal methods concentrate on designing sophisticated CNN or Transformer architectures to recover distorted and missing texture. In this paper, we try to integrate location information and physical model into off-the-shelf CNN or Transformer architectures to help improve their performance. Specifically, we notice that existing methods deploy a preprocessing sub-network to generate a binary or soft mask to indicate the raindrop location, which will increase the network parameters and computational complexity. In contrast, a location-aware learning branch is embedded to teach the encoder in the training phase with the capability of perceiving the position of the raindrops. Note that this location-aware learning branch can be removed during the inference process (achieving performance improvements at no cost). Furthermore, instead of directly reconstructing the raindrop-free image (i.e., background scene), we devise a physics-based reconstruction scheme to first learn the transparency matrix and the raindrop layer. The latent background layer is then reversely derived based on the physical model. By combining the above-mentioned components, we propose our location-aware learning and physics-based reconstruction (LLPR) framework for this challenging ill-posed problem. We also collect a real-world raindrop-degraded image dataset, which is challenging for single-image raindrop removal (SIRR) methods. Extensive experimental results demonstrate the effectiveness and generality of our LLPR framework, achieving superior performance against state-of-the-art SIRR methods. The code will be made available upon acceptance.","published":"2026-09-25T03:36:46Z","updated":"2026-09-25T03:36:46Z","authors":["Zewei He","Xingyu Liu","Xing Luo","Guizhong Fu","Zixuan Chen","Yu Chen","Jinlei Li","Zhe-Ming Lu"],"categories":["cs.CV"],"headline":"New method improves removing raindrops from single photos","summary":"Raindrops on cameras or windows blur and distort background scenes, making it hard to restore clear images. Researchers developed a new approach that teaches an AI to recognize raindrop positions during training without slowing down actual use, and uses a physics-based model to separate raindrop effects from the background. Their method, tested on a challenging real-world dataset, shows better results than existing techniques, although findings are based on image data and computational experiments rather than real-time use."},{"id":"2609.30757v1","title":"Maltsev Constraint Satisfaction Problems and Deterministic Logspace With Counting","abstract":"In this article, we prove that the problem of solving $\\operatorname{CSP}(\\mathbf{A})$, where $\\mathbf{A}$ is a finite relational template which admits a Maltsev polymorphism is in a specific complexity class DET, which is related to the complexity of computing the determinant of a matrix with integer entries. Such a class is intimately related to well-studied MOD-logspace classes in the theory of computational complexity. To prove this fact, we develop a new algorithm for solving syntactically simple binary instances of Maltsev constraint satisfaction problems, rather different from the well-known Bulatov-Dalmau algorithm, which does not require the explicit use or knowledge of a Maltsev polymorphism of the template but, rather, utilizes a graph whose vertices are 2-generated subuniverses of $\\mathbb{A}$, where $\\mathbb{A}$ is the Maltsev algebra parametrizing $\\operatorname{CSP}(\\mathbf{A})$. The theoretical importance of this algorithm is reflected in two facts: (1) it places the problem $ \\operatorname{CSP}(\\mathbf{A})$ in a complexity class related to the deterministic logspace with counting, which, in itself, has a strong connection to a variety of standard algorithmic problems in linear algebra, and (2) it only makes use of the relational structure of the template without the need for the explicit use of a compatible Maltsev polymorphism, depending entirely on the strong ``symmetry\" of constraints compatible with such polymorphisms and the knowledge of 2-generated subuniverses of $\\mathbb{A}$.","published":"2026-09-25T03:34:15Z","updated":"2026-09-25T03:34:15Z","authors":["Dejan Delic","Ali Syed"],"categories":["cs.CC"],"headline":"New Algorithm Links Certain Puzzle Problems to Efficient Computation","summary":"Researchers studied a type of problem called Maltsev Constraint Satisfaction Problems, which involve finding solutions under specific symmetrical rules. They developed a new algorithm that places these problems within a computational class related to efficiently computing matrix determinants, connecting them to known problems in linear algebra. Importantly, their method does not require detailed knowledge of the underlying symmetrical rules but uses the structure of related mathematical objects. This result is theoretical and does not include practical applications or empirical testing."},{"id":"2609.30756v1","title":"Selective Amortization of Full-Budget Counterfactual Reasoning for Visual Token Communication","abstract":"Generative image communication transmits compact semantic tokens under a limited packet budget, where token selection directly affects the final reconstruction quality after the complete packet is decoded. However, accurately estimating the terminal value of every candidate token requires repeated receiver-side reconstruction, resulting in substantial encoder-side computation. To address this problem, we propose ACV-Gate, an adaptive candidate evaluation framework that learns to approximate full-budget counterfactual evaluation and selectively assigns exact evaluations to the most informative candidates. Specifically, a set-aware student is trained using terminal advantages and regrets to predict candidate rankings directly, while a selective refinement mechanism evaluates only a bounded candidate set containing both Local-MDL and direct actions; cost-based thresholds further enable explicit control of the average evaluation workload. Experiments on CIFAR-10 show that ACV-Gate consistently improves reconstruction quality while substantially reducing candidate evaluations; at 0.20 bpp, the primary adaptive configuration improves PSNR over LocalMDL by 0.636 dB with only 2.13 candidate evaluations per image, corresponding to 27.60% of the calls required by the Exact-Full expert. Matched-candidate comparisons, synchronized GPU measurements, and evaluations on STL-10 and 384 *384 scale transfer further demonstrate consistent quality computation trade-offs, with particularly pronounced gains at low bit rates. These results show that combining terminal-value learning with selective candidate evaluation provides an effective and controllable mechanism for allocating encoder computation in packet-constrained generative image communication.","published":"2026-09-25T03:31:35Z","updated":"2026-09-25T03:31:35Z","authors":["Qinglei Qi","Zhihe Liang","Fengzhan Jing","Shenao Zhu","Lei Zhang","Chenyang Zhang","Shuqing He","Jia Guo"],"categories":["cs.AI"],"headline":"Improving Image Token Selection to Save Computing Power","summary":"The researchers address the problem of efficiently selecting visual tokens for image communication under limited data budgets, which normally requires heavy computation to predict how each token affects image quality. They developed ACV-Gate, a method that learns to estimate which tokens matter most and only fully evaluates a small, important subset. Tests on image datasets showed their approach improves image quality while reducing computations, especially at low data rates. These findings are based on experiments and simulations with no direct real-world deployment yet."},{"id":"2609.30755v1","title":"Training-Free Bottleneck Width Planning for Convolutional Autoencoders","abstract":"Multiscale Spectral Rate-Distortion (MS-SRD) estimates the bottleneck channels required at user-supplied spatial cuts from training images and a normalized mean-squared error (NMSE) bound, without fitting a neural network. Its covariance-tail rule is exact for shared linear block-convolutional autoencoders under squared error. A nested-scale dominance result motivates reporting the activation-parameter Pareto frontier alongside the minimum-latent candidate. At NMSE <= 0.01 on thirteen grayscale datasets, its latent-size prediction has 0.84% mean absolute percentage error against nonlinear patch-autoencoder boundaries; ten predictions are exact and the remaining three differ by one channel. In a four-dataset deployable comparison, MS-SRD matches all retrospective external widths and all four selected models pass, without training a selector; a 46-fit validation grid and four Least-Volume fits each pass on two datasets. In a skip-closed U-shaped autoencoder at the same bound, five predictions are exact, nine are within one channel, and every failing prediction is one channel short. Experiments at looser bounds show progressively larger nonlinear savings.","published":"2026-09-25T03:30:00Z","updated":"2026-09-25T03:30:00Z","authors":["Guannan Guo"],"categories":["cs.CV"],"headline":"Estimating Autoencoder Bottleneck Size Without Training","summary":"Researchers developed a method called Multiscale Spectral Rate-Distortion (MS-SRD) to predict the necessary size of a key part of convolutional autoencoders, known as the bottleneck, using only training images and an error limit. This approach requires no neural network training and matches or closely approximates the best sizes found by more complex methods across several datasets. Their results are based on theoretical analysis and simulations with various image datasets, but real-world performance beyond these tests is not addressed."},{"id":"2609.30754v1","title":"Genuine Multi-Entropy of Fully Symmetric Gaussian States","abstract":"We study the tri-partite genuine Rényi multi-entropy of fully symmetric Gaussian states as a function of the squeezing parameter for different total number of bosonic modes under different subsystem configurations. Exact results for a few modes are obtained, and we conjecture an asymptotic formula at the limit of large squeezing parameter as a function of the replica index $n$, while also finding subleading subsystem-size dependent terms for $n=3$ under a specific tripartition. We analytically show that the genuine $n=2$ Rényi multi-entropy of any pure bosonic Gaussian states is zero for any number of modes and for any tripartition. Furthermore, to investigate the dependence on subsystem size, we plot the curve of genuine Rényi multi-entropy as a function of the number of modes in the subsystems while keeping the total number of modes fixed. We discuss differences and similarities between this result and studies of the multi-entropy curve for Haar random states. In particular, we identify a subsystem size configuration related with the multi-entropy time in an evaporating black hole for which the behavior of the genuine multi-entropy changes.","published":"2026-09-25T03:29:42Z","updated":"2026-09-25T03:29:42Z","authors":["Hugo A. Camargo","Mitsuhiro Nishida"],"categories":["hep-th","cond-mat.stat-mech","quant-ph"],"headline":"Researchers analyze complex entropy in special symmetric quantum states","summary":"The researchers study a specific type of quantum entropy called genuine R\u0000e9nyi multi-entropy in fully symmetric Gaussian states, focusing on how it changes with a parameter called squeezing and different subsystem sizes. They provide exact results for small cases and propose a formula for large squeezing values, while also examining how subsystem size affects this entropy. Their work is theoretical, based on mathematical analysis and plotting curves, and highlights differences from previous studies on random quantum states; they note connections to theoretical models of black hole evaporation."},{"id":"2609.30753v1","title":"Skill Profiling with Attributable Reasoning (SPAR): A Wearable Analysis System for Boxing","abstract":"A punch is a ballistic, full-body action driven by a kinetic chain running from the legs through the trunk to the arm, where a small sequencing error separates a scoring strike from a miss. Wearable sensors can capture this movement in the gym, but most deployable systems only classify which punch was thrown rather than assess how well it was thrown. We present Skill Profiling with Attributable Reasoning (SPAR), an eight-IMU garment and pressure-insole system that classifies each punch as expert or novice and treats an explanation of that prediction as feedback. Feedback is only useful if the person receiving it can act on it, so SPAR explains the prediction at three tiers, a per-joint attribution for the analyst, a counterfactual over kinetic-chain layers for the coach, and a plain-language narrative of the two for the athlete. Across 17 participants and 4,713 punches, SPAR reaches a leave-one-participant-out AUC of 0.842 (95% CI [0.769, 0.907] over participants). A frozen time-series foundation model encodes the joint-angle and plantar-force series, and a small transformer trained on the cohort classifies the encoding. We audit the two quantitative tiers and report six themes from a thematic analysis of interviews with six practicing boxing coaches.","published":"2026-09-25T03:28:50Z","updated":"2026-09-25T03:28:50Z","authors":["Nibraas Khan","Hanchen David Wang","Enya Bullard","Ritam Ghosh","Ruj Haan","Aarav Agrawal","Meiyi Ma","Nilanjan Sarkar"],"categories":["cs.HC","cs.LG"],"headline":"New Wearable System Analyzes and Explains Boxing Punch Quality","summary":"Researchers developed SPAR, a wearable sensor system designed to not only identify different types of boxing punches but also evaluate how well each punch is executed by distinguishing expert versus novice technique. The system uses sensors on the body and feet to track motion and force, then provides feedback with explanations tailored for analysts, coaches, and athletes in simple language. Tested with 17 participants throwing thousands of punches, SPAR showed promising ability to assess punch skill levels, with feedback design informed by interviews with boxing coaches."},{"id":"2609.30752v1","title":"Differentiable RNA Secondary Structure Extraction for Deep Learning","abstract":"Many deep learning approaches to RNA secondary structure prediction have recently been proposed. They typically output a weight matrix $W$ where $W_{ij}$ is an arbitrary weight for base $i$ pairing with base $j$. Converting this matrix to a predicted secondary structure or base-pairing probability matrix typically involves ad hoc and problematic downstream algorithms. Despite the importance of this conversion step, which we refer to as structure extraction, it has received relatively little attention in the literature. In this work, we analyze how the congruence between training and extraction methods affects prediction performance. To do this, we compare four extraction algorithms: a Nussinov-like dynamic programming method, maximum-weight graph matching and the greedy extraction algorithms used by SPOT-RNA and RiNALMo. These are evaluated on outputs from the pretrained RiNALMo model and three toy models trained in this paper: a differentiable Nussinov-like model, a binary cross-entropy (BCE) baseline, and a model that incorporates a novel symmetric doubly stochastic matrix (SDSM) normalization algorithm during training which allows it to output base-pairing probability matrices directly, without a separate extraction step. This SDSM normalization algorithm is differentiable and can be added inline to any deep learning model during training and evaluation. We find that the performance of each extraction method depends strongly on how the corresponding model was trained. Considering the toy models themselves, the SDSM model showed the strongest overall performance: it outperformed the BCE baseline under all four extraction algorithms and produced pre-extraction outputs closest to the ground truth. These results suggest that SDSM normalization is a tractable alternative to traditional structure extraction.","published":"2026-09-25T03:28:29Z","updated":"2026-09-25T03:28:29Z","authors":["Tyler Illman","Max Ward","Marcell Szikszai","Ryan K. Krueger"],"categories":["cs.LG"],"headline":"New Method Improves RNA Structure Prediction in Deep Learning","summary":"Researchers studied how different ways of interpreting deep learning outputs affect predicting RNA secondary structures. They compared multiple extraction methods, including a new approach using a symmetric doubly stochastic matrix (SDSM) normalization that produces usable pairing probabilities directly. Their experiments showed that this new SDSM method consistently performed better than traditional approaches and baselines in predicting RNA base pairings. These findings suggest that SDSM normalization may offer a better alternative to existing steps that convert deep learning outputs into RNA structures, though results are based on trained models and simulations."},{"id":"2609.30751v1","title":"Backbone-Adaptive Evidence Routing for Robust Pairwise LLM Judging","abstract":"Pairwise language-model judges can gather evidence through direct comparison, reasoning, or reference-based verification, but no single protocol is best across benchmarks and judge backbones. We introduce Backbone-Adaptive Evidence Routing (BAER), which adapts the evidence mechanism while preserving candidate symmetry: swapping the two responses may reverse the preference but cannot change its strength. BAER separates each expert's signed preference from candidate-invariant reliability and builds three symmetric heads: evidence stacking, reliability-based expert routing, and candidate-blind reference verification. Development data select one head for each benchmark--backbone condition, and that choice is frozen before testing. Across four benchmarks and two 8B judge backbones, BAER achieves the highest test accuracy among the compared methods in all eight conditions, with full prediction coverage and gains of 0.87--7.32 points over the strongest external baseline. The results show that adapting how evidence is gathered is more reliable than fixing one judging protocol everywhere.","published":"2026-09-25T03:26:52Z","updated":"2026-09-25T03:26:52Z","authors":["Zeyan Li","Jing Peng","Jianfeng Xu"],"categories":["cs.AI"],"headline":"Adaptive Evidence Methods Improve AI Language Judging Accuracy","summary":"Researchers studied how AI systems judge language responses by comparing them in different ways, like direct comparison or reference checks. They created a method called Backbone-Adaptive Evidence Routing (BAER) that adapts how evidence is gathered to fit different tasks while keeping fairness between compared responses. Testing on multiple benchmarks showed BAER generally performed better than fixed methods, suggesting flexible evidence gathering can enhance AI judging reliability. These findings are based on controlled tests using specific AI models and datasets."},{"id":"2609.30750v1","title":"Sparsely connected rank-inspired neural network","abstract":"We propose a Sparsely Connected Rank-Inspired Neural Network (SCRINN) for function approximation and the numerical solution of partial differential equations (PDEs). By introducing a structured sparse connectivity architecture and an orthogonality- and normalization-guided loss function, SCRINN constructs an approximately orthonormal neural basis while promoting progressive rank growth of its evaluation matrix and reducing the number of trainable parameters. A two-stage framework is developed, in which an approximately orthonormal neural basis is first constructed, and the output weights are subsequently determined using least-squares methods. Numerical experiments on function approximation and a variety of steady-state, time-dependent, and nonlinear PDEs demonstrate that SCRINN consistently achieves high approximation accuracy with substantially lower training costs. The code is publicly available on GitHub.","published":"2026-09-25T03:26:01Z","updated":"2026-09-25T03:26:01Z","authors":["Yunqing Huang","Wu Su","Nianyu Yi","Peimeng Yin"],"categories":["math.NA"],"headline":"New Sparse Neural Network Improves Solving Mathematical Equations","summary":"Researchers developed a new neural network called SCRINN to better approximate functions and solve complex mathematical equations known as partial differential equations (PDEs). Their method uses a special sparse design and a unique training process that reduces the number of parameters while maintaining accuracy. Tests showed SCRINN achieves precise results with less training effort, but findings are based on computational experiments rather than real-world testing."},{"id":"2609.30749v1","title":"ORCA: Evaluating LLMs on Data Science Code Translation","abstract":"Data Science Code Translation (DSCT) is the process of converting code between data science libraries while preserving functional equivalence and enabling interoperability across data science ecosystems. While Large Language Models (LLMs) have demonstrated considerable progress in Data Science Code Generation (DSCG), their performance in DSCT remains insufficiently studied. To address this gap, we introduce ORCA, a comprehensive benchmark with two complementary settings: ORCA-MAIN, which comprises 1,600 carefully curated grounding-level tasks across 3 representative domains: Data Querying, Data Manipulation, and Deep Learning; and ORCA-PROJECT, which contains 200 translation tasks over complete data science projects across 7 data science task types. Each task is accompanied by annotated reference translations and test cases for validating functional equivalence. We further incorporate a multi-stage quality verification process that thoroughly verifies task correctness and test case robustness. Experimental results demonstrate challenges in DSCT, with even frontier LLMs showing limited performance. Specifically, Claude-Opus-4.6 achieves a success rate of 56.92% on ORCA-MAIN and 33.67% on ORCA-PROJECT, indicating considerable room for improvement in DSCT. We also observe a clear directional preference in DSCT, where translation is consistently easier when the source code expresses the task through more explicit, fine-grained operations. Motivated by this, we propose an intent-augmented method, in which the model first infers source-code intent and then uses it as additional context for translation, achieving average absolute success-rate gains of 4.80% and 5.33% on ORCA-MAIN and ORCA-PROJECT, respectively.","published":"2026-09-25T03:24:27Z","updated":"2026-09-25T03:24:27Z","authors":["Xiaolong Li","Jinyang Li","Bowen Qin","Ge Qu","Nan Huo","Xiaohan Xu","Shipei Lin","Reynold Cheng"],"categories":["cs.AI"],"headline":"Testing AI on translating data science code between libraries","summary":"Researchers studied how well advanced language models convert code among different data science tools, aiming to keep the code's function the same. They created ORCA, a new test with many tasks, to measure model accuracy in translating code for data queries, manipulation, and deep learning. Their results show current models often struggle, with success rates around 34-57%. They also found translations work better when the original code is explicit, and improving model awareness of code intent helps boost accuracy slightly."},{"id":"2609.30748v1","title":"Pfaffian Geometry and Indicial Singularities of Third-Order Linear ODEs","abstract":"We study third-order linear differential equations through their associated Pfaffian and projective foliations. In the regular-singular setting, we introduce a Fuchs-adapted Pfaffian model in which the indicial polynomial appears directly in the special-fiber singular scheme, and Frobenius resonance acquires a projective geometric interpretation.","published":"2026-09-25T03:24:26Z","updated":"2026-09-25T03:24:26Z","authors":["Víctor León","Bruno Scárdua"],"categories":["math.DG","math.DS"],"headline":"New geometric approach to certain third-order differential equations","summary":"The researchers investigate third-order linear differential equations using a geometric framework called Pfaffian and projective foliations. They developed a new model that connects an important polynomial related to singularities directly to the geometry of the problem. Their work offers a fresh way to understand resonance phenomena in these equations through geometric concepts. These findings are theoretical and focus on the mathematical structure rather than practical applications."},{"id":"2609.30747v1","title":"Beyond the Last Truffula Tree: SustainAI - A Water-Aware, Closed-Loop Framework for Environmentally Accountable AI","abstract":"As artificial intelligence (AI) becomes embedded in everyday life, its environmental footprint, particularly water consumption remains largely invisible. While energy and carbon impacts are widely recognized, the substantial freshwater demands of data center cooling and electricity generation receive little attention. To address this gap, we introduce SustainAI, a water-aware, closed-loop framework incorporating environmental accountability into AI deployment. SustainAI integrates real-time water metering, a hallucination-aware penalty model, and a water-aware routing algorithm that accounts for regional water stress. Evaluated via Small Language Models (SLMs) extracting health misinformation, results reveal an 11-fold variation in water footprint across geographically distributed data centers (0.0477 mL to 0.5360 mL per inference). Across 1,335 inference runs, the system consumed approximately 399 mL of water but produced only 240 correct outputs, demonstrating that substantial resources are spent on inaccurate responses. Crucially, SustainAI extends beyond technical optimization through a Care by Design lens, framing AI sustainability around relational ethics, regional equity, and ecological stewardship. By combining water monitoring, adaptive accountability, and Care by Design principles, SustainAI provides a practical foundation for integrating ethical care and environmental responsibility into AI infrastructure design and lifecycle management.","published":"2026-09-25T03:24:12Z","updated":"2026-09-25T03:24:12Z","authors":["Farnaz Farid","Tashfia Towkee","Sania Nasreen","Sami bin Azad"],"categories":["cs.CY","cs.AI","cs.DC","cs.HC"],"headline":"New Framework Makes AI's Water Use More Visible and Responsible","summary":"Researchers developed SustainAI, a system that tracks and reduces the water used by AI technologies, especially for data center cooling. By measuring water use in real-time and adjusting where AI work is done based on regional water stress, SustainAI aims to improve environmental accountability. Their tests showed large differences in water consumption depending on location and highlighted that many AI outputs are inaccurate despite significant water use. The framework also emphasizes ethical care and regional fairness in managing AI's environmental impact."},{"id":"2609.30746v1","title":"Mechanism-Aware Ensemble Conditioning for Data-Limited Emulation of Extreme Events","abstract":"Extreme events in chaotic systems are difficult to learn from short trajectories because they are controlled by transient finite-time instability rather than by frequently observed bulk dynamics. We propose a mechanism-aware conditioning plug-in framework that turns a nudged coarse ensemble into a non-intrusive sensor of local instability geometry. In the small-noise regime, the ensemble covariance aggregates the same finite-time deformation kernels that govern local instability, providing a Jacobian-free proxy for the local amplification structure around a synchronized coarse trajectory. A small FiLM module injects statistics of this ensemble geometry into an otherwise unchanged backbone while leaving the coarse simulator unchanged. We demonstrate this interface in two distinct pipelines: a Transformer-style residual-attention corrector for a controlled low-dimensional chaotic system and a probabilistic recurrent STORN corrector for topographic two-layer quasi-geostrophic (QG) flow. In the low-dimensional benchmark, ensemble covariance directions co-activate with OTD modes and FiLM conditioning improves 99th-percentile exceedance-frequency errors over an identical no-context Transformer baseline. In QG, a fixed ensemble-conditioned FiLM-STORN model trained on only \\(50\\) time units substantially improves long-horizon rare-event statistics in the data-limited regime, including density-tail errors, exceedance frequencies, and spatial exceedance-area distributions relative to an unconditioned STORN trained on the same data; on averaged high-threshold exceedance diagnostics, it also outperforms the baseline STORN trained with $20$ times more high-resolution data. These results show that local instability geometry is not merely interpretable post hoc, but an actionable conditioning signal for data-efficient rare-event emulation.","published":"2026-09-25T03:23:41Z","updated":"2026-09-25T03:23:41Z","authors":["Isabella S. Thiel","Juan Bello-Rivas","Yannis G. Kevrekidis","Themistoklis P. Sapsis"],"categories":["cs.LG","math.DS","physics.ao-ph"],"headline":"New method helps predict rare chaotic events with limited data","summary":"Researchers addressed the challenge of learning extreme events in chaotic systems from short data sequences, which is hard because these events depend on brief instability phases rather than common patterns. They developed a technique that uses local instability information as a conditioning signal to improve predictive models without changing the underlying simulator. Tests on simplified chaotic models and complex fluid flow simulations showed this approach better captures rare event statistics using much less data than traditional methods. The findings are based on simulation experiments and focus on improving rare-event prediction efficiency."},{"id":"2609.30745v1","title":"Anatomy-Aware Dexterity-Driven Design Optimization of Surgical Continuum Robots","abstract":"Performing complex medical procedures with continuum robots requires careful selection of their geometric design parameters. The robot should have high dexterity in the specific anatomical environment of its procedure. This work presents a design optimization method that considers both dexterity and anatomy. We introduce the Reachable Volumetric Dexterous Solid Angle (RVDSA) metric as our objective, which measures the ability of a robot's end effector to reach the points in a goal volume from different directions via collision-free paths from a start configuration. We present a computationally efficient motion planner to compute this objective function for a given robotic design, and we use an asymptotically optimal simulated annealing optimizer to compute an optimized design. We applied our new method to optimize the design of a bimanual dexterous sheaths robot for performing procedures on cancerous polyps in colon anatomies, achieving a 78% higher RVDSA on average than optimizing for 3D voxel coverage alone.","published":"2026-09-25T03:18:39Z","updated":"2026-09-25T03:18:39Z","authors":["Tony Qin","Peter Connor","Khoa Dang","Carter Hatch","Caleb Rucker","Robert J. Webster","Ron Alterovitz"],"categories":["cs.RO","cs.AI"],"headline":"New Method Improves Design of Surgical Robots for Specific Anatomy","summary":"The researchers addressed the challenge of designing flexible surgical robots that work well within the unique shapes of human anatomy. They developed a new way to measure how well a robot can move and reach within a target area safely, introducing a metric called RVDSA. Using computer simulations, they optimized a robot design for colon procedures and found their method increased reachability by 78% compared to older approaches, focusing on both movement ability and anatomical constraints."},{"id":"2609.30744v1","title":"Asymptotic Theory for Combining Dependent $p$-Values for Global Hypothesis Testing","abstract":"Combining $p$-values is a fundamental procedure in global hypothesis testing. In modern high-dimensional settings, however, component $p$-values often exhibit complex dependence and rely on asymptotic approximations rather than exact finite-sample uniform distributions. This paper establishes a unified asymptotic theory for weighted transformation statistics that decouples marginal finite-sample approximation error from the joint dependence structure. We also provide sufficient conditions based on conditional probability bounds to verify the joint-tail conditions. Utilizing this framework, we derive explicit dimension-growth and correlation rates for test statistics operating under asymptotic Gaussian and chi-square calibrations. For non-exact finite-sample statistics, we analyze standardized weighted sums, demonstrating how Cramér moderate deviations control relative tail error. Analytical examples demonstrate why both marginal and joint conditions are mathematically indispensable for valid global inference under dependence, and numerical experiments confirm that our asymptotic framework maintains accurate finite-sample size control at extreme significance levels.","published":"2026-09-25T03:15:29Z","updated":"2026-09-25T03:15:29Z","authors":["Haoyi Yang","Lingzhou Xue"],"categories":["math.ST"],"headline":"New Theory Improves Combining Dependent p-Values in Big Data Tests","summary":"Researchers developed a mathematical framework to better combine p-values—measures used to test overall scientific hypotheses—when those p-values are related or dependent, which often happens in large data studies. Their theory separates errors from individual tests and from their dependence patterns, providing conditions that help ensure valid combined results. The framework, supported by analytical examples and simulations, helps maintain correct error rates even in complex, high-dimensional settings, though conclusions are based on asymptotic (large sample) approximations rather than exact small-sample results."},{"id":"2609.30743v1","title":"From S3Q Theory to Implementation: Towards an Architecture for Machine Qualia","abstract":"A key challenge in machine consciousness research is translating theoretical models into computational-level implementations. In this paper, we address this challenge by proposing a five-layer implementation architecture for the S3Q (Simulated, Situated, Structurally Coherent) theory of consciousness. Rather than introducing novel formalisms, the architecture composes published computational primitives into a single pipeline. S3Q identifies three jointly necessary conditions for qualia: (1) grounded sensorimotor situatedness, (2) internal simulation via a world model, and (3) structural coherence between predictions and observations. No existing computational system implements all three simultaneously. We map each S3Q tenet to specific, compatible computational machinery and specify how these components interface within a single representation pipeline that operates on continuous, differentiable, per-object slot vectors, along with a developmental bootstrap sequence and falsifiable predictions for the composed system that no subset of the architecture produces in isolation. The model suggests that a basic sense of \"self\" develops by linking actions to their outcomes, and that behavior falls into three patterns (hesitation, curiosity, or avoidance) depending on how unexpected an outcome is and whether it is experienced as positive or negative. Each prediction is individually falsifiable, providing the field with a testable framework to advance our understanding of machine consciousness.","published":"2026-09-25T03:12:20Z","updated":"2026-09-25T03:12:20Z","authors":["Tetiana Grinberg","Katrina Schleisman","Patryk Laurent","Bogdan Udrea","Minda Myers","Brian Aufderheide","Luis El Srouji","Doyle Groves","Kevin Schmidt"],"categories":["cs.AI"],"headline":"Building a Computer Model for Machine Consciousness","summary":"Researchers tackle the challenge of turning a theory of consciousness, called S3Q, into a workable computer system. Their design combines three key ideas: sensing and interacting with the environment, simulating the world internally, and matching predictions to what is observed. They propose a step-by-step architecture that links these ideas and suggests a basic sense of self can emerge when actions connect to outcomes. Their approach offers specific, testable predictions to help future studies on machine consciousness."},{"id":"2609.30742v1","title":"Optimal Prophet Inequalities for Gain from Trade","abstract":"We initiate the study of prophet trading, an online trading model in which a trader interacts with sellers and buyers who arrive in a uniformly random order and whose prices are drawn independently from a common known distribution. The trader aims to maximize the expected gain from trade (GFT), with performance measured against an omniscient prophet that knows the realized arrival order and all prices in advance. Unlike the classical prophet inequality problem, the trader must make both purchasing and selling decisions while managing the inventory, which makes both the prophet benchmark and the analysis of online algorithms substantially more intricate. For arbitrary numbers of buyers and sellers, we propose a simple threshold-based algorithm and prove a distribution-free competitive ratio of~2, which is the best possible. Our main technical contribution is an exact characterization of the inventory process induced by threshold trading. By expressing the algorithm's expected GFT in terms of the cumulative holding probability at buyer arrivals, we derive an exact formula for this inventory term, which serves as the foundation of our analysis and yields sharper guarantees in several important special cases. For balanced trading (with $n$ sellers and $n$ buyers), we improve the competitive ratio to $(2n^2-n)/((n+4^{-n}-1)(n+1))$, which is strictly less than 2 for every $n>1$. For the single-seller case, we sharpen the analysis of the fixed-threshold algorithm to obtain an asymptotic competitive ratio of $2e^2/(e^2+1)\\approx1.76$. Furthermore, by exploiting the additional structure of this setting, we design an adaptive-threshold algorithm whose competitive ratio approaches $e/(e-1)\\approx 1.58$. Due to a symmetry property of our general algorithmic idea, the same $1.76$-competitive guarantee also holds for the single-buyer case.","published":"2026-09-25T03:12:06Z","updated":"2026-09-25T03:12:06Z","authors":["Xujin Chen","Xiaodong Hu","Changjun Wang","Qingjie Ye"],"categories":["math.OC","cs.DS"],"headline":"New strategies improve online trading against a perfect predictor","summary":"Researchers studied a model where a trader buys and sells from random buyers and sellers with known price patterns, aiming to maximize profit compared to a perfect predictor who knows all prices and order in advance. They developed a simple algorithm setting price thresholds and proved it performs within twice the profit of the perfect predictor, the best possible for general cases. For balanced and special cases with one seller or buyer, they improved performance guarantees with more detailed analysis and adaptive strategies. The results focus on theoretical performance bounds without real-world testing."},{"id":"2609.30741v1","title":"From Mono to Stereo: Accelerating Binocular Gaussian Splatting via Reprojection and Selective Patching","abstract":"Binocular rendering requires two nearby views of the same scene and therefore repeats substantial visibility and shading work. We present a 2D Gaussian Splatting (2DGS) pipeline that fully renders a dominant-eye RGB image and an alpha-weighted depth proxy, reprojects that image to the affiliated eye, and repairs uncovered pixels. Small interior gaps are interpolated, whereas larger disoccluded regions are identified as regions of interest (ROIs) and selectively re-rendered. The depth proxy reuses the alpha-blending weights computed during dominant-eye rasterization, avoiding a separate depth-rendering pass. An adaptive ROI generator localizes the required updates using reprojected image boundaries and optional connected center-hole detection. On DTU, Tanks and Temples, and MipNeRF-360, the method reduces the measured time of a sequential two-pass binocular reference by 15.5\\% to 28.8\\% and peak GPU memory by 6\\% to 11\\%. The corresponding affiliated-eye quality degradation is at most 1.3 dB PSNR, 0.02 SSIM, and 0.02 LPIPS, representing a measurable trade-off that requires application-specific perceptual validation. These results establish a practical efficiency-quality trade-off for controlled static-scene stereo rendering and motivate future evaluation under continuous motion and on physical VR hardware.","published":"2026-09-25T03:11:46Z","updated":"2026-09-25T03:11:46Z","authors":["Hongfei Zhu","Ling Zhou"],"categories":["cs.CV"],"headline":"Faster 3D stereo image rendering by reusing and fixing parts of one view","summary":"Researchers addressed the problem of rendering two close views of the same 3D scene, which usually repeats expensive calculations. They developed a method that fully renders one main eye's image with depth information, then reuses and adjusts it for the other eye by fixing small gaps and selectively re-rendering only necessary parts. Their tests show this approach speeds up rendering by about 15-29% and lowers memory use moderately, with minimal quality loss. The findings are based on controlled, static scenes and suggest further testing is needed for moving images or VR devices."},{"id":"2609.30740v1","title":"Energy convexity and uniformity of the $H$-surface flow in $\\mathbb{R}^{3}$ with Dirichlet boundary condition","abstract":"We prove that the energy functional associated with surfaces of prescribed mean curvature in $\\mathbb{R}^3$ exhibits a convexity property when restricted to maps from the unit $2$-disk having small Dirichlet energy and a fixed boundary value, provided that the $3$-form $H \\cdot \\text{vol}_{\\mathbb{R}^3}$ has a primitive satisfying certain bounds. Under milder assumptions on $H:\\mathbb{R}^3 \\to \\mathbb{R}$, we show that an analogous convexity estimate holds along the heat flow of the functional (the $H$-surface flow) when the initial Dirichlet energy is sufficiently small. This is done first for classical solutions, and then extended by approximation to weak solutions using a quantitative uniqueness result adapted from previous work on the harmonic map heat flow. As a consequence of the convexity estimate, we show that the $H$-surface flow with small-energy initial map of class $C^0 \\cap W^{1, 2}$ on the unit $2$-disk converges uniformly to a unique limit at infinite time, which solves the corresponding stationary problem (the $H$-surface system) with the same boundary value.","published":"2026-09-25T03:10:57Z","updated":"2026-09-25T03:10:57Z","authors":["Da Rong Cheng","Longzhi Lin","Xin Zhou"],"categories":["math.DG"],"headline":"Researchers study shape-changing surfaces with energy stability","summary":"The researchers investigated a type of geometric surface in three-dimensional space that changes over time according to a rule involving its curvature and boundary constraints. They proved that when the initial surface has low energy, a certain energy measure behaves in a stable, 'convex' way during this evolution. This finding allowed them to show the evolving surfaces settle down uniformly over time to a unique final shape meeting the boundary conditions. Their conclusions rely on mathematical assumptions and apply primarily to surfaces with small initial energy."},{"id":"2609.30739v1","title":"SEA-CLIP-Tiny: Efficient Multilingual Text-Vision Embedding for Southeast Asian Languages","abstract":"Multilingual text-vision embedding models are essential for cross-lingual image-text retrieval, but Southeast Asian languages remain poorly supported due to the region's linguistic diversity and limited data and computing resources. In this paper, we introduce SEA-CLIP-Tiny, a compact multilingual text-vision embedding model for Southeast Asia with fewer than 50M parameters. Our model adapts a CLIP-KD-style framework to Southeast Asian multilingual settings through regional data curation and multilingual teacher guidance. Experiments across seven Southeast Asian languages show that SEA-CLIP-Tiny achieves the strongest average retrieval performance among the evaluated student models, reaching 12.9%, 31.5%, and 42.2% at R@1, R@5, and R@10, respectively. Compared with MobileCLIP2, it improves average R@10 by 12.1 points while using 38.4% fewer parameters and lower measured CPU latency. These results highlight the importance of region-aware training for efficient multilingual text-vision models in Southeast Asia.","published":"2026-09-25T03:10:30Z","updated":"2026-09-25T03:10:30Z","authors":["Puja Ahmad Habibi","Faiz Assabil Firdaus","Ashvanth S","Ekapol Chuangsuwanich","Pume Tuchinda","Peerat Limkonchotiwat"],"categories":["cs.CL"],"headline":"New Compact Model Improves Image-Text Matching for Southeast Asian Languages","summary":"Researchers developed SEA-CLIP-Tiny, a smaller and efficient model designed to link images and text across seven Southeast Asian languages. They tailored the training data and guidance specifically for this diverse region to better support underrepresented languages. Tests showed their model performed better than others while using fewer resources, highlighting the benefit of focusing on regional language features in such technology."},{"id":"2609.30738v1","title":"Beyond Mean Attention: Diversity-Aware, Layer-Wise Scoring for KV Cache Eviction","abstract":"KV cache eviction methods such as SnapKV and PyramidKV rank tokens solely by mean attention over a small observation window. We study a unified score, $μ_i+λ_1σ_i+λ_2\\mathrm{corr}(i,S)$, adding attention dispersion across window queries and redundancy relative to selected tokens. For $λ_2<0$, the score penalizes similarity to selected tokens as in maximal marginal relevance (MMR), without extra forward passes. To test whether this relevance-diversity balance should vary with depth, we compare fixed global coefficients with three-segment and quadratic profiles. Only these depth profiles are searched on a development split under a $\\sinh$ reparameterization. On all 16 English LongBench datasets with Mistral-7B at a budget of 64 entries per layer, a single global diversification constant improves 13 of 16 datasets (macro +1.1); the gain holds at budget 32 and narrows at 128. Per-dataset search finds no detectable layer structure on most datasets; on passage retrieval it finds a large one: a mid-layer sign flip that rewards similarity and is worth +9.6 over the baseline at budget 64 and, without re-tuning, +13.2 over the global constant at budget 128. Ablations attribute the gain to the redundancy term; replaying every accepted search state on the held-out test set separates genuine structure from tuning noise.","published":"2026-09-25T03:08:00Z","updated":"2026-09-25T03:08:00Z","authors":["Tianfang Xie","Wei Zhu"],"categories":["cs.CL","cs.LG"],"headline":"New method improves how AI decides which memory to keep","summary":"Researchers studied how AI systems select important past information to remember, focusing on balancing relevance and diversity of tokens. They proposed a new scoring method that considers both average attention and redundancy, testing it on language tasks with the Mistral-7B model. Their findings show improvements on most test datasets, especially by adjusting scoring across model layers, though some gains appeared only in certain tasks like passage retrieval. Results are based on controlled experiments without real-world application tests."},{"id":"2609.30737v1","title":"Strain-controlled topology and quantum control at a cubic Ce$^{3+}$ crystal-field crossing","abstract":"Field-tuned crystal-field crossings provide a route to synthetic topology in localized rare-earth degrees of freedom when symmetry-resolved perturbations can independently control the crossing states. Here we show that a field-induced crossing in cubic Ce$^{3+}$ has precisely this structure. For $B\\parallel[001]$, the crossing is protected by a twofold rotation, while the $T_{2g}$ shears $ε_{xz}$ and $ε_{yz}$ break that protection and mix the two states through orthogonal pseudospin components. Together with magnetic-field detuning, these controls generate an isolated diabolical point with unit-magnitude Chern charge in a three-dimensional parameter space. Using the CeTe crystal-field scale gives an ideal single-ion crossing near $35.5$ T. A calibrated point-charge calculation gives a representative projected shear coupling $|α_{T_{2g}}|\\simeq18.1$ meV per unit tensor strain, corresponding to gaps of $0.036$ and $0.072$ meV for tensor shears of $0.10\\%$ and $0.20\\%$. We identify the associated elastic and magnetic signatures, discuss the limitations of the electrostatic estimate in a hybridizing Ce compound, and determine the adiabatic and coherence requirements for cyclic geometric control. The result is a symmetry-based framework that connects high-field crystal-field reconstruction, strain response, and parameter-space topology in rare-earth systems.","published":"2026-09-25T03:05:16Z","updated":"2026-09-25T03:05:16Z","authors":["A. Ghosh","J. T. Haraldsen"],"categories":["cond-mat.mtrl-sci","cond-mat.other"],"headline":"Researchers explore strain and magnetic control of crystal states in rare earths","summary":"The researchers studied how magnetic fields and strain (stretching or shearing) can precisely control certain energy states in a rare-earth material containing cubic Ce$^{3+}$ ions. They found that applying a magnetic field along a specific direction creates a protected crossing of energy levels, which strain can then mix, producing a unique topological feature called a diabolical point. Their theoretical calculations estimate the strain needed to open small energy gaps and discuss physical signatures and challenges for controlling these states, but results are based on models and require further experimental validation."},{"id":"2609.30736v1","title":"Recent advances in poled lithium niobate","abstract":"Lithium niobate is a versatile material for both classical and quantum photonics, recognized for its outstanding electro-optic and nonlinear optical properties. Through a process known as poling, periodic ferroelectric crystal domains can be engineered to enable quasi-phase-matched frequency conversion, efficient modulation, and the generation of quantum light sources. The emergence of lithium niobate on insulator technology has further enhanced its suitability for scalable integrated photonics, offering ultra-low optical losses and strong light confinement while retaining the material's inherent advantages. Here, the techniques used to fabricate and characterize periodically poled lithium niobate are reviewed. Key developments are discussed, offering insights into the future of domain engineering of lithium niobate.","published":"2026-09-25T03:05:01Z","updated":"2026-09-25T03:05:01Z","authors":["Kibret A. Messalea","Tim Weiss","Yang Yang","Hamed Arianfard","Brett C. Johnson","Alberto Peruzzo"],"categories":["physics.optics"],"headline":"How Engineered Lithium Niobate Improves Light Technologies","summary":"Researchers reviewed recent progress in shaping lithium niobate crystals to better control light for classical and quantum uses. By creating regular patterns in the crystal through poling, they enable improved frequency changes, light modulation, and quantum light generation. The study focuses on fabrication and analysis methods of these engineered crystals and highlights advances that could guide future improvements in light-based technologies."},{"id":"2609.30735v1","title":"Praxis: Distilling Physical Interaction Priors from Egocentric Videos for Generalizable Whole-Body Manipulation","abstract":"Mobile humanoid manipulation requires both reaching a usable workspace and preserving precise hand-object interactions as object poses and contact conditions change. Learning these behaviors from limited task-specific data remains challenging. To bridge this gap, we introduce Praxis, a whole-body manipulation framework that combines physical interaction priors from one-shot egocentric video demonstrations with closed-loop posture calibration and online perception. The framework coordinates three stages: vision-language-guided navigation toward target objects, closed-loop posture calibration to align the arm-hand workspace, and dexterous manipulation with synchronized upper- and lower-body control. Online visual feedback re-grounds demonstrated interaction geometry under new object poses and scene configurations, while tactile feedback adapts hand motions to actual contact conditions. Each manipulation skill is specified by one human demonstration, without task-specific manipulation-policy retraining. Experiments across five long-horizon manipulation tasks demonstrate spatial, visual, and cross-object generalization, as well as recovery from external physical disturbances across all three stages.","published":"2026-09-25T03:04:09Z","updated":"2026-09-25T03:04:09Z","authors":["Shuliang He","Ruiyan Xu","Bo Yue","Hengming Zhang","Huayi Zhou","Shuai Wang","Wei-Shi Zheng","Guiliang Liu"],"categories":["cs.RO"],"headline":"New Method Helps Robots Learn Whole-Body Tasks from Videos","summary":"Researchers developed Praxis, a system that helps humanoid robots perform complex tasks by learning from just one first-person video demonstration. It combines vision, body posture adjustments, and touch feedback to handle changing object positions and conditions without retraining for each task. Tests showed robots could navigate, adjust, and manipulate objects in different scenarios, even recovering from physical disturbances. The results are based on controlled experiments and focus on combining sensory feedback to improve robot interaction with objects."},{"id":"2609.30734v1","title":"Learning What to Skip: Counterfactual Credit Assignment for Efficient Multi-Agent LLM Workflows","abstract":"Multi-agent LLM workflows use planning, execution, verification, and summarization to improve task performance, yet the value of each component depends on the state already produced. Executing every component can waste computation or overwrite a correct intermediate answer. We formulate component omission as counterfactual credit assignment: full-workflow logs reveal the executed trajectory's reward, while controlled skip interventions reveal the consequences of omitting a future step. We introduce Learning What to Skip (LW2S), which learns action-specific safety models from these interventions and combines held-out calibration with domain-native guards to select skips. When an early skip is rejected, the controller can continue execution and reconsider a later component. Across mathematical reasoning, multiple-choice QA, and code generation with two instruction-model families, LW2S reduces recorded token cost while matching or improving aggregate full-workflow accuracy in the evaluated settings. Scale-up and second-topology experiments further examine component redundancy, while shared-error cases reveal why agreement alone is insufficient for skip selection. These findings connect efficient workflow execution to learning the conditional utility of individual components.","published":"2026-09-25T03:02:57Z","updated":"2026-09-25T03:02:57Z","authors":["Jinfeng Xu","Zheyu Chen","Ziyue Peng","Zheng Lin","Shuo Yang","Jinze Li","Zheng Xing","Mengran Li","Victor C. M. Leung"],"categories":["cs.AI"],"headline":"Researchers Teach AI When to Skip Steps for Efficiency","summary":"The researchers studied how multi-step AI processes can waste time by doing unnecessary parts. They created a method called LW2S that learns when it's safe to skip certain steps by testing what happens if a step is left out. Using this approach, their tests in math, question answering, and code writing showed they could reduce effort while keeping or improving accuracy. The findings help understand how to make AI workflows more efficient by deciding which parts really add value."},{"id":"2609.30733v1","title":"Amplify What You Gaze At: Target Saliency Boosting in Text-to-Image Generation","abstract":"Text-to-image generation has advanced in controlling what, where, and how objects appear, yet how visual attention is distributed among objects remains largely unexplored. In this paper, we introduce Target Saliency Boosting, a new task aimed at boosting the visual saliency of a specific object during text-to-image generation without requiring any visual priors. Our key insight is that visual saliency is inherently relative: boosting the saliency of a target object also depends on the global saliency distribution across all objects in the scene. Based on this insight, we propose GazeME, a lightweight framework that uses saliency-marked prompts, inserting learnable marker tokens around object descriptions to indicate which objects to visually emphasize or suppress. To learn these markers, we construct a saliency-semantics dataset that associates objects in image--prompt pairs with object-level saliency scores, and propose Saliency Prior Marker Activation (SPMA), a saliency-aware stochastic marker activation strategy that exploits relative saliency relationships for robust training. During inference, GazeME automatically inserts appropriate markers into the prompt, thereby directly enhancing the visual saliency of the target object. Extensive experiments demonstrate that GazeME effectively boosts target saliency while preserving both semantic alignment and image quality.","published":"2026-09-25T03:01:29Z","updated":"2026-09-25T03:01:29Z","authors":["Shengqi Dang","Zhengxi Yu","Feilin Han","Xingyu Lan","Nan Cao"],"categories":["cs.CV"],"headline":"Boosting Focus on Objects in AI-Generated Images","summary":"Researchers studied how to make AI-generated images highlight specific objects more clearly. They developed a method called GazeME that uses special tags in text prompts to tell the AI which objects to emphasize or tone down, without needing extra visual information. Their tests showed GazeME can successfully make chosen objects stand out in images while keeping the descriptions accurate and images high-quality. These findings are based on controlled experiments with their new approach."},{"id":"2609.30732v1","title":"TR-SSQP: A Trust-Region Method for Constrained Stochastic Optimization under Heavy-Tailed Noise","abstract":"We consider stochastic nonlinear optimization problems with deterministic equality constraints. While unconstrained stochastic optimization is well understood, the interplay between optimality and feasibility in the constrained setting poses significant challenges. Moreover, existing theoretical guarantees for constrained stochastic methods predominantly rely on bounded-variance assumptions, leaving the heavy-tailed noise regime largely unexplored. To address this gap, we propose a novel trust-region method within the stochastic sequential quadratic programming framework, termed TR-SSQP. Our method employs a normal-tangential decomposition in the step computation to balance optimality and feasibility. In addition, we incorporate a normalization mechanism in the design of the trust-region radius, together with Polyak momentum for gradient estimation, ensuring stable updates without gradient clipping. When the trust-region radius and the momentum parameter decay at appropriate rates, we establish global almost-sure convergence of the method. To the best of our knowledge, this is the first asymptotic convergence result for constrained stochastic optimization under heavy-tailed noise. We demonstrate the promising performance of the proposed method through extensive numerical experiments, including comparisons among its variants and with existing constrained stochastic optimization methods.","published":"2026-09-25T03:00:50Z","updated":"2026-09-25T03:00:50Z","authors":["Haoxuan Wang","Yuchen Fang","Sen Na"],"categories":["math.OC","cs.LG","stat.CO","stat.ML"],"headline":"New method tackles hard constrained optimization with heavy-tailed noise","summary":"The researchers studied optimization problems where constraints must be met, but random noise with extreme values challenges existing methods. They developed a new approach called TR-SSQP that balances finding good solutions while respecting constraints using a special step calculation and momentum technique. Their work includes mathematical proof of convergence under difficult noise conditions and shows promising results in tests compared to current methods, though findings come from theoretical analysis and simulations."},{"id":"2609.30731v1","title":"Structured Bayesian Modeling of Dynamic Receptive1 Fields in Salamander Retinal Ganglion Cells","abstract":"Neurons in the visual system are selective for specific spatial and temporal stimulus features, described by their \\emph{receptive field}. Estimating one means a coefficient per pixel per time bin from few trials -- a high-dimensional problem requiring regularization. Sparse regularizers such as the LASSO handle the dimension but select pixels independently at each time point, with nothing to keep the region coherent in space or smooth in time; it can fragment or reorganize discontinuously even when the true response evolves smoothly, a failure since this evolving pattern is what a receptive-field estimate should capture. We formulate dynamic receptive-field estimation as a high-dimensional Bayesian problem: a Poisson model combining a Gaussian Markov random field in space with an autoregressive process in time, so the estimated field is smooth and coherent across space and time. On recordings from $155$ salamander retinal ganglion cells, fitting this model independently per neuron recovers a coherent surface, where a pixel-level Poisson-LASSO comparison instead returns a fragmented one. Summarizing each neuron's surface by its space-averaged temporal response and clustering these curves with a model-based functional-clustering procedure, BIC selects three balanced temporal-response phenotypes ($85$, $32$, $38$ neurons), against a degenerate grouping from clustering the raw surfaces. A simulation study with known ground truth confirms the same pattern, with the model beating an unregularized Poisson GLM, LASSO, and the elastic net on recovery and estimation accuracy, though LASSO controls false positives better. The per-neuron field identification, its contrast with LASSO, and the functional-clustering population typing constitute this paper's contribution.","published":"2026-09-25T03:00:14Z","updated":"2026-09-25T03:00:14Z","authors":["Alokesh Manna"],"categories":["cs.NE","stat.ME"],"headline":"New Bayesian Model Better Maps Salamander Vision Neurons","summary":"Researchers studied how neurons in salamander eyes respond to visual signals, aiming to improve estimates of their receptive fields—the specific patterns neurons detect over space and time. They developed a new statistical method combining spatial and temporal information to produce smoother, more coherent maps than previous independent pixel methods. Testing on real neuron data and simulations, their approach provided clearer patterns and grouped neurons into meaningful types, although some trade-offs in error types remain."},{"id":"2609.30730v1","title":"Photon spheres, elliptic surfaces, and Seiberg-Witten geometry","abstract":"We study null geodesics in Reissner-Nordström-de Sitter spacetime and in its $ω=-2/3$ Kiselev deformation through the elliptic curves defined by the radial equation. The relation of these curves to the non-critical $E_7$ Seiberg-Witten family yields differential equations for finite-distance light deflection, and explains the boundary source term through a suitable gauge of the Seiberg-Witten differential. This geometry separates the extremal point $Q^2=M^2$ from the Argyres-Douglas point $Q^2=9M^2/8$. We also determine the Kiselev critical and horizon loci, and show that, at fixed nonzero angular momentum, the Argyres-Douglas and ultracold loci meet only at zero Killing energy.","published":"2026-09-25T02:59:33Z","updated":"2026-09-25T02:59:33Z","authors":["Cordell Blankenship","Andreas Malmendier","Michael T. Schultz"],"categories":["gr-qc","math.AG"],"headline":"Researchers link black hole light paths to complex geometry","summary":"The researchers studied how light moves near certain charged black holes by examining mathematical curves called elliptic curves related to the black holes' radial equations. They connected these curves to a complex framework known as Seiberg-Witten geometry, which helped them derive equations describing how light bends near these objects. Their work also distinguishes different special points related to the black hole's charge and mass, and identifies critical zones in a related model, but results are theoretical and based on advanced mathematical analysis."},{"id":"2609.30729v1","title":"Input-Layer Starvation: Why Per-Layer Pruning Breaks IoT Intrusion Detectors","abstract":"Intrusion detectors for small Internet-of-Things (IoT) devices are usually compressed by pruning and judged by overall accuracy. We show that this hides a severe class-level failure, find its cause, and give low-overhead prevention and repair. On CICIoT2023, a two-layer convolutional detector pruned with uniform layer-wise magnitude pruning at 80% sparsity loses 16 points of accuracy but half of its macro-F1, the mean per-class F1 (0.542 to 0.271 over five independently trained models); 17 of 34 classes are materially damaged. Remaining weight count does not explain it: a perceptron and a transformer pruned to the same or fewer weights lose at most 0.096. The first layer does. It has 192 weights; uniform pruning leaves 38, 46% of its 64 filters lose every input weight, and fine-tuning under that starvation leaves the running means of the first normalisation layer displaced by up to 0.8 standard deviations in a few surviving channels, on which the deployed model collapses. Protecting those 192 weights, or pruning globally at the same sparsity, prevents the collapse (loss 0.013); recomputing the normalisation statistics on unlabelled training data, with no weight changed, repairs it (loss 0.039) and returns the false-alert rate to 33% (dense 29%). Damage shows a strong increasing dose-response in first-layer sparsity, starving a perceptron's input layer reproduces the collapse, and the pattern holds on TON_IoT. The failure is misattribution and false alerts, not silent evasion: on validation-selected blind spots, uniformly pruned detectors misattribute 72% of the traffic, against 50% with the first layer protected and 47% for the dense model.","published":"2026-09-25T02:56:27Z","updated":"2026-09-25T02:56:27Z","authors":["Md Anas Biswas"],"categories":["cs.CR","cs.LG"],"headline":"Why Pruning Layers Hurts IoT Intrusion Detection Accuracy","summary":"Researchers studied how simplifying intrusion detection models for small IoT devices by pruning network layers can cause serious problems. They found that aggressively pruning the first layer causes many detection errors and a big drop in performance. However, protecting this layer or adjusting certain statistics after pruning can prevent or fix these issues. Their findings, based on tests with specific datasets and models, highlight a hidden flaw in common pruning methods and suggest ways to improve intrusion detectors for IoT devices."},{"id":"2609.30728v1","title":"Learning Polarization Image Restoration with General Restoration Priors","abstract":"Polarization imaging captures distinctive surface and geometric cues that benefit a wide range of vision tasks. However, real-world polarization acquisition is often affected by multiple coupled degradations, making image restoration essential for practical polarization vision. Existing methods are largely tailored to specific degradations and remain constrained by the limited scale and quality of polarization data. To address these limitations, we develop an all-in-one polarization restoration framework for diverse and composite degradations. We first study the impact of different polarization representations on restoration performance and identify the normalized Stokes representation as an effective choice for separating intensity and polarization information. Accordingly, we devise a dual-branch architecture that separates intensity and polarization modeling. To overcome the limitations of polarization-specific training, the intensity branch leverages pretrained general restoration priors and a mixture-of-experts extension for composite degradations, while its restoration knowledge is adaptively distilled into the symmetric polarization branch via a cross-domain feature transform. In addition, we establish a composite-degradation polarization benchmark to support all-in-one restoration research. Extensive experiments on public datasets and our proposed benchmark demonstrate the effectiveness of the proposed method.","published":"2026-09-25T02:55:50Z","updated":"2026-09-25T02:55:50Z","authors":["Chenggong Li","Jinhao Liu","Caiyun Wu","Yidong Luo","Junchao Zhang","Degui Yang"],"categories":["cs.CV","eess.IV"],"headline":"New Method Improves Polarization Image Restoration Amid Multiple Problems","summary":"Researchers studied ways to improve images taken with polarization cameras, which capture unique surface details but often suffer from various quality problems in real-world use. They developed a new approach that separates brightness and polarization data, using a dual system where one side benefits from general image restoration knowledge and shares this with the polarization side. Their experiments on public and new tests show this method works well, though the findings focus on demonstration rather than real-world deployment."},{"id":"2609.30727v1","title":"Deep Pseudo-Proximal Map: A Self-Supervised Data-Fitting Agent for Iterative Reconstruction","abstract":"Iterative algorithms for inverse imaging problems split reconstruction into alternating subproblems, one of which requires evaluating the data-fitting proximal map. In many practical cases, this proximal map has no analytic solution and so must be approximated by an inner loop solver at every iteration, compounding the cost of the outer reconstruction loop. To address this, we propose the pseudo-proximal map (PPM), a reformulation of the data-fitting proximal map as the minimum mean square error estimate of a synthetic probabilistic model. We implement the deep PPM as a self-supervised neural network trained only on sampled Gaussian noise, requiring no ground-truth training images. The deep PPM can be trained for any operator for which the forward model $A$ and its transpose $A^T$ can be evaluated, with provable equivalence to the proximal map when $A$ is linear. We validate the deep PPM on Gaussian deblurring and 4x super-resolution, where the proximal map has an analytic solution, and on X-ray computed tomography (XCT), where no analytic solution exists. Used as the data-fitting agent within an iterative reconstruction method, the deep PPM reproduces the reference reconstruction to within 1% NRMSE for all three operators, and for XCT, it replaces the inner conjugate-gradient loop with a single network evaluation that is 18x faster.","published":"2026-09-25T02:54:52Z","updated":"2026-09-25T02:54:52Z","authors":["Haley Duba-Sullivan","Emma J. Reid","Charles A. Bouman","Gregery T. Buzzard"],"categories":["eess.IV"],"headline":"New AI Method Speeds Up Complex Image Reconstruction Tasks","summary":"Researchers developed a new technique called deep pseudo-proximal map (deep PPM) to improve iterative image reconstruction, a process used in imaging tasks like deblurring and X-ray scans. Their method uses a self-trained neural network that learns from noise rather than real images and works without needing exact mathematical solutions. Tests show it closely matches standard results while significantly reducing computation time, especially in cases where traditional solutions are slow or unavailable."},{"id":"2609.30726v1","title":"Almost-orthogonal Strichartz estimates and radial improvements","abstract":"We extend the Schatten-duality principle from orthonormal systems to (almost-orthogonal) Bessel families and apply it to prove sharp and improved Bessel-family Strichartz estimates for the Kadomtsev--Petviashvili, Zakharov--Kuznetsov, and radial Schrödinger equations. In particular, for the Schrödinger equation, we establish an improved estimate in a larger radial range of space-time exponents, not from dispersive estimates as in the KP/ZK models, but from direct robust Schatten estimates, using some vector-valued multi-product trace formula and multilinear weighted fractional integral formula. To the best of our knowledge, this is the first radial improvement for systems, and the almost-orthogonal estimates are also new to the literature. We also derive necessary conditions.","published":"2026-09-25T02:54:48Z","updated":"2026-09-25T02:54:48Z","authors":["Hongzhou Ji","Liping Xu","An Zhang"],"categories":["math.CA"],"headline":"New Mathematical Estimates Improve Understanding of Certain Wave Equations","summary":"Researchers studied special mathematical tools called Strichartz estimates, which help analyze solutions to wave and quantum equations like the Schrödinger equation. They extended previous principles to more general systems called almost-orthogonal Bessel families and found sharper, improved estimates, particularly for cases with radial symmetry. Their work uses complex new formulas and provides necessary conditions for these improvements. These results are theoretical and based on advanced mathematics, offering new insights rather than practical applications."},{"id":"2609.30725v1","title":"Analyzing and Mitigating Cost-Inefficient Behaviors in Coding Agents","abstract":"Although effective, coding agents often incur substantial monetary costs. Their recurring cost-inefficient behaviors remain underexplored. We conduct the first study of behavioral cost inefficiencies in coding agents, analyzing 1,200 trajectories from Claude Code and Mini-SWE-Agent across four configurations on SWE-bench Verified. We identify three cost-inefficient behaviors: subsumed retrieval, similar script generation, and test re-execution. We then evaluate three mitigation strategies: structure-aware retrieval, agent-synthesized skills, and developer-designed skills, over 10k trajectories on held-out SWE-bench Verified and Pro tasks. Our main findings are: (1) The three behaviors affect 79.00\\%--98.00\\% of coding tasks and account for up to 22.75\\% of task cost. (2) Structure-aware retrieval can introduce retrieval overhead and alter agent delegation, causing inconsistent improvements in retrieval efficiency and cost increases of up to 28.14\\%. (3) Agent-synthesized skills tend to produce low-level, trace-specific guidance, limiting their effectiveness and generality. (4) In contrast, developer-designed skills provide high-level, trace-agnostic guidance, reducing cost by up to 41.73\\%, roughly twice the maximum gain from agent-synthesized skills.","published":"2026-09-25T02:54:18Z","updated":"2026-09-25T02:54:18Z","authors":["Yiran Hu","Nan Jiang","Shanchao Liang","Anik Dey","Yi Wu","Lin Tan"],"categories":["cs.AI","cs.SE"],"headline":"Reducing costly behaviors in AI coding agents","summary":"Researchers studied inefficient and costly behaviors in AI coding tools by analyzing over 1,200 coding sessions. They identified common costly actions like repeated retrieval and test reruns, affecting most coding tasks and adding significant expense. The team tested three ways to reduce these costs and found that developer-designed guidance was most effective, cutting costs by up to 42%, while other methods had limited or mixed results. These findings come from analyzing simulations and may help improve AI coding efficiency."},{"id":"2609.30724v1","title":"EviDETR: Preserving Query-Relevant Temporal Evidence for Moment Retrieval and Highlight Detection","abstract":"Joint video moment retrieval and highlight detection requires identifying query-relevant temporal segments while estimating clip-level saliency, yet DETR-style pipelines do not explicitly preserve query-relevant evidence throughout encoding, decoding, and cross-task prediction. We propose EviDETR, an evidence-preserving framework with three components. Semantic-aware Feature Reweighting (SFR) enhances query-relevant clip representations through saliency estimation and cross-modal interaction. A Temporal Top-2 Mixture-of-Experts (TTop2MoE) decoder performs query-adaptive refinement via sparse expert routing. MR-to-HD (MR2HD) fusion transfers span-level retrieval evidence to clip-level highlight prediction through confidence-weighted multi-scale aggregation. Using CLIP+SlowFast features, EviDETR achieves 69.29 R1@0.5, 54.77 R1@0.7, and 48.41 Avg. mAP for moment retrieval on QVHighlights, together with 41.83 HD-mAP and 68.33 HIT@1. Strong results on TACoS and Charades-STA further demonstrate cross-dataset transferability.","published":"2026-09-25T02:53:54Z","updated":"2026-09-25T02:53:54Z","authors":["Haoran Sun","Yufan Li","Qichen Zhang","Haoran Zhao","Shuqi Wang"],"categories":["cs.CV"],"headline":"New Method Improves Video Search and Highlight Identification","summary":"Researchers developed EviDETR, a system designed to better find specific moments and important highlights in videos based on a spoken or written query. It preserves important information related to the query throughout the process using enhanced features and a specialized decoder, then links retrieved moments to clips to identify highlights. Tested on several video datasets, EviDETR showed improved accuracy, though these findings are based on simulated experiments with specific data and tools."},{"id":"2609.30723v1","title":"Concurrent Parameter Learning and Current Control for Large-scale Grid-following Inverter-based Resources","abstract":"The integration of large-scale inverter-based resources (IBRs) into the grid presents two key challenges in terms of optimal power generation and transmission system stability. This work proposes concurrent transmission parameter learning and optimal power generation and current control for the grid-following IBRs. The proposed approach estimates transmission parameters online using the current measurements, enabling the IBRs to adapt their control actions to evolving grid conditions. The learned parameter estimates are incorporated into an optimization-based power distribution strategy to determine the optimal power setpoints. In parallel, a current control scheme is designed to regulate the grid-following (GFL) IBRs current output based on the estimates and received active and reactive power references from the high-level optimizer, thereby improving the dynamic response to the grid events. The effectiveness of the proposed framework is validated through a simulation in a MATLAB-Simulink environment. From the results, the impact of the concurrent estimator and current controller during the grid events can be seen.","published":"2026-09-25T02:49:29Z","updated":"2026-09-25T02:49:29Z","authors":["Satish Vedula"],"categories":["math.OC"],"headline":"New Method Helps Manage Large-Scale Inverter Power on the Grid","summary":"Researchers explored challenges of integrating many inverter-based power sources into the electrical grid, focusing on stability and efficient power use. They developed a method that continuously learns grid conditions and adjusts power output and current control in real-time. Using computer simulations, they showed their approach could improve system response during grid events, but findings are limited to these simulation tests and need real-world validation."},{"id":"2609.30722v1","title":"TrafficImag: A Benchmark for Counterfactual Roadside Traffic Video Generation","abstract":"Existing roadside traffic datasets support perception, forecasting, and visual question answering, but they do not evaluate counterfactual video generation, in which a selected actor is modified and the generated future should remain consistent with road topology and unrelated traffic. We introduce TrafficImag, the first benchmark for counterfactual roadside traffic video generation. TrafficImag combines a large-scale roadside dataset (9,022 annotated images, 7,043 deduplicated video clips, and 31,145 actor-centered history-future samples) with an executable protocol that supports behavior reasoning, intervention-aware image editing, and conditional video generation. Each intervention is represented as an actor-level program describing the target actor, intended behavior, legal route, interaction order, and temporal constraints, enabling a unified evaluation interface across heterogeneous foundation models. TrafficImag evaluates four complementary validity dimensions: initial-state correctness, route and behavior validity, interaction consistency, and non-target preservation, and considers an end-to-end counterfactual successful only when all four are satisfied. Across state-of-the-art foundation models, the strongest reasoner reaches 80.4% macro F1, the complete condition interface raises end-to-end success from 23.3% to 55.0% for the best generator. Oracle studies further show that conditional video execution is the primary remaining bottleneck. TrafficImag provides a reproducible benchmark for evaluating and diagnosing counterfactual traffic video generation beyond perceptual video quality.","published":"2026-09-25T02:49:06Z","updated":"2026-09-25T02:49:06Z","authors":["Xiangyu Li","Tianyi Wang","Zhihao Dou","Christian Claudel","Zhaomiao Guo"],"categories":["cs.CV","cs.AI"],"headline":"New Benchmark Tests Imagined Traffic Videos with Changed Drivers","summary":"Researchers created TrafficImag, a new system to test how well computers can generate videos of traffic scenes when one driver's behavior is changed while everything else stays realistic. They combined a large set of traffic videos with a method to check if the changes follow real road rules and interactions. Their tests found current models struggle to fully succeed, especially in creating detailed future videos. The work highlights challenges in simulating traffic behavior and offers a way to measure progress."},{"id":"2609.30721v1","title":"When 10,000 Windows Are Not 10,000 Tests: Auditing Statistical Confidence in Sliding-Window Time-Series Classification","abstract":"Sliding-window classifiers are often evaluated on thousands of overlapping test windows, even though neighboring predictions share observations and remain nested within recordings and subjects. Subject-disjoint evaluation prevents one form of leakage but does not make those test windows independent. We present a practical audit that maps three claims - performance on observed recordings, future recordings from observed subjects, and unseen subjects - to explicit aggregation rules and established dependence-robust inference. At 75% overlap, controlled simulations give 16.9% Type-I error for IID observed-record inference and 7.2% for session-centered Bartlett-HAC: a substantial improvement with residual miscalibration. Audits of frozen WISDM and HARTH predictions show that nearly fourfold growth in test rows yields only 1.75-1.94-fold variance-equivalent information growth. At that overlap, fixed-record paired Accuracy-difference intervals are 1.22-1.66 times the IID widths; this inflation is not universal at zero overlap. On HARTH, paired Accuracy-difference intervals include zero across three overlap settings, whereas Macro-F1 favors MiniROCKET. Independent recomputation, common-session checks, class-level results, and separately seeded calibration make the audit's scope and limitations inspectable. The resulting workflow distinguishes additional predictions from additional independent evidence.","published":"2026-09-25T02:49:00Z","updated":"2026-09-25T02:49:00Z","authors":["Xinze Shi","Litian Zhang","Binrui Shi"],"categories":["cs.LG"],"headline":"Checking Accuracy Claims in Overlapping Time-Series Tests","summary":"The researchers studied how sliding-window tests for time-series data often use many overlapping segments, which aren't truly independent. They developed a method to better evaluate performance claims for observed data, future data from the same people, and new people by adjusting for these overlaps. Their analysis shows that simply having more overlapping tests doesn't proportionally increase reliable information and highlights how traditional accuracy estimates can be overly optimistic. The findings come from simulations and audits of existing datasets, with a transparent approach to understand the method's limits."},{"id":"2609.30720v1","title":"Optimal Personalized Subspace Learning for Multi-view Tensor Observations","abstract":"In this work, we model the observed multi-view tensors by decomposing the underlying signal in each view into two components: (i) the shared component that captures common dynamics across all views, and (ii) the private component that accounts for view-wise unique variations. To decouple the shared and private components, we introduce a novel Tucker personalized subspace principal component analysis (TPS-PCA) approach for tensors, which admits a one-step closed-form solution and serves as an ideal surrogate for our extended tensor-version personalized PCA (TP-PCA), adapted from the seminal work by \\cite{shi2024personalized}. The theoretical analysis reveals that the proposed TPS-PCA estimators reach the minimax lower bound in terms of view-wise tensor decoupling, whereas the TP-PCA estimators only achieve a rate of average decoupling error across views, which is still slower than that of the TPS-PCA estimators. Extensive numerical experiments are conducted on synthetic and real datasets, demonstrating the wide applicability of the proposed method in fields including power management, financial analysis, and activity recognition.","published":"2026-09-25T02:48:49Z","updated":"2026-09-25T02:48:49Z","authors":["Kangxiang Qin","Zeyu Li","Xinbing Kong","Wang Zhou"],"categories":["stat.ME"],"headline":"New method separates shared and unique data features across multiple views","summary":"Researchers developed a mathematical approach called TPS-PCA to analyze complex multi-view data by breaking it into shared features common to all views and private features unique to each. Their method provides a precise solution supported by theoretical analysis showing it performs better than previous approaches in separating these components. They tested the method on both simulated and real data, indicating it may be useful in areas like power management and finance, though results are based on this specific approach and theory."},{"id":"2609.30719v1","title":"Werracle: Sub-Cent Intra-Block AI Reflex Oracles and Flash-Loan Circuit Breakers for EVM Smart Contracts","abstract":"Contemporary on-chain artificial intelligence (AI) encounters an intractable Von Neumann memory and latency wall. Storing static floating-point neural weight matrices inside Ethereum Virtual Machine (EVM) storage costs millions of gas, rendering direct on-chain inference impossible. While Zero-Knowledge Machine Learning (ZK-ML) offloads matrix tensor multiplications to off-chain provers, it introduces fatal constraints: 10 to 300 seconds of SNARK proving latency and 250,000 to 500,000 gas per proof verification. Because decentralized finance (DeFi) exploits - such as uncollateralized flash-loan attacks, predatory sandwich MEV, and toxic loss-versus-rebalancing (LVR) flow - occur atomically inside a single block, ZK-ML oracles cannot react in time. Here, we present Werracle, a production-grade, zero-storage on-chain AI decision oracle fitting inside a single 32-byte EVM storage slot (bytes32). Leveraging foundational procedural Mandelbrot escape dynamics (z_{n+1} = z_n^2 + c) established by Dagli et al. (arXiv:2609.25498), Werracle derives continuous non-linear decision hyperplanes from a 24-byte coordinate triplet Theta = (c_x, c_y, zoom). Implemented in pure Solidity bytecode using fixed-point Q16.16 arithmetic (WerrMath.sol), Werracle evaluates a 16-point Pareto micro-grid in only 21,438 gas (under 0.0005 USD on Layer-2 rollups like Base and Arbitrum) with sub-millisecond execution latency. We demonstrate real-world DeFi efficacy via WerracleFeeHook.sol, a Uniswap v4 dynamic swap fee governor that measures orderbook turbulence on-the-fly and atomically adjusts liquidity provider fees between 0.05% and 0.50%. The protocol is formally verified against a 1,000-test cryptographically sealed deterministic verification suite (100.0% pass rate) with telemetry permanently disabled, operating live on a dedicated EVM devnet sandbox (Chain ID 4242).","published":"2026-09-25T02:45:03Z","updated":"2026-09-25T02:45:03Z","authors":["Volkan Dağlı","Zerrin Dağlı","Dağhan Dağlı"],"categories":["cs.CR","cs.AI","cs.DC"],"headline":"New Tiny AI Oracle Enables Fast Decisions in Blockchain Finance","summary":"Researchers developed Werracle, a compact AI oracle that fits into a single 32-byte storage slot on the Ethereum blockchain. Unlike previous AI methods that were slow or costly, Werracle uses mathematical patterns to make quick decisions with low gas fees and near-instant execution. The team tested Werracle with a live financial tool that adjusts trading fees on Uniswap, showing it works efficiently and reliably in real blockchain settings."},{"id":"2609.30718v1","title":"NEMSim: Learning Control-Conditioned Multi-Event Physical Dynamics via Executable Event-Mechanism Priors","abstract":"High-fidelity simulation of control-conditioned multi-event physical systems is computationally expensive, especially across broad control spaces and long trajectories. In these systems, macroscopic evolution emerges from localized discrete events whose intensities and effects depend on process controls and evolving local states, while the available system knowledge is typically expressed as event-attribute descriptions. Purely data-driven surrogates must infer these event effects from limited trajectory coverage, which can hinder generalization to unseen control regimes. Physics-guided methods instead primarily build on equation-level constraints or differentiable solvers rather than discrete event-rule priors. We therefore propose NEMSim (Neural Event-Mechanism Simulator), which compiles predefined event-attribute descriptions into an executable transition structure linking control-dependent event intensities, prior-guided mechanism attribution, and state-dependent responses. To enable evaluation of control-conditioned multi-event dynamics with explicit system knowledge, we construct a 3D KMC-based benchmark pairing high-fidelity trajectories with explicit event rules, standardized splits, and evaluation protocols. Across three settings, NEMSim reduces Avg. RMSE by 58.9%-81.3% relative to the strongest baseline in each setting. It also remains best in the data-efficiency study with training-data fractions down to 10%. Mechanism analyses further show that these gains arise from executable rule integration rather than prior access or architecture alone.","published":"2026-09-25T02:44:59Z","updated":"2026-09-25T02:44:59Z","authors":["Junsong Yu","Junjie Xie","Pengwei Liu","Dong Ni"],"categories":["cs.LG"],"headline":"New Simulator Improves Physical System Predictions Using Event Rules","summary":"Simulating complex physical systems with many interacting events and controls is usually very slow and difficult to generalize. The researchers created NEMSim, a new simulator that uses predefined event descriptions as rules to better predict how systems evolve under different controls. Their tests show NEMSim predicts outcomes more accurately and efficiently than other methods by integrating explicit event rules, not just relying on data or mathematical equations. However, their findings are based on simulations and a specific benchmark they developed."},{"id":"2609.30717v1","title":"RecToolBench: Benchmarking Recommendation-Specific Tool Orchestration under Fuzzy User Intent","abstract":"Recent advances in agentic recommender systems are shifting recommender systems from passive filtering engines to instruction-following agents that use external tools to resolve user intent. However, existing benchmarks often assume explicit user intent, simplified tool environments, or isolated function calls, leaving realistic tool orchestration for recommendation underexplored. To bridge this gap, we propose RecToolBench, a Model Context Protocol (MCP)-based benchmark for evaluating tool-using recommender agents under fuzzy user instructions. RecToolBench contains more than 1,200 executable tasks across three recommendation domains, 13 MCP servers, and 32 tools, spanning single-tool calls, parallel tool calls, sequential tool chains, and hybrid tool orchestration. We construct RecToolBench with a scalable synthesize--fuzzify--judge pipeline that generates executable fuzzy recommendation tasks, and evaluates agent trajectories using rule-based execution checks and rubric-based LLM evaluation. Experiments on representative LLMs show that syntactically valid tool calls do not guarantee successful recommendations. Models struggle with semantic parameter grounding, multi-step evidence integration, and grounded final recommendations, especially as orchestration complexity increases. Our results identify tool orchestration under fuzzy user intent as a major bottleneck for agentic recommender systems. Our data and code are available at https://github.com/ShawnChenn/RecToolBench.","published":"2026-09-25T02:44:07Z","updated":"2026-09-25T02:44:07Z","authors":["Xiao Chen","Yicheng Zhao","Yingying Wu","Zhendong Chu","Changyi Ma","Qingsong Wen","Xuan Song"],"categories":["cs.IR"],"headline":"New Benchmark Tests Recommendation Tools Handling Vague User Requests","summary":"Researchers created RecToolBench, a new test setup to evaluate recommendation systems that use multiple tools when user requests are unclear. They designed over 1,200 tasks requiring tool coordination and measured system performance in complex scenarios. Their findings show current systems often fail to fully understand vague instructions and combine multiple steps effectively, revealing challenges in making recommendation tools work well together under realistic conditions."},{"id":"2609.30716v1","title":"Words Speak Louder Than Order: A Behavioral Evaluation of Gemma 4","abstract":"When a language model receives two conflicting documents as input, how does it decide which one to prioritize? Does it rely on how the sources are framed or the presentation order of the documents? We evaluated this behavior on Google's pre-trained Gemma 4-e4b model across a targeted behavioral suite (n = 13 items, 784 forward passes in short, single-turn contexts) using a completely counterbalanced experimental design. This setup allowed us to mathematically isolate the specific effects of source framing and reading position, while ensuring the model's natural vocabulary biases were canceled out. Across ten test conditions, we discovered the following: 1. Source framing heavily overpowers reading position. When directly competing, the semantic framing of a source (such as presenting it as an official guideline or a fresh update) had a significantly stronger impact on the model's final answer than the presentation order of the document. 2. The model favors the first document it reads, but this bias is highly variable. While the model consistently demonstrated a primacy effect (preferring the first document presented), the actual strength of this bias fluctuated by at least a factor of 5 based solely on the surface wording. 3. Overall structural repetition, not short copy-cues, drives positional bias. The model's preference for the first document is not a mechanical reaction to short, repetitive trigger phrases, such as \"is [Answer]\". However, the primacy effect does increase significantly when the two competing documents are structurally identical, using word-for-word verbatim templates. Introducing variation in the overall wording between the two sources reduces this positional bias.","published":"2026-09-25T02:43:40Z","updated":"2026-09-25T02:43:40Z","authors":["Amanda Fitch"],"categories":["cs.CL","cs.AI"],"headline":"How a Language Model Chooses Between Conflicting Documents","summary":"Researchers studied Google's Gemma 4 language model to see how it picks between two conflicting documents. They found that the way a source is described (like an official guideline) influences the model's choice much more than the order in which the documents are read. While the model tends to favor the first document, this preference changes a lot depending on wording. The study also showed that repeated document structure increases this bias, but varying the wording reduces it. These findings come from controlled tests with simulated inputs."},{"id":"2609.30715v1","title":"RoboMonitor: Label-Efficient Runtime Monitoring of Robot Task Execution via Predictive Representation Learning","abstract":"Learned robot policies produce actions, but their outputs alone do not establish whether execution is progressing as intended. Robot execution monitoring requires identifying the current execution phase, detecting failures, and recognizing task completion from observations available during execution. Training such monitors requires annotations that are scarce in datasets collected for robot-policy learning. We present RoboMonitor, a label-efficient vision--language execution monitor that learns from these datasets before introducing monitoring supervision. We pre-train on 25 hours of multi-camera trajectories spanning 12 manipulation tasks and two robot embodiments, using action-conditioned future-feature prediction, inverse dynamics, and masked-present prediction. We then transfer the learned visual and context encoders to a causal monitor and apply temporal supervised fine-tuning (Temporal SFT), which combines supervision throughout each observation window with consistency objectives within and across overlapping windows. At deployment, RoboMonitor requires only the task instruction and camera observations. On a four-task monitoring benchmark, RoboMonitor trained with 52 labeled episodes achieves 93.1% mean phase accuracy and 85.9% macro recall over two fine-tuning seeds, exceeding Qwen3-VL and Robometer trained with the same monitoring supervision. Its phase accuracy also exceeds that of both Qwen3-VL and Robometer trained with 100 episodes. A Qwen3-VL ablation shows that Temporal SFT reduces mean spurious phase switching from 15.23% to 4.95%. In closed-loop deployment, the integrated system completes 39 of 40 simulated Toolbox Sorting trials and 35 of 40 real-world Reel Packing trials, with no false recovery triggers observed.","published":"2026-09-25T02:43:24Z","updated":"2026-09-25T02:43:24Z","authors":["Abhiroop Ajith","Gokul Narayanan","Kyle Coelho","Tingji Zhao","Yash Shahapurkar","Brian Zhu","Melih Erdogan","Ted Krubasik","Constantinos Chamzas","Eugen Solowjow"],"categories":["cs.RO"],"headline":"New RoboMonitor tracks robot tasks efficiently with little training data","summary":"Researchers developed RoboMonitor, a system that watches robot actions to check if tasks progress correctly, fail, or finish. It learns mostly from existing robot videos before training on limited labeled examples, improving accuracy in recognizing task phases compared to other systems. The monitor uses cameras and task instructions without extensive manual labeling. Tests showed high accuracy on multiple tasks and reliable performance in both simulations and real robot setups, though findings are based on specific benchmarks and controlled trials."},{"id":"2609.30714v1","title":"CRC-Router: Risk-Constrained Routing for Medical Agentic AI Systems","abstract":"Agentic AI systems are increasingly being explored in medical imaging to improve throughput and reduce clinician workload; however, safe deployment remains challenging because autonomous errors may propagate into downstream clinical decisions. A central requirement is therefore not only strong predictive performance, but also a reliable routing mechanism that determines when the system should proceed autonomously and when a case should be escalated for further review. To address this gap, we propose CRC-Router, a risk-constrained, uncertainty-aware routing module that is applicable to both conventional medical prediction models and agentic medical AI systems. CRC-Router combines multiple complementary uncertainty signals with the predictive score to construct a per-finding routing feature vector, maps this vector to an estimated wrong-accept risk using a lightweight per-finding risk model, and then applies Conformal Risk Control (CRC) to calibrate acceptance thresholds under a user-specified risk target. Instantiated on chest X-ray multi-finding triage using the NIH ChestX-ray14 dataset, CRC-Router achieves the strongest empirical risk--coverage trade-off among the evaluated baselines, both as a standalone routing layer and as a plug-in module integrated with the state-of-the-art MedRAX agent. These results demonstrate both the effectiveness of CRC-Router in selective medical automation and its modular, model-agnostic compatibility with existing predictive and agentic medical pipelines. Code is publicly available at https://github.com/XLIAaron/CRC-Router","published":"2026-09-25T02:42:15Z","updated":"2026-09-25T02:42:15Z","authors":["Xueyang Li","Mingze Jiang","Gelei Xu","Jun Xia","Ching-Hao Chiu","Mengzhao Jia","Danny Z. Chen","Yiyu Shi"],"categories":["cs.AI"],"headline":"New method helps medical AI decide when to ask for human review","summary":"Researchers developed CRC-Router, a tool designed to improve safety in medical AI systems by deciding when the AI should handle cases autonomously and when to escalate them for human review. It uses multiple signals to estimate the risk of errors and adjusts its decisions to keep mistakes within a user-defined limit. Tested on chest X-ray data, CRC-Router showed better balance between risk and coverage compared to other methods, supporting safer and more reliable use of AI in healthcare."},{"id":"2609.30713v1","title":"Parameter Estimation for Unnormalized Discrete Models via Empirically Localized Deformed Bregman Divergence","abstract":"Estimation of parameter of probabilistic models is an important task in the field of machine learning.For models of discrete variables, calculation of the normalization constant of model is sometimes difficult and a lot of researches have been done to avoid the calculation of the normalization constant. In this paper, we tackle with the difficulty by combining a technique of empirical localization and a deformed Bregman divergence.The technique of empirical localization makes it possible to drastically reduce computational cost of the calculation of the normalization constant, and in addition, appropriate choice of the deformation for the Bregman divergence can invest the proposed estimator with various kinds of favorable statistical properties, such as efficiency or robustness against outlier noise.","published":"2026-09-25T02:41:32Z","updated":"2026-09-25T02:41:32Z","authors":["Takashi Takenouchi"],"categories":["stat.ML","cs.LG"],"headline":"New Way to Estimate Parameters in Complex Discrete Models","summary":"Researchers studied how to estimate parameters in probabilistic models involving discrete variables, where calculating a certain normalization factor is often difficult. They combined a method called empirical localization with a mathematical tool named deformed Bregman divergence to reduce computational cost and improve properties like efficiency and robustness to noise. Their approach addresses challenges in parameter estimation, based on theoretical and computational techniques described in the paper."},{"id":"2609.30712v1","title":"Disjunctive Submodular Functions: Envelopes and Applications to Inventory and 0-1 Quadratic Optimization","abstract":"This paper considers convex envelopes of disjunctive submodular functions---functions that are lattice family submodular over faces of a hypercube---and constructs the first strongly polynomial algorithm for their separation when there are two facial disjunctions. Submodular functions, whose convex envelopes are characterized by the Lovász extension, have occupied a fundamental role in constructing relaxations for combinatorial and nonlinear optimization problems. However, disjunctive submodular function envelopes have not been explored besides the use of ellipsoid algorithm, which remains practically intractable. Our algorithm is derived in three steps by expressing the disjunctive function as a minimum of two extended submodular functions, introducing a variable lifting technique, and constructing the sublinear envelope in the lifted space. The paper also makes several other contributions. First, we provide a disjunctive formulation for the case where each submodular function admits a linear programming formulation. Second, we derive the closed-form sublinear envelope characterization for intersecting submodular functions, yielding new structural insights into a multi-product inventory sales maximization problem. Third, we fully characterize the convex envelope of a bilinear function defined over a cycle graph in the original variable space. Finally, we show computationally that the cycle inequalities close approximately 60\\% of the gap for complete and Hadamard graphs, over 30\\% of the gap for complete bipartite graphs, and over 80\\% of the gap for sparse graphs such as cactus and Halin graphs. The resulting relaxations are also more efficient to solve than previous extended space formulations.","published":"2026-09-25T02:38:16Z","updated":"2026-09-25T02:38:16Z","authors":["Zhongqi Wu","Taotao He","Mohit Tawarmalani"],"categories":["math.OC"],"headline":"New method improves optimization for inventory and quadratic problems","summary":"Researchers studied complex mathematical functions called disjunctive submodular functions, important for solving certain optimization problems. They developed the first efficient algorithm for handling these functions when combining two parts, avoiding slower previous methods. Their work includes new formulas and insights for inventory sales models and quadratic functions on graphs, showing their approach can significantly improve problem-solving efficiency in tests. The findings are based on mathematical theory and simulations, offering better tools but not guaranteed solutions for all cases."},{"id":"2609.30711v1","title":"Recommendation World Models for Future-State Control","abstract":"Sequential recommendation optimizes which items to rank, while each displayed slate also shapes subsequent feedback and user state. We study how a trained ranker can support decisions about these future consequences. We introduce UA-TWM, a utility-anchored world-model interface that constructs nearby slate actions, estimates their target-relevant consequences, and selects an alternative subject to utility constraints. The reference slate serves as a fallback when no alternative qualifies. A logged-replay instantiation combines utility and target-gain estimates with calibrated failure-risk prediction; a closed-loop instantiation uses one-step state-action prediction and updates its decisions after observed feedback. We evaluate transfer across twelve sequential backbones on MovieLens-25M and KuaiRand-Pure, and repeated target-directed interaction in KuaiSim. Attaching the interface improves Recall@20, NDCG@20, and future-state alignment for every matched logged backbone. Selection ablations reveal the utility and risk costs of aggressive target pursuit, while closed-loop diagnostics isolate the contribution of action-conditioned prediction. Local consequence modeling thus enables target-aware selection around a trained sequential ranker.","published":"2026-09-25T02:37:48Z","updated":"2026-09-25T02:37:48Z","authors":["Jinfeng Xu","Zheyu Chen","Ziyue Peng","Jianheng Tang","Zheng Lin","Jing Yang","Puzhen Wu","Zheng Xing","Victor C. M. Leung"],"categories":["cs.IR"],"headline":"New Method Helps Recommendations Consider Future User Reactions","summary":"Researchers studied how recommendation systems that suggest item lists can better predict how these suggestions influence future user preferences and feedback. They developed UA-TWM, a method that estimates the likely effects of different item lists on user satisfaction and adjusts recommendations accordingly while managing risk. Testing on movie and content datasets showed improvements in relevance and alignment with future user states. The findings come from simulations and logged data analyses, focusing on improving recommendation decisions by modeling future consequences around existing ranking methods."},{"id":"2609.30710v1","title":"A new type of skin thickness in high-spin isomers to constrain equation of state of spin-polarized nuclear matter","abstract":"We explore experimental probes for constraining the equation of state (EOS) of spin-polarized nuclear matter, where spins of nucleons are aligned along a particular direction. For this purpose, we calculate the $12^+$ isomeric state of the \\ce{^{52}Fe} nucleus with the relativistic point-coupling model. We argue that the spin skin thickness of this high-spin state, that is, the difference between the radius of the spin-up density and that of the spin-down density, strongly correlates with the spin slope parameter of the EOS. This is in analogy to the well known linear correlation between the slope parameter of the symmetry energy and the neutron skin thickness of finite nuclei. We show that this correlation is retained even when one considers the difference of the radii between the $12^+$ and the ground states, even though the correlation is much weaker than in the case of the difference between the spin-up and the spin-down radii. We also discuss the mean-square charge radii of the $12^+$ state. We find that by taking the square of the radii the dependence on the spin slope parameter becomes much stronger, and thus it can serve as a good probe of the spin slope parameter given a high resolution of laser-spectroscopy experiments.","published":"2026-09-25T02:37:42Z","updated":"2026-09-25T02:37:42Z","authors":["Toi Tachibana","Kouichi Hagino","Kenichi Yoshida","Qiang Zhao"],"categories":["nucl-th","nucl-ex"],"headline":"New skin thickness idea helps study spin-aligned nuclear matter","summary":"Researchers studied a special high-spin state of iron-52 nucleus to better understand how nuclear matter behaves when particles' spins align. They used a model to calculate differences in size between regions with spins pointing up versus down, finding it relates to a key property called the spin slope parameter. Their results suggest measuring certain radius differences, especially with precise laser techniques, could help constrain theories about spin-polarized nuclear matter. The findings are theoretical, based on calculations with a specific model."},{"id":"2609.30709v1","title":"VLALight: Lightweight Vision-Language-Action Models for Emergency-Aware Traffic Signal Control","abstract":"Traffic signal control (TSC) is essential for mitigating urban congestion. Recent advances in vision-language models (VLMs) enable richer interpretation of intersection scenes, opening new opportunities for visual-context-aware TSC. However, the loose coupling and repeated information conversion between modules can lead to the loss of fine-grained visual details, while sequential inference introduces substantial latency. To address these limitations, we propose VLALight, a lightweight end-to-end vision-language-action framework that directly maps intersection observations and signal-phase information to discrete signal actions. To handle the multi-view nature of TSC, VLALight combines multiple directional camera views into a unified visual input and uses textual instructions to establish their correspondence with traffic movements and signal phases. This design enables direct action prediction with a compact 0.5 B-parameter model, without intermediate image-to-text descriptions or handcrafted traffic-state representations. Experiments show that VLALight delivers the best emergency-vehicle service of all compared methods, reducing pooled emergency waiting time by 21.1% over the cascaded VLMLight while running in real time on local hardware and generalizing to unseen intersection topologies and traffic-flow patterns.","published":"2026-09-25T02:36:44Z","updated":"2026-09-25T02:36:44Z","authors":["Kemou Jiang","Maonan Wang","Xingchen Zou","Jiayue Zhu","Yuhang Fu","Sicheng Wang","Xi Chen","Yirong Chen","Zhiyong Cui"],"categories":["cs.CV","cs.AI"],"headline":"New Model Helps Traffic Lights Respond Faster to Emergencies","summary":"Researchers created VLALight, a compact model that uses camera views and traffic signals to control traffic lights more effectively, especially for emergency vehicles. Unlike previous methods, it combines multiple camera views directly and uses text instructions to connect visual input with traffic actions, resulting in faster decision-making without converting images to text first. Tests showed it reduced emergency vehicle waiting times by over 20%, worked in real time, and adapted well to new intersections, according to the researchers."},{"id":"2609.30708v1","title":"Combining General and Domain-Specific Pretext Tasks for Brain MR Image Segmentation","abstract":"A key challenge in medical image analysis is the scarcity of large annotated datasets for specific populations and diseases. As deep learning models rely heavily on labeled data, effective transfer learning strategies are needed to reduce the dependence on manual annotations. Self-supervised learning has emerged as a promising approach for developing foundation models by enabling the learning of transferable feature representations from large-scale unlabeled medical imaging datasets. In this study, we investigate voxel-level brain age prediction as a domain-specific self-supervised pretext task and compare it with image inpainting, a widely used non-domain-specific alternative. We further propose a multitask self-supervised pretraining framework that jointly optimizes both objectives to learn complementary neuroimaging representations. The pretrained models are evaluated on three downstream magnetic resonance image segmentation tasks: multiple sclerosis lesion segmentation, ischemic stroke lesion segmentation, and cortical brain structure segmentation. Overall, the proposed multitask pretraining framework consistently outperformed the single-task pretrained models and training from scratch across most experimental settings, demonstrating the benefit of combining domain-specific and general self-supervised learning pretext tasks for the development of generalizable neuroimaging foundation models.\\ Code Availability: The source code used in this study is publicly available at https://github.com/TasneemN/Combining-General-and-Domain-Specific-Pretext-Tasks-for-Brain-MR-Image-Segmentation/","published":"2026-09-25T02:36:20Z","updated":"2026-09-25T02:36:20Z","authors":["Tasneem Nasser","Susanne Schmid","Roberto Souza","Naser El-Sheimy"],"categories":["cs.CV","cs.AI"],"headline":"New Method Improves Brain MRI Analysis Using Combined Learning Tasks","summary":"Researchers addressed the problem of limited labeled medical images by exploring self-supervised learning, which uses unlabeled data to train models. They tested two approaches: predicting brain age at each voxel (domain-specific) and repairing missing image parts (general). By combining these tasks in a new training method, they improved model performance on brain MRI segmentation tasks like lesion and brain structure identification. The study shows combining different learning tasks helps create better models for analyzing brain images."},{"id":"2609.30707v1","title":"A Monolithic Discontinuous Galerkin Framework for Darcy Optimal Control with Radon-Measure Tracking and Pointwise Control Constraints","abstract":"We study an elliptic optimal control problem governed by Darcy's equation in heterogeneous porous media, with pointwise box constraints on the control. The objective functional is formulated via a Radon measure, which allows the desired pressure state to be tracked on observation sets of varying dimension, including points, curves, and subdomains, within a single formulation. The state and adjoint equations are discretized by a symmetric interior penalty discontinuous Galerkin method, yielding a locally mass-conservative approximation that is robust across strong permeability discontinuities, while the control is approximated by piecewise constants. The state, adjoint, and control are retained as primary unknowns in a single coupled optimality system and solved monolithically by a primal-dual active set strategy. We establish stability and well-posedness of the discretization and derive a priori $L^2$ error estimates for both variables. The principal difficulty is the reduced regularity of the adjoint state induced by the measure-valued tracking data. The analysis controls the resulting adjoint-control coupling through an intermediate adjoint driven by the continuous optimal state. Numerical experiments confirm the predicted convergence rates for point, curve, and subdomain observations, exhibit mesh-independent primal-dual active set iteration counts, and demonstrate local mass conservation.","published":"2026-09-25T02:32:47Z","updated":"2026-09-25T02:32:47Z","authors":["SeongHee Jeong","Sanghyun Lee"],"categories":["math.NA","math.OC"],"headline":"New Method Improves Control of Fluid Flow in Porous Materials","summary":"Researchers developed a mathematical approach to better control fluid flow in complex porous materials, using Darcy's equation with constraints on control values. They introduced a way to track pressures at various locations, such as points or curves, within a single framework. Their method uses advanced numerical techniques to ensure accurate, stable solutions even with abrupt changes in material properties, and tests show it performs as predicted. However, findings are based on theoretical analysis and simulations."},{"id":"2609.30706v1","title":"LAVOIR: Teaching a Single-Pass Decision Encoder When and What to Ask with Amortized Value of Information","abstract":"\"System One\" decision models such as TypeSafe's Jev and its open counterpart Laya answer typed questions about a text in a single forward pass with calibrated probabilities, but they cannot ask for missing information: when a first message does not say what separates two departments, they guess. We present LAVOIR (Laya with Value-Of-Information Routing), which places the candidate pieces of missing information (slots) in the input next to the answer options, so that one forward pass returns both the decision distribution and, for every slot, the expected gain in the probability of the correct decision if the user were asked about it. VOI targets need no human labels: gold decisions come from schema rules, an LLM only verbalizes messages and answers, a model from another family checks every text, and pairing each message with several profiles makes regression on realized gains estimate the expected gain. A Gini-impurity cap bounds the predicted value by what a calibrated model can still gain. In a controlled study, decisions on seen schemas are statistically indistinguishable from the Bayes ceiling. The final model's question policy matches a greedy oracle VOI policy on seen schemas (AUC 0.799 vs. 0.797), and with at most 0.5 questions per conversation it is 14.1 points more accurate than never asking. On real ABCD conversations, one real exchange raises accuracy by 8.3 points where LAVOIR asks and leaves it unchanged where it does not; on SGD the cap lowers the asking rate from 93% to 8.6%. On Laya's twelve benchmarks LAVOIR is above Laya's reported scores on seven, and it answers a question in 31 ms (median, GH200).","published":"2026-09-25T02:31:02Z","updated":"2026-09-25T02:31:02Z","authors":["Furkan Yilmaz","Habibe Aleyna Tasdemir","Muhammed Faruk Gozay"],"categories":["cs.AI","cs.CL"],"headline":"New method helps AI decide when and what questions to ask","summary":"Researchers developed LAVOIR, an AI approach that improves decision-making by estimating the value of asking additional questions before answering. Unlike previous models that guess missing details, LAVOIR predicts how much asking about specific information can increase accuracy, without needing extra human-labeled data. Tests show it asks fewer questions but improves accuracy notably compared to always guessing, matching the performance of ideal question-asking strategies on certain tasks. These findings are based on designed experiments and simulations with no claims beyond tested scenarios."},{"id":"2609.30705v1","title":"The Price of Thought: Does Test-Time Reasoning Pay in LLM Trading?","abstract":"While inference-time reasoning in large language models (LLMs) promises better decision making, its higher computational cost may not yield better economic outcomes. Yet reasoning controls are rarely evaluated as economic interventions, where changes in model outputs must translate into better portfolios after trading costs. We conduct a controlled study of representative LLMs from the DeepSeek, GPT, and Gemini families. We vary reasoning effort while holding information available at each formation date, prompts, output formats, and portfolio construction fixed. Our evaluation covers a full year of U.S. equities under three input conditions: numerical, identifiable news, and masked news. It includes more than 800,000 asset predictions and repeated model generations. Across all three model families, additional reasoning does not produce a reliable improvement in net portfolio returns. For DeepSeek, where we examine the full progression from no reasoning to maximum reasoning, performance is nonmonotonic. Repeated generations also produce unstable treatment effects and portfolio selections, even when overall scores remain similar. These findings show that additional reasoning can change financial decisions without reliably improving their economic value, motivating validation for each task before deployment.","published":"2026-09-25T02:28:01Z","updated":"2026-09-25T02:28:01Z","authors":["Jiayi Chen","Guiling Wang"],"categories":["cs.AI"],"headline":"Extra Reasoning in AI Trading Models May Not Boost Profits","summary":"Researchers studied whether having large language models (LLMs) do more reasoning during stock trading predictions leads to better financial results. They tested various LLMs over a year of U.S. stock data with different types of inputs and controlled setups. Their analysis found that more reasoning did not consistently improve portfolio profits after costs, and sometimes made decisions less stable. The researchers suggest carefully validating AI reasoning for each specific use before relying on it in real trading."},{"id":"2609.30704v1","title":"From Visual Search to Movement Control: A Priority Field for Artificial Agents","abstract":"Human spatial attention is widely conceptualized as being guided by a priority map that integrates perceptual salience, current goals, and past experiences. Here, we extend priority-based computation to movement control in artificial agents. We first introduce a lightweight model of visual search based on an integrated priority map. Trained on human saccades, it reproduced key behavioral patterns, including oculomotor suppression and history-driven selection. Extending the search model, we equipped an artificial agent with a priority field and evaluated its performance in a reach-avoid task that required reaching a goal destination while avoiding moving obstacles. Compared with alternative architectures, priority-field agents trained more efficiently and performed better in unseen, complex scenarios, even from simple demonstrations. Adding a simple memory mechanism also produced human-like, history-driven effects in anticipating the likely location of the upcoming goal. These findings suggest that priority-based computation may provide a promising foundation for movement control in artificial agents.","published":"2026-09-25T02:24:36Z","updated":"2026-09-25T02:24:36Z","authors":["Han Zhang","Zhong Cao"],"categories":["cs.RO"],"headline":"New Model Helps Artificial Agents Decide Where to Look and Move","summary":"Researchers developed a simple model that mimics how humans focus their attention based on what stands out and past experiences. They trained this model using human eye movements and found it reproduced key behaviors. Applying it to artificial agents, the model helped them navigate tasks like reaching a target while avoiding obstacles more efficiently and adaptively. The research suggests using this priority-based approach could improve movement control in artificial systems, though these results come from controlled tasks and simulations."},{"id":"2609.30703v1","title":"SAGE: Source-Anchored Guidance via Frequency Equalization for Hierarchical RGB-T Alignment and Fusion","abstract":"Spatial misregistration and cross-modal discrepancies often cause ghosting, structural blurring, and content imbalance in RGB-T fusion. Existing methods typically decouple appearance adaptation, geometric alignment, and information fusion, limiting dependency propagation across stages. We propose Source-Anchored Guidance via Frequency Equalization for Hierarchical RGB-T Alignment and Fusion (SAGE), a unified framework integrating frequency equalization, hierarchical alignment, and subband fusion. SAGE employs invertible joint encoding and source-specific low-frequency modulation to derive structural and gain guidance while preserving source information. Hierarchical frequency collaborative alignment estimates global affine geometry from low-frequency approximations and transfers geometric and contextual cues to high-frequency correlation reasoning for reliability-aware residual refinement. Guided subband fusion jointly aggregates the aligned frequency coefficients under propagated source and alignment guidance, coordinates complementary low- and high-frequency information, and reconstructs the fused image through the inverse wavelet transform. Extensive experiments on RGB-T datasets with real-world and synthetic misalignments demonstrate consistently competitive performance in alignment and fusion, validating the effectiveness of source-anchored guidance for weakly registered RGB-T images.","published":"2026-09-25T02:22:00Z","updated":"2026-09-25T02:22:00Z","authors":["Timing Li","Yiming Sun","Boan Tao","Xiyuan Gao","Haifang Cao","Pengfei Zhu"],"categories":["cs.CV","cs.AI"],"headline":"New Method Improves Combining RGB and Thermal Images","summary":"Researchers addressed issues in combining color (RGB) and thermal (T) images, which often suffer from misalignment causing blurring and ghosting. They developed a method called SAGE that better aligns and fuses these images by using frequency-based techniques and hierarchical processing steps. Their experiments on various datasets showed that SAGE improves alignment and fusion quality, supporting the idea that their approach helps handle imperfectly matched RGB-T images. The findings are based on tests with both real and simulated data."},{"id":"2609.30702v1","title":"From Bilateral Trade to Matching Markets: Sharp Gains from Trade","abstract":"We study gains from trade in matching markets with independent private values and costs, Bayesian incentive compatibility, interim individual rationality, and no expected budget deficit. A second-best guarantee for finite bilateral trade extends without loss to matching markets with independent Borel priors, arbitrary downward-closed feasibility, and finite expected first-best gains. For bounded buyers with monotone hazard rates and arbitrary bounded sellers, we determine the exact worst-case ratio of second-best to first-best gains, approximately $0.72490721$. For binary buyers and sellers with at most $m$ types, we determine the exact ratio for every $m$, including $8/9$ when $m=2$ and a limit of $4/5$ as $m$ grows. Both families of bounds are tight already in bilateral trade.","published":"2026-09-25T02:20:38Z","updated":"2026-09-25T02:20:38Z","authors":["Zhengyang Liu","Ying Qin","Zihe Wang"],"categories":["cs.GT"],"headline":"Researchers quantify trade benefits in complex matching markets","summary":"The researchers analyze how much benefit people can gain from trading in markets where buyers and sellers are matched based on their preferences and private costs. They extend results known for simple two-party trades to more complex multi-party matching situations, providing exact ratios that compare achievable trade benefits under realistic constraints to an ideal scenario. Their findings highlight the limits of gains from trade in these settings, supported by mathematical calculations rather than real-world experiments."},{"id":"2609.30701v1","title":"Propagation delays and regional intensity changes in lensed hotspot images","abstract":"Propagation delays cause an image recorded at a single observer time to combine radiation emitted at different stages of the source evolution. Comparable changes in total intensity can accompany distinct and even opposite changes in apparent image size. Using regional intensities, centroids, and covariances, we apply the law of total covariance to separate changes in regional intensity weights from changes in internal widths and centroid separation. Ray tracing simulations of a finite Gaussian hotspot moving along a prescribed strong field trajectory show two events with comparable attenuation in screen integrated intensity but opposite changes in second moment size. In the contracting event, the regional weights move away from balance; in the expanding event, they move toward balance even as the centroids approach each other. Differential propagation delays therefore drive these opposite size responses by redistributing intensity between spatially separated image regions.","published":"2026-09-25T02:18:56Z","updated":"2026-09-25T02:18:56Z","authors":["Yong Li","Xiao-Gang Lan"],"categories":["gr-qc","astro-ph.IM"],"headline":"How light delays affect images of moving hotspots near black holes","summary":"Researchers studied how delays in light reaching an observer can mix signals emitted at different times from a moving hotspot near a strong gravitational field. Using simulations, they found that similar changes in total brightness can be linked to opposite changes in the apparent size of hotspot images. Their work shows that these size changes happen because light delays shift how brightness is distributed across different parts of the image, revealing complex effects in gravitational lensing scenarios."},{"id":"2609.30700v1","title":"Coarse geometry of metric measure spaces","abstract":"Using ideas from optimal transport theory, we introduce a notion of measured coarse equivalence for metric measure spaces and define a corresponding variant of the uniformly finite homology of Block and Weinberger, called weighted $\\ell^\\infty$ homology, for large-scale doubling metric measure spaces. We prove that this homology is invariant under measured coarse equivalence and that the vanishing of its zeroth homology is equivalent to weighted non-amenability. The proofs combine techniques from optimal transport theory and the disintegration of measures.","published":"2026-09-25T02:16:53Z","updated":"2026-09-25T02:16:53Z","authors":["Takayuki Okuda","Takashi Shioya"],"categories":["math.MG"],"headline":"New Method Explores Large-Scale Geometry of Measured Spaces","summary":"Researchers used ideas from optimal transport theory to study large-scale properties of spaces equipped with both distances and measures. They created a new mathematical tool, called weighted ℓ∞ homology, to analyze these spaces and showed it remains unchanged under certain equivalences. Their work connects this tool to a property called weighted non-amenability. The findings are theoretical, relying on advanced mathematical concepts and proofs, and apply specifically to a type of space known as large-scale doubling metric measure spaces."},{"id":"2609.30699v1","title":"\"If You're Not Doing It, Somebody Else Is\": Active Negotiation and the Invisible Labor of Sustained LLM Use","abstract":"Large language models (LLMs) have become fixtures of academic work even as their users describe them as degrading their writing, thinking, and skills. Dominant adoption frameworks read continued use as evidence of satisfaction, and cannot explain continued use of a distrusted tool. We interviewed 36 graduate student workers, balanced between English-as-a-foreign-language (EFL) and non-EFL speakers, and introduce the Active Negotiation framework: a model of sustained LLM use as a recurring cycle of risk, mitigation, and justification. A failure surfaces a risk, mitigation labor addresses it, and a justification renders the residual risk tolerable until the next failure reopens the cycle. The cycle runs across three dimensions: practical, auditing output; internal, auditing one's own cognition and identity; and social, managing how peers and institutions perceive use. EFL participants invoke linguistic parity as a further justification. We reframe continued adoption as compliance sustained by invisible labor.","published":"2026-09-25T02:13:25Z","updated":"2026-09-25T02:13:25Z","authors":["Matt Viana","Patrick Erickson","Shomir Wilson","Dana Calacci"],"categories":["cs.HC"],"headline":"Why People Keep Using AI Writing Tools They Don’t Fully Trust","summary":"Researchers studied why graduate students keep using large language models (LLMs) for writing even though they feel these tools harm their skills. By interviewing 36 students, they developed a framework showing that users constantly manage risks by checking output, reflecting on their thinking, and handling social judgments. Especially for non-native English speakers, continued use also involves striving for language equality. The study finds that ongoing LLM use involves hidden effort to manage problems, not simple satisfaction."},{"id":"2609.30698v1","title":"MM-VeriAgent: Learning to Use Extensive Tools to Verify Multimodal Misinformation with Reinforcement Learning","abstract":"Real-world multimodal misinformation often involves mixed forgery sources, requiring sample-specific detection strategies. Existing tool-augmented methods rely on predefined workflows or inference-time planning, limiting adaptability or increasing inference cost. To address this issue, we introduce \\textbf{MM-VeriAgent}, which learns to verify mixed-source multimodal misinformation with tools. We first build \\textbf{MM-VeriTools}, a specialized toolkit for misinformation detection agents. By benchmarking various candidate models and methods on the sub-tasks required by mixed-source detection, we select the strongest for textual, visual, and cross-modal forgery analysis and encapsulate them as callable tools with a unified interface. On top of this toolkit, we train the LVLM agent with reinforcement learning to teach it how to use these tools to better solve mixed-source detection. Since many of the tools are specialized models whose online execution at every rollout severely limits RL efficiency, we further introduce \\textbf{Tool-Execution Cache}, which pre-executes candidate tool calls and reuses their cached outputs during training. This preserves multi-step rollouts while reducing online tool execution, largely improving the training efficiency.Experiments on MMFakeBench demonstrate substantial accuracy gains over the base model without explicit tool search at inference time. Ablation and efficiency analyses further validate the learned tool-use policy and show that Tool-Execution Cache reduces online tool executions during training.","published":"2026-09-25T02:12:16Z","updated":"2026-09-25T02:12:16Z","authors":["Peipei Li","Shuhan Xia","Shengyang Liu","Zekun Li","Ran He"],"categories":["cs.CV"],"headline":"Training AI to Verify Mixed-Source Misinformation Using Special Tools","summary":"The researchers address the challenge of detecting misinformation that combines text and images from different fake sources, which existing methods struggle to handle efficiently. They created a new system called MM-VeriAgent that learns how to use a set of specialized detection tools through trial-and-error training. Their tests show improved accuracy in spotting fake content, and they developed a method to speed up training by reusing tool results, though these findings are based on experimental setups and specific benchmarks."},{"id":"2609.30697v1","title":"The second gap for self-shrinkers with constant norm of the second fundamental form","abstract":"Let $X: M^{n}\\to \\mathbb{R}^{n+1}$ be a complete self-shrinker with constant squared norm of the second fundamental form $S$. In this paper, we prove that if $S\\leq \\frac{10}{7}$, then $S=1$ or $S=0$, and the self-shrinker is isometric to either a round sphere $\\mathbb{S}^n(\\sqrt{n})$ with the center at the origin, or a cylinder $\\mathbb{S}^k(\\sqrt{k})\\times \\mathbb{R}^{n-k},~~1\\leq k\\leq n-1$, or a plane through the origin.","published":"2026-09-25T02:11:20Z","updated":"2026-09-25T02:11:20Z","authors":["Pengpeng Cheng","Tongzhu Li"],"categories":["math.DG"],"headline":"Researchers identify specific shapes in geometric self-shrinkers","summary":"The researchers studied special shapes called self-shrinkers, which change size but keep form, in mathematic spaces. They focused on cases where a certain geometric measurement stays constant and found that if this measurement is below a specific threshold, the shapes must be either a perfect sphere, a cylinder, or a flat plane. Their results are theoretical and based on mathematical proofs about these idealized geometric objects."},{"id":"2609.30696v1","title":"Learning Vision-Based Agile Gap Traversal: Differentiable Simulation with a Warm-Started Critic","abstract":"Traversing narrow gaps is challenging for autonomous quadrotors, especially when control commands come directly from high-dimensional visual observations. Existing end-to-end methods often rely on behavior cloning or full-rollout backpropagation through time (BPTT) via differentiable simulation, which can limit policy performance or incur high training costs. We propose a two-stage reinforcement learning framework for more efficient ego-centric visuomotor gap-traversal policy training, leveraging quasi-analytical policy gradients (QPG) via differentiable simulation and critic warm-starting. The framework utilizes QPG to avoid backpropagation through visual rendering, reducing computation and memory costs while improving sample efficiency. In the first stage, an expert actor and critic are trained using privileged observations, including gap geometry. Unlike prior gap-traversal approaches, our training utilizing QPG does not require resetting the agent along optimized reference trajectories. In the second stage, a visual policy is trained using binary gap masks from two ego-centric cameras and low-dimensional observations, while its privileged critic is warm-started from the first stage. This substantially improves training efficiency and traversal success compared with cold-starting the critic or using full-rollout BPTT. Our framework does not require retraining the expert actor when system parameters change, enabling more efficient generalization across drone platforms than state-of-the-art visual gap-traversal methods based on action supervision. The learned visual policy also generalizes to gaps with unseen shapes. Extensive real-world experiments further demonstrate robust gap traversal using binary masks rendered online. Beyond gap traversal, the proposed framework is generic and can be extended to other visuomotor robot learning tasks.","published":"2026-09-25T02:07:30Z","updated":"2026-09-25T02:07:30Z","authors":["Nuthasith Gerdpratoom","Tianchen Sun","Yichao Gao","Lin Zhao"],"categories":["cs.RO","eess.SY"],"headline":"New Method Helps Drones Navigate Narrow Gaps Using Visual Input","summary":"The researchers address the challenge of autonomous drones flying through narrow gaps using only visual data. They developed a two-step learning process combining efficient simulation techniques and a pre-trained evaluation system to train the drone’s control policy without heavy computation. Their approach improves training speed and success rates in both simulated and real-world tests, and the trained system can adapt to different drones and previously unseen gap shapes. However, the conclusions are based mainly on simulations and specific experiment setups."},{"id":"2609.30695v1","title":"Multi-Objective Human-in-the-Loop Bayesian Optimization of a Lower-Limb Exoskeleton","abstract":"Human-in-the-loop optimization (HILO) is a common approach for optimizing the control of assistive devices to account for the wearer's unique biomechanics and subjective preferences. However, despite research suggesting that a person may have a different prioritization of objectives depending on time-varying factors such as the environment, their mood, or energy levels, existing HILO approaches only consider a single objective or enforce a fixed weighting on a set of objectives. Neither approach is capable of representing an individual's preferences over objectives. In this work, we propose Multi-Objective Human-in-the-loop Bayesian Optimization (MO-HILBO), which builds on explicit multi-objective Bayesian optimization to efficiently infer a personalized set of Pareto-optimal controllers. We compare our approach with an existing multi-objective HILO method and experimentally demonstrate MO-HILBO on a lower-limb exoskeleton across two objectives: metabolic cost (efficiency) and ordinal human feedback (comfort). We find that MO-HILBO (1) discovers Pareto-optimal controllers, and (2) that the pairwise ordering of points on the Pareto front itself is consistent with validation trials. Lastly, we open-source mohilo, a Python package for running both HILO and MO-HILBO on wearable devices: https://dynamicmobility.github.io/mohilo/.","published":"2026-09-25T02:03:29Z","updated":"2026-09-25T02:03:29Z","authors":["Neil Janwani","Matthew T. Lerner","Aaron J. Young","Maegan Tucker"],"categories":["cs.RO","cs.HC"],"headline":"New Method Optimizes Exoskeletons for Comfort and Efficiency","summary":"Researchers developed a new approach called MO-HILBO to improve how assistive devices like lower-limb exoskeletons are personalized. Unlike earlier methods that focus on just one goal or use fixed goal priorities, this technique considers multiple goals, such as comfort and energy efficiency, and adapts to individual preferences. The researchers tested MO-HILBO experimentally and found it could effectively identify optimal settings balancing these goals. They also made their software publicly available for others to use."},{"id":"2609.30694v1","title":"Training-Free Contextual ASR via SpeechLLM-Based Error-Aware Selective Retrieval","abstract":"Recognition of domain-specific and low-frequency terms remains challenging for automatic speech recognition (ASR). Although contextual biasing can improve their recognition, directly providing a large terminology dictionary introduces many irrelevant biasing terms. Retrieval-based contextual biasing addresses this issue by selecting candidate terms from an external dictionary, but querying many recognized words requires numerous dictionary lookups and may yield poorly targeted candidates. We propose a training-free contextual ASR framework in which a pretrained speech large language model (SpeechLLM) jointly generates an ASR hypothesis and localizes error spans likely to involve domain-specific terms. Only the localized spans are used to retrieve phonologically similar terms from an external terminology dictionary. The same SpeechLLM then re-recognizes the audio conditioned on the first-pass hypothesis and the retrieved terms, without task-specific model training. To assess applicability across domains, we evaluate the framework on medical, air traffic control, and financial speech. The proposed method substantially reduces dictionary queries while improving the recall and ranking of relevant terminology candidates and second-pass ASR performance across all three domains.","published":"2026-09-25T02:00:44Z","updated":"2026-09-25T02:00:44Z","authors":["Natsuo Yamashita","Ai Nemoto","Ryosuke Koichi","Masaaki Yamamoto"],"categories":["cs.SD","eess.AS"],"headline":"New Method Improves Speech Recognition of Rare Terms Without Extra Training","summary":"Researchers tackle the problem of recognizing uncommon or specialized words in automatic speech recognition (ASR). They developed a method that uses a pretrained speech language model to first identify possible errors related to these terms, then selectively retrieves similar words from a dictionary to improve understanding—all without additional training. Tests in medical, air traffic control, and financial fields show that this approach reduces unnecessary dictionary searches and enhances the accuracy of recognizing specialized vocabulary."},{"id":"2609.30693v1","title":"Convergence rates in the periodic homogenization of vanishing-viscous Hamilton-Jacobi equations","abstract":"We study convergence rates when periodic homogenization and vanishing viscosity occur at the same scale in Hamilton--Jacobi equations whose momentum Hessians are positive definite at every point. For bounded Lipschitz initial data and a class of smooth Hamiltonians, we establish an optimal rate $O(\\varepsilon|\\log\\varepsilon|)$ on fixed time intervals. The second-order term plays a key role: the elliptic cell problem provides higher derivative bounds for the effective Hamiltonian $\\bar{H}$ and its Legendre dual $\\bar{L}$. The proof is purely PDE, based on corrector expansions and viscosity comparison, while its guiding idea comes from the control interpretation of the Hopf--Lax formula. For a fixed target $(x,t)$, a minimizing origin selects a characteristic velocity and its dual momentum, these guide the construction of smooth comparison profiles, whose gradients supply the momentum argument of the first-order corrector. Thus we never need to differentiate the effective solution, even at points where it is nonsmooth.","published":"2026-09-25T01:58:31Z","updated":"2026-09-25T01:58:31Z","authors":["Kai Qin"],"categories":["math.AP"],"headline":"Researchers analyze precise convergence in complex mathematical equations","summary":"The researchers investigate how quickly solutions converge when two effects—periodic averaging and vanishing viscosity—occur together in specific mathematical equations called Hamilton-Jacobi equations. They focus on cases with smooth conditions and bounded initial data, establishing an optimal convergence rate involving a small parameter ε. Their approach uses advanced mathematical techniques related to partial differential equations and control theory, relying on constructing comparison functions without needing to differentiate nonsmooth solutions. The findings are theoretical and apply to this particular setting."},{"id":"2609.30692v1","title":"LUMO (Lightweight Unified Multilingual Orchestrator): A Privacy Preserving Offline Voice Assistant","abstract":"Reliable voice interaction is essential in environments with limited internet connectivity and strong privacy. However, most existing voice assistants depend on cloud-based services, which leads to latency issues, dependency on internet access, and privacy vulnerabilities. This research presents LUMO (Lightweight Unified Multilingual Orchestrator), a privacy preserving offline voice assistant designed for edge computing environments. This system integrates local Automatic Speech Recognition (ASR), locally deployed quantized Large Language Model (LLM), and Text-to-Speech (TTS) synthesis into a fully offline pipeline running on a Raspberry Pi 5 with 8 GB RAM. To enable efficient operation on resource constrained hardware, the language model is compressed using 4-bit GGUF quantization, which reduces memory usage while preserving practical conversational capability. Existing edge based voice assistants Mycroft provides partial offline functionality without a generative LLM, with an approximate latency of ~5 s and power consumption of ~12 W, while Rhasspy supports full offline operation but lacks generative capabilities, with ~3 s latency and ~11 W power usage. In contrast, LUMO achieves a Word Error Rate (WER) of 6.8% for short English utterances in low noise conditions, an end-to-end response latency of 2.0-4.0 s, and a lower peak power consumption of approximately 9.0 W. The system also achieves effective offline recognition for Bangla speech, supporting multilingual accessibility in low resource settings. By operating entirely offline, LUMO provides strong data privacy, reduced need for cloud connectivity, and suitability for privacy sensitive edge execution such as rural healthcare, education, and disaster response scenarios.","published":"2026-09-25T01:57:42Z","updated":"2026-09-25T01:57:42Z","authors":["Md. Mehedi Hasan Naeem","Mst. Kamrunnahar Ruma","Nafiza Anjum","Shakila Sultana","Md. Sujan Ali"],"categories":["cs.LG"],"headline":"LUMO: An Offline Voice Assistant Protecting Privacy and Working Without Internet","summary":"The researchers developed LUMO, a voice assistant that works entirely offline, addressing problems of internet reliance and privacy found in typical cloud-based assistants. They combined speech recognition, a compact language model, and speech synthesis to run on low-power hardware like a Raspberry Pi. Tests showed LUMO responds quickly, uses less power, and understands English and Bangla well in quiet settings. Findings suggest LUMO may help users in places with weak internet and strong privacy needs, though results come from controlled experiments on limited hardware."},{"id":"2609.30691v1","title":"ADF-EA: A Unified Execution Assurance System for Agent Device Foundation","abstract":"Agents based on large language models (LLMs) can access heterogeneous devices through tools and APIs, but reliable execution must account for unmet effects, uncertain outcomes, and changing prerequisites. A command may be acknowledged without producing its intended effect, while missing feedback may obscure an action that has already succeeded. We present Agent Device Foundation--Execution Assurance (ADF-EA), an architecture that connects agent planning and device execution through shared capability contracts. Device Capability Contracts (DCCs) unify invocation conditions, intended effects, evidence requirements, and recovery rules across heterogeneous interfaces. Agents use these contracts to plan, while the runtime applies the same semantics to authorize actions, verify effects, and govern continuation and completion. Persistent execution state retains verified progress, unresolved outcomes, and remaining budgets across plan revisions, enabling observation-based recovery, authorized retries, and necessary state repair. We formalize the execution lifecycle and establish conditional soundness properties for completion and recovery authorization. Evaluations span multiple LLMs, five agent frameworks, and simulated process-control, household, and robotic manipulation domains. Compared with direct invocation and existing execution-checking approaches, ADF-EA reduces false completion and unnecessary repetition, supports necessary state repair, prevents calls to unavailable capabilities, and preserves permitted task completion and recovery. These results demonstrate DCCs as a reusable semantic foundation for agent autonomy across heterogeneous devices, unifying capability-based planning, evidence-grounded execution, and authorized recovery within one architecture.","published":"2026-09-25T01:56:12Z","updated":"2026-09-25T01:56:12Z","authors":["Xuechun Li","Jiaxin Liang","Jie Li","baolong Li","Jue Wang","Peng Yuan","Hang Huang"],"categories":["cs.MA"],"headline":"New system helps AI agents reliably control diverse devices","summary":"Researchers developed ADF-EA, an architecture that helps AI agents using large language models better manage and verify actions when controlling different devices. It uses shared 'Device Capability Contracts' to clearly define what actions require, their expected results, and how to handle problems or missing feedback. Tested with various AI models and simulated tasks, this system helped reduce errors like false task completion and repeated actions, supporting better recovery and state management. The findings show promise but are based on simulations and formal models."},{"id":"2609.30690v1","title":"Threat-Aware Energy-Efficient Deployment for Dynamic UAV Networks: A Multi-Agent RL Approach","abstract":"Ensuring operational safety in threat-prone environments remains a critical challenge for multi-UAV networks serving as aerial base stations. This paper proposes an efficient framework to maximize global energy efficiency (EE) while promoting safe operation through threat-aware clustering and reward-based safety enforcement. The proposed framework is executed in three steps. First, a threat-aware K-means (TAKM) algorithm determines the minimum required UAVs and computes safe initial placements. Second, an optimal matching stage assigns physical UAVs to these centroids to minimize energy expenditure. Third, a threat-aware multi-agent twin delayed deep deterministic policy gradient (MATD3) algorithm dynamically optimizes trajectories, power, and user associations. Simulation results show that the proposed framework achieves zero observed safety violations in the considered scenarios while achieving superior EE and faster convergence than other learning methods and non-clustering baselines. Compared to heuristic optimization, the proposed framework outperforms the greedy particle swarm optimization (GPSO) and achieves performance comparable to that of the optimized PSO (OPSO), while incurring significantly lower online deployment computational complexity. Furthermore, the proposed framework demonstrates effective generalization to unseen user distributions, large UAV fleets, and different threat geometries, while maintaining zero safety violations.","published":"2026-09-25T01:52:43Z","updated":"2026-09-25T01:52:43Z","authors":["Faisal Al-Kamali","Hussein A. Ammar","Francois Chan","James H. Bayes","Yasser Gadallah","Mohamed H. Ahmed"],"categories":["cs.IT","cs.AI","cs.LG","eess.SP"],"headline":"Smart Drones Use AI to Stay Safe and Save Energy in Risky Areas","summary":"Researchers developed a new approach to help multiple drones work safely as flying base stations in dangerous environments. Their method uses a special step-by-step strategy to place drones, assign tasks, and adjust routes and power dynamically to avoid threats while using energy efficiently. Simulations show this approach prevents safety problems and performs better or faster than other methods, even adapting well to different conditions, though these findings are based on computer models rather than real-world tests."},{"id":"2609.30689v1","title":"Comment on \"Symmetric Pseudo-Random Matrices\"","abstract":"In 2018, Soloveychik, Xiang and Tarokh considered a pseudo-random symmetric circulant matrix constructed from binary Golomb sequences of length $n=2^m-1$, and claimed a proof that its empirical spectral distribution converges almost surely to the semicircle law as $n$ grows to infinity by the method of moments. In this comment note we show that their argument contains several technical flaws and inappropriate applications of technical lemmas. Instead we demonstrate that the eigenvalues of the matrix are simply a normalized twisted Kloosterman sum varying over the multiplicative character, to which we apply Katz's result directly to establish the asymptotically semicircle spectral distribution deterministically.","published":"2026-09-25T01:52:24Z","updated":"2026-09-25T01:52:24Z","authors":["Chin Hei Chan"],"categories":["math.PR","cs.IT","math.NT"],"headline":"Researchers correct a math proof on pseudo-random matrix behavior","summary":"The researchers reviewed a 2018 study that claimed a certain type of pseudo-random matrix's patterns follow a known semicircle distribution as its size grows. They found mistakes in the original proof and instead linked the matrix's properties to a mathematical concept called twisted Kloosterman sums. Using an established result by Katz, they provided a more direct and reliable demonstration that the matrix's behavior approaches the semicircle distribution. Their work focuses on improving the theoretical understanding of these matrices through careful mathematics."},{"id":"2609.30688v1","title":"On the Limits of Univariate Deep Learning for Significant Wave Height Forecasting","abstract":"This study conducts a systematic hyperparameter search across five deep learning architectures, DLinear, LSTM, PatchTST, ResAttLstm, and Mamba2, and nine context lengths (1-168 h) for single-station significant wave height (Hs) forecasting on NDBC buoy 41009, followed by re-evaluation of the best configurations on a 47-buoy, 37-year corpus. The five families converge to a common performance level on the multi-buoy evaluation (between-family SD = 0.0014 m^2, 0.8% of the grand mean), a spread dwarfed by the 4.83x cross-dataset MSE shift between buoy corpora. All multi-buoy trials beat persistence (mean skill +0.062), but no architecture consistently outperforms the others. On the single-buoy experiment, skill peaks at 12-24 h where five trials fall below persistence, per-family Q4/Q3 test MSE ratios range from 2.4 to 2.6, and deep models underperform persistence for the most extreme 1% of waves. These findings are consistent with the interpretation that persistence already captures the dominant linear-inertial signal in univariate Hs, and that architecture engineering under this univariate input setting has reached diminishing returns: cross-buoy variance, not model class, dominates forecast error. Future work should prioritise atmospheric covariates, zero-shot cross-buoy transfer, and decomposition of Hs into swell and wind-sea components. By establishing a rigorous reference baseline for what univariate Hs models can and cannot achieve, this study provides a benchmark against which future multivariate and physics-informed approaches can be calibrated, and offers practical guidance for lightweight buoy-level forecasting in mid-latitude storm-dominated and swell-mixed environments.","published":"2026-09-25T01:51:57Z","updated":"2026-09-25T01:51:57Z","authors":["Yilin Zhai","Hongyuan Shi","Zaijin You"],"categories":["physics.ao-ph","cs.LG","physics.comp-ph"],"headline":"Limits of Single-Data Deep Learning for Wave Height Forecasting","summary":"Researchers tested five deep learning methods to predict significant wave height using only past wave data from buoys. They found all methods performed similarly, with no model clearly better, and that simple persistence predictions (assuming future conditions equal current) were hard to beat, especially for extreme waves. Their results suggest that using only one type of data limits forecasting accuracy, and future studies should include additional weather information and separate wave types to improve predictions. This work sets a baseline for what single-data models can achieve."},{"id":"2609.30687v1","title":"Particle Trajectories Beneath Fully Nonlinear Waves Generated by Horizontal Seabed Motion","abstract":"We investigate fully nonlinear water waves and fluid-particle dynamics generated by the horizontal motion of a seabed obstacle with prescribed time-dependent velocity. The governing equations are the full Euler equations, formulated in a time-dependent conformal domain that simultaneously maps the moving free surface and seabed onto fixed boundaries. The main contribution of this work is a Lagrangian formulation for computing particle trajectories in the resulting genuinely unsteady conformal domain. We derive a closed-form trajectory system in the canonical domain in which the time dependence of the conformal map is entirely represented by the real and imaginary parts of an analytic function that can be evaluated spectrally from the surface. We apply the formulation to waves generated by horizontal submarine landslide motion and characterize particle displacements throughout the fluid as functions of the initial particle position and Froude number. Our findings identify distinct regions in which particle motion is predominantly associated with the moving seabed, the generated wave, or the combined action of both. The results reveal a transition in the dominant mechanism driving particle motion: at low Froude numbers, particle displacements are primarily associated with the moving seabed, whereas at high Froude numbers the generated wave becomes increasingly dominant, particularly near the free surface and away from the obstacle path.The numerical predictions are benchmarked against laboratory data available in the literature, showing good agreement and providing a quantitative assessment of the model accuracy. Moreover, laboratory topographies beyond those considered here can be readily incorporated into the numerical framework through a Hermite interpolation procedure.","published":"2026-09-25T01:47:50Z","updated":"2026-09-25T01:47:50Z","authors":["João Vitor P. Poletto","David Andrade","Marcelo V. Flamarion","Roberto Ribeiro-Jr"],"categories":["physics.flu-dyn","math.NA"],"headline":"Studying Particle Movement Under Waves from Moving Seabeds","summary":"Researchers studied how water particles move under complex waves caused by the horizontal motion of a seabed obstacle, like a submarine landslide. They developed a new mathematical method to track particles in this changing environment and ran simulations to see how different forces affect particle movement. Their results showed that at slower seabed motions, particles move mainly with the seabed, while at faster speeds, wave motion dominates—especially near the water surface. They tested their model against lab data and found good agreement, supporting their approach."},{"id":"2609.30686v1","title":"A 2D autocorrelation-based frequency estimator reflecting spatial tissue distribution to improve Ultrasound H-scan tissue characterization","abstract":"H-scan is a promising quantitative ultrasound technique that estimates the frequency content of backscattered signals and maps the estimated frequencies onto a red/blue color scale to reflect underlying tissue properties. Although it relies on matched filters tuned to different frequencies, the broad spectral bandwidth of ultrasound produces noisy, granular displays. Here, we introduce an adaptive frequency estimator designed to suppress the noise within homogeneous regions while preserving sharpness across tissue boundaries. The method combines 2D autocorrelation with a matched filter. In the first stage, a matched-filter-based estimation yields an a priori map of the spatial distribution of frequencies. The local heterogeneity of these estimates then defines a 2D weighting function that guides a second estimation stage. Drawing on the concept of Loupas's blood velocity estimator, we apply autocorrelation over a 2D spatial kernel to recover the axial frequency components, employing a weighted summation that accounts for the spatial frequency distribution within the kernel. We benchmarked the proposed estimator against conventional approaches, including the short-time Fourier transform, the H-scan matched filter, and standard autocorrelation, using both Field II simulations and in vivo data from human subjects with hepatic steatosis. In simulation, our adaptive estimator reduced the noisy texture in homogeneous regions while retaining clear boundary delineation, whereas the other estimators could achieve only one of these objectives. Applied to the in vivo human liver, the estimator improved H-scan image quality by lowering noise and enhancing the discrimination of steatotic liver from adjacent gallbladder and skin layers.","published":"2026-09-25T01:45:48Z","updated":"2026-09-25T01:45:48Z","authors":["Jihye Baek","Thurston Brevett","Dongwoon Hyun","Ahmed El Kaffas","Kevin J. Parker","Jeremy J. Dahl"],"categories":["physics.med-ph"],"headline":"New Method Enhances Ultrasound Imaging of Liver Tissue","summary":"Researchers developed a new technique to improve ultrasound images by reducing noise while keeping clear details where different tissues meet. Their method uses a two-step process combining frequency estimation and spatial analysis to better map tissue properties. Tested on simulations and human liver data, the approach produced clearer images that helped distinguish fatty liver tissue from nearby structures, suggesting it could improve ultrasound-based tissue characterization."},{"id":"2609.30685v1","title":"Positive scalar curvature on products of noncompact manifolds","abstract":"We prove that the product of three connected noncompact smooth manifolds without boundary admits a complete metric of uniformly positive scalar curvature. The proof uses the additivity of Morse indices on products and an orientation-free version of Das's open Morse--surgery construction. We also give an example of two open manifolds whose product admits no complete metric of nonnegative scalar curvature.","published":"2026-09-25T01:45:43Z","updated":"2026-09-25T01:45:43Z","authors":["Lizhi Chen","Kuntao Jin","Milan Jovanovic"],"categories":["math.DG"],"headline":"New findings on positive curvature in combined open spaces","summary":"Researchers studied shapes called noncompact manifolds, which are unbounded and have no edges. They showed that if you take the product (a way of combining) of three such spaces, you can find a way to measure distances so that the combined shape has uniformly positive curvature everywhere. Their method involves advanced mathematical tools related to shape features and surgeries. They also found examples where combining two such spaces cannot produce this positive curvature. These results come from theoretical proofs, not physical experiments."},{"id":"2609.30684v1","title":"PixSim: a calibrated open-source simulator of instant-payment fraud, recovery and interdiction under analyst capacity constraints","abstract":"Brazil's Pix settles about 5.9 billion instant, irreversible transfers a month. A fraudulent transfer can be recovered only while the funds remain in a traceable account, and in 2025 the Central Bank's recovery mechanism (MED) returned 9% of accepted contested value. Interdiction therefore has to happen before settlement, by routing each transaction to pass, human review or block, under a finite analyst team and a regulatory hold window. To our knowledge no public simulator jointly models irreversible settlement, a regulated recovery mechanism, downstream fund dispersal and capacity-constrained review. We present PixSim, an open-source simulator of the Pix rail with these elements, calibrated to Banco Central do Brasil open data, with every parameter sourced, calibrated to one published observable, or registered as an assumption. With the model frozen, full-scale runs reproduce the 2025 recovery rate within 0.006 and its decomposition within 0.02; the February-April 2026 window is reported as a misfit and the May 2026 tracing regime as a projection. On a benchmark with a payer-side scorer, four reference policies and ten scenarios, within the simulated mule model: recovery after settlement is constrained by dispersal speed; staffing by the arrival profile cuts a fixed rule's alert expiry from 52% to 2% at constant hours; halving the team removes a fixed threshold-and-block rule's advantage over a queue-aware rule, on loss and on loss plus false-block harm (+0.106 of victim value, positive on all twenty paired seeds), while a reversal at two thirds of the team was not confirmed on independent seeds; and a synthetic scorer of held-out AUC 0.82 cuts lost value by about a quarter. Code and data: https://doi.org/10.5281/zenodo.22948895","published":"2026-09-25T01:44:15Z","updated":"2026-09-25T01:44:15Z","authors":["Bashir Zeimarani","Alireza Khatib","Somayeh Mousavinasr","Carlos Maurício Serodio Figueiredo"],"categories":["cs.LG","cs.CE","q-fin.RM"],"headline":"New Simulator Models Instant Payment Fraud and Recovery Limits","summary":"Researchers developed PixSim, an open-source computer model simulating Brazil's Pix instant payment system and its fraud recovery process under realistic conditions. The simulator accounts for limited human analysts, irreversible transactions, and fund tracing challenges. Calibrated to real data from Brazil's central bank, PixSim closely reproduces known 2025 recovery rates and explores how staff levels and review policies affect fraud detection and recovery. These results are based on simulation modeling and depend on assumptions documented by the researchers."},{"id":"2609.30683v1","title":"A Large-Scale Empirical Study of Modern Phishing Email Content","abstract":"Phishing remains one of the most pervasive threats to Internet users, and email remains its predominant delivery channel. Email content is the attack surface of phishing: it is what the victim reads and what automated defenses inspect. Yet the composition of modern phishing content is poorly measured. Prior work has characterized dimensions such as theme, call-to-action (CTA), and impersonation, but not at scale, and their associations and temporal changes remain unclear, owing to small or source-specific corpora, bag-of-words topic models, and a focus on text alone. We present a content-focused measurement study of 2.9M distinct real-world phishing emails collected over 13 months (June 2025 - June 2026) in collaboration with the Anti-Phishing Working Group (APWG). We treat each email as a composite artifact comprising message text and its attachments: 272K images, 143K PDFs, and 57K calendar invitations. Using an LLM pipeline validated against human-annotated samples, we analyze these components along three dimensions (theme, CTA, and impersonation), examine the associations among them, and measure longer-term change against a historical dataset. We find that attackers diversify what they use to deceive but converge on how victims should respond: no theme exceeds 21.3% of emails, while a single CTA, URL navigation, accounts for 73.0%. CTA and impersonation choices are conditioned on theme. Attachments play three roles: images supplement the message text, PDFs substitute for it by carrying the pretext, and calendar invitations reinforce it by replicating interaction endpoints into a persistent medium. Over the longer term, the dominant CTA for invoice-themed phishing shifted from URL navigation to offline communication, rising from 6.7% in 2015 to 46.9% in 2025.","published":"2026-09-25T01:41:03Z","updated":"2026-09-25T01:41:03Z","authors":["Jaehwan Park","Woonghee Lee","Fujiao Ji","Doowon Kim"],"categories":["cs.CR"],"headline":"Study Analyzes Content of 2.9 Million Phishing Emails","summary":"Phishing emails are a common online threat, but researchers know little about their detailed content. The researchers studied nearly 3 million real phishing emails, including texts and attachments like images and PDFs, over 13 months. They found a wide variety of deceptive themes, but most emails urged victims to click links, and attachment types served different roles in the scams. Over time, some phishing tactics changed, such as shifting from link clicks to offline contact for invoice scams. These findings come from analyzing real emails with advanced language tools."},{"id":"2609.30682v1","title":"Structure-Guided Masked Autoencoders for Ultra-High Resolution Scientific Image Understanding","abstract":"Self-supervised pre-training with Vision Transformers, including Masked Autoencoders (MAE), is difficult to apply to gigapixel scientific images. Random masking is poorly matched to the structured, multi-scale morphology of scientific data, while uniform tokenization produces prohibitively long sequences that make $O(N^2)$ attention impractical. We propose SGMA, a structure-guided masked autoencoding framework for ultra-high-resolution scientific images. SGMA couples two components: a content-adaptive quadtree tokenizer that compresses gigapixel images into a fixed-length sequence, and a structure-conditioned masking process that biases reconstruction toward spatially informative regions. To stabilize this process across scales, we introduce Damped Accumulation (DA), which aggregates signal-dependent responses across the tree into a structure canvas used to guide masking. The resulting pre-training task preserves fine microstructure while remaining compatible with standard ViT encoders and MAE-style reconstruction. Across electron microscopy, whole-slide optical microscopy, and X-ray CT datasets, SGMA consistently outperforms MAE baselines. It achieves 95.68% Dice on the 8K x 8K x 28K SpringXCT dataset, improving over the same-architecture MAE baseline by +13.00 points, and 83.21% Dice on the 32K^2 WSI PAIP dataset, improving by +16.84 points, while providing up to a 24.8x inference speedup.","published":"2026-09-25T01:39:06Z","updated":"2026-09-25T01:39:06Z","authors":["Enzhi Zhang","Du Wu","Rui Zhong","Cong Ma","Isaac Lyngaas","Amir Koushyar Ziabari","Xiao Wang","Peng Chen","Tao Luo","Toshio Endo","Fumiyoshi Shoji","Kento Sato","Kentaro Uesugi","Takayuki Nonoyama","Ryuji Kiyama","Masahiro Yoshida","Masaru Tezuka","Tetsuya Ishikawa","Satoshi Matsuoka","Masaharu Munetomo","Mohamed Wahib"],"categories":["cs.CV"],"headline":"New Method Helps Computers Understand Ultra-High-Resolution Scientific Images","summary":"Researchers developed a technique called SGMA to improve how computers analyze very large scientific images, like those from microscopes and medical scans. They designed a way to focus on important image regions while reducing the huge data size, making the process faster and more accurate. Tests show SGMA outperforms previous methods, though the findings are based on specific image datasets and computer simulations."},{"id":"2609.30681v1","title":"Long-time reduction for the biharmonic nonlinear Schrödinger equation","abstract":"In this article, we study the reduction of the one-dimensional biharmonic nonlinear Schrödinger equation to the cubic nonlinear Schrödinger equation in the vanishing higher-order dispersion limit. In the intermediate regularity regime $0<s<2$, where the energy conservation law does not control the $H^s$--norm, we prove the long-time $L^2$--convergence of $H^s$--solutions with an exponential-in-time approximation bound. The argument relies mainly on persistence of regularity. This result provides a simple example of its use in limit problems without relying on higher-order conservation laws.","published":"2026-09-25T01:37:02Z","updated":"2026-09-25T01:37:02Z","authors":["Younghun Hong","Junyeong Jang"],"categories":["math.AP"],"headline":"Simplifying a Complex Wave Equation Over Time","summary":"Researchers studied a complicated wave equation used in physics and showed how it simplifies to a more common equation when a certain effect becomes very small. They focused on cases where usual energy measures don’t apply and proved that solutions remain close for a long time, with precise error limits growing exponentially in time. Their work illustrates a method to handle such simplifications without depending on complex conservation rules, based on keeping track of solution smoothness."},{"id":"2609.30680v1","title":"A Log-Star Comparison Between Expectation Threshold and Fractional Expectation Threshold","abstract":"We proved a log-star comparison between the expectation threshold $q(\\mathcal F)$ and the fractional expectation threshold $q_f(\\mathcal F)$ for any nontrivial increasing family $\\mathcal F$ on a finite ground set $V$ of size $|V|=N$. Specifically, we show that $$q_f(\\mathcal F)\\le64\\log_2^*(N+2)\\,q(\\mathcal F),$$ where we define $\\log_2^* x$ to be the least integer $k\\ge0$ such that applying $\\log_2$ repeatedly $k$ times gives a number at most $1$.","published":"2026-09-25T01:35:56Z","updated":"2026-09-25T01:35:56Z","authors":["Xuan Fang","Tianyu Wang"],"categories":["math.CO","math.PR"],"headline":"Researchers Compare Two Mathematical Thresholds Using Log-Star Function","summary":"The researchers examined two related mathematical quantities called expectation threshold and fractional expectation threshold, which apply to certain collections of sets. They proved an inequality showing the fractional expectation threshold is at most the expectation threshold multiplied by a function called the log-star of the set size, which grows very slowly. Their result provides a precise relationship between these thresholds for any such collection on a finite set. This finding is theoretical and based on mathematical proof rather than experiments or applications."},{"id":"2609.30679v1","title":"Design-Ignoring versus Design-Respecting World Models for Epidemiology","abstract":"World models for epidemiology learn from records shaped by study designs, including assignment, sampling, measurement, and related processes. A model may therefore reconstruct observed trajectories while learning an intervention contrast that depends on how records were collected. We formalize study design as constraints on a world model latent-world interface, action mechanism, observation likelihood, and target readout. This distinguishes design-respecting models, which encode these constraints, from design-ignoring models, which fit selected records without relating assignment and observation to the intended intervention question. Using a large scale cluster-randomized test-negative trial, we hold the latent structure and fitting settings fixed and compare a design-ignoring case-count model with a design-respecting test-negative observation model. Both models achieve comparable factual reconstruction. Yet across 500 paired resampling experiments at different relative sampling intensity, median contrast changes are substantial for the design-ignoring model, versus merely marginal for the design-respecting model. The same qualitative separation holds when sampling also varies across clusters. Thus, factual reconstruction alone does not establish design alignment; testing whether fitted contrasts preserve design-implied observation-process invariances provides a sharper evaluation.","published":"2026-09-25T01:33:18Z","updated":"2026-09-25T01:33:18Z","authors":["Xiangyu Yu","Weiyu Liu"],"categories":["stat.ME","stat.AP","stat.ML"],"headline":"New Study Explores Better Models for Understanding Epidemics","summary":"Researchers studied how computer models learn from health data shaped by how studies are designed, such as who is sampled and how measurements are taken. They compared models that ignore these design details with ones that carefully include them, using data from a large trial on infectious disease. While both models matched observed data well, the design-respecting models gave more stable and reliable estimates when conditions changed. This suggests that considering study design in models can improve their usefulness for interpreting interventions, based on the researchers' simulations."},{"id":"2609.30678v1","title":"Simulation-Efficient Analog Circuit Yield Optimization via Monte Carlo Zeroth-Order Gradient Estimation","abstract":"Yield optimization under process variation is expensive because each candidate design must be evaluated across many Monte Carlo SPICE samples. The resulting finite-sample yield is also piecewise constant in the design parameters, providing little local information for optimization. We introduce zeroth-order Monte Carlo stochastic gradient descent (ZO-MC-SGD), a black-box method that converts continuous specification margins into stochastic descent directions. Each update evaluates opposite design perturbations under shared process samples, allowing a small simulation batch to estimate a local direction without differentiating SPICE or fitting a global surrogate model. A Spearman rank-correlation test checks that the margin-based loss orders designs consistently with empirical yield. We prove that the estimator is unbiased for a Gaussian-smoothed surrogate and derive variance and sample-complexity bounds with no explicit dependence on process dimension. Across five analog circuit benchmarks with up to 30 design variables and 42 process variables, ZO-MC-SGD reaches a mean yield of 0.95 on four circuits within 50--200 simulations and the empirical yield ceiling on the fifth. Relative to the best of five black-box and learning-based baselines, it reduces the required simulation budget by up to a factor of eight.","published":"2026-09-25T01:32:26Z","updated":"2026-09-25T01:32:26Z","authors":["Liyan Tan","Yequan Zhao","Ben F. Jamroz","Ari Feldman","Zheng Zhang"],"categories":["cs.CE"],"headline":"New Method Speeds Up Analog Circuit Design Testing","summary":"Researchers addressed the problem of optimizing circuit designs, which is costly because many simulations are needed to test how variations affect performance. They developed a method called ZO-MC-SGD that efficiently estimates design improvements using fewer simulations by smartly comparing small design changes under the same conditions. Their tests on several circuit examples showed that this approach achieves high quality results with much less simulation effort compared to existing methods. The findings are based on simulation experiments and mathematical analysis, without involving real-world circuit fabrication."},{"id":"2609.30677v1","title":"Electrochemical Growth of Full Volume Meissner Effect Superconducting BKBO","abstract":"In this work we show the results of electrochemical synthesis of BKBO crystals using three different experimental setup configurations. The setups differ between each other by varying level of control over physical variables and progressing from a two-electrode to a three-electrode configuration using highly oriented platinum counter electrode. By systematic analysis of the temperature dependence of the magnetic susceptibility at the superconducting transition of the collected crystals we observe that an order of magnitude sharper superconducting transition is achieved for the most comprehensive three-electrode setup. In addition, the level of chemical substitution can be controlled by the value of the overpotential versus reference electrode. We demonstrate that with this technique a full volume Meissner effect can be achieved with a sharp transition temperature for the superconducting crystal.","published":"2026-09-25T01:30:56Z","updated":"2026-09-25T01:30:56Z","authors":["Oleksandr Foyevtsov","Kateryna Foyevtsova","Mohamed Oudah","Dan Bizzotto","George Sawatzky"],"categories":["cond-mat.supr-con"],"headline":"Improved Method to Grow Better Superconducting BKBO Crystals","summary":"Researchers studied different setups to grow BKBO crystals, which are special materials that can become superconductors. By using a setup with three electrodes, they were able to better control the process, resulting in crystals with more precise superconducting transitions and stronger magnetic effects. Their findings suggest that this method allows improved control over the crystal properties, but the results are based on experimental synthesis and magnetic measurements."},{"id":"2609.30676v1","title":"Can a Robot Read Braille? - Learning to Adapt Contact via Imitation Learning for Tactile Braille Recognition","abstract":"For people who are blind, touch provides an essen-tial channel for accessing written information through Braille. Bringing a similar capability to robots requires them not only to recognize tactile patterns, but also to actively establish physical contact that makes those patterns readable. Yet existing robotic Braille readers largely focus on recognition after contact, leaving contact establishment itself insufficiently addressed. We present an adaptive-contact framework for robotic tactile Braille reading that assesses contact quality and physically corrects unsuitable contact before recognition and reconstruc-tion. Multi-Head Policy Learning uses expert-guided contact-adjustment demonstrations to jointly learn contact acceptability and pose corrections. During deployment, the robot iteratively evaluates and re-establishes contact, retaining reliable tactile observations for pose-aware fusion and Braille reconstruction. Across 20 physical Braille plates used for learning and eval-uation, the proposed approach achieves 94.0% tactile quality and 88.6% tactile reconstruction on the ten online-evaluation plates. These results demonstrate the importance of actively establishing readable contact, rather than relying solely on recognition under imperfect tactile observations, for reliable robotic Braille reading.","published":"2026-09-25T01:30:23Z","updated":"2026-09-25T01:30:23Z","authors":["Xi Chen","Yunlong Shan","Sihan Chen","Jun Hu","Zhongxuan Li","Shiyao Zhang","Sichao Liu","Zhong Zhao","Kosta Jovanovic","Peng Zhou"],"categories":["cs.RO"],"headline":"Robot Learns to Read Braille by Adjusting Its Touch","summary":"Researchers developed a robot that can read Braille by not just recognizing patterns, but by actively adjusting its touch to get clear contact with the Braille surface. They trained the robot using expert demonstrations, enabling it to judge and improve its touch before reading. Tests on physical Braille samples showed the robot achieved high accuracy in sensing and reconstructing the Braille patterns, highlighting that good physical contact is essential for reliable robotic Braille reading."},{"id":"2609.30675v1","title":"On the support of double Grothendieck polynomials","abstract":"We prove that the support of every double Grothendieck polynomial is an $M^\\natural$-convex set. Our main new tool is a rigidity result for the $K$-classes of multiprojective varieties with rational singularities.","published":"2026-09-25T01:30:22Z","updated":"2026-09-25T01:30:22Z","authors":["Yairon Cid-Ruiz"],"categories":["math.AG","math.AC"],"headline":"Researchers Study Shapes in Double Grothendieck Polynomials","summary":"The researchers investigated a mathematical property called the 'support' of double Grothendieck polynomials, complex objects in algebra. They showed that this support forms a special type of shape known as an M-natural-convex set. To achieve this, they developed a new method involving a rigidity property related to specific geometric objects called multiprojective varieties that have a certain type of singularities. Their findings are theoretical and contribute to the understanding of these mathematical structures."},{"id":"2609.30674v1","title":"SkillRefine: Cross-Source Skill Induction and Execution Validation for LLM Agents in Refinery Planning Software","abstract":"Operating industrial planning software such as AspenTech PIMS (Process Industry Modeling System) requires an LLM agent to combine structural knowledge, procedural knowledge from expert records, and constraints revealed only during execution. These evidence sources are heterogeneous and individually incomplete: documentation describes tables and interfaces but omits task-level coordination, expert CASE records expose multi-table modification patterns without explicit schema grounding, and execution feedback reveals latent constraints only when a plan is executed. We present \\textsc{SkillRefine}, a framework that exploits the Documentation--Practice Gap between documented structure and expert practice to extract candidate coordination patterns, ground them against table definitions and COM specifications, and compile them into progressively disclosed skill packages with provenance. The library is then refined through label-free compliance screening, oracle-based match decomposition over table, row, column, and value dimensions, and signal-conditioned trajectory attribution for localized repair. We evaluate \\textsc{SkillRefine} on PIMS-Bench, a benchmark built from two AspenTech PIMS demonstration models, using disjoint construction and held-out test tasks. On the held-out test set, \\textsc{SkillRefine} achieves absolute component match F1 gains of 14\\%--30\\% across four LLM backbones, with the largest gains on complex multi-table coordination tasks.","published":"2026-09-25T01:29:20Z","updated":"2026-09-25T01:29:20Z","authors":["Dongxiao Liu","Boren Zhu","Yuwen Ding","Linghui Li","Lei Li","Xiaoyong Li"],"categories":["cs.CE"],"headline":"New Method Improves AI Planning in Industrial Software","summary":"The researchers studied how large language models (LLMs) can better operate complex industrial software used for refinery planning. They developed SkillRefine, a system that combines information from documentation, expert records, and execution feedback to create and improve skill packages for the AI to follow. Tested on a benchmark from AspenTech software, SkillRefine showed significant improvements in matching task components, especially for complex tasks involving multiple tables. The results are based on simulations with specific software models and may not fully generalize."},{"id":"2609.30673v1","title":"Realignment of ferroelectric nematic by photoinduced electric field","abstract":"Liquid crystal displays (LCDs) employ thin pixels of a paraelectric nematic, confined between two glass plates with transparent indium tin oxide (ITO) electrodes. A voltage applied to the electrodes realigns the molecules, changing the optical appearance of the pixel. Here we demonstrate that a similar molecular realignment can be triggered by low-power light irradiation when the paraelectric nematic in the cell with ITO electrodes is replaced with a ferroelectric nematic (NF). Irradiation creates an intrinsic electric field that is strong enough to realign the NF but too weak to affect the paraelectric nematic. The extraordinary sensitivity of the ITO/NF pair to light irradiation creates a platform for optical control of ferroelectric polarization and applications in photonics, sensing, smart windows, and beyond.","published":"2026-09-25T01:25:10Z","updated":"2026-09-25T01:25:10Z","authors":["Sathyanarayana Paladugu","Oleksandr Kurochkin","Ruslan Kravchuk","Mykola Kravets","Volodymyr Sashuk","Bijaya Basnet Vassili G. Nazarenko","Sergij V. Shiyanovskii","Oleg D. Lavrentovich"],"categories":["cond-mat.soft"],"headline":"Light-Induced Realignment in Ferroelectric Liquid Crystals","summary":"The researchers studied how molecules in ferroelectric nematic liquid crystals can be realigned using low-power light instead of electrical voltage, which is normally needed for liquid crystal displays. They found that light creates an internal electric field strong enough to change the molecular alignment in ferroelectric nematics but not in standard paraelectric nematics. This shows potential for controlling these materials with light, based on their testing of molecular realignment under light irradiation."},{"id":"2609.30672v1","title":"Highly Scalable Selectorless Cryogenic Memory Array Using Ferroelectric Josephson Field-Effect Transistors","abstract":"Scalable memory systems that satisfy the temperature, speed, and energy requirements of cryogenic environments are essential for the development of large-scale quantum computers. They may also benefit high-performance computing and space applications. However, existing cryogenic memory technologies often suffer from limited scalability, low operating speed, and/or high power consumption, restricting the scalability of target applications. Ferroelectric Josephson field-effect transistors (Fe-JoFETs), which combine ferroelectric polarization with the superconducting properties of Josephson junctions, offer a promising solution. The ferroelectric layer enables nonvolatile storage capability, while the Josephson junction supports high-speed, energy-efficient operations. In this work, we leverage Fe-JoFETs to develop a highly scalable, ultra-low-power, nonvolatile cryogenic memory array that does not need additional selector devices for random access. Moreover, the superconducting component of Fe-JoFET provides a binary decision during read, eliminating the need for sensing peripheral circuitry. We first develop a physics-based Verilog-A compact model for Fe-JoFETs and use it to verify the functionality of the proposed memory array. By eliminating both selector and sensing circuitry, the proposed memory architecture offers higher scalability than existing technologies. The ultra-low-power operation of this memory also makes it compatible with strict power budgets of cryogenic applications.","published":"2026-09-25T01:24:05Z","updated":"2026-09-25T01:24:05Z","authors":["Saheeb Ahmad","Shamiul Alam"],"categories":["cs.ET"],"headline":"New Low-Power Memory Design for Cryogenic Quantum Computers","summary":"Researchers address the challenge of creating memory systems that work efficiently in very cold environments needed for quantum computers. They developed a new type of memory array using ferroelectric Josephson field-effect transistors (Fe-JoFETs), which combine nonvolatile storage with superconducting properties for fast, energy-saving performance. Their computer model shows this design can scale well without needing extra devices for memory access or sensing, potentially improving memory systems for cryogenic applications. However, these results are based on simulations rather than physical prototypes."},{"id":"2609.30671v1","title":"A Vehicle-Integrated Approach to Digital Twin Deployment for Bridges","abstract":"The project A Vehicle Integrated Approach to Digital Twin Deployment for Bridges is developing a scalable framework for bridge condition assessment using vibration measurements from a sensorised inspection vehicle. It combines vehicle bridge interaction modelling, state estimation, machine learning and surrogate modelling to reduce reliance on permanent sensor networks. The Old Ada Bridge in Japan is the principal case study because it provides direct and vehicle-based field measurements across multiple experimentally introduced damage states. Progress has been made across three streams. First, forward and inverse Fourier Neural Operator (FNO) models have been demonstrated on a benchmark beam, enabling rapid response prediction and identification of damage location and severity. Second, an inspection vehicle optimisation framework has been developed to select vehicle mass and tyre suspension stiffness that maximise separation between healthy and damaged responses, integrating contact-point reconstruction, damage assessment, Kriging and particle swarm optimisation. Third, an Augmented Kalman Filter-based virtual sensing framework has been developed to estimate moving loads and reconstruct displacement, velocity and acceleration at unmeasured locations from sparse sensors. The next phase will extend and integrate these methods using the Old Ada Bridge model and field data. FNO models will be transferred to the truss structure and evaluated using intact and damaged measurements. Vehicle optimisation will be tested for transferability, while virtual sensing will be extended to the coupled vehicle bridge system to estimate road roughness, vehicle parameters, moving forces and structural responses. Ultimately, these components will form an integrated vehicle-driven digital twin for rapid simulation, response reconstruction and damage assessment under realistic conditions.","published":"2026-09-25T01:21:41Z","updated":"2026-09-25T01:21:41Z","authors":["Mehri Alamdari","Zihao Liu","Jiaji Wang","Kawabe Daigo","Chul Woo Kim"],"categories":["eess.SP"],"headline":"New Method Uses Vehicles to Monitor Bridge Health","summary":"Researchers are developing a system that uses sensors on inspection vehicles to assess bridge conditions by measuring vibrations, reducing the need for permanent sensors on bridges. They tested parts of this approach on Japan's Old Ada Bridge, creating models to predict damage and optimize vehicle settings for better data. The project combines different techniques to estimate forces and bridge responses from limited sensor data. Future work will integrate these methods to create a 'digital twin'—a virtual bridge model for quick damage assessment in real conditions."},{"id":"2609.30670v1","title":"TRACE: Temporal Audit and Condition-aware Evaluation of Streaming Video Understanding","abstract":"Streaming video understanding requires models to interpret evidence as it arrives, yet current evaluations often report task scores without specifying when evidence becomes valid, how visual history is maintained, or how responses are triggered. As a result, similar scores may correspond to different workloads, failure modes, and operational behavior. We introduce TRACE (Temporal Audit and Condition-aware Evaluation), a condition-aware benchmark and evaluation framework that makes these factors explicit. TRACE combines temporally audited visual tasks with evidence timing and instruction-dependent trigger annotations, a unified causal Core--Adapter protocol that controls information availability while recording actual history processing and response events, and multidimensional reporting of answer quality, timeliness, response-selection behavior, workload, completion, and reliability. On 1,240 records from 517 videos, we evaluate eight publicly available models or systems in eight configurations. We find that nearly identical QA accuracy can mask substantial differences in completion, answer validity, and generation workload, while proactive performance separates into response quality, response delay, false alarms (responses emitted while no target window is currently valid and a later one remains), and missed target windows. These results show that streaming-video performance should be interpreted as execution-conditioned system behavior rather than a single score. Our benchmark and code can be accessed at \\href{https://github.com/om-ai-lab/trace-bench}{https://github.com/om-ai-lab/trace-bench}.","published":"2026-09-25T01:21:25Z","updated":"2026-09-25T01:21:25Z","authors":["Yibo Ma","Qianqian Zhang","Peng Liu","Tiancheng Zhao"],"categories":["cs.CL","cs.CV"],"headline":"New Method Evaluates Streaming Video Understanding More Fully","summary":"Researchers highlight that current tests of streaming video understanding often miss when and how models process information and give answers. They developed TRACE, a new way to evaluate video models that tracks timing, history use, and response behavior in detail. Testing eight models on many videos, they found similar accuracy scores can hide big differences in how and when models respond, suggesting video understanding should be judged by detailed behavior, not just overall scores."},{"id":"2609.30669v1","title":"A Comprehensive Study of the Long-Period Cataclysmic Variable V630 Cassiopeiae","abstract":"We present a comprehensive review of V630 Cassiopeiae, which is a very long-period cataclysmic variable. Based on new observations combining spectroscopy and high-resolution photometry provided by TESS and precise distance data from GAIA, we confirm and extend previously described characteristics. The light curves show differential behaviors between different epochs, attributed to changes in the disk's brightness distribution, likely due to non-axial structures such as precessing eccentric disks. The spectroscopy reveals complex Ha line profiles, with dual components and an emission distribution that suggests a large, asymmetric, and possibly precessing disk, rather than a classic shock structure in the matter-flow impact region. Doppler tomography indicates that the emission originates at the periphery of the disk, distributed in an asymmetric pattern rather than a compact point, complicating the interpretation of radial velocities and limiting their use for accurately determining masses. Through photometric and spectroscopic analysis, we established a donor mass estimate with an approximate value of 0.38 Ms at the adopted distance of 2409 pc, with significant uncertainty related to distance and other systematic factors. Finally, using the parameters of V630 Cas, binary evolution simulations were conducted with MESA incorporating CARB magnetic braking prescription to identify an evolutionary sequence that explains its properties.","published":"2026-09-25T01:19:12Z","updated":"2026-09-25T01:19:12Z","authors":["Ivan Mora Zamora","Gagik Tovmassian","Juan Echevarria","Diogo Belloni"],"categories":["astro-ph.SR"],"headline":"Study Reveals Complex Behavior of Long-Period Star System V630 Cassiopeiae","summary":"Researchers studied V630 Cassiopeiae, a star system with a long orbital period where one star pulls material from the other, forming a disk. Using new observations and data from space telescopes, they found the disk's brightness and shape change over time, showing complex and uneven patterns rather than simple, stable structures. Their analysis indicates challenges in measuring star masses accurately, and they used simulations to explore how the system might have evolved, though uncertainties remain due to distance and other factors."},{"id":"2609.30668v1","title":"Long-time Korteweg-de Vries approximation for the Fermi-Pasta-Ulam-Tsingou system","abstract":"We prove that, as the lattice spacing $h$ tends to zero, general solutions to the infinite Fermi--Pasta--Ulam--Tsingou (FPUT) system can be approximated in $L^2$ by two counter-propagating Korteweg--de Vries (KdV) waves on time intervals of order $\\log(1/h)$. This resolves an open question raised by the first author and collaborators~\\cite{HKY2021}. Our proof combines the FPUT conservation law with an $h$-uniform local well-posedness theory in $L^2$ and persistence of Sobolev regularity. We also introduce a frequency-localized auxiliary equation to overcome the difficulty of comparing the Fourier restriction norms associated with the FPUT and KdV flows.","published":"2026-09-25T01:17:04Z","updated":"2026-09-25T01:17:04Z","authors":["Younghun Hong","Junyeong Jang"],"categories":["math.AP"],"headline":"Researchers link Fermi-Pasta-Ulam-Tsingou system to KdV wave patterns","summary":"The researchers studied a mathematical model called the Fermi-Pasta-Ulam-Tsingou (FPUT) system, which describes vibrations in a chain of particles. They showed that as the spacing between particles gets very small, solutions to this system can be closely approximated by two specific wave patterns known as Korteweg-de Vries (KdV) waves, over certain timescales. Their proof uses advanced math techniques to address previous open questions, but it applies to theoretical conditions and simulations rather than physical experiments."},{"id":"2609.30667v1","title":"StarWM: Self-Supervised Trained Attention Routing for Robust World Models","abstract":"A robust world model must strike the balance between faithfully capturing environmental dynamics and abstracting away from irrelevant content. While reconstruction-based world models ensure faithful supervision, they misallocate representational capacity by pixel area rather than dynamics relevance for visual tasks, which can cause task-irrelevant content to dominate the learned representation. Alternatively, reconstruction-free methods avoid this bias but risk discarding possibly relevant information. We propose StarWM, which uses a cross-attention module trained on self-supervised dynamics to decide where reconstruction applies. A dual-stream decoder then restricts reconstruction to the attended regions, with stop-gradient barriers preventing interference between the two objectives. These components allows reconstruction to supervise the visual content of attended regions without contaminating the latent with non-predictive information. On DeepMind Control with dynamic video backgrounds, default (reward-free) StarWM achieves the strongest performance under random-frame distractors and substantially outperforms reconstruction-based baselines under sequential video. In addition, its reward-augmented variant matches or exceeds reconstruction-free methods on sequential video, achieving the highest overall return across all distractor regimes. Mechanistic probing confirms StarWM preserves state attributes with near-perfect fidelity through long-horizon imagination while systematically discarding distractors.","published":"2026-09-25T01:14:35Z","updated":"2026-09-25T01:14:35Z","authors":["Zeqiang Zhang","Fabian Wurzberger","Maximilian Otte","Daniel Schmid","Sebastian Gottwald","Arne Peter Raulf","Daniel Alexander Braun"],"categories":["cs.CV","cs.LG"],"headline":"New method improves AI understanding of visual environments","summary":"Researchers developed StarWM, an AI model that better focuses on important parts of visual scenes while ignoring distractions. It uses a special attention system to decide which areas to reconstruct in training, balancing detail and relevance. Tests showed StarWM outperforms existing methods in complex video environments with distracting backgrounds, preserving important state information over time. The findings are based on simulation results in controlled tasks and highlight a way to improve AI models for understanding dynamic visual data."},{"id":"2609.30666v1","title":"Stability of oblique sonic shocks in steady supersonic potential flow past a wedge","abstract":"When a uniform steady supersonic oncoming flow impinges on a straight wedge, if the wedge angle is smaller than the detachment angle, two types of steady oblique shocks satisfying the entropy condition will form in the flow field: weak shocks with supersonic or subsonic downstream flow, and strong shocks with subsonic downstream flow. There have been many results on the stability of oblique shocks with subsonic or supersonic downstream flow. However, much less is known about the stability of oblique shocks with sonic downstream flow, since the flow behind the shock wave is very sensitive to disturbances. This paper investigates the stability of oblique sonic shocks under the assumption of a straight wedge with non-uniform incoming flow. We reduce the problem to a degenerate hyperbolic free boundary problem. A local Lipschitz-continuous solution with sonic-supersonic downstream flow to the free boundary problem is constructed, using the classical Ascoli-Arzelà theorem and a diagonal argument. The main difficulty of this free boundary problem lies in the hyperbolic degeneracy on the sonic line. By adopting a weighted characteristic decomposition method, we establish a delicate estimate of the downstream flow near the sonic line.","published":"2026-09-25T01:14:12Z","updated":"2026-09-25T01:14:12Z","authors":["Geng Lai"],"categories":["math.AP"],"headline":"Researchers Study Stability of Sonic Shocks on Wedges in Supersonic Flow","summary":"The researchers examined how shock waves, specifically sonic shocks, behave when supersonic air hits a wedge-shaped object. They focused on shocks where the airflow behind changes from sonic (speed of sound) to supersonic, which are sensitive and less understood. By simplifying the problem into a mathematical model, they constructed a solution describing this flow, overcoming challenges related to changes in flow type near the shock. Their findings are based on theoretical analysis and mathematical methods, not physical experiments."},{"id":"2609.30665v1","title":"On Standard perturbations of the Affine twist map","abstract":"For any $t\\in\\mathbb{R}$, consider the Affine twist map $\\rm{Aff_t}:\\mathbb{T}^2\\to\\mathbb{T}^2$ given by $$\\rm{Aff_t}(x,y)=(x+y \\text{ mod 1}, y+t \\text{ mod 1}).$$ The map $\\rm{Aff_t}$ clearly possesses an invariant foliation by horizontal curves. If $t$ is rational, each leaf is periodic, and if $t$ is irrational, the orbit of each leaf is dense in the torus. In both cases, the vertical rotation set of the proper lift of $\\rm{Aff_t}$ to the vertical cylinder is reduced to $\\{t\\}.$ Now, for $k\\in\\mathbb{R},$ define $$f_{k,t}(x,y):=(x+y+k\\sin (2πx)\\text{ mod 1},y+k\\sin (2πx)+t\\text{ mod 1}).$$ We show that: 1) when $t=p/q$ for integers $p$ and $q>0$, KAM theory implies the existence of a constant $k_{p/q}>0$ such that for $|k|<k_{p/q}$, the vertical rotation set of an adequate lift of $f_{k,p/q}$ is just $\\{p/q\\}.$ 2) when $t$ is irrational, for any $k\\neq 0$ the vertical rotation set of the adequate lift of $f_{k,t}$ is a non-degenerate interval which contains $t$ in its interior. In other words, it is not easy to build area-preserving twist maps whose vertical rotation sets are reduced to a single irrational number. From Theorem A of \\cite{eujul}, such a map needs to have an invariant foliation by Lipschitz graphs over the horizontal coordinate. In particular, all its iterates must satisfy a twist condition. This is precisely what does not hold for $f_{k,t}$, for all non-zero values of $k$.","published":"2026-09-25T01:12:17Z","updated":"2026-09-25T01:12:17Z","authors":["Salvador Addas Zanata"],"categories":["math.DS"],"headline":"Study of rotation behavior in perturbed Affine twist maps","summary":"The researchers investigated how small changes affect the long-term twisting patterns of a mathematical function called the Affine twist map, which acts on a doughnut-shaped surface. They analyzed cases when a key parameter is rational and irrational. Their findings suggest that for rational values, slight perturbations preserve a simple rotation pattern, but for irrational values, any non-zero change leads to more complex behaviors with a range of rotations. The work highlights challenges in creating such systems with single rotation patterns for irrational parameters, supported by theoretical reasoning and existing mathematical results."},{"id":"2609.30664v1","title":"Rigidity of stationary and uniformly rotating planar Euler flows in the $C^1$ Yudovich class","abstract":"We prove a rigidity theorem for stationary and uniformly rotating solutions of the two-dimensional incompressible Euler equation with vorticity $ ω_0\\in C^1(\\mathbb{R}^2)\\cap L^1(\\mathbb{R}^2) \\cap L^\\infty(\\mathbb{R}^2). $ If the angular velocity $Ω$ satisfies \\[ Ω\\leq\\frac12\\inf_{\\mathbb{R}^2}ω_0 \\qquad\\text{or}\\qquad Ω\\geq\\frac12\\sup_{\\mathbb{R}^2}ω_0, \\] then $ω_0$ is radially symmetric, in particular, every stationary vorticity of one sign in the class $C^1(\\mathbb{R}^2)\\cap L^1(\\mathbb{R}^2) \\cap L^\\infty(\\mathbb{R}^2)$ must be radially symmetric about some point. Furthermore, for $Ω\\neq0$, the center is necessarily the origin; The proof is based on the analysis for level-set geometry of a normalized stream function. A componentwise Bernoulli formula, together with the isoperimetric inequality and Pohozaev identities, yields a nonnegative defect measure encoding both geometric defects and topological branching. An analysis at infinity forces this measure to vanish, reducing the problem to a semilinear elliptic equation with a bounded nonnegative Borel nonlinearity. Radial symmetry then follows from a theorem of P.-L.\\ Lions.","published":"2026-09-25T01:11:42Z","updated":"2026-09-25T01:11:42Z","authors":["Boquan Fan","Yuchen Wang","Chunjing Xie","Weicheng Zhan"],"categories":["math.AP"],"headline":"Researchers find symmetry in certain steady 2D fluid flows","summary":"The researchers studied steady or uniformly spinning solutions of a two-dimensional fluid flow described by the incompressible Euler equation. They showed that under specific conditions relating to the fluid's rotation rate and vorticity, the fluid's rotation pattern must be perfectly symmetrical around a point. Using advanced mathematical tools, they proved this symmetry is necessary for these flows within a particular smoothness and boundedness class. Their results rely on theoretical analysis and existing mathematical theorems about elliptic equations."},{"id":"2609.30663v1","title":"Symmetric Kahane--Salem--Zygmund Inequalities and the Supremum Norm","abstract":"The Kahane--Salem--Zygmund inequality provides unimodular $m$-linear forms with small supremum norm. We consider its dimension-free symmetric constant $C_m^{\\mathrm{sym}}$ over the real scalar field, defined by the requirement that, for every $n$, some real symmetric unimodular $m$-linear form $A:(\\ell_\\infty^n)^m\\to\\mathbb R$ satisfy \\[ \\|A\\|\\le C_m^{\\mathrm{sym}}n^{(m+1)/2}. \\] For complex scalars, Boas obtained under permutation symmetry an upper bound of order at most $\\sqrt{m\\log m}\\,\\sqrt{m!}$; over the real scalar field, an elementary argument gives the sharper order $\\sqrt m\\,\\sqrt{m!}$. We prove \\[ C_m^{\\mathrm{sym}}\\ge \\left(\\sqrt{\\frac2e}+o(1)\\right)\\frac{\\sqrt{m!}}m, \\] using the square-free Walsh spectrum of the diagonal polynomial. In the opposite direction, we establish \\[ C_m^{\\mathrm{sym}}\\le C_0\\sqrt{m!}, \\] where $C_0$ is absolute, removing the factor $\\sqrt m$ from the real upper bound. This estimate follows from a geometric argument in which rigidity for Gram permanents reduces the relevant configurations to sets controlled by Gaussian width. For unrestricted unimodular forms, we also obtain a quantitative rectangular estimate from truncated Hadamard matrices; in equal dimension, the normalized minimum is at most $1+o(1)$ whenever $m=o(n^{5/6})$.","published":"2026-09-25T01:05:30Z","updated":"2026-09-25T01:05:30Z","authors":["Daniel M. Pellegrino","Anselmo B. Raposo","Eduardo V. Teixeira"],"categories":["math.FA"],"headline":"Researchers study size limits of special symmetric mathematical forms","summary":"The researchers investigated the size limits, or supremum norms, of special symmetric multilinear forms—mathematical objects that take multiple inputs and produce a number—over real numbers. They improved lower and upper bounds for a particular constant describing these limits, using advanced mathematical techniques involving polynomials and geometric arguments. Their findings refine previous estimates but are theoretical and focus on properties within mathematical analysis rather than practical applications."},{"id":"2609.30662v1","title":"LLM Parkinsonism: Executive-Control Failure, Token-Inefficient Persistence, and an Uncertainty-Aware Global Executive Control Architecture for Autonomous Language-Model Agents","abstract":"Large language models (LLMs) can plan, use tools, write code, and execute long-horizon workflows, yet strong local competence does not guarantee project-level executive control. Agents may continue acting after the original objective is satisfied, producing low-value refinements, repeated verification, and repairs to self-created complexity. We use LLM Parkinsonism as a narrowly defined, non-clinical metaphor for this pattern of persistent action despite diminishing task-level value. We argue that the problem is not explained by autoregressive next-token prediction alone, but more directly by concentrating proposal generation, scope interpretation, progress assessment, and stopping authority within the same self-conditioned loop. We therefore introduce Global Executive Control (GEC) v0.2, an uncertainty-aware governance architecture that separates action generation from project-level control. In a 24,000-episode matched-candidate benchmark under a common 40,000-token ceiling, a first-candidate baseline achieved 67.42% hard-goal success, a candidate-set local control achieved 96.53%, and GEC achieved 96.57%. The candidate-set control shows that access to multiple candidate actions explains most of the success gain; relative to that control, GEC preserved success while reducing mean token use from 19,782 to 12,574 (36.4%) and restricted mean tokens to completion at the 40,000-token ceiling from 16,136 to 13,114 (18.7%), while eliminating measured pre-completion drift and sharply reducing gross complexity. Governance-overhead sensitivity remained favorable through an additional 500 synthetic governance tokens per cycle. These mechanistic simulations support explicit governance of scope, evidence, resource use, and stopping, while live-model validation remains necessary.","published":"2026-09-25T01:00:15Z","updated":"2026-09-25T01:00:15Z","authors":["Dongsheng Xiao","Zeyuan Wang","Xuzhe Xia","Bo Zhao","Yankai Cao"],"categories":["cs.AI"],"headline":"Researchers propose better control for language model task persistence","summary":"The researchers studied a problem where advanced language models keep working on tasks even after the main goal is met, leading to unnecessary actions and complexity. They introduced a new control system, called Global Executive Control, that separates action generation from overall project management to improve efficiency. Their computer-based tests showed this system maintained success rates while using fewer steps and reducing wasted effort, but real-world testing with actual models is still needed to confirm these findings."},{"id":"2609.30661v1","title":"Population loss in shallow ReLU networks: Bias & families of critical points","abstract":"The main result presented is a formula for the population loss in the student-teacher kernel model that is applicable to shallow ReLU networks with bias. This extends previous work of Choo and Saul (2009) and Brutzkus and Globerson (2017). The formula makes essential use of Owen's T-function. The necessary theory of the T-function is given and a high precision coding using MPFR for the T-function, based on an algorithm of Komelj (2023), is available on request. It is shown that various families of spurious minima described in past papers of Arjevani and the author extend to biased networks and that the loss is always strictly decreased when bias is added. The change in landscape geometry caused by adding bias appears to be relatively mild. Only the simplest examples are described in this paper where it is assumed that the number of inputs is equal to the number of neurons (this restriction is for reasons of length). A review of relevant previous results on unbiased networks is included. Aside from Gaussian statistics, the main mathematical tools and ideas come from analytic geometry (analytic and subanalytic sets, the Curve Selection Lemma).","published":"2026-09-25T00:58:48Z","updated":"2026-09-25T00:58:48Z","authors":["Michael Field"],"categories":["cs.LG"],"headline":"New formula explains errors in shallow neural networks with bias","summary":"Researchers developed a formula to measure prediction errors (population loss) in simple neural networks that use a specific activation function (ReLU) and include bias terms. They extended previous work and used advanced math tools, including Owen's T-function, to understand how adding bias affects the network's performance. Their findings show that adding bias consistently reduces error, though it only slightly changes the mathematical landscape of solutions. The work focused on simple cases and builds on earlier studies without bias terms."},{"id":"2609.30660v1","title":"Geometry as Thermodynamics:Entropy Stationarity and Horizon Residues","abstract":"Thermodynamic descriptions of gravity involve both variational conditions for spacetime dynamics and analytic structures associated with horizons. We examine their relation while keeping their assumptions distinct. First, we present an explicit derivation of the Einstein equation from the established null-vector entropy functional, including the null constraint, the matter term, and the integration constant associated with the cosmological constant. Second, for an analytic static spherical geometry with a nondegenerate Killing horizon, we define a meromorphic radial one-form whose residue is the inverse of twice the surface gravity. Euclidean regularity then fixes the Hawking temperature. Combining this residue with the Einstein--Hilbert Noether charge gives a normalized contour representation of the Wald entropy. The construction reproduces the Schwarzschild temperature, entropy, and Smarr relation, but does not constitute an independent microscopic derivation of the area law. We establish the limits of a stronger identification between pole structure and dynamics: an asymptotically flat family can retain the Schwarzschild horizon residue, area, surface gravity, and mass while violating the vacuum Einstein equation, and an extremal charged solution possesses a higher-order pole. We also show why exponentiating an unspecified entropy functional does not produce a universal entropy residue. The resulting framework separates entropy stationarity, horizon analyticity, and charge normalization, providing explicit consistency tests for further thermodynamic interpretations of gravitational singularities.","published":"2026-09-25T00:58:31Z","updated":"2026-09-25T00:58:31Z","authors":["Wen-Xiang Chen"],"categories":["gr-qc"],"headline":"Researchers Explore Thermodynamics Concepts in Gravity and Black Holes","summary":"The researchers studied how ideas from thermodynamics relate to gravity and the geometry around black hole horizons. They derived Einstein's equation using an entropy-based approach and analyzed mathematical structures near black hole horizons to connect temperature and entropy. Their work reproduces known black hole properties but clarifies that it does not provide a new fundamental explanation for black hole entropy. They also identified limitations in linking these mathematical features directly to gravity's dynamics, offering tests for future theories."},{"id":"2609.30659v1","title":"Image Reconstruction from Phase with Untrained Neural Priors","abstract":"Fourier phase encodes important spatial image structure, but recovering an image without measured spectral magnitude requires additional constraints and leaves absolute intensity ambiguous. We propose a projection-based two-stage framework that combines Fourier-phase and spatial-support constraints with an image-specific neural prior. The first stage alternates constraint enforcement with regularized neural-prior updates, while the second performs phase/support refinement alone with guaranteed convergence. We evaluate two neural-prior implementations on the same 77 microscopy images and compare them with a constraint-only baseline. After 500 final refinement passes, the best-performing variant achieves 31.41 dB pooled PSNR, 35.75 dB mean PSNR, and 0.9531 mean SSIM, improving pooled PSNR by 1.51~dB and reducing pooled MSE by 29.3% relative to the baseline. The results demonstrate the benefit of combining neural guidance with explicit constraint refinement at the evaluated iteration budget, while showing that lower phase residual alone does not guarantee greater reconstruction accuracy.","published":"2026-09-25T00:55:19Z","updated":"2026-09-25T00:55:19Z","authors":["Ene Meco","Ahmet Enis Cetin"],"categories":["eess.IV","cs.CV"],"headline":"New method improves image recovery using phase data and neural networks","summary":"The researchers tackled the challenge of reconstructing images when only phase information is available, which normally makes exact intensity unclear. They developed a two-step approach combining known image constraints with neural networks trained specifically for each image. Testing on microscopy images showed their method produced clearer reconstructions than relying on constraints alone. Their findings suggest that blending neural guidance with traditional constraints improves results, though smaller phase errors don't always mean better images."},{"id":"2609.30658v1","title":"DiffusionShadow: Diffusion-based Shadow Caching for Neural Volume Rendering","abstract":"Implicit neural representations (INRs) have gained momentum in scientific visualization due to their compactness and scalability to large datasets, making them well suited for integration with direct volume rendering (DVR). However, real-time volume rendering of INR with advanced illumination effects, such as shadows, remains computationally expensive, as evaluating shadow terms via ray marching is costly. Alternatively, precomputing and storing shadows for many lighting directions is prohibitive in both memory and storage. To address this, we introduce a diffusion-based shadow caching framework that compresses a vast set of pre-calculated shadow INRs into a single diffusion model. Rather than focusing on generalizing to unseen directions, our method effectively memorizes and reconstructs a dense set of pre-trained lighting conditions on the fly. We first encode a collection of shadow coefficient volumes as shadow INRs, and then train a diffusion model conditioned on lighting direction to predict the corresponding shadow INR weights at inference time. This design integrates directly with standard INR renderers without additional runtime sampling. Experiments show that our approach achieves faster rendering than traditional methods while bypassing the massive storage bloat of independent INRs, producing shadows that closely match most of the reference results.","published":"2026-09-25T00:55:05Z","updated":"2026-09-25T00:55:05Z","authors":["Kai-Chen Tung","Qi Wu","David Bauer","Mengjiao Han","Silvio Rizzi","Kwan-Liu Ma"],"categories":["cs.GR","cs.LG"],"headline":"New Method Speeds Up Real-Time Shadows in 3D Volume Rendering","summary":"The researchers tackled the problem of slow, costly rendering of shadows in 3D scientific visualizations using implicit neural representations (INRs). They developed a diffusion-based system that stores many pre-calculated shadow effects in a compact model, which can quickly generate shadows for different lighting directions without huge storage needs. Their experiments show this method renders shadows faster than traditional approaches and produces results similar to detailed reference shadows, though the findings are based on simulations and specific testing."},{"id":"2609.30657v1","title":"Prompt Injection Detection for Email Agents Through Attack Chain Modeling","abstract":"Large language model email assistants are particularly vulnerable to indirect prompt injection because untrusted email content can be retrieved into the model context and influence subsequent tool use. Existing prompt injection detectors mainly formulate this problem as binary malicious text classification, which overlooks the important factor that harmful agent behavior often arises through a sequence of stages. We propose a detection framework that models this attack chain by combining a text detector, verifiers specific to each stage, explicit rule-based risk signals, user intent and action consistency analysis, and a logistic decision policy. To support this framework, we derive attack chain labels from prompt injection datasets, evaluate the proposed framework under random splits, temporal phase transfer, conditional stage transfer, cross-dataset transfer, and conduct ablation studies on multiple benchmarks. Results show that random train test splits substantially overestimate robustness under distribution shift, while later tool argument stages are more predictable than earlier stages in the framework. We also show that training on harmless emails that resemble attacks helps reduce false alarms while preserving the ability to detect real attacks. Across five binary benchmarks, our framework achieves a mean F1 score of 0.406 under the strict threshold setting policy, compared with 0.216 for the strongest of five pretrained detectors evaluated without additional training. These results highlight the value of combining attack stage predictions with checks for conflicts between the user's request and instructions in retrieved emails. Our experiments also demonstrate the importance of training with challenging benign examples to balance attack detection and false alarms.","published":"2026-09-25T00:51:56Z","updated":"2026-09-25T00:51:56Z","authors":["Ahmad Hashmi","Dhyey Patel","Yunting Yin"],"categories":["cs.CR","cs.CL"],"headline":"New Method Detects Complex Email Attacks on AI Assistants","summary":"Researchers studied how AI email assistants can be tricked by harmful instructions hidden in emails. They developed a detection system that looks at the stages of such attacks, combining text analysis, rule checks, and user intent. Testing on various datasets showed their method better identifies attacks and reduces false alarms compared to existing detectors, but results are based on simulated data and highlight challenges in handling shifting email content over time."},{"id":"2609.30656v1","title":"Component Benchmark: Hierarchical Model Profiling for Large-scale Recommendation Systems","abstract":"Large-scale recommendation models pose distinct, under-explored profiling challenges. Most recommendation model architectures are structurally heterogeneous, intermixing memory-bandwidth-bound operations, small compute-bound dense layers, dynamic shapes from jagged categorical features, and low-arithmetic-intensity operations. Recommendation models evolve rapidly as modeling engineers experiment with compositions, often written without visibility into hardware execution characteristics. Standard profiling tools offer either end-to-end throughput or operator-level traces, but cannot attribute performance to the submodules that practitioners reason about. We present Component Benchmark (CB), a profiling system that independently characterizes each submodule performance in a hierarchical manner, providing a tree-structured, interactive visualization that brings performance clarity to ML practitioners. At its core, CB provides a simple yet extensible, submodule-based benchmarking framework with a plugin architecture that enables hierarchical performance analysis. These large-scale recommendation models are TB-scale, run on thousands of GPUs and ingest 100B examples per day. We demonstrate CB's effectiveness on common open sourced models and discuss how CB has been leveraged to accelerate modern recommendation model performance analysis and optimization.","published":"2026-09-25T00:51:36Z","updated":"2026-09-25T00:51:36Z","authors":["Dharak Kharod","Yuzhen Huang","Zhou Wang","Jackie Xu","Fuzail Khan","Jacky Zhou","Hao Yan","Lidong Zhao","Xizhou Feng","Yvonne Liu","Karthik Jayaraman","Praveen Ramachandran","Vishwa Karia","Yashasvi Makin"],"categories":["cs.IR"],"headline":"New Tool Helps Understand Performance of Big Recommendation Models","summary":"Researchers developed a system called Component Benchmark (CB) to improve how we analyze the performance of large recommendation models used in machine learning. These models are complex and run on many GPUs, making it hard for existing tools to show detailed performance linked to specific parts of the model. CB breaks down performance into smaller submodules and visualizes it hierarchically, helping engineers better understand and optimize these large-scale systems. The findings come from testing CB on open models and focus on performance analysis, not on direct improvements."},{"id":"2609.30655v1","title":"Certificate-Carrying Distributed Model Predictive Control on Product Manifolds with $\\mathrm{SO}(3)$","abstract":"This paper studies constraint certification in synchronous distributed model predictive control (DMPC) when neighboring predictions change between sampling instants. Before the parallel local solves, each agent communicates a shifted prediction and an announced update budget. A hard trajectory trust region makes that budget enforceable, while an edge-wise feasibility cap computed from the shifted packets keeps the fallback feasible without using any current optimizer output. Distance and relative-attitude constraints are tightened with explicit Lipschitz constants and two budget layers: one accounts for the simultaneous neighbor update and the other retains a checkable shift reserve. We prove hard pairwise constraint satisfaction and recursive feasibility under stated nominal-execution and terminal assumptions, give the additional residual caused by execution error, and derive a local practical value-decrease bound. A spacecraft formation example uses hard terminal and pairwise constraints, a geodesic relative- attitude constraint on $\\SO$, and reproducible terminal-set checks. Comparisons with fixed, trajectory-only, and windowed online margins show that the proposed budget reduces conservatism while preserving a positive shifted-feasibility margin.","published":"2026-09-25T00:48:48Z","updated":"2026-09-25T00:48:48Z","authors":["Shengjun Zhang","Tingyi Liu","Lei Xu","Tao Yang"],"categories":["eess.SY","math.OC"],"headline":"New Method Ensures Safe Coordination in Distributed Control Systems","summary":"The researchers address the problem of ensuring safety constraints when multiple controllers predict and adjust their actions asynchronously. They propose a method where each controller shares planned updates with a guaranteed 'update budget' and uses mathematical bounds to keep all constraints satisfied without constant re-optimization. Their approach, tested in spacecraft formation control simulations, reduces overly cautious restrictions compared to previous methods while maintaining safe and feasible operation under ideal conditions. The findings are theoretical and demonstrated through simulation examples."},{"id":"2609.30654v1","title":"A Cyanopolyyne-rich but COM-poor Massive Protostar: The First Hot Carbon Chain Chemistry Source G28.28-0.36","abstract":"We present molecular emission line data from the massive young stellar object (MYSO) G28.28-0.36 (G28.28) obtained with the Atacama Large Millimeter/submillimeter Array Band 3. Cyanopolyynes (HC$_3$N and HC$_5$N) and three complex organic molecules (COMs; CH$_3$OH, CH$_3$CN, and CH$_3$CHO) are detected from the MYSO G28.28. In addition, strong emission regions of cyanopolyynes are identified between G28.28 and a nearby ultracompact H II region. The HC$_5$N emission is coincident with the dust continuum peak, where an excitation temperature of 100 K is derived from CH$_3$CN. These results suggest that the Hot Carbon Chain Chemistry (HCCC) mechanism produces cyanopolyynes in the hot region around G28.28. We find that G28.28 exhibits a unique chemical feature: cyanopolyynes are abundant, but COMs are deficient, unlike the other MYSOs studied previously. These results imply that G28.28 is a counterpart of the Warm Carbon Chain Chemistry (WCCC) low-mass source L1527. G28.28 is the first HCCC source identified so far.","published":"2026-09-25T00:46:07Z","updated":"2026-09-25T00:46:07Z","authors":["Kotomi Taniguchi","Masao Saito","Prasanta Gorai","Olli Sipilä","Kazuhito Dobashi","Paola Caselli","Tirupati Kumara Sridharan","Tomomi Shimoikura","Jonathan C. Tan"],"categories":["astro-ph.GA","astro-ph.SR"],"headline":"Scientists find unusual molecule-rich young star G28.28-0.36","summary":"Researchers studied a massive young star called G28.28-0.36 using a powerful telescope and detected unusual chemicals called cyanopolyynes, while common complex organic molecules were rare. They suggest a special chemical process named Hot Carbon Chain Chemistry creates these molecules near the star. This star uniquely has many cyanopolyynes but few complex organics, differing from others studied before, making it the first known example of this chemical environment in a massive young star."},{"id":"2609.30653v1","title":"Fully 3GPP-Compatible Long-Range Sensing for LEO-ISAC: A Window-Grid Processing Framework","abstract":"This paper addresses the long-range sensing problem in bistatic low-Earth-orbit integrated sensing and communication (LEO-ISAC) systems. Conventional OFDM-based sensing schemes fail in LEO-ISAC due to excessive propagation delays, which cause cross-symbol misalignment and unequal signal durations. We propose a window-grid processing framework that leverages the deterministic target geometry to resolve both challenges as a pure receiver-side processing: the transmitted sensing signal is 3GPP-compatible. Simulations demonstrate meter-level ranging accuracy for two targets at bistatic ranges beyond 640 km with 100.8 MHz bandwidth.","published":"2026-09-25T00:44:37Z","updated":"2026-09-25T00:44:37Z","authors":["Yi Geng","Xun Fan","Chenfeng Xu","Yu Zhao"],"categories":["cs.IT"],"headline":"New Method Enhances Long-Range Sensing for Satellite Systems","summary":"The researchers studied how to improve sensing over long distances in low-Earth orbit satellite communication systems, which conventional methods struggle with due to signal timing issues. They developed a new processing approach that works on the receiver side without changing the transmitted signal, keeping it compatible with current standards. Their simulations showed this method can accurately measure distances to targets over 640 km away, though these results are based on computer models, not real-world tests."},{"id":"2609.30652v1","title":"Recursive Self-Improvement via On-Policy Distillation for Reasoning","abstract":"On-policy distillation (OPD) trains a student model by having it generate trajectories, then matching its next-token predictions with an external teacher's next-token predictions. This provides dense, token-level supervision to the student. On-policy self-distillation (OPSD) eliminates the need for the external teacher. Specifically, a second frozen copy of the student model, now given the ground truth in its context, serves as the teacher. The student model only receives the problem and learns to mimic the privileged teacher model, while the teacher remains frozen throughout training. Previous work showed that freezing the teacher is useful for training stability, but we argue that this can prevent the teacher from incorporating the improvements learned by the student during training. Our primary contribution is to address this limitation with a recursive framework built around two complementary components. First, we let the privileged teacher co-evolve with the student so that revision learned in one round can guide the next, a process we refer to as Dynamic Co-Evolution (DCE). Second, because stronger revision can also make responses too verbose and self-critical, we additionally train on shorter, verified rewrites of the model's own on-policy responses. We call this complementary objective Self-Refined Concise Learning (SRCL). Overall, our comprehensive evaluations show that DCE+SRCL outperforms OPSD across multiple model scales and four competition-level mathematics benchmarks. Specifically, on Qwen3-8B, DCE+SRCL reaches 65.97% Average@12, outperforming OPSD by 35.62 percentage points while reducing mean output length by 7.80% relative to DCE alone.","published":"2026-09-25T00:42:31Z","updated":"2026-09-25T00:42:31Z","authors":["Shangjian Yin","Zehao Zhao","Kavosh Asadi","Rui Liu","Yuchen Lu","Shike Mei","Hang Cui","Luke Simon","Zhouxing Shi","Hamed Firooz"],"categories":["cs.CL"],"headline":"Improving AI Reasoning by Letting Teacher and Student Learn Together","summary":"Researchers studied how to train AI models to solve problems better by having a student model learn from a teacher model. Traditionally, the teacher is fixed, but this can limit progress. The researchers introduced a method where the teacher model evolves alongside the student, allowing improvements to guide future learning. They also trained the models to give shorter, clearer answers. Their approach, tested on math problem benchmarks, showed better performance and more concise responses compared to previous methods. Results are based on simulations with AI models and focus on training techniques."},{"id":"2609.30651v1","title":"Verification of Compiler-to-Accelerator Mappings for Machine Learning Accelerators","abstract":"To meet the performance needs of modern machine learning (ML) applications, ML compiler frameworks support compiler-to-accelerator mappings that offload parts of application code to operations in specialized hardware accelerators. However, most of these frameworks do not verify these mappings down to the hardware level, potentially resulting in functional mismatches. In this paper we propose BOLT, the first framework for formally verifying the correctness of compiler-to-accelerator mappings for coarse-grained intrinsics in ML accelerators, with respect to a formal hardware semantics. BOLT does not require additional information from the compiler, and verifies the functional equivalence of the application code and the code for the mapped hardware accelerator intrinsic, including handling of complex loop nests and tensor data layouts in hardware. It effectively utilizes a pattern of *aligning* software loops with the hardware, followed by *relating* corresponding data layouts, to enable verification using well-aligned product programs. To support these steps, we propose two custom templates --- the sync-skeleton and the layout-sketch --- to guide users in aligning loops and specifying data layout relationships, respectively. We have developed a proof-of-concept prototype for BOLT and use it to successfully verify the correctness of several complex mappings for two recent open-source ML accelerators.","published":"2026-09-25T00:41:32Z","updated":"2026-09-25T00:41:32Z","authors":["Akash Gaonkar","Mike He","Yi Li","Bo-Yuan Huang","Andrew Cheung","Vishal Canumalla","Gus Henry Smith","Zachary Tatlock","Grigory Fedyukovich","Sharad Malik","Aarti Gupta"],"categories":["cs.PL"],"headline":"Checking Machine Learning Compiler Mappings to Hardware Accelerators","summary":"Researchers developed BOLT, a new method to verify that parts of machine learning programs correctly map onto specialized hardware accelerators. They focus on ensuring the software's code and the hardware instructions behave the same, even with complex loops and data layouts. Their prototype successfully checked several mappings, helping to catch potential mismatches without requiring extra compiler information. However, their work is based on formal verification methods and tested on recent open-source hardware designs."},{"id":"2609.30650v1","title":"Causal Retention in Interactive Agents: Interface Factorization and Selective Adaptation","abstract":"Task performance need not determine which intervention mechanism an agent retains. We study causal retention: whether a frozen learned state answers a mechanism-probe map fixed independently of training, including action, context, direct target, value, and delay. For finite structural causal model classes, the optimal probe error is a Bayes decision risk. It vanishes exactly when every learning-interface fiber lies within one probe-answer fiber; any state obtained by post-processing that interface inherits the same lower bound. A posterior-coverage theorem characterizes budgeted retesting, while an exact edit decomposition shows that the shifted set is the unique support of an error-free target update. Causal Core implements these conditions through evidence-gated writing, readout filtering, temporal credit, hidden-context setup, and local diagnostic updates. Experiments cover finite causal systems, continuous simulators, an official TD-MPC2 world model, and Qwen2.5-7B-Instruct. A frozen Qwen last-layer probe reaches 0.958 balanced accuracy on source mechanisms but 0.583 on changed delays; the gated mechanism state reaches 1.000 and accepts only 0.056 of synchronized-readout candidates. In TD-MPC2, five target states per actuator recover effect-sign accuracy from 0.057 to 0.948 without degrading stable responses. Causal retention is therefore distinct from task sufficiency and source-domain decodability.","published":"2026-09-25T00:40:21Z","updated":"2026-09-25T00:40:21Z","authors":["Shengjun Zhang","Tingyi Liu","Dong Xie","Yunlong Dong","Xiang Wang","Cheng Zeng"],"categories":["cs.LG","cs.AI","stat.ML"],"headline":"Researchers study how agents retain causes beyond task success","summary":"The researchers explore 'causal retention,' which looks at whether a learned agent's internal state reliably reflects cause-and-effect relations, independent of how well it performs a task. They develop theoretical tools and an implementation called Causal Core to measure and improve this kind of retention, testing it in various computer simulations and models. Their findings suggest that an agent's understanding of causes can differ from just performing well, showing that cause retention is a distinct and measurable property."},{"id":"2609.30649v1","title":"A VLA Study of the Disturbed Massive Cluster PLCK G165.7 + 67.0 and Its Two Narrow Angle Tail Galaxies","abstract":"Since spectroscopic measurements provide only the line-of-sight velocity, transverse motions are difficult to constrain. As a result, transverse velocities have so far been measured only for local galaxies. Radio galaxies whose jets are bent back by ram pressure into nearly parallel Narrow-Angle Tails (NATs) offer a way to extend such measurements to higher redshifts. Here, we present proprietary 15 GHz (Ku-band) and archival 6 GHz (C-band) continuum observations from the Very Large Array (VLA) of seven radio galaxies in the galaxy cluster field PLCK G165.7+67.0 (G165). VLA imaging resolves two of the cluster galaxies into NATs whose tails both extend toward the Southwest, away from the cluster center of mass. Spectral index maps made at matched angular resolution resolve a steepening gradient consistent with radiative aging along the length of the tails. Physical properties of the NATs are measured and used to constrain two independent models: a Mach cone model and a nonrelativistic hydrodynamic flow model based on Euler's equation. The models returned space velocities of $\\sim 2000 ~\\rm km~ \\rm s^{-1}$ for both galaxies. One NAT is the brightest cluster galaxy (BCG) and has a measured radial velocity of $-3300 ~\\rm km ~\\rm s^{-1}$, resulting in an even higher space velocity. The high inferred 3D velocity of the BCG may be explained if it is intercepted close to core passage.","published":"2026-09-25T00:39:46Z","updated":"2026-09-25T00:39:46Z","authors":["Aurora Wilde","Eric F. Jimenez-Andrade","Brenda L. Frye","Patrick S. Kamieneski","Kevin C. Harrington","Megan Donahue","Min S. Yun","Anton M. Koekemoer","Marcel Neeleman","Massimo Pascale","Nicholas Foo","Reagen Leimbach","Seth Cohen","Rogier A. Windhorst"],"categories":["astro-ph.GA"],"headline":"Studying fast-moving galaxies in a distant galaxy cluster","summary":"Researchers used radio telescopes to study two special galaxies with bent jets, called Narrow-Angle Tails (NATs), in a distant galaxy cluster named PLCK G165.7+67.0. By analyzing the shape and radio waves of these jets, they estimated the galaxies' speeds in three dimensions, finding very high velocities around 2000 km/s. These results come from applying physical models to their radio observations, helping measure motions otherwise hard to detect, but are based on interpretation of telescope data rather than direct measurement."},{"id":"2609.30648v1","title":"Random walk in a non-homogeneous random environment on some random trees and the non-negative integers","abstract":"In this paper, we consider a random walk in a non-homogeneous random environment on two types of random trees and non-negative integers. The random trees here have progeny distributions that depend on the parent's generation. We provide a sharp threshold for the type of the random walk in a random environment on the random trees and the non-negative integers, up to the critical case. Then, we find the speed of the random walk on the non-negative integers and on random spherically symmetric trees. We use the method based on the relationship between the mean exit time from a vertex and speed. Meanwhile, we discuss several special cases that could be derived from those results. We also include many limiting formulas for sequences and series in the appendix, which are helpful for this study and others.","published":"2026-09-25T00:39:06Z","updated":"2026-09-25T00:39:06Z","authors":["Tamer Oraby","András Telcs"],"categories":["math.PR"],"headline":"How Random Walks Behave on Changing Random Trees and Numbers","summary":"Researchers studied how a random walker moves through environments that change unpredictably, focusing on certain random tree structures and the counting numbers. They identified clear boundaries that determine different movement types and calculated the walker's average speed in these environments. Their work is based on mathematical relationships between the time it takes to leave a point and the overall speed, with special cases and helpful formulas included, but it remains theoretical and focused on specific random models."},{"id":"2609.30647v1","title":"Conditional Predictive Sufficient Statistics for Visual Representation Learning","abstract":"A useful visual representation is a statistic of the observed past that retains the latent factors shared with the future and discards patch-private noise. We formalize this requirement as a conditional predictive sufficient statistic (CPSS). Under a shared-factor model of image patches, the mutual information between the past and the next patch equals the information the past carries about the shared factor, up to a remainder that the next patch itself fails to reveal. Predicting the next patch embedding with a cosine loss is maximum likelihood for a von Mises-Fisher model of that embedding's direction, and is therefore a tractable surrogate for the predictive information. The same population loss is also minimized by a constant embedding, so stop-gradient does not by itself select the sufficient statistic; it only blocks the symmetric gradient that implements the constant solution in one step. The regression target is a shallow embedding, which forces the network output back into that shallow range and leaves the sufficient statistic in intermediate blocks. Small causal Transformers on MNIST and CIFAR-10 are used as diagnostics, not as a leaderboard. On MNIST the future shift and the stop-gradient move probe accuracy by tens of points, and the CPSS readout peaks before the output. On CIFAR-10, with the same short budget and no augmentation, every objective lands near a linear classifier on pixels. What still matches the derivation is the geometry: the CPSS output is a worse readout than its best intermediate block, next-pixel regression does not pay that penalty, and removing the stop-gradient collapses the effective rank of the embedding even when the pretext loss looks perfect.","published":"2026-09-25T00:36:49Z","updated":"2026-09-25T00:36:49Z","authors":["Yuzhou Hong"],"categories":["cs.CV"],"headline":"New method to improve visual representations by predicting future image parts","summary":"Researchers studied how to create better visual representations that capture meaningful features shared between past and future image parts while ignoring noise. They introduced a concept called conditional predictive sufficient statistic (CPSS) and tested it using simple neural networks on image datasets MNIST and CIFAR-10. Their experiments showed that CPSS can identify useful internal features before the final output, but fully matching theory remains challenging, especially on complex images. The findings rely on simulations and reveal subtle effects of training methods like stop-gradient."},{"id":"2609.30646v1","title":"Amortized Bayesian Disease Mapping and Boundary Detection on Heterogeneous Spatial Graphs","abstract":"Spatial disease maps help public-health researchers identify geographic inequalities, but standard Bayesian smoothing can obscure localized disparities when neighboring communities have sharply different socioeconomic or behavioral profiles. Analysts therefore need to determine where smoothing should be interrupted and repeat that analysis as maps, adjacency structures, and outcomes change. We develop a covariate-informed Bayesian boundary model and an amortized posterior approximation trained across heterogeneous areal graphs. The model distinguishes local interruptions in smoothing from broader residual spatial dependence; the trained approximation handles maps with different numbers of regions. Simulations examine posterior calibration, boundary-probability recovery, replicated-data behavior, and MCSE-controlled agreement with prior-matched MCMC. In contrast to traditional approaches that analyze these data separately, we demonstrate the effectiveness of using a single trained deep learning network to analyze respiratory hospitalizations in Greater Glasgow; lung cancer incidence in California; and tracheal, bronchial, and lung cancer mortality in South Korea, comprising 58 to 241 regions. Selected boundary density is greatest in Glasgow and lowest in South Korea despite substantial residual spatial dependence in both, showing that local interruption and broader spatial persistence need not vary together. Across all three applications, edge-level boundary probabilities agree substantially with dataset-specific analyses, although posterior spread and thresholded boundary sets differ. These results support reusable Bayesian boundary analysis across the evaluated disease-map class and identify the validation needed before deployment to new applications.","published":"2026-09-25T00:35:28Z","updated":"2026-09-25T00:35:28Z","authors":["Luca Aiello","Sudipto Banerjee"],"categories":["stat.ME","stat.AP"],"headline":"New Method Identifies Boundaries in Disease Maps More Accurately","summary":"Researchers developed a new statistical approach to better spot sharp changes in disease rates between neighboring areas, which traditional methods can miss. They trained a deep learning model that works across different maps and regions, tested it with simulations, and applied it to health data from Glasgow, California, and South Korea. Their findings show the method can distinguish local boundaries from broader patterns, suggesting a reusable tool for public health, though careful validation is needed before using it in new cases."},{"id":"2609.30645v1","title":"Special Cohen--Macaulay sheaves on partial resolutions of rational surfaces singularities","abstract":"We introduce the categories $\\CM(X)$ and $\\SCM(X)$ of reflexive sheaves on a minimal partial resolution $f\\colon X \\to \\Spec R$ of a rational surface singularity $\\Spec R$. The main result of this paper establishes that $\\SCM(X)$ possesses a natural Frobenius structure, serving as a geometric counterpart to the algebraic Frobenius structure on special Cohen--Macaulay $R$-modules, introduced by Iyama--Wemyss and Iyama--Kalck--Wemyss--Yang. Utilizing this geometric framework, we establish an exact equivalence between $\\SCM(X)$ and the category of special Cohen--Macaulay $R$-modules equipped with a specific exact structure, which induces a triangle equivalence between their stable categories. Consequently, this provides a direct, geometric proof of the Iyama--Kalck--Wemyss--Yang equivalence and yields a Buchweitz-type equivalence $\\underline{\\SCM}(X) \\simeq D_{\\sg}(X)$.","published":"2026-09-25T00:34:39Z","updated":"2026-09-25T00:34:39Z","authors":["Hokuto Uehara"],"categories":["math.AG","math.AC","math.RT"],"headline":"Researchers link geometric and algebraic structures in surface singularities","summary":"The researchers study special types of mathematical objects called reflexive sheaves on surfaces with certain singular points. They show that these objects have a natural structure that mirrors known algebraic structures on related modules. By building this geometric framework, they prove a precise equivalence between these geometric objects and algebraic modules, confirming and providing a new geometric proof for previous algebraic results. Their work remains theoretical, focusing on abstract mathematical categories and equivalences."},{"id":"2609.30644v1","title":"MR. POP: Multi-Robot Parallel Optimizing Planner for Almost-Surely Asymptotically Optimal Planning","abstract":"Finding globally optimal paths remains a fundamental challenge in multi-robot motion planning. Despite acceleration of almost-surely asymptotically optimal (a.s.a.o.) planners via CPU-based parallelism, achieving both probabilistic convergence guarantees and strong computational performance, these algorithms still struggle to scale to multi-robot settings. As such, we introduce MR. POP, a GPU-based a.s.a.o. multi-robot planner based on dRRT and the AO-x meta-algorithm. MR. POP uses large-scale GPU-based SIMT-parallelism to simultaneously run hundreds of roadmap construction and tree search iterations with underlying parallel nearest neighbor search and collision checking operations. We show that this enables MR. POP to become the only planner achieving a 100% solve rate while being faster than state-of-the-art a.s.a.o. planners in multi-robot systems up to 35-DOF. MR. POP also raises the success rate of downstream motion optimizers (e.g., from 4% to 72%), by creating high-quality, diverse seeds that help avoid local minima.","published":"2026-09-25T00:33:33Z","updated":"2026-09-25T00:33:33Z","authors":["Chih H. Huang","Roy Xing","Brian Plancher","Zachary Kingston"],"categories":["cs.RO"],"headline":"New Planner Speeds Up Multi-Robot Path Finding on GPUs","summary":"The researchers address the challenge of finding the best paths for many robots moving together, which is usually slow and hard to scale. They developed MR. POP, a new planning method that uses graphics card technology to run many calculations in parallel. Their study shows MR. POP solves all tested problems faster than existing methods for systems with up to 35 movement dimensions and improves success rates for later optimization steps by providing better starting solutions."},{"id":"2609.30643v1","title":"MARCEDES: Score-based causal discovery under non-Gaussianity with continuous optimization","abstract":"We consider the problem of learning the underlying causal directed acyclic graph (DAG) structure corresponding to a structural equation model (SEM) with non-Gaussian errors. Motivated by an intentionally misspecified non-Gaussian SEM with all Laplace errors, we first introduce the mean absolute residual risk, defined over the space of all real matrices, and show that, asymptotically, the risk of the true weighted causal DAG matrix is strictly smaller than that of any other matrix. Nevertheless, to enhance generality and account for high-dimensional and finite-sample settings, we further incorporate row-specific sparsity penalties along with a soft DAG constraint to derive a continuous score function over the space of real matrices. Accordingly, we propose a score-based DAG learning method, named MARCEDES, formulated as an unconstrained score minimization problem, which can be efficiently solved using gradient-based optimization techniques, thereby circumventing the challenges associated with constrained optimization. Furthermore, we develop a computational algorithm to handle the non-smoothness of the score objective and to enable optimal tuning of row-specific sparsity penalties under a generalized Bayes framework. Finally, we demonstrate the efficiency and improved performance of the proposed method over existing approaches through an extensive simulation study.","published":"2026-09-25T00:30:04Z","updated":"2026-09-25T00:30:04Z","authors":["Anamitra Chaudhuri","Anirban Bhattacharya","Yang Ni"],"categories":["stat.ML","cs.LG","stat.CO","stat.ME"],"headline":"New Method to Discover Cause-Effect Networks Using Data Patterns","summary":"Researchers tackled the challenge of identifying cause-effect relationships in data with certain types of random noise. They developed a new approach called MARCEDES that transforms this problem into an easier mathematical task using continuous optimization. Their method uses a scoring system that favors simpler, more meaningful network structures and can be solved efficiently with computers. Simulations suggest MARCEDES performs better than existing methods, though results are based on modeled data rather than real-world experiments."},{"id":"2609.30642v1","title":"A Framework for Identifying, Categorizing, and Explaining Bias in AI-Generated Code","abstract":"As Large Language Models (LLMs) become integrated into software development workflows, concerns regarding unintentional biases in AI-generated code. Although evidence suggests these biases exist, limited research has systematically identified, categorized, and explained them. This study investigates bias in AI-generated code and evaluates whether LLMs can reliably identify and explain it through a taxonomy-driven framework. We extended an existing dataset of biased AI-generated Python code and manually annotated snippets with bias categories and human-authored justifications to establish a ground-truth dataset. Using this dataset, we evaluated proprietary and open-source LLMs as automated bias detection and justification systems through ICL. Finally, we analyzed similarity between LLM-generated explanations and human-authored justifications using structured justification and code identification metrics. Our findings demonstrate that LLMs can effectively support code bias identification and explanation. Gemini achieved 80.14% classification accuracy, with 84.0% precision and 95.7% recall, while the best open-source alternative, Qwen3-coder, achieved 82.45% accuracy, 68.64% precision, and 80.22% recall. Additionally, the models achieved justification similarity scores of 80.4% and 80.14%, respectively, relative to human-authored reasoning, and code identification similarity scores of 86.0% and 87.82%. These results suggest that LLMs can detect biased logic in generated Python code and produce explanations that substantially align with expert interpretations.","published":"2026-09-25T00:22:51Z","updated":"2026-09-25T00:22:51Z","authors":["Manaal Basha","Aimee M. Ribeiro","Gema Rodriguez-Perez"],"categories":["cs.SE","cs.AI"],"headline":"Researchers Study Bias in AI-Generated Coding and Its Detection","summary":"This study looks at unintended biases that can appear in computer code written by AI language models. The researchers created a detailed dataset of biased Python code examples labeled with categories and human explanations. They then tested several AI models to see if they could automatically identify and explain these biases. Their findings suggest that some AI models can recognize biased coding patterns and provide explanations that closely match expert human reasoning, although this work is based on specific datasets and models tested."},{"id":"2609.30641v1","title":"Deciphering the VHE gamma-rays spectra of Extreme High-Frequency peaked BL Lac Objects through a Photohadronic Model","abstract":"Extreme high-frequency peaked BL Lac objects (EHBLs) have their synchrotron peak above $\\sim 1$ keV and their inverse Compton peak extends up to several TeVs. Due to low luminosity of EHBLs, these objects are often undetected, and only a small number of them are thus known. As a result, they are poorly understood. Using a photohadronic model and including extragalactic background light correction we analyzed 22 very high-energy (VHE) gamma-ray spectra of 15 EHBL sources observed in different periods. We find that the photohadronic model used can adequately describe the observed VHE spectral energy distribution and furthermore, we classify them according to the value of the photon spectral index, $δ$. The redshift of the EHBL HESS J1943+213 is also tightly constrained using the same photohadronic model. Within the sample, a plurality of flaring epochs fall in the high-emission state which is followed by the low-emission state and the very-high-emission state respectively. These fractions of flaring epochs characterize only the present sample, not the EHBL population at large. Due to the anomalous behavior of the VHE gamma-ray spectra of the EHBL H 2356-309, we suggest it to be a new subclass of EHBL having hard and very soft intrinsic spectra during different flaring epochs.","published":"2026-09-25T00:20:52Z","updated":"2026-09-25T00:20:52Z","authors":["Sarira Sahu","D. I. Páez-Sánchez","A. U. Puga Oliveros","M. E. Iglesias Martínez","Jose Guerra Carmenate","R. de J. Pacheco-Aké","G. Sánchez-Colón","Subhash Rajpoot","P. Fernández de Córdoba","Gaetano Lambiase"],"categories":["astro-ph.HE"],"headline":"Studying Extreme Cosmic Gamma Rays with a New Light Model","summary":"Researchers studied rare cosmic objects called Extreme High-Frequency peaked BL Lac objects, which emit very high-energy gamma rays but are hard to detect and understand. Using a specialized photohadronic model, they analyzed 22 gamma-ray spectra from 15 sources, finding the model fits observed data well and allows classification based on photon behavior. They also identified a potential new subclass due to unusual spectra, though their findings apply only to their sample and are based on model analysis rather than direct observation."},{"id":"2609.30640v1","title":"Towards more plausible point-identifying assumptions in two-sample Mendelian randomization","abstract":"Two-sample Mendelian randomization (MR) is a widely applied methodology in epidemiology. In two-sample MR, summary data (typically, regression coefficients and standard errors) quantifying the association between multiple genetic variants and the exposure and the outcome are used in an instrumental variable framework aimed at estimating the causal effect of the exposure on the outcome. Most two-sample MR methods were developed under data-generating models where the association of for each candidate genetic instrument with the exposure, as well as the causal effect of the exposure on the outcome, are constant in the additive scale. These assumptions are useful because they imply that, had all genetic variants been valid IVs, they would all estimate the same causal parameter - namely, the constant causal effect. We refer to this condition as summary-level homogeneity. However, these are rather strong homogeneity conditions which may raise concerns about the plausibility of these methods in practice. In this paper, we show that summary-level homogeneity is implied by the following conditions: the causal effect is additive linear, but not necessarily constant across, all strata of the population; and uncorrelatedness between heterogeneity in the causal effect and in the association between each genetic variant and the exposure. Under these conditions, typical two-sample MR methods can be interpreted as estimators of the average causal effect. These results clarify that point-identifying assumptions required for two-sample MR methods are weaker than previously anticipated, which contributes to their plausibility and interpretation in at least some practical applications.","published":"2026-09-25T00:12:33Z","updated":"2026-09-25T00:12:33Z","authors":["Fernando Pires Hartwig","Neil Martin Davies","George Davey Smith"],"categories":["stat.ME"],"headline":"Researchers Clarify Assumptions Behind Genetic Studies on Cause and Effect","summary":"Scientists study a method called two-sample Mendelian randomization, which uses genetic information to estimate whether one factor causes another in health research. They explain that previous methods assumed strict conditions about how genes affect traits. Their work shows these conditions can be weaker than thought, meaning the methods might still give meaningful average effects even if real-world conditions vary. Their findings help make the approach more believable and easier to interpret, but the conclusions are based on theoretical analysis rather than new experimental data."},{"id":"2609.30639v1","title":"A nonoverlapping spectral additive Schwarz method for interior penalty discontinuous Galerkin discretizations of Anisotropic Elliptic Problems","abstract":"We design and analyze a nonoverlapping additive Schwarz preconditioner for interior penalty discontinuous Galerkin discretizations of anisotropic elliptic problems. The preconditioned method coupled with a Krylov subspace iteration is shown to be independent of the highly discontinuous (and anisotropic) jump coefficients as well as the subdomain size. To increase efficacy, various auxiliary spaces are considered to reduce the size of the coarse grid operator. We demonstrate how to modify the additive Schwarz preconditioner such that it is applicable to the nonsymmetric IPDG schemes. Several numerical experiments verify the theory and validate the robustness of the preconditioner.","published":"2026-09-25T00:06:15Z","updated":"2026-09-25T00:06:15Z","authors":["Maurice S. Fabien","Sijing Liu","Marcus Sarkis"],"categories":["math.NA"],"headline":"Improving math tools for solving complex anisotropic problems","summary":"The researchers developed a new mathematical method to efficiently solve challenging equations that describe anisotropic elliptic problems, where properties vary in different directions. They created a special preconditioner that works well even when coefficients suddenly change and tested it with different settings to improve performance. Their tests support that this method is stable and effective, though their results come from mathematical analysis and computer experiments rather than real-world applications."},{"id":"2609.30638v1","title":"Structure of irreducible homomorphisms to/from free and injective modules","abstract":"Let $R$ be a commutative noetherian local ring. We extend the work of the author and Takahashi by investigating the structure of irreducible monomorphisms originating from free modules in the category of finitely generated $R$-modules. In the case where $R$ is complete, we further study the structure of irreducible monomorphisms and epimorphisms to and from injective modules in the category of artinian $R$-modules.","published":"2026-09-25T00:01:58Z","updated":"2026-09-25T00:01:58Z","authors":["Saeed Nasseh"],"categories":["math.AC"],"headline":"Researchers analyze specific maps between algebraic modules over rings","summary":"The researchers studied special kinds of mathematical functions called irreducible homomorphisms involving free and injective modules, which are algebraic structures over a type of ring known as a commutative noetherian local ring. They extended previous work by exploring the structure of these functions from free modules and also examined them in the context of injective modules over complete rings. Their findings clarify theoretical properties within abstract algebra but are focused on mathematical theory without practical applications included."},{"id":"2609.30637v1","title":"Quasi-equivalence of relabeled positroid cluster structures","abstract":"We prove the Fraser--Sherman-Bennett conjecture on the quasi-equivalence of positroid cluster structures arising from admissibly relabeled plabic graphs. We construct a tilting module from the boundary necklace and obtain a Frobenius categorification of the relabeled cluster structure. By identifying localized characters of rank-one modules with Plücker coordinates, we match the stable images of the mutable face modules with the summands of a reachable cluster-tilting object in the stable category. The induced derived equivalence yields a unimodular transformation of relative index lattices intertwining the full extended exchange matrices. These comparisons show that the identity map on the coordinate ring is a quasi-cluster isomorphism between the two cluster structures.","published":"2026-09-24T23:55:11Z","updated":"2026-09-24T23:55:11Z","authors":["Yilin Wu"],"categories":["math.RT","math.RA"],"headline":"Researchers confirm a mathematical conjecture on cluster structures","summary":"The researchers addressed a complex problem involving positroid cluster structures, which are mathematical objects related to networks called plabic graphs. They proved a conjecture that these structures remain essentially the same under certain relabeling operations by constructing specific mathematical modules and showing equivalences between them. Their work provides a theoretical confirmation connecting different configurations, but it is focused on abstract mathematical properties and does not include practical applications."},{"id":"2609.30636v1","title":"The supersymmetric Scarf I Hamiltonian with reflections and its discrete para-Bannai--Ito version","abstract":"The para-Bannai--Ito polynomials are orthogonal on a Bannai--Ito bi-lattice that is \\emph{two-sided} --- it extends on both sides of a central value $c_0$ --- and depend, besides the two grid parameters, on an isospectral deformation parameter. Their Jacobi matrix $J$ is invariant under the reversal of the sites at one value of that parameter only, the persymmetric point, and we determine its continuum limit there. Because the grid is two-sided, the natural object is the shifted operator $Q=J-c_0$, and the limit is taken about the centre of the band: $Q$ contracts to a first--order Dunkl operator, a supercharge, and $H=Q^2$ to the generalized Pöschl--Teller (Scarf I) Hamiltonian, with little $-1$ Jacobi eigenfunctions --- supersymmetric quantum mechanics with reflections, the reflection entering the square root of the Hamiltonian. The second grid parameter adds a reflection term to the Hamiltonian itself, and the bi-lattice is, up to a shift, the spectrum of the limiting supercharge. The convergence of the eigenvectors is proved directly on the polynomials. On this grid the persymmetric model transfers states perfectly, under an arithmetic condition on the parameters, but never revives fractionally, revival being the work of the deformation; both properties pass to the limit. These results are for an even number of sites; for an odd number of sites the limit is instead the matrix supersymmetric pair of Pöschl--Teller Hamiltonians, with the reflection in the Hamiltonian as a Dunkl potential. The bispectral pair of the discrete model goes over to the Dunkl realization of the anticommutator algebra of the little $-1$ Jacobi polynomials.","published":"2026-09-24T23:43:50Z","updated":"2026-09-24T23:43:50Z","authors":["Stéphane Z. Beaulac","Nicolas Crampé","Quentin Labriet","Lucia Morey","Marlon Josue Rivera Valladares","Luc Vinet"],"categories":["math-ph"],"headline":"Researchers link special polynomials to quantum mechanics models with reflections","summary":"The researchers studied para-Bannai–Ito polynomials, which relate to a two-sided grid structure, and examined a key matrix associated with them. They showed that in a certain limit, this matrix corresponds to a quantum system known as the Scarf I Hamiltonian, including a reflection component in its square root, connecting to supersymmetric quantum mechanics. Their findings, based on mathematical analysis and polynomial behavior, describe how these discrete models approximate continuous quantum operators, with results differing slightly for even or odd system sizes."},{"id":"2609.30635v1","title":"Generating Functions for Casimir Invariants of Simple Lie Algebras","abstract":"We consider generalised Casimir operators associated with irreducible representations for finite-dimensional, semisimple, complex Lie algebras. Using Klimyk's formula and Okubo's formula, we derive a rational ordinary generating function for the Casimir operator eigenvalues for simple Lie algebras which does not rely on knowledge of the Weyl group. We also directly reproduce generating function expressions for the classical Lie algebras originally due to Perelomov and Popov.","published":"2026-09-24T23:43:01Z","updated":"2026-09-24T23:43:01Z","authors":["Michael Flattery","Michael P. Tuite"],"categories":["math.RT","hep-th"],"headline":"New way to calculate special numbers in advanced algebra","summary":"The researchers studied special mathematical objects called Casimir operators, related to symmetrical structures called simple Lie algebras. They developed a new formula to generate important numbers related to these operators without needing complex group information usually required. Their work also confirmed earlier methods for classical cases, but the results are theoretical and focused on abstract algebraic calculations."},{"id":"2609.30634v1","title":"In-Context Binding Capacity in Language Models","abstract":"How many assignments can a language model recall before it loses track of which value belongs to which entity? We measure this limit using continuous recall curves for 12 models at or below 3B parameters and a threshold sweep over 30 open models up to 12B. On the continuous curves, the load at which recall falls halfway to chance follows $K_{50}=cN^α$, with $α=0.820$ and $R^2=0.73$. The broader sweep shows an eightfold range associated with pretraining recipe, although the continuous curves show no detectable recipe effect after controlling for scale, with few modern models in the fit. We derive why interference can lower measured capacity by reducing single-binding recall even when the load-dependent recall profile is unchanged. Direct task training also exceeds the extrapolated zero-shot law, but different measurement criteria prevent interpreting that comparison as a capacity gain. Its formation times follow a power-law form in two independent codebases, conditional on runs that succeed. Together, these results characterize capacity at the model's query interface. Bounds on joint recall and a decomposition of policy errors connect this measurement to working memory and instruction following, without treating recall as a measure of alignment. The controlled task also provides a baseline for testing whether binding limits constrain world-state tracking; the present experiments do not measure state updates or downstream transfer.","published":"2026-09-24T23:41:03Z","updated":"2026-09-24T23:41:03Z","authors":["Manas Venkata Sai Ravulapalli","Samrath Singh Chadha"],"categories":["cs.LG"],"headline":"Researchers Study How Much Info Language Models Can Remember Together","summary":"The researchers investigated how many different assignments a language model can remember simultaneously before mixing up which detail belongs to which item. They tested various models, finding a pattern in memory limits related to model size but no clear effect from training methods after adjusting for scale. Their work helps describe the models' ability to recall multiple pieces of information at once, relating to memory and following instructions. However, these experiments do not show how well models update information over time or apply it later."},{"id":"2609.30633v1","title":"Stable initialization without the CLT","abstract":"Successful training of deep neural networks is highly dependent on the distribution of the initial weights. If the weights are too large, network training blows up; if they are too small, the model fails to learn features. Stable initialization is the optimal moderation between these two extremes. The conventional theory of random networks uses the Central Limit Theorem to control inter-neuron dependencies, which introduces distributional approximation error and coupling between layers. For networks with sine activations, we derive the uniform-phase initialization, which obviates distributional approximation and fully decouples the layers. Ours is the first work to use the sine function's periodic symmetry. Models trained with the uniform-phase initialization outperform the state of the art in neural representation tasks like image and audio fitting. We find that our untuned models are competitive with the best-tuned baselines from previous work and support $μ$P width scaling.","published":"2026-09-24T23:39:49Z","updated":"2026-09-24T23:39:49Z","authors":["Simon Kuang","Kyle Chickering","Xinfan Lin"],"categories":["cs.LG"],"headline":"New way to start neural networks improves training stability","summary":"Researchers looked at how to set initial weights in neural networks, which is crucial to avoid training failures or poor learning. They developed a method called uniform-phase initialization for networks using sine functions, which avoids common mathematical approximations and allows network layers to work independently. Their approach, based on the sine wave's symmetry, showed better performance in tasks like image and audio fitting compared to existing methods, though results are based on specific models and training scenarios."},{"id":"2609.30632v1","title":"A Novel $q$-Derivative Framework with Applications to $q$-Deformed Thermodynamics and Leakage Suppression in Superconducting Qubits","abstract":"We propose a new $q$-derivative operator built directly from Jackson's $q$-number formulation, designed to preserve the structural properties of standard differential calculus while incorporating deformation effects. By analyzing $q$-deformed Heisenberg algebras, we demonstrate that this formulation maintains the consistency of thermodynamic quantities-such as internal energy, particle number, and specific heat-in dilute gas limits without requiring ad-hoc chain-rule modifications. Furthermore, we explore the physical implications of algebraic deformation using the Biedenharn-Macfarlane realization, showing how $q$-deformation induces intrinsic anharmonicity in quantum oscillator spectra and affects multi-level quantum systems ($d \\ge 3$). Applying this algebraic scheme to superconducting transmon qubits, we derive analytical pulse-shaping corrections that generalize the Derivative Removal by Adiabatic Gate (DRAG) technique, offering a robust method to suppress computational leakage in ultra-fast quantum logic operations.","published":"2026-09-24T23:36:57Z","updated":"2026-09-24T23:36:57Z","authors":["André A. A. Marinho","Gisele B. Freitas","Clovis A. C. Filho"],"categories":["quant-ph","cond-mat.stat-mech"],"headline":"New math framework aids quantum thermodynamics and qubit control","summary":"Researchers developed a new math tool called a q-derivative that adapts standard calculus to account for certain deformations. They applied this to quantum systems, showing it keeps thermodynamic properties consistent without extra tricks. Using this approach, they proposed improved methods to reduce errors in superconducting quantum bits during fast operations. These findings come from theoretical work exploring algebraic models and their effects on quantum systems."},{"id":"2609.30631v1","title":"Adapting Personalized Speech Enhancement for Low-Latency Audio-Visual Target-Speaker Extraction","abstract":"Online audio-visual target-speaker extraction aims to remove competing voices while preserving speech quality and bounding lookahead. Existing extractors are built and evaluated for separation on synthetic mixtures, leaving listening quality and meeting behavior largely untested. We introduce Audio-Visual Personalized Voice Quality Enhancement (AV-PVQE), which approaches these requirements from the other direction. We start from a personalized speech enhancement model that reconstructs a requested voice at high quality but confuses the target in 46% of two-speaker mixtures despite clean enrollment. Adding mouth features at its speaker-conditioning input and jointly fine-tuning the visual and reconstruction networks reduces this rate to 1.6%, with no future frames and 20 ms of algorithmic delay. Compared with an online autoregressive audio-visual extractor, AV-PVQE yields separation gains on two synthetic benchmarks and larger gains on recorded meetings, and keeps its advantage on excerpts with more speakers than the fine-tuning mixtures. In personalized P.835 listening tests on two meeting corpora, it improves overall quality over this extractor by 0.57 and 0.63 MOS, with similar mean rating relative to the starting model. Preservation and rejection tests show that it keeps the target intact when no competing voice is present and suppresses competing speech when the target is absent.","published":"2026-09-24T23:36:16Z","updated":"2026-09-24T23:36:16Z","authors":["Rayhan Rashed","Senja Filipi","Ross Cutler"],"categories":["eess.AS","cs.CV","cs.SD"],"headline":"Improving Clear Voice Extraction Using Audio and Visual Cues","summary":"The researchers addressed the challenge of isolating a specific person's voice from overlapping conversations in real-time audio and video. They developed a method called AV-PVQE that combines personalized speech enhancement with visual information from the speaker's mouth movements to better distinguish the target voice. Their tests, including synthetic and real meeting recordings, suggest this approach reduces confusion between speakers and improves audio quality with very low delay, though the results are based on specific benchmarks and listening tests described by the researchers."},{"id":"2609.30630v1","title":"Quantitative Decay for Linear Parabolic Equations","abstract":"We establish sharp temporal lower bounds for the full spatial $L^2$-norm of energy solutions to $\\partial_tu-Δu=vu$ on $\\mathbb{R}^n$, with $n\\geq3$. For global solutions, nonvanishing at a single time implies the lower bound $ce^{-Ct}$ when $v$ is small in the scale-invariant space $L^\\infty_tL^{n/2}_x$, and $ce^{-Ct^2}$ throughout the subcritical range $v\\in L^\\infty_tL^p_x$, with $p>n/2$, without smallness. The linear time exponent is optimal already for the free heat equation, while we show that the quadratic exponent is optimal for bounded complex potentials on $\\mathbb{R}^3$. No additional regularity of the potential is required. The proof uses a convexified mixed Carleman estimate whose norms are designed together with the final time truncation. Integration over spatial centers removes the heat-kernel weight and yields quantitative comparisons of the full solution norms. To prove quadratic sharpness in the Euclidean energy class, we construct a time-dependent function $g$ that localizes Meshkov's periodic parabolic example while keeping the induced potential bounded across the zeros of the periodic profile.","published":"2026-09-24T23:35:40Z","updated":"2026-09-24T23:35:40Z","authors":["Igor Kukavica","Qi Xu"],"categories":["math.AP"],"headline":"New Insights on How Solutions to Heat-Like Equations Fade Over Time","summary":"Researchers studied how solutions to certain heat-related mathematical equations decrease in size over time. They found precise lower limits on the rate of this decay depending on specific conditions of the equation's parameters. Their work uses advanced mathematical tools and confirms that these decay rates are the best possible under the given assumptions, based on theoretical analysis and constructed examples. These results apply to equations on multi-dimensional spaces and do not require extra smoothness of the involved functions."},{"id":"2609.30629v1","title":"FRESHLATENT: Channel-Aware Latent Adaptation for Resource-Constrained Embodied VLM Perception","abstract":"Mission-critical UAVs increasingly rely on split vision-language model (VLM) perception under tight onboard-resource and wireless-communication constraints. However, corruption of transmitted intermediate features creates a deployment mismatch for clean-trained split interfaces, while stronger channel-aware codecs can impose substantial onboard cost. We present FreshLatent, a lightweight channel-aware latent adapter that trains a power-normalized encoder-decoder through wireless corruption while keeping the surrounding VLM frozen. We formulate deployment around a mission-conditioned perception requirement and embedded interface cost, linking channel quality and communication budget to the operating conditions under which perception remains usable. At 0 dB and the tightest communication budget, FreshLatent improves gIoU and cIoU over clean split compression by 20.79 and 20.87 points, respectively. At the most adverse evaluated SNR (0 dB), across all three communication budgets, FreshLatent recovers 63.5-69.1% of the gIoU improvement achieved by a much heavier, range-trained feature-JSCC codec. On an NVIDIA Jetson AGX Xavier in 10-W mode, FreshLatent uses 37-40x fewer encoder parameters, 7.7-9.9x lower edge-interface latency, and 8.8-10.0x lower edge-interface energy than the heavier codec. Together, these results show that lightweight channel-aware adaptation can recover a substantial fraction of the robustness of a much larger communication interface while broadening quality-valid operation under constrained wireless conditions.","published":"2026-09-24T23:32:35Z","updated":"2026-09-24T23:32:35Z","authors":["Rajat Bhattacharjya","Minwoo Kim","Arnab Sarkar","Tamoghno Das","Sing-Yao Wu","Eli Bozorgzadeh","Marco Levorato","Nikil Dutt"],"categories":["eess.SP","cs.CV","cs.DC","cs.LG","cs.RO"],"headline":"Lightweight method improves drone vision under poor wireless conditions","summary":"Researchers addressed challenges faced by drones using vision-language models when wireless communication is unreliable and onboard resources are limited. They developed FreshLatent, a small adapter that adjusts transmitted data to cope with wireless interference without heavily increasing onboard computing costs. Their tests show FreshLatent significantly improves perception accuracy and efficiency compared to existing methods, especially under very poor signal quality, while requiring much less computational power and energy on typical drone hardware."},{"id":"2609.30628v1","title":"OpenHail: An Event-Driven Gymnasium Environment for Electric Ride-Hailing Fleet Control","abstract":"Machine-learning policies have attracted increasing interest for ride-hailing fleet control in recent years. Reinforcement learning, in particular, requires a structured simulation environment that specifies observations, actions, rewards, and decision epochs for training and evaluation. For electric fleets, this environment must also capture the interaction among stochastic demand, vehicle operations, and capacitated charging infrastructure. We present OpenHail, an open-source Gymnasium environment for joint control of electric ride-hailing fleets. Its fixed-size observation--action interface exposes request assignment, repositioning, and charging to a single policy. The event-driven simulator represents requests with pickup deadlines, vehicle job queues, battery dynamics, and finite-capacity charging facilities with first-in--first-out queues. A configurable decision-epoch mechanism separates internal simulator events from policy interactions, supporting event-driven, periodic, hybrid, and policy-requested control within the same operational model. The software provides seeded instances, feasible-action utilities, evaluation tools, operational metrics, and baseline policies. The source code is available at https://github.com/tommaso-schettini/openhail.","published":"2026-09-24T23:31:38Z","updated":"2026-09-24T23:31:38Z","authors":["Tommaso Schettini","Nicholas D. Kullman","Jorge E. Mendoza"],"categories":["cs.LG","math.OC"],"headline":"OpenHail Simulates Electric Ride-Hailing Fleet Control for Research","summary":"Researchers developed OpenHail, an open-source simulation tool that helps study how to control fleets of electric ride-hailing vehicles. The simulator models realistic factors like ride requests with deadlines, vehicle battery use, and limited charging stations. It allows testing and training of machine-learning policies under different control timings. The tool aims to support research by providing a detailed, flexible environment but is based on simulation rather than real-world trials."},{"id":"2609.30627v1","title":"Contrast-Aware Annotation for Statistical Graphics: The ggtwotone Package for ggplot2","abstract":"Readable annotations are essential for interpreting statistical graphics, yet conventional single-color lines and labels can lose visibility when they cross backgrounds with heterogeneous luminance. We introduce a contrast-aware annotation framework implemented in the ggtwotone R package for ggplot2. The framework combines dual-stroke rendering, adaptive text-color selection, and perceptually guided highlight palettes to improve annotation visibility across light and dark regions. A shared contrast-adjustment engine supports WCAG- and APCA-based contrast criteria, reducing the need for manual color adjustment. The package provides contrast-aware geoms for segments, curves, paths, mathematical functions, regression overlays, and text while remaining compatible with standard ggplot2 workflows. A simulation-based evaluation across heterogeneous background colors demonstrates improved worst-case contrast for dual-stroke annotations and adaptive text selection, while also identifying trade-offs in perceptual separation as the number of requested highlight colors increases. Applications to statistical graphics and scientific images illustrate the framework in practical visualization settings. Together, these tools provide a reproducible approach for incorporating contrast considerations directly into graphical annotation.","published":"2026-09-24T23:29:05Z","updated":"2026-09-24T23:29:05Z","authors":["Beenu Sareena","Heike Hofmann"],"categories":["stat.CO"],"headline":"New Tool Makes Graph Annotations Easier to See on Complex Backgrounds","summary":"Researchers developed a software package called ggtwotone for the R programming environment that improves the visibility of labels and lines on graphs, especially when backgrounds vary in brightness. Their method uses techniques like dual strokes and adaptive text colors to ensure annotations stand out on both light and dark areas. Tests using computer simulations showed better contrast and highlighted some trade-offs, offering a practical way to make graph annotations easier to read without manual color adjustments."},{"id":"2609.30626v1","title":"Frequency-Modulated Piezoelectric Haptic Display","abstract":"We present a frequency-modulated (FM) haptic display based on piezoelectric vibrating actuators. Existing haptic displays commonly encode haptic intensity through the deformation amplitude of individual haptic pixels. Although amplitude-modulated (AM) approaches have enabled compact haptic pixels, independently controlling the deformation amplitude of a large number of pixels can require increasingly complex and bulky driving systems, posing challenges for scaling toward high-density, large-area wearable displays. To address this scaling challenge, we investigate an FM design principle in which haptic intensity is encoded through vibration frequency. We further develop a \\textit{Shared-Source Frequency Modulation} (SSFM) structure in which multiple haptic pixels are powered by a common power amplifier while their vibration spectrum are controlled individually, reducing the need for independent high-power amplification at each pixel. A proof-of-concept piezoelectric haptic display was built and evaluated on rendering spatial and temporal haptic patterns through volunteer tests. The results show that participants reliably distinguished spatial and temporal patterns encoded using FM principles within the investigated operating range. These findings demonstrate the feasibility of FM-based distributed haptic rendering and suggest a potential pathway toward more compact driving architectures for future high-density, large-area wearable haptic displays.","published":"2026-09-24T23:25:51Z","updated":"2026-09-24T23:25:51Z","authors":["Boyuan Liang","Lingfeng Sun","Masayoshi Tomizuka"],"categories":["cs.RO"],"headline":"New Vibration Method Could Improve Wearable Touch Devices","summary":"Researchers explored a new way to create touch sensations using vibrations that change frequency instead of intensity. They built a prototype device using piezoelectric materials and tested it with volunteers, who could tell apart different touch patterns made this way. This approach might help design simpler, smaller systems for large wearable touch displays, but the findings are based on initial tests and need further work."},{"id":"2609.30625v1","title":"Audio LLMs Know When They Can't Hear You","abstract":"Audio large language models allow users to interact with the model through speech. When an input recording is too degraded, the model may misinterpret the user's query and respond based on an incorrect transcription. In this paper, we study model-conditional transcription reliability: whether an Audio LLM can recognize when its own transcription is unreliable. We first prompt the Audio LLM to assess whether its own transcription would be reliable, and find that the model is a poor judge of its own transcription reliability: in most cases, it predicts that its transcription will be reliable. We find that existing approaches, including speech quality predictors, audio LLM generation uncertainty, and transcript-conditioned WER estimation, provide limited signals for detecting transcription failures. In contrast, we discover that transcription reliability is strongly represented in the model's audio-encoder representations. Based on this observation, we devise a lightweight reliability predictor that operates on representations extracted by the frozen audio encoder and predicts the reliability class before generation. The reliability predictor can trigger a clarification request from the user when their voice query is predicted to be unreliable, while allowing reliable queries to proceed without modifying the underlying Audio LLM. Our predictor achieves 81.10% in-domain and 78.09% cross-domain macro-F1 scores, outperforming the strongest baselines by 10.33 and 11.93 points, respectively. Finally, we show that reliability labels can transfer across Audio LLM families, and that transfer performance is closely related to the alignment of their model-specific reliability boundaries.","published":"2026-09-24T23:22:40Z","updated":"2026-09-24T23:22:40Z","authors":["Amirhosein Javadi","Richa Dixit","Mehrdad Farajtabar","Minsik Cho","Devang Naik","Mohammad Samragh"],"categories":["cs.AI"],"headline":"New Method Helps Audio Language Models Know When They Mishear You","summary":"The researchers studied whether audio language models (Audio LLMs) can tell when their speech transcriptions are unreliable. They found that the models themselves are generally poor at judging their transcription quality. Existing methods to detect errors also perform poorly. However, the team discovered that certain internal audio features can predict transcription reliability well. Using this, they created a simple tool that warns users when their spoken input might be misunderstood, improving detection accuracy without changing the original models. These findings come from controlled evaluations and comparisons across models."},{"id":"2609.30624v1","title":"The para--Racah Hamiltonian as an exactly solvable discretization of the Pöschl--Teller box","abstract":"The para--Racah Hamiltonian is the Jacobi matrix $J$ of the para--Racah polynomials, taken at the persymmetric value of their isospectral deformation parameter: a mirror--symmetric tridiagonal matrix whose spectrum is a \\emph{quadratic} bi--lattice, the squares $(s+a)^2$ and $(s+c)^2$ of two interlaced arithmetic progressions of unit step, and which carries out fractional revival when the grid parameters $a$ and $c$ satisfy a Diophantine condition. As the number of sites grows, $J$ contracts to the trigonometric \\emph{Pöschl--Teller} Hamiltonian on a finite interval. We establish this twice: on the recurrence relation, and on the polynomials themselves, for which it is the confluence Racah~$\\to$~Jacobi. Both grid parameters survive: $c-a$ becomes a scalar centrifugal term at the interior point fixed by the mirror --- the exact analogue of the para--Krawtchouk origin --- while $a+c-1$ sets two confining singular walls at the ends, which turn the spectrum into the quadratic pair $(t+a)^2,(t+c)^2$ with Jacobi eigenfunctions $P_t^{(a+c-1,\\,\\mp(c-a))}$, the two sublattices converging to two families that differ only by the sign of the second Jacobi parameter. The mirror symmetry of $J$ becomes the reflection about the midpoint; it grades the two families and carries the fractional revival to the limit. Algebraically, the bispectral pair of the discrete model, which generates the Racah algebra $\\mathfrak{r}_1$, becomes a pair of generators of the Jacobi algebra $\\mathfrak{j}_1$ --- one rung down the Askey--Wilson hierarchy --- the dynamical algebra of the Pöschl--Teller Hamiltonian.","published":"2026-09-24T23:22:33Z","updated":"2026-09-24T23:22:33Z","authors":["Stéphane Z. Beaulac","Nicolas Crampé","Quentin Labriet","Lucia Morey","Marlon Josue Rivera Valladares","Luc Vinet"],"categories":["math-ph"],"headline":"Researchers link para-Racah matrix to Pöschl–Teller quantum model","summary":"The researchers studied the para-Racah Hamiltonian, a special symmetric matrix related to para-Racah polynomials with a spectrum formed by two interlaced quadratic sequences. They showed that as the matrix size grows, it closely approximates the known Pöschl–Teller Hamiltonian, a mathematical model for quantum particles in a finite box. Their work connects discrete and continuous systems through polynomial limits and algebraic structures, supporting a theoretical bridge between exactly solvable discrete models and a continuous quantum system."},{"id":"2609.30623v1","title":"CraftTrace: Unflattening Videos into Malleable, Creation-Inspired Structures for Generative Editing","abstract":"Recent generative video editing models enable video content modification (e.g., changing a character) but target short clips. Extending them to full multi-shot videos requires tedious work to locate relevant content across shots, segment it into clips, craft context-aware editing prompts for each clip, and repeatedly articulate complex editing intent. To address this, we explore an interaction paradigm for editing through underlying video structures (e.g., scripts, scenes, characters, shots, and their relationships). We present CraftTrace, an interactive prototype that transforms a video into a malleable, multilevel structure for generative editing. Users work in task-centric workspaces to modify elements or reshape relationships, while an AI agent translates and propagates changes across the video. A user study and expert review show that this structure helps users understand videos, formulate and refine editing intent, and explore alternatives, supporting rapid prototyping during early-stage exploration and full video post-production.","published":"2026-09-24T23:18:23Z","updated":"2026-09-24T23:18:23Z","authors":["Boyu Li","Yuqian Zhou","Duotun Wang","Ding Li","Zhe Lin","Nanxuan Zhao","Zeyu Wang","Lin-Ping Yuan","Hongbo Fu"],"categories":["cs.HC"],"headline":"New Tool Helps Edit Videos by Organizing Their Story Elements","summary":"Researchers created CraftTrace, a system that breaks down videos into parts like scenes and characters to make editing easier. Instead of editing short clips one by one, users can modify story elements, and an AI applies changes throughout the whole video. Their study suggests this method helps people understand and edit videos more efficiently, especially for complex, multi-scene projects."},{"id":"2609.30622v1","title":"Bagged Martingale Posteriors: Calibrated Uncertainty Quantification for Predictive Resampling","abstract":"Martingale posteriors and related predictive resampling methods replace the likelihood--prior pair used within Bayesian inference with a predictive model for future observations. These methods are simple to implement and increasingly popular due to their computational efficiency, but little is known about their ability to accurately quantify uncertainty. In this work, we study the concentration and calibration properties of the martingale posterior for general functionals, and show that credible sets can systematically undercover if the predictive algorithms are not carefully tuned. We propose a simple remedy: the bagged martingale posterior. Rather than starting every predictive path from the observed sample, we start the paths from random bootstrap resamples of the data and then amalgamate the resulting draws. Critically, this approach incurs no additional {simulation} cost compared to standard predictive resampling, and delivers conservatively calibrated credible sets. This scalability enables application to a range of challenging examples, including sparse high-dimensional regression and nonparametric conditional quantile models.","published":"2026-09-24T23:14:27Z","updated":"2026-09-24T23:14:27Z","authors":["Hui Wang","Edwin Fong","David T. Frazier"],"categories":["stat.ME","math.ST"],"headline":"New method improves uncertainty estimates in predictive models","summary":"Researchers studied a technique called martingale posteriors, which use predictions for future data instead of traditional Bayesian methods for drawing conclusions. They found that existing methods can underestimate uncertainty if not tuned correctly. To fix this, they introduced the 'bagged martingale posterior,' which starts predictions from multiple resampled versions of the data to better capture uncertainty. Their approach is computationally efficient and shows promise in complex settings like high-dimensional regression, but the findings are based on theoretical and simulation studies."},{"id":"2609.30621v1","title":"Assessing the Transferability of General-Purpose MachineLearning Interatomic Potentials for Heterogeneous Catalysis with HetCat26","abstract":"Foundation machine learning interatomic potentials (MLIPs) promise near-density functional theory (DFT) accuracy across broad areas of chemistry and materials science. However, their performance in describing systems and processes relevant to heterogeneous catalysis remains underexplored. Here, we introduce HetCat26, a collection of benchmark tests designed to assess pre-trained MLIPs across key aspects of catalytic modeling, including surface energetics, metal-metal oxide interactions, adsorption, and catalytic reaction networks. Evaluating fifteen foundation models, we find that performance on existing general materials benchmarks is only weakly predictive of performance on HetCat26; transferability to heterogeneous catalysis cannot be inferred from these general benchmarks. Across the benchmark tests, current models describe surface energetics and, perhaps surprisingly, reaction barriers well, whereas larger errors are observed for adsorption, and DFT site preferences are often not reproduced. Two models, eSEN-30M-OAM and MACE-MH-1-OMAT, nevertheless achieve high accuracy across the properties evaluated. Beyond model benchmarking, HetCat26 highlights the importance of training data consistency: PBE and PBE+U calculations should not be mixed within a training set. Overall, we identify challenges limiting the transferability of current foundation MLIPs to heterogeneous catalysis and, more broadly, to chemical reactions at interfaces, providing guidance for the development of the next generation of foundation models.","published":"2026-09-24T23:14:14Z","updated":"2026-09-24T23:14:14Z","authors":["Alexandre Peuch","Giaan Kler-Young","Kaifeng Niu","Jinwoo Hwang","Manos Mavrikakis","Angelos Michaelides","Fabian Berger"],"categories":["cond-mat.mtrl-sci"],"headline":"Testing AI Models for Catalysis Chemistry Accuracy","summary":"Researchers created HetCat26, a set of tests to evaluate how well advanced AI models predict chemical behaviors important for catalysis, such as surface reactions and molecule adsorption. They found that models performing well on general materials tests do not always do well on catalysis tasks, revealing limits in current AI methods. Two models performed better than others, and the study highlights the need for consistent training data and suggests challenges remain before these AI tools can reliably model complex catalysis processes."},{"id":"2609.01931v1","title":"Agent Flight Recorder: Tamper-Evident Audit Trails with On-Chain Anchoring for Long-Horizon Tool-Using Agents","abstract":"Long-horizon agents execute thousands of actions, resulting in sequential failures rather than isolated errors. When a coding agent deletes a production database or a prompt injection spreads across agents, the incident raises questions of causality, authority, and non-repudiable third-party verification. The Agent Flight Recorder captures each agent action as a structured, canonically serialized event binding eight semantic fields from intent through execution to provenance. Hash chaining and Merkle batching provide tamper evidence and compact inclusion proofs. For cross-organizational disputes where no party's infrastructure qualifies as neutral ground, periodic on-chain anchoring of epoch roots lets any verifier with the disclosed payload and Merkle proof check the record independently, without pre-agreeing on a trusted intermediary. The on-chain footprint is minimal: each anchor stores a 32-byte epoch root and a back-pointer, and no event content touches the chain. We evaluate the system across five cumulative ablation configurations on synthetic agent workloads. The full system adds ~48 microseconds median per-event latency and 512 bytes per event. L2 anchoring costs $2.30 per 100K events at 100-event epochs. The full integrity stack detects edit, delete, reorder, and fork tampering at 100% with zero false positives. Structured forensic queries achieve 1.0 precision on guardrail and delegation lookups where unstructured text search yields 0.013 and 0.077 respectively.","published":"2026-09-01T22:59:08Z","updated":"2026-09-01T22:59:08Z","authors":["Laurent Bindschaedler","Quentin Botha","Christoph Siebenbrunner"],"categories":["cs.CR","cs.MA"],"headline":"A flight recorder could reveal tampering with AI agents’ activity logs","summary":"AI agents can take thousands of actions, making it hard to reconstruct what went wrong. The researchers built a “flight recorder” that logs those actions and uses digital fingerprints, with small proofs stored on a blockchain, to let others check whether the record was changed. In tests using simulated agent activity, it detected every tested type of tampering with no false alarms and added only a tiny delay to each recorded action."}],"total":1131,"about":"Saved arXiv papers, not the complete arXiv catalogue. AI summaries are based on abstracts. Preprints may not have been peer reviewed."}