# Research Papers — 2026-09 · bundle page 23 of 266 > Full-text markdown of papers #1101–1150 of 13298 published in 2026-09, newest first. AI-readable mirrors of the full 13298-paper 2026-09 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-09/page/23 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/22.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/24.md) ### Learning from Viable Failure Prefixes: Milestone Viability Potential Policy Optimization for Long-Horizon LLM Agents Authors: Qi Zhou, Yuanfan Li Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37111v1 Long-horizon LLM agents require reinforcement learning methods that can assign credit to intermediate decisions under sparse and delayed rewards. Existing group-based methods such as GRPO and GiGPO alleviate this issue by comparing rollout returns or repeated anchor states, but they still fail when the compared returns have no variation. We identify this failure mode as zero-credit failure: during early training, many failed rollouts contain useful prefixes, yet existing methods assign them no task-discriminative advantage. To address this issue, we propose Milestone Viability Potential Policy Optimization (MVPO), a potential-routed policy optimization algorithm that learns from viable failure prefixes. MVPO estimates prefix potential over Union-Find viability regions, repairs zero-credit groups with potential-difference advantages, and attenuates the potential branch according to relative performance progress. Experiments with Qwen2.5-1.5B-Instruct show that MVPO outperforms eight strong baselines, including GRPO and GiGPO. Under the same training length, MVPO improves over the GiGPO baseline by +4.4 success points on ALFWorld and +5.3 on WebShop, while adding only 0.16%-0.20% advantage-construction overhead. --- ### Designing a Boundary Negotiating Artifact for Collaborative Socio-Technical Sense-Making in AI Regulatory Sandboxes Authors: Idoia Landa-Oregi, Tom Deckenbrunnen, Alessio Buscemi, Daniele Pagani, German Castignani Published: 2026-09-29 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2609.37109v1 The rapid, unpredictable advancements in AI system capabilities has seen regulators take adaptive and experimental approaches to policymaking. Established in other domains as instruments balancing regulation with innovation, regulatory sandboxes are seen as solutions for AI regulation. However, analyses mostly focus on the legal and institutional design of AI Regulatory Sandboxes (AIRSes). With the legal framework leaving the socio-technical interpretation to stakeholders, this creates a gap on the sense-making required to fulfill the AIRS purpose. In this paper, we approach this by designing a Boundary Negotiating Artifact as a way to mediate meaning in AIRSes. Through Research-through-Design we iteratively develop a tool, providing an interface for the different stakeholders to collaborate in AI assessment. We then position it as technical backbone in established AIRS frameworks, structuring the collaborative sense-making of the involved stakeholders. We further report the insights gained from our design process leaving the qualitative evaluation for future work. --- ### Waypoint-1.5: A Real-Time Video World Model for Consumer Hardware Authors: Rajit Rajpal, Shahbuland Matiana, Liew Wei Pyn, Anmol Agarwal, Ryan Craig, Andrew Lapp, Mithun Hunsur, Sami BuGhanem, Scottie Fox, Aaron Sanders Carson Poole, Irene Park, Dave Rossi et al. Published: 2026-09-29 Categories: cs.CV, cs.HC arXiv: https://arxiv.org/abs/2609.37107v1 We present Waypoint 1.5, a real-time diffusion world model for interactive video generation on consumer-grade hardware. Unlike general video diffusion models, interactive world models (iWMs) must respond to dense user controls under strict latency and throughput constraints. Waypoint 1.5 is pre-trained on 100,000 hours of diverse, control-aligned video game data across hundreds of games, and generates playable video conditioned on full keyboard and mouse input. The model includes two resolution variants that run across a wide spectrum of consumer hardware. To characterize this unique setting, we distinguish rendered FPS, latent FPS, and control rate. We describe the data pipeline, architecture, training methodology, and runtime system behind Waypoint 1.5. We evaluate interactivity through latency and throughput. Finally, we discuss the safety and ethics considerations unique to iWMs. --- ### VACE: Validation-Gated Alternating Co-Evolution of Agent Models and Harnesses Authors: Jiexing Qi, Yu He, Jun Liu, Qichen Huang, Shaohua Hu, Zhan Dang, Guohua Chen, Rui Yang, Wen Jiang, Yang Liu, Tao Lyu, Fangming Li Published: 2026-09-29 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.37105v1 Language model agents can be improved by updating their model weights or refining the harness that guides task execution. These components are coupled: weight updates change how the model uses the harness, while harness updates change the trajectories used for training. We propose VACE, Validation-Gated Alternating CoEvolution, which alternates agentic reinforcement learning with trajectory-driven harness refinement. After each RL stage, VACE reuses the collected trajectories to propose a harness revision and evaluates the incumbent and candidate with the updated model held fixed. The candidate guides subsequent training only if it improves validation performance. With Qwen3.5-9B, VACE achieves 45.26% test accuracy on OfficeQA and a mean partial-credit score of 75.19% on AutomationBench, exceeding weight-only RL by 6.43 and 9.09 percentage points and ungated alternation by 4.59 and 6.95 points, respectively. Across 44 harness proposals, 17 reduce validation performance at the updated checkpoint and are rejected before subsequent RL training, highlighting the importance of validation gating. --- ### What Does Post-Training Change in Multilingual Reasoning? Authors: Hongyang Li, Xiao Li, Caesar Wu, Grégoire Danoy, Pascal Bouvry Published: 2026-09-29 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.37104v1 Open-source reasoning models provide unequal access to reasoning capability across languages. When a model can solve a problem but cannot deliver a complete solution in the user's language, language becomes an access barrier rather than merely a source of performance variation. We audit Qwen3 checkpoints on competition-mathematics tasks in eleven languages. Across the ten non-English languages, only 15.4-17.9% of problems receive a correct, terminating solution with visible reasoning in the requested language in any of 16 samples, compared with 92.9% in English. To identify the source of this disparity, we evaluate thirteen endpoints from one model family, spanning released checkpoints, multilingual supervised fine-tuning (SFT) at two scales, controlled SFT ablations, and three reinforcement-learning (RL) reward formulations. We jointly track correctness, language adherence, termination, and delivery efficiency. The dominant bottleneck shifts across post-training stages. Released models often reason in English. Multilingual SFT restores target-language reasoning, but accuracy declines across multilingual, English-only, and single-language SFT runs, showing that this cost is not specific to multilingual mixing; non-English reasoning traces additionally become prone to non-terminating loops. RL restores termination in both arms at no cost in accuracy, but only the arm whose reward includes a language term delivers: rewarding correctness alone returns the model to English. Together, these stages establish a constructive post-training path from English-pivoted capability to multilingual reasoning that is reliably delivered. --- ### V2X-WAM: A Cooperative World Action Model for End-to-End Autonomous Driving Authors: Junwei You, Weizhe Tang, Can Wang, Yan Zhao, Jun Hua, Haotian Shi, Wei Zhang, Lin Wang, Bin Ran Published: 2026-09-29 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2609.37098v1 Vehicle-infrastructure cooperation can complement onboard sensing with broader and more informative observations of the traffic environment, providing valuable support for end-to-end autonomous driving. However, existing cooperative driving methods mainly exploit roadside information to enhance the representation of the current scene, while the future consequences of prospective driving actions are rarely modeled explicitly. This limits the ability of the planner to anticipate how its decisions may interact with the evolving traffic environment. To address this issue, we propose V2X-WAM, a cooperative world action model that tightly couples cooperative scene understanding, action generation, and future-world reasoning. V2X-WAM constructs a reliability-aware spatiotemporal representation from vehicle- and infrastructure-side observations, while compressing infrastructure information into a compact quantized message for efficient communication. Based on the resulting cooperative representation, a multimodal planner generates prospective trajectories, which explicitly condition future occupancy and dynamic-flow prediction. The predicted world consequences are then fed back to refine the planned trajectory, forming a closed interaction between action and future-world evolution. Experiments on a large-scale real-world cooperative driving dataset demonstrate that V2X-WAM consistently improves planning accuracy and safety over representative end-to-end cooperative driving methods, while achieving stronger future-world prediction and substantially lower communication overhead. Ablation studies further validate the effectiveness of the proposed design. --- ### Breaking the Illusion of Review Reliability under Static Evaluation: SCOPE Fuzzing for LLM-based Scientific Reviewers Authors: Zhuo Chen, Hao Zeng, Jiawei Liu, Guoxiu He, Le Cai, Liu Haotan, Li Wenbo, Yong Huang, Wei Lu Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37097v1 The rapid growth of submissions and reviewing workload has accelerated the use of large language models (LLMs) in peer review. Prior studies suggest that LLM-based reviewers can penalize content perturbations, such as overclaiming, indicating a certain degree of reliability. Yet these conclusions are largely based on a narrow set of perturbation strategies instantiated with static templates, providing limited evidence of actual reliability. In this paper, we construct a three-level evaluation framework covering perturbations to surface presentation, argumentative logic, and value judgment. Experiments on representative LLM-based reviewers reveal two limitations of static evaluation: stratified vulnerability, where perturbation effects depend on whether the paper's original review score is high or low, and perturbation undercoverage, where a single template misses vulnerabilities exposed by diverse realizations. To address these limitations, we propose SCOPE-Fuzzer, a strategy-aware fuzzer that combines feedback-driven strategy selection with adaptive mutation of paper content. By iteratively probing reviewers with dynamic perturbations, SCOPE-Fuzzer consistently uncovers vulnerabilities overlooked by static evaluation and other baselines. --- ### Why MLLMs Struggle to Count: Overcoming Individuation and Aggregation Bottlenecks with ConvStack Authors: Liwei Che, Yihao Quan, Sen Fang, Hongyi Wang, Ranjay Krishna, Ruixiang Tang, Vladimir Pavlovic Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37096v1 Multimodal Large Language Models (MLLMs) consistently struggle with fine-grained visual counting, yet the underlying causes remain poorly understood. In this work, we present a mechanistic analysis of this failure mode, identifying two critical bottlenecks inherent to the global attention pipeline of MLLMs. First, we reveal an individuation bottleneck stemming from image patchification: because Vision Transformers process patches independently, they struggle to group fragmented geometric features across boundaries into distinct object representations. Second, we identify a collapse in the subsequent counting aggregation process, where representation separation rapidly diminishes as numerosity increases due to attention compression. Identifying and formalizing these twin bottlenecks constitutes our first major contribution. To overcome them, we propose ConvStack, a lightweight architecture that operates directly in the visual token space to explicitly aggregate and inject local spatial structures via zero-initialized residual connections. By explicitly addressing the individuation bottleneck, ConvStack provides unambiguous geometric evidence for downstream aggregation. Remarkably, by fine-tuning exclusively on counting tasks, the model achieves substantial improvements in dense object counting and broader spatial understanding benchmarks, without compromising on general visual capabilities. --- ### Task-Oriented Visual Feature Compression via Residual Vector Quantization for Device-Edge Multimodal Inference Authors: Luning Pang, Cheng Yuan, Jiawei Shao, Mingtao Huang, Yuan Shen Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37090v1 Large multimodal models (LMMs) support diverse visual understanding and reasoning tasks but are often impractical to run entirely on resource-constrained devices. Device-edge co-inference reduces device computation, yet transmitting visual data over bandwidth-limited uplinks can introduce substantial delay. Task-oriented feature compression (TOFC) reduces the payload through feature aggregation and entropy coding. However, continuous-feature coding remains costly, and query-agnostic aggregation may discard task-relevant local evidence. We propose query-guided task-oriented feature compression (Q-TOFC) for device-edge multimodal inference. Q-TOFC employs residual vector quantization (RVQ) to encode each merged feature as a compact sequence of codebook indices, reducing its representation cost and allowing more features to be transmitted. It further incorporates query relevance into feature aggregation and uses a quantization error compensation adapter to mitigate the distortion introduced by discrete quantization. Experiments on seven multimodal benchmarks show that Q-TOFC reduces the visual payload by 53.6% relative to TOFC while maintaining comparable average normalized task performance. End-to-end latency evaluations further demonstrate lower latency under bandwidth-constrained uplinks. --- ### Real2Gym: Building Gyms from Videos, Bringing Skills to Robots Authors: Kerui Ren, Yingxiang Xu, Kaiwen Song, Linning Xu, Bo Dai, Mulin Yu, Tao Lu Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37089v1 Real-world videos provide rich demonstrations of manipulation, but turning them into reusable robot skills requires visually aligned environments, executable physical interactions, and mechanisms for learning from experience. We introduce Real2Gym, an agentic Real2Sim2Real framework that turns human and robot demonstrations into interactive simulation gyms and brings skills acquired in simulation to physical robots. The Real2Sim module reconstructs editable scenes, aligns objects and cameras with the input, validates demonstrated or retargeted actions through native physics execution, and generates task-conditioned variations with action-feasibility checks. Within these environments, the agent generates executable code for manipulation stages, observes their outcomes, and distills successful attempts and failures into reusable task procedures, object-relative motions, and recovery strategies. Through a shared perception-and-control interface, these skills guide subsequent execution in simulation and on real robots, with motions adapted to current observations and no updates to the underlying model weights. Extensive evaluations demonstrate that Real2Gym enables high-fidelity simulation environment reconstruction, outperforming GPT-6 Astra Direct Mode by 16.7% in success rate with approximately 74.9% fewer policy-execution tokens across these environments, while exceeding it by 33.3% in physical robot execution success rate across four tasks on a real Franka robot. --- ### ARGOS: Reinforcement Learning-Driven Multidimensional Elasticity for Service Orchestration in the Computing Continuum Authors: Javier Mateos-Bravo, Sergio Laso, Juan Luis Herrera, Ilir Murturi, Pantelis Frangoudis, Schahram Dustdar Published: 2026-09-29 Categories: cs.DC, cs.AI arXiv: https://arxiv.org/abs/2609.37085v1 Data-intensive services in the Computing Continuum must balance analytics quality, resource usage, and cost across heterogeneous nodes with limited and uneven capacity. This balance becomes especially difficult when resource scaling reaches capacity limits, because changes in demand and cluster pressure must then be absorbed without violating client-defined quality ranges. Existing orchestrators mainly adapt resources, placements, or replicas, while analytics requirements such as coverage, sample, and freshness remain fixed. This article presents ARGOS, the Adaptive Reinforcement Learning-Driven Governance for Orchestrated Services, an end-to-end controller that formulates multidimensional elasticity as a per-request Markov decision process over analytics quality and cluster pressure, supported by capacity-aware admission. ARGOS is evaluated under controlled workloads and time-varying multi-tenant arrivals on a heterogeneous cluster. Across the controlled scenarios, the deep reinforcement learning policies consistently outperform the non-learning baselines and approach the independently tuned best-fixed reference. A separate live evaluation reports improvements over the static midpoint under realistic and saturated arrivals, with no recorded CPU or memory violations but remaining coverage violations. These results support deep reinforcement learning as an adaptive mechanism for multidimensional elasticity when resource scaling alone is insufficient. --- ### Identifying ODEs from Unstructured Data with Causal Representation Learning Authors: Alessandro Trenta, Riccardo Massidda, Davide Bacciu, Sara Magliacane Published: 2026-09-29 Categories: cs.LG, cs.AI, stat.ML arXiv: https://arxiv.org/abs/2609.37083v1 We study the problem of recovering the governing ODE of a dynamical system from unstructured, high-dimensional observations such as images. Existing methods for ODE discovery typically assume direct measurements of the variables, or do not provide theoretical guarantees on the learned variables and equations. While Causal Representation Learning (CRL) methods provide guarantees on identifying variables from high-dimensional observations up to component-wise diffeomorphisms, we show that in general these variables cannot be used directly as input to equation discovery methods, which typically assume that the variables will lead to sparse equations. So we introduce SParse Equivalent Equation Discovery AutoEncoder (SPEED-AE), a framework that combines a pretrained CRL method with a component-wise autoencoder that learns transformations of variables that are amenable to sparse ODE discovery. We show that for polynomial ODEs, this additional step allows us to restrict the identifiability of each variable from polynomial to monomial diffeomorphisms. Experiments on Lotka-Volterra, Lorenz, and a two-pendulum system show that SPEED-AE improves on the disentanglement of the CRL methods and that it recovers ODEs that are closest to the ground truth, while achieving state-of-the-art forecasting performance. --- ### Traverse: Learning When to Remember, Reset, and Redirect for Long-Horizon Web Search Authors: Jingyuan Ma, Lynx Aster, He Zhang, Siyao Song, Weijie Yuan, Zhe Zhang, Kai Jia, Zhifang Sui Published: 2026-09-29 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.37082v1 Long-horizon information-seeking agents often accumulate noisy or misleading context, causing early mistakes to persist and making recovery increasingly difficult. We introduce an autonomous search harness in which the agent manages its own search process through three states: Rubric, Answer, and Verify. The agent first defines criteria for a valid answer, searches under these criteria, and then independently verifies the result before deciding whether to terminate or continue searching. It is further equipped with a Seal Memory tool that enables active context management. Training this behavior with reinforcement learning, however, can induce Seal Collapse, resulting in unstable training and preventing the agent from reliably learning when and how to use its memory tools. We solve this with a simple strategy that trains only the final segment after context management. Our 35B model achieves 72.83 on BrowseComp, outperforming comparable open-source systems, and consistently improves over the base model across BrowseComp-ZH, xbench, DeepSearchQA, WideSearch, financial investigation, and product search. Ablations show that autonomous compression outperforms automatic compaction and validate our RL design. --- ### LDM-is-AE: Latent Diffusion Model is an Auto-Encoder for End-to-End Image Generation Authors: Zhengqiang Zhang, Lingchen Sun, Rongyuan Wu, Qiaosi Yi, Xiangtao Kong, Chaodong Xiao, Lei Zhang Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37080v1 Latent Diffusion Models (LDMs) typically adopt a two-stage pipeline: an auto-encoder (AE) is first pre-trained to define a latent space, then a diffusion model is trained to perform denoising within it. Such a two-stage design introduces a representation mismatch, as the latent space is optimized for reconstruction rather than adapting the denoising dynamics. We reveal that the LDM itself is an AE, and consequently present LDM-is-AE, an end-to-end one-stage LDM training framework that eliminates the need for a separately trained tokenizer. Our key observation is that the LDM backbone actually performs a latent-to-feature-to-latent transformation at each denoising step, which can be interpreted as an internal decoding--encoding process. Leveraging this structure, we split the DiT backbone into two reciprocal components, DiT-E (i.e., DiT Encoding) and DiT-D (i.e., DiT Decoding), and impose image-space supervision on the intermediate features across all timesteps. Our model encourages the internal representation to align with the image domain throughout denoising, thereby establishing an explicit latent-to-image-to-latent path. At the zero-noise timestep, our model further performs an image-to-latent-to-image mapping, corresponding to an auto-encoding process. As a result, LDM-is-AE jointly learns latent representations and denoising dynamics in an end-to-end manner, yielding a diffusion-native latent space tailored to the generation process. Experiments demonstrate that LDM-is-AE exhibits highly competitive generation performance, achieving an FID of 1.80 and 1.90 on 256x256 and 512x512 class-conditional image generation, respectively. --- ### ZeroDiff: Zero-Shot Time Series Reconstruction via Informed-Prior Diffusion Authors: Yingda Fan, Dan Lu, Xiaowei Jia Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.37078v1 Time series modeling increasingly demands high-quality supervision, yet target observations remain scarce - exogenous inputs are broadly available, but target measurements are often unavailable due to cost, infrastructure, or accessibility constraints. Can models trained on observed locations reconstruct target time series where measurements have never been collected? We term this zero-shot time series reconstruction. A naive approach - directly mapping exogenous inputs to targets - can yield predictions at unobserved locations, but without target signals, such models fail to capture the intrinsic dynamics of the target variable, producing overly smooth outputs that underestimate extremes. This reveals systematic errors that call for explicit modeling and calibration. We propose ZeroDiff, which constructs an informed prior from exogenous variables alone, then learns to calibrate reconstruction errors through diffusion - training on observed locations and generalizing to unobserved ones. Experiments across diverse real-world datasets demonstrate significant improvements over existing approaches. Our code is available at https://github.com/YingdaFan/ZeroDiff-ICML2026. --- ### UnlearningSoup: Is Repeated Tuning Necessary for Large Language Model Unlearning? Authors: Puning Yang, Qizhou Wang, Junchi Yu, Bo Han, Xiuying Chen Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.37076v1 Large language models trained on vast corpora inherently risk memorizing harmful content that may later re-emerge in their outputs. To mitigate this issue, existing unlearning methods typically rely on training-based parameter updates, such as gradient ascent and its variants, to delete targeted content while preserving other knowledge. However, balancing the competing goals of forgetting and retention makes hyperparameter choices for these methods particularly difficult, often requiring repeated tuning to obtain a strong model that still leaves substantial room for improvement and transfers poorly across models and datasets. To address this challenge, we investigate whether unlearning runs exhibit exploitable structure in weight space, and observe that models from different runs still lie in a shared evaluation-performance basin. This suggests that stronger models may be recovered through an unlearning-tailored soup strategy, reducing the need for repeated tuning for further improvement or new settings. Motivated by this, we propose UnlearningSoup, a unified framework that provides two strategies: EfficientSoup uses binary-search-based interpolation to quickly discover a well-performing model in the early stage, where repeated tuning would otherwise make strong model selection costly. PerformanceSoup uses reweighted souping to efficiently unlock the remaining performance potential in the later stage, where repeated tuning becomes increasingly inefficient. Extensive experiments across diverse datasets and models show that UnlearningSoup delivers 2.4x to 3.3x efficiency gains in hyperparameter selection, while consistently improving performance across settings. --- ### Context without Commitment: Robust Dense Correspondence under Non-Rigid Deformation Authors: Yuzhen He, Sara Homscheid Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37071v1 Non-rigid point-cloud registration aims to find the corresponding target point for each point on a deforming source surface. Point-level matching keeps the full target cloud available, but correspondence becomes ambiguous when different regions have similar local geometry. Regional or coarse-to-fine methods provide broader spatial context, but an incorrect regional match can exclude the correct correspondence before dense matching. We propose CoCo-Reg, which uses regional patches to enrich dense point features without allowing patch predictions to restrict the final point-level search. CoCo-Reg constructs farthest-point-sampled patches, exchanges geometric information within and between source and target, supervises patch similarity using identity-corrected point overlap, and projects the resulting regional information back to dense point features. The final registration stage still scores the full target cloud before global point-level candidate selection. On 726 held-out ModelNet10 objects across nine deformation levels, two established learning-based baselines obtain mean correspondence errors of 0.1993 and 0.1921, whereas CoCo-Reg obtains 0.0547. Relative to its point-level baseline, this is a 72.6\% reduction. CoCo-Reg achieves lower correspondence error on 92.3\% of paired test objects and reduces the mean fraction of points with error above 0.1 from 47.3\% to 17.3\%. Chamfer distance and HD95 decrease in the same direction, and CoCo-Reg remains lower across all tested deformation levels. These results support using regional context for dense non-rigid correspondence without imposing a hard patch-level restriction on the final search. Because evaluation uses one checkpoint per method, the reported gains characterize the complete systems rather than the isolated causal contribution of an individual component. Code will be made publicly available. --- ### Predictive Safety Curricula for Robust Legged Locomotion Authors: Ivan Ovinnikov, Pascal Sutter, Christian Gehring, Jordis Herrmann Published: 2026-09-29 Categories: cs.RO, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.37070v1 Rare but consequential failures can persist in learned locomotion policies for legged robots even when average task performance is high, in part because standard curricula primarily adapt task difficulty rather than the distribution of safety-critical experience. We introduce Predictive Safety Curricula (PSC), a framework for allocating locomotion training experience using learned predictions of future safety cost. PSC trains a distributional safety critic from policy rollouts and uses its predictions to prioritize both terrain contexts and previously encountered randomized events. The resulting curriculum modifies the training distribution while leaving the task reward and policy-optimization loss unchanged. We evaluate PSC in controlled rough-terrain locomotion and in production locomotion systems. PSC improves reliability relative to standard terrain progression, advantage-based replay, and learning-progress curricula, with the largest gains on difficult terrain and under degraded observations. The same allocation principle transfers to two production locomotion stacks. On ANYmal-D hardware, PSC reduces shank-collision incidence by $63\%$ relative to the learning-progress curriculum across three matched training seeds, with a reduction in every seed. On a production stair-climbing platform, PSC eliminates observed shank collisions in the evaluated hardware trials. These results show that learned predictions of future safety cost can provide an effective signal for allocating training experience toward rare failure modes and improving locomotion reliability. --- ### FACT: Fidelity-Aware Construction of Articulated Twins Authors: Kuixiang Shao, Chuansen Nie, Yinuo Bai, Jiayuan Gu, Jingyi Yu Published: 2026-09-29 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2609.37067v1 Visually plausible articulated assets may still fail during contact interactions or exhibit inaccurate motion. We present FACT (Fidelity-Aware Construction of Articulated Twins), an agentic framework that progressively constructs articulated twins to improve geometry, contact, and dynamic fidelity. The agent drives an evidence--diagnosis--revision loop on a shared editable representation, selecting measurements and model edits using quantitative feedback, while numerical tools execute and validate the updates. It reconstructs editable articulated geometry from images through feature planning, targeted measurements, and diagnostic refinement. On this reference, it repairs collision proxies through task-aware local repartitioning before fidelity-constrained compression. Finally, it constructs response models from passive-response videos, using simulation residuals to guide model revision and constrained physical parameter fitting. Experiments show that FACT improves geometric reconstruction over baselines, enables more reliable interaction with simpler collision proxies, and better reproduces held-out physical responses than direct parameter inference. --- ### Beyond Compression: Diagnosing How Post-Training Changes Mathematical Reasoning Authors: Hongyang Li, Yiming Zhu, Xiao Li, Caesar Wu, Said Mammar, Pascal Bouvry Published: 2026-09-29 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.37066v1 Post-training is central to mathematical reasoning in modern large language models (LLMs), but endpoint pass@1 alone underidentifies what has changed. Gains may reflect newly reachable solutions, cheaper sampling of latent solutions, surface robustness, or memorisation. We compare three post-training paths under a common diagnostic readout: our sufficiently trained off-policy distillation trajectories, released Qwen3 off-policy-plus-on-policy distillation endpoints, and a released DeepSeek-Math endpoint trained with Group Relative Policy Optimisation (GRPO). Our probe uses cross-surface pass@K over verbatim prompts, paraphrases, numerical isomorphisms, and translations, plus consistency, distribution-shape, and verified supervised-fine-tuning (SFT) membership analyses. We find two regimes. On easier AMC problems, large-K ceilings are near saturation, so post-training mainly compresses sample cost. On harder AIME problems, post-training expands the large-K ceiling over the base model: sufficient off-policy distillation already raises this ceiling, Qwen3 released endpoints raise it further, and DeepSeek-Math GRPO does not dominate sufficient off-policy distillation at large K. English-dominant distillation improves non-English reasoning but preserves language-tier gaps. A controlled-overfit audit finds limited sensitivity in current SFT-membership probes. Compression is one regime of post-training, not a universal explanation. --- ### RL-PaO: Prediction as Action in Decision Making under Uncertainty Authors: Jiahui Feng, Dafang Zhao, Zheng Chen, Zhengmao Li, Lingwei Zhu Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.37065v1 Decision-making under uncertainty often relies on predicted parameters, yet accurate prediction does not necessarily lead to good operational decisions. Aligning prediction with downstream optimization requires learning from the consequences of the decisions those predictions induce. We introduce RL-PaO, a reinforcement learning framework that integrates system formulation, optimization, and decision execution into a single environment. This yields a Markov decision process in which prediction is regarded as action: it shifts the environment to produce subsequent context and reward that explicitly aligns prediction error with realized cost, and learning the optimal policy does not require differentiating through the black-box solver. We evaluate RL-PaO on day-ahead energy scheduling using real historical data. On the test year, RL-PaO achieves the lowest annual cost among the non-oracle baselines, achieving on average $10\%$ cost reduction. Moreover, RL-PaO is capable of further analyses to provide strong interpretability both from the policy evolution perspective and the cost-accuracy trade-off. --- ### vSkipper: Translating Dynamic Layer Skipping into LLM Serving Gains Authors: Wei Da, Yavuz Ferhatosmanoglu, Evangelia Kalyvianaki Published: 2026-09-29 Categories: cs.DC, cs.LG arXiv: https://arxiv.org/abs/2609.37062v1 Dynamic layer skipping reduces LLM computation by allowing each token to execute only a subset of the model's layers. However, existing skippers rely on specialized generation loops and do not integrate with modern serving engines. As a result, fewer executed layers do not necessarily translate into lower serving latency: FlexiDepth skips 8 of Llama-3-8B's 32 layers on average, yet its standard generation loop decodes 14.6--21.0% more slowly than the base model. We present vSkipper, a virtualization layer that makes dynamic layer skippers pluggable in serving engines while preserving continuous batching, fixed-shape batches, paged KV caching, and captured decode graphs. At each routed layer, vSkipper groups tokens by the skipper's decision and uses routed execution only when predicted to be profitable. We implement vSkipper in SGLang and evaluate the released FlexiDepth checkpoint against upstream SGLang under identical prompts, arrivals, output lengths, and launch settings. At the knee of upstream's load curve, vSkipper reduces mean end-to-end latency by 36.8% on GSM8K and 13.6% on BBH. Under saturation, it increases request throughput by 11.3% and 7.4%. Serving adds no statistically resolved quality loss beyond the checkpoint's own. Across synthetic skip policies, two Qwen3 skippers, and three GPUs, we demonstrate reuse without workload-specific tuning. To our knowledge, vSkipper is the first system to realize serving-efficiency gains from per-token interior layer skipping within a modern LLM serving engine. The code is open-sourced as an SGLang fork at https://github.com/AKafakA/sglang-vskipper/tree/vskipper-ref --- ### A Comprehensive View of Fairness through Distributional Stability Authors: Gayane Taturyan, Charlotte Laclau, Stephan Clémencon Published: 2026-09-29 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.37061v1 We view fairness as a property of distributional stability. Rather than assessing a predictor under a fixed data distribution, we study how its predictions change under perturbations that modify the composition of protected groups. A predictor is fair if it remains stable under such shifts. Under this perspective, several classical notions of fairness arise as stability with respect to specific perturbations, with the associated unfairness gap given by a Lipschitz constant of a prediction-rate functional. This formulation also yields guarantees that hold uniformly over a range of demographic compositions at test time, without requiring knowledge of the deployment distribution. It leads to a learning procedure based on convex combinations of reweighted predictors, formulated as a second-order cone program, for which we establish generalization bounds. Experiments on standard benchmarks illustrate the approach. --- ### Message Passing Does More with Less for In-Context Learning on Graphs Authors: Dooho Lee, Jinmo Lee, Minho Jeong, Kijung Shin, Jaemin Yoo Published: 2026-09-29 Categories: cs.LG, cs.SI arXiv: https://arxiv.org/abs/2609.37057v1 Achieving strong performance with graph neural networks (GNNs) typically requires training and hyperparameter tuning for each dataset, incurring repeated costs and effort. Graph in-context learning (ICL) avoids this by using a single pretrained model to predict unknown node labels directly from labeled context nodes. Existing approaches, however, rely on dense attention across nodes, making inference increasingly expensive as graphs grow. In this work, we present Ephris, a new graph in-context learner built on sparse message passing, scaling linearly with the number of node-feature entries and graph edges. Ephris is pretrained entirely on synthetic graphs generated from structural causal models with diverse graph structures and relational dynamics, exposing the model to varied dependencies among topology, features, and labels. We evaluate Ephris on 51 node-classification datasets against 15 extensively tuned GNNs and existing graph ICL methods under both high- and low-label train/validation/test splits. Across both settings, Ephris ranks first on all four aggregate measures: Elo, improvability, average rank, and accuracy. Its inference cost remains comparable to training a single GNN once, while being over 10 times faster than previous graph ICL models. Together, these results advance the performance-runtime Pareto frontier, demonstrating that strong graph ICL does not require dense attention. Code and model weights are available at https://github.com/nums-ai/ephris. --- ### Evolving Towards Better Codes: LLM-Guided Search for High-Distance Binary Linear Codes Authors: Amal Seddas, Vladyslav Shashkov, Maryna Viazovska, Emmanuel Abbe Published: 2026-09-29 Categories: cs.IT, cs.AI, cs.NE arXiv: https://arxiv.org/abs/2609.37056v1 Evolutionary program search driven by large language models (LLMs) has produced record-breaking constructions for open problems in combinatorics and beyond. We apply this approach to the longstanding problem of improving the best-known bounds for binary linear codes. Building on the EvoTune evolutionary framework and the ShinkaEvolve codebase, we introduce LinCodeEvolve, which evolves code-construction programs against an exact minimum-distance evaluator. A strategy loop combines diversity-driven search and expert supervision: when progress plateaus, new strategies are used to redirect the search. LinCodeEvolve discovers seven record-breaking codes, $(172,21,66)$, $(173,20,68)$, $(176,21,68)$, $(181,21,70)$, $(184,21,72)$, $(189,22,72)$ and $(200,21,77)$, six of which have concise quasi-cyclic descriptions. With standard code modification techniques, they improve $22$ entries of the tables. Every code is verified by exhaustive enumeration. These results suggest that LLM-guided search can help find improved codes and complement existing methods in coding theory. --- ### Spatial-OPSD: Self-Improving Spatial Reasoning via Label-Free Self-Distillation Authors: Zhenyu Liu, Zhangquan Chen, Keyi Chen, Mingze Sun, Xiang An, Haodong Jing, Ruqi Huang Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37055v1 Vision-language models (VLMs) increasingly operate in embodied and spatially grounded settings, where accurate understanding of depth, viewpoint, and three-dimensional relations is essential. However, improving spatial reasoning typically relies on ground-truth answers, answer-derived rewards, or other forms of task-specific supervision. We introduce Spatial-OPSD, a label-free self-improvement framework that instead exploits spatial structure naturally available from perception and reconstruction tools. During training, a privileged teacher receives automatically obtainable spatial priors, such as depth, reconstructed 3D relations, and camera geometry, while the student observes only the original visual-language input. On trajectories sampled by the student itself, the teacher provides dense token-level supervision, allowing the student to internalize spatial knowledge without ground-truth answer labels or privileged information at inference time. To extend this supervision beyond a single round, we adopt a round-wise recursive training scheme: the teacher remains frozen within each round to provide a stable learning target, and the improved student initializes both teacher and student in the next round, where privileged spatial priors re-establish an informative teacher--student asymmetry. This enables repeated self-improvement while avoiding a rapidly moving teacher during optimization. Across four VLM families, a single round of Spatial-OPSD consistently improves the five-benchmark average, while three rounds further push a strong spatially specialized model to the open-source frontier, achieving the highest average among the open models and the best results on three of five spatial reasoning benchmarks. Our code is available at https://github.com/vermouth599/Spatial-OPSD. --- ### actr: aligning thoughts and responses for multilingual safety in reasoning llms Authors: Xianhui Zhang, Jian Yu, Chengyu Xie, Chenhang Cui, Shuyi Miao, Pengyang Shao, Yu Zheng, Fei Shen, Tat-Seng Chua Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37054v1 Ensuring the safety of reasoning large language models (LLMs) across languages is essential for their reliable deployment. However, when exposed to jailbreak attacks in non-high-resource languages, these models may generate unsafe responses even when their reasoning traces identify safety risks. To address this issue, we propose aligning cross-lingual thoughts and responses (ACTR), a framework that improves multilingual safety alignment by strengthening the use of existing safety reasoning. Specifically, we first present the think gap score (TGS) to compare the normalized contributions of reasoning traces to attention outputs during response generation across languages, and use reasoning- trace substitution to measure the cross-lingual safety gap. Next, using a corpus of jailbreak queries, we assess neuron importance through changes in response representations caused by neuron masking and compare the high-importance neuron sets obtained with reasoning enabled and disabled to identify safety think neurons that support the use of safety reasoning. Finally, we devise neuron-selective consistency optimization (NSCO), which uses a frozen judge model to reward agreement between the safety categories of reasoning traces and responses while updating only the parameters associated with the selected neurons, requiring no human-annotated responses or preference data. Across two reasoning models, ACTR achieves lower average attack success rates than the evaluated state-of-the-art methods on AdvBench-X and MultiJail, with safety gains extending to unseen languages, while preserving or improving average performance on multilingual knowledge and mathematical reasoning tasks and limiting false refusals of benign requests. Warning: this paper contains examples with unsafe content. --- ### MatToolBench: Benchmarking Multimodal Agents in Real-World Materials Science Workflows Authors: Mei Wu, Rui Xie, Runyu Zhang, Yuqiang Li, Tianfan Fu, Bo Chen, Kai Yu, Xin Chen, Lu Chen Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37053v1 Multimodal GUI agents have achieved impressive results on general software benchmarks, yet their ability to operate professional scientific software remains largely unexplored. In materials science, sparse domain-specific web data, specialized interfaces, and tacit workflow conventions create blind spots that general-purpose pretraining cannot readily bridge. We present MatToolBench, the first real-environment benchmark for evaluating multimodal GUI agents on professional materials science software, comprising 204 tasks across 10 tools in three modalities: GUI operation, OriginPro scripting, and code-based database queries, all executed inside a Windows 11 VM. Each task is decomposed into fine-grained sub-criteria by domain experts, enabling interpretable partial-credit scoring; the GUI component of our multi-level evaluation pipeline achieves an average F1 of 0.98. For OriginPro figure-generation tasks, we further conduct a human-LLM agreement study to validate the use of a multimodal judge for secondary aesthetic assessment. Our experiments show that strong performance on general benchmarks does not transfer to professional scientific workflows, and that this gap is not a visual-grounding problem alone: failures arise from domain-specific operational knowledge, sparse pretraining coverage of scientific software, weak cross-tool artifact handoff, and critical states exposed only visually. Even the best model reaches only 25% success rate on GUI tasks and 45% on code tasks. MatToolBench therefore serves as a challenging diagnostic benchmark and real-environment testbed for data-scarce, knowledge-intensive scientific workflows. --- ### OmniRoute: Mapping Temporal Semantic Evidence to Audio-Visual Token Budgets for Efficient Omnimodal Large Language Models Authors: Yuchen Deng, Zidang Cai, Feidiao Yang, Yufei Wang, Jie Wang, Hai-Tao Zheng, Yuxing Han Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37052v1 Omnimodal large language models (Omni-LLMs) encode audio and visual streams into temporally interleaved token sequences for multimodal reasoning. However, processing long audio-visual token sequences incurs substantial prefill costs. Existing compression methods have made progress, but often overlook temporal changes in audio-visual semantic relevance. Motivated by temporal variation and local continuity, we propose OmniRoute, a training-free, two-stage compression framework. First, Temporal Evidence-Guided Budgeting (TEGB) derives chunk-wise modality preferences and initial leading-modality budgets from semantic relevance and local content variation. Second, Budget-Constrained Semantic Compression (BCSC) compresses the leading modality and then calibrates the follower's retention target using the actual retained fraction. For video, it combines spatiotemporal grouping with query-guided selection; for audio, it selects tokens based on encoder attention and query relevance, then merges residual tokens into context anchors under visual guidance. Experiments on four representative benchmarks demonstrate a better trade-off between inference efficiency and performance than competitive baselines. The code and interface will be released to facilitate further research. --- ### NHO: A Neural Hamiltonian Operator for Anchor-based Region Localization and Dense Correspondance Authors: Jing Li, Yawei Luo, Xiangze Meng, Ying Li, Tieru Wu, Rui Ma Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37048v1 Non-rigid partial-to-full shape correspondence from sparse anchors requires identifying the corresponding region on the full surface and recovering dense correspondences between the partial shape and that region. We present NHO, which combines sparse anchors with the intrinsic geometry of the partial shape to learn a neural Hamiltonian operator whose localized eigenspace encodes both the region support and intrinsic coordinates for dense correspondence. NHO parameterizes the Hamiltonian potential as an intrinsic neural field and optimizes it using anchor evidence together with spectral and geometric constraints. To resolve the spatial ambiguity left by sparse anchors, we introduce reciprocal refinement between operator estimation and correspondence recovery. At each round, the current eigenspace provides spectral coordinates and restricts matching to its induced support, while geometrically reliable correspondences provide additional evidence for updating the potential. After refinement, aggregated eigenfunction energy yields the final localization, and the recovered map initializes dense correspondence refinement. Experiments demonstrate competitive accuracy on both tasks and robustness to uniform scaling and rotation. --- ### Multi-Depth Temporal Fusion for Feedforward, Locally Trained Spiking Neural Networks Authors: Aidin Attar, Eleonora Cicciarella, Michele Rossi Published: 2026-09-29 Categories: cs.NE, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.37047v1 We propose a new spiking neural network (SNN) design to process static images and event streams using time-to-first-spike (TTFS) latencies. Our key research question is which architectural choices best accommodate local and online learning in multi-layer convolutional SNNs. This question is addressed via an original framework combining residual-like connections with multi-depth feature aggregation and consensus. The full SNN pipeline features an early-vision front end, to convert raw visual data into sparse spike latencies, a four-layer convolutional backbone trained layerwise with unsupervised spike-timing-dependent plasticity (STDP), a deterministic Multi-Depth Temporal Fusion (MDTF) and a final classifier trained with reward-modulated spike-timing-dependent plasticity (R-STDP). Rather than replacing early features in deeper layers, the proposed MDTF preserves early temporal evidence, adding sparse residual events from intermediate layers, and incorporating deeper features only when they agree in time with earlier representations. The resulting architecture is experimentally validated across MNIST, Fashion-MNIST, CIFAR-10, and N-MNIST, delivering strong classification performance under a fully local learning regime. Selective multi-depth fusion significantly outperforms traditional STDP/R-STDP baselines on higher-variability visual tasks (achieving +18.2 pp on Fashion-MNIST and +29.2 pp on CIFAR-10). Furthermore, activity-budget analyses show that the network retains high accuracy even when removing a large fraction of late or weak spike events, confirming its high data efficiency and reduced event-processing requirements. The codebase is publicly available at github.com/aidinattar/multi-depth- temporal-fusion-snn. --- ### Speed in the Blind Spot: An Interpretability Analysis of Dynamic Perception in VLMs for Autonomous Driving Authors: Katharina Winter, Stefan Englmeier, Fabian B. Flohr Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37046v1 Vision-Language Models are increasingly used in autonomous-driving systems, yet their ability to recover dynamic physical state from visual input remains insufficiently characterized. We study velocity understanding as a controlled diagnostic across three tasks: surrounding-agent speed, current ego speed, and short-horizon future ego-speed proposal. On nuScenes, we evaluate open-weight general-purpose and PhysicalAI VLMs, together with the driving-oriented Alpamayo-1.5 Vision-Language-Action model, using multiple input and output formulations. We combine verbal evaluation with temporal perturbations, counterfactual ego-speed hints and linear probes of hidden representations. The tasks exhibit distinct failure modes. Surrounding-agent speed is weakly encoded in an agent-specific form, whereas current ego speed is often internally accessible but poorly verbalized: continuous probes achieve 4.7-5.8 km/h MAE compared with 10.2-16.8 km/h MAE for verbal outputs. Multiple frames provide inconsistent verbal gains to single frame inputs, and frame order is rarely exploited. Under non-optimized QLoRA, task-specific adaptation improves both task-relevant latent speed representations and verbal readout, but continuous surrounding-agent speed estimation remains weak, while most future-speed gains survive frame shuffling, indicating limited temporal grounding. Driving specialized Alpamayo-1.5 shows stronger latent representations for surrounding-agent and future ego speed, while current ego-speed decodability is comparable and substantial probe-verbal gaps remain. Thus, driving specialization can strengthen motion representations but does not guarantee stronger encoding across both scene and ego states or reliable readout. The results show that plausible planning outputs do not necessarily imply reliable recovery or temporal grounding of the underlying dynamic state. --- ### Learning from Think-Mode Advantage via On-Policy Distillation Authors: Wanqi Ren, Jianxiang Wang, Danxuan Liu, Linyi Ding, Yuan Zhang Published: 2026-09-29 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.37044v1 Explicit intermediate reasoning gives large language models (LLMs) a stronger problem-solving mode. We study learning from this think-mode advantage via on-policy distillation (OPD). OPD preserves student-generated trajectories and provides dense token-level teacher targets at student-visited prefixes. Privileged reasoning is used during distillation rather than student inference. Uniform ThinkOPD, a natural think-enabled OPD baseline, conditions a fixed teacher on one shared think trace and uniformly distills every sibling student response. Although its prefixes are on-policy, the trace need not follow a route compatible with every complete response: the same privileged trace can induce different teacher-student discrepancies even when responses reach the same outcome. We summarize this interaction with trace-response divergence (TRD) and introduce ThinkOPD, which routes supervision at the response level by combining group-relative reward gain with a TRD-based compatibility proxy. Final response weights are normalized within each rollout group. Across mathematical reasoning and code generation, ThinkOPD outperforms Uniform ThinkOPD in both same-model settings and both cross-model teacher-student pairs, and it exceeds representative rationale and self-distillation baselines in a controlled comparison. Controlled interventions show that outcome benefit and the TRD-based proxy provide complementary routing signals in this setting. Think-enabled OPD provides a controlled setting for studying how teacher advantage becomes transferable along student responses. --- ### Iterative Exact Discrete Guidance for Energy-Based Sampling Authors: Yuwen Qian, Yidong Ouyang, Zhengyan Wan, Hongyuan Zha Published: 2026-09-29 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2609.37043v1 Sampling from unnormalized distributions over large discrete state spaces becomes difficult when a multimodal target is far from a tractable reference. We introduce Iterative Exact Discrete Guidance (IEDG), a population-exact, trajectory-wise guidance framework for unnormalized discrete targets. Rather than learn the full reference-to-target correction in one step, IEDG introduces a global Boltzmann tilt along an annealing trajectory. Each stage learns a stage-local posterior correction for an incremental Boltzmann tilt of the current source, while the resulting corrections are accumulated relative to a fixed analytic posterior. At the population optimum, exact stage posteriors recover the correct reverse dynamics, whose exact simulation reproduces the target distribution. IEDG chooses stage increments by relative effective sample size (rESS), which controls Rényi-2 displacement and locally adapts the step size to the thermodynamic geometry of the annealing path. Our stagewise total-variation analysis shows that limited overlap amplifies Bregman fitting error by $1/\sqrt{\mathrm{rESS}}$, while posterior, simulation, and truncation errors enter additively. IEDG improves all distribution-level errors over the neural baselines on ordered, exactly enumerated Ising $4\times4$, while substantially reducing one-shot errors on Ising/Potts $16\times16$ across thermodynamic regimes and attaining the best neural-sampler result on several reported local-statistic and phase-coverage metrics. On Max-Cut, its best-of-512 and average-sample ratios exceed all the baselines. Code and artifacts are available at https://github.com/StillFantasy123/iterative-exact-discrete-guidance. --- ### GleanVID: Complementary Token Selection for Efficient Video Large Language Models Authors: Shuo Yang, Changbai Li, Rui Tang, Xinyu Zhao, Linlin Yang, Baochang Zhang Published: 2026-09-29 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.37042v1 Video Large Language Models (VideoLLMs) have achieved strong video understanding capabilities but incur substantial inference overhead due to the large number of visual tokens. Existing VideoLLM token compression methods largely rely on selection-independent scoring, overlooking cross-frame complementarity and consequently retaining redundant evidence across frames. Instead, we view video token selection as a progressive evidence accumulation process. It aims to retain visual evidence that is individually informative and collectively complementary under a limited token budget. Building on this insight, we introduce GleanVID, a training-free inference acceleration framework for VideoLLMs. Specifically, GleanVID first allocates the global token budget across frames according to temporal novelty and then selects tokens by jointly considering local representativeness and subspace complementarity, thereby preserving richer and less redundant visual evidence. Extensive experiments across diverse VideoLLMs and benchmarks demonstrate that GleanVID consistently achieves state-of-the-art performance. Notably, with only 25% of visual tokens, GleanVID preserves 98.6% of Qwen3-VL's original performance while reducing its prefill latency by 44.7%. On LLaVA-OV-7B, GleanVID at a 25% retention ratio even slightly surpasses the original model. --- ### Beam Search as Test-Time Self-Distillation via Counterfactual Contexts Authors: Su Ee Tan, Xiaotong Ji, Rasul Tutunov, Haitham Bou-Ammar, Matthieu Zimmer Published: 2026-09-29 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.37041v1 Self-Distillation Fine-Tuning (SDFT) enables a language model to act as its own teacher: by conditioning on a demonstration, the model produces an implicit reward via pointwise mutual information, which guides on-policy learning without external supervision. However, SDFT operates at training time: it requires gradient updates and access to expert demonstrations, making it inapplicable at inference. We propose test-time self-distillation, a decoding-time method that extracts a steering signal from the self-distillation framework without any parameter updates, reward models, or training data. Our key insight is that counterfactual contexts, i.e. fixed textual templates that hypothetically prime the model for excellent versus poor reasoning, can substitute for the demonstration. The log-odds ratio of a candidate answer under these two counterfactual conditions defines a new reward signal. We derive the optimal KL-regularized policy under this reward, which takes the form of a Gibbs reweighting of the base distribution. Crucially, this reweighting is global: it cannot be decomposed into independent per-token operations without ignoring future trajectory quality. We therefore approximate the target distribution via beam search. Experiments on mathematical reasoning (MATH500), code generation (HumanEval), and graduate-level science QA (GPQA) across multiple model scales show that test-time self-distillation improves over standard sampling, low temperature, beam search and power sampling baselines on average, demonstrating that the self-distillation principle can be operationalized at inference time. --- ### Selecting The Most Informative Tokens in Natural Language Autoencoders Authors: Federico Torrielli, Gianluca Barmina, Andrea Blasi Núñez, Amon Rapp, Luigi Di Caro, Peter Schneider-Kamp, Lukas Galke Poech Published: 2026-09-29 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.37040v1 Natural language autoencoders translate a language model's internal activations into readable explanations. Explaining every token position is costly. Which positions should an auditor inspect to understand a potential threat? We study this question across $4.7$ million explanations on prompt injection and concealment. We compare signals from model computation with a ranker trained only on chat structure. Chat structure usually selects more relevant explanations than the computational signals, without requiring a model forward pass for position selection. On three of four datasets, explaining just $5\%$ of positions retains nearly all of the success rate from explaining every position, where success means obtaining an explanation about the threat. The benefit varies with the audit task. We also show that pretrained verbalizers recover words that models have learned to conceal through fine-tuning, without additional verbalizer training. These results identify where auditors can concentrate explanation generation and show that useful explanations can extend beyond the model a verbalizer was trained to describe. --- ### NowcastDiT: Diffusion Transformers are Effective Precipitation Nowcasters Authors: Haoran Xu, Xingzhuo Guo, Yuchen Zhang, Jincheng Zhong, Jianmin Wang, Mingsheng Long Published: 2026-09-29 Categories: cs.LG, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2609.37038v1 Precipitation nowcasting demands accurate short-term forecasts under strong spatiotemporal variability. Diffusion models are well suited to modeling complex precipitation distributions, yet existing approaches often introduce increasingly specialized designs, leaving the capability of a standard diffusion architecture underexplored. We show that a standard Diffusion Transformer already provides a simple and scalable foundation for precipitation nowcasting, with domain-specific requirements accommodated naturally within its design space. Based on this principle, we develop NowcastDiT and instantiate this flexibility through two complementary adaptations: a dynamics-aware noise prior for temporally coherent forecasts, and end-to-end reinforcement learning with timestep-aware rewards for meteorological skill. Experiments on SEVIR and MRMS benchmarks show that NowcastDiT achieves state-of-the-art performance in both perceptual quality and meteorological skill. These results suggest that standard DiT can serve as an effective foundation for precipitation nowcasting. --- ### High-Resolution Dynamic Functional Connectivity Generation with Graph-Variate Flow Matching Authors: Om Roy, Yashar Moshfeghi, Keith Malcolm Smith Published: 2026-09-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.37037v1 High-resolution dynamic functional connectivity (DFC) can reveal rapidly evolving brain-network interactions, but short temporal windows yield noisy, often low-rank covariance estimates. Graph-Variate Dynamic (GVD) connectivity addresses this by modulating fast instantaneous interactions with stable trial-level support. This suppresses spurious fluctuations and emphasizes persistent, informative connections. We show that the Hadamard construction lifts low-rank instantaneous connectivity from the positive-semidefinite to the positive-definite cone, keeping high-resolution trajectories on the SPD manifold without ridge regularisation or post-hoc projection. We introduce GVD-CFM, a class-conditional generative model for high-resolution dynamic connectivity. Each trial is represented as SPD GVD matrices on a product Riemannian manifold, then mapped through a global log-Euclidean diffeomorphism and an invertible temporal DCT basis. A Transformer-based conditional flow models all spectral modes jointly and generates the full trajectory non-autoregressively in Euclidean coordinates while preserving exact correspondence with valid SPD sequences. Retaining the full DCT basis also enables decoding on denser temporal grids without retraining. Across multiple EEG motor-imagery datasets, GVD-CFM delivers the strongest overall results for held-out distributional fidelity, temporal-dynamics preservation, and synthetic-to-real classification. It also remains computationally efficient relative to strong raw-signal and direct GVD-space generative baselines. GVD-CFM therefore provides a practical framework for realistic, temporally coherent, high-resolution brain-network generation with preserved manifold structure and resolution-flexible decoding from a single trained model. --- ### Watch-Think-Interact: Bootstrapping Long-Horizon Multi-Turn Streaming Video Reasoning with Reinforcement Learning Authors: Ziheng Huang, Yicheng Bao, Xueheng Li, Zhenkun Gao, Bangwei Liu, Kunquan Li, Yuxiang Shen, Bangyan Li, Xuejiao Wang, Changbo Wang, Gaoqi He Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37035v1 Streaming video assistance requires models to answer asynchronous questions from an observed prefix under a fixed context budget. Existing approaches model response timing or compress history, but an online state formed before future questions are known can omit visual details before later questions reveal their relevance; the retained state alone cannot recover them. We introduce Watch-Think-Interact (WTI), a closed-loop framework for multi-question streaming video reasoning. WTI maintains compact natural-language memory entries tagged with source-video time ranges; these entries support direct reasoning when sufficient and otherwise anchor selective recall of finer visual evidence. For each question, WTI answers when current context and memory suffice, continues watching when required evidence has not appeared, or recalls a relevant past interval and decides again after incorporating the returned chunks, without replaying the full observed history. To train this behavior, we construct WTI-82K, comprising 82,335 timed questions across 4,812 causally aligned trajectories, and develop Stream-GDPO to optimize complete multi-question streaming rollouts using trajectory-level feedback for response timing, source-video recall, and memory updates. WTI achieves state-of-the-art aggregate performance among the compared open-source streaming baselines, reaching 83.3% on StreamingBench and 73.6% weighted overall accuracy on OVO-Bench. --- ### FedLAFP: Low-Rank Aggregation Meets Full-Rank Personalization in Federated Fine-Tuning Authors: Mengjun Yi, Huaian Gu, Yinghao Ai, Furao Shen, Jian Zhao Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37033v1 Federated parameter-efficient fine-tuning enables clients to adapt pre-trained models without sharing raw data or communicating the full model, but statistical heterogeneity makes a single global adapter insufficient for personalized prediction. Existing personalized methods typically use the same low-rank structure for both shared and private adaptation, overlooking their distinct requirements for aggregation and personalization. We propose FedLAFP, a role-aware framework that couples a compact, globally aggregated LoRA branch with a client-private, full-rank-capable RandLoRA branch. The shared branch provides an efficient interface for transferring common knowledge, whereas the private branch combines fixed random low-rank bases with learned scaling coefficients to provide expressive client-specific adaptation without additional communication. Client- and layer-specific mixing coefficients jointly fuse the two branches, and only the shared LoRA parameters are exchanged. A controlled linear study supports this role assignment: LoRA yields more aligned client updates and lower aggregation error, while RandLoRA more accurately recovers client-specific residuals. Experiments across four visual recognition benchmarks show that FedLAFP consistently outperforms local-only and federated LoRA baselines, achieving an average personalized accuracy of $86.93\%$ and exceeding the best baseline average by $1.30$ percentage points. --- ### UniBuild: Unified Building Mapping From Multi-Source Optical Remote Sensing Imagery With Detail Decoding and Geometry Regularization Authors: Wei Huang, Chenying Liu, Yilei Shi, Xiao Xiang Zhu Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37031v1 Building extraction from optical remote sensing (RS) imagery is fundamental to urban mapping, yet existing methods are often dataset-specific and generalize poorly to unseen domains. Their practical use is also limited by insufficient detail recovery and weak geometric regularization, leading to blurred boundaries, irregular shapes, and merged adjacent buildings. To address these issues, we propose UniBuild, a unified building extraction framework for multi-source RGB optical RS imagery. First, a unified multi-dataset training scheme is constructed over heterogeneous RGB optical datasets to learn transferable building representations across sensors and resolutions. Second, a novel detail-preserving HR-DPT decoder is designed to integrate high-level semantic features with high-resolution spatial features, enhancing building detail recovery. Third, geometry-aware regularization is introduced through a structure-tensor-based direction-aware loss for boundary direction consistency and a saddle-aware loss for suppressing false activations in narrow inter-building gaps under low-resolution conditions. We train and evaluate UniBuild on multi-source RGB optical datasets, including 10 public high-resolution datasets and two self-collected low-resolution datasets. Experiments show that UniBuild consistently improves building-region accuracy, boundary sharpness, and adjacent-building separation across diverse datasets. It also generalizes well to unseen domains and supports practical building extraction from RGB optical RS imagery up to 10\,m resolution. The predicted masks can be further converted into GIS-compatible building footprints through simple polygonization. The trained model and inference code are released at https://github.com/zhu-xlab/UniBuild. --- ### MotionInsight: Diagnosing Object Motion Deficiencies in Generated Videos Authors: Jiahao Zhan, Yongrui Ma, Qunliang Xing, Xuanyu Zhang, Jingqi Tong, Junlin Li, Li zhang, Shijie Zhao, Tianfan Xue Published: 2026-09-29 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.37030v1 Despite rapid progress in video generation models, they still exhibit obvious motion deficiencies, often manifested as incorrect object motion. However, most existing video quality evaluations focus on aesthetic quality or text-video alignment. To address this gap, we study object-centric motion fidelity assessment, evaluating target objects along object consistency, motion continuity, and physical plausibility. To achieve this, we first introduce VidMotion, a diagnostic dataset of 6,879 videos with designated moving objects and fine-grained annotations including dimension-wise scores and failure causes. We further propose MotionInsight, a diagnostic evaluator that shifts assessment from implicit RGB-frame observation to explicit motion-space diagnosis. By constructing motion-aware representations, MotionInsight makes subtle motion deficiencies more observable. We also introduce motion-specific rewards during GRPO to transform observed motion into a diagnostic assessment. Experiments demonstrate that MotionInsight provides an effective basis for diagnosing object motion deficiencies, producing human-aligned scores along three dimensions and grounded explanations. --- ### LongSpark: Efficient speculative decoding with a fixed-cost parallel drafter Authors: Hao-Yuan He, Peng-Fei Liu, Si Shen, Ming Li Published: 2026-09-29 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.37029v1 Speculative decoding accelerates autoregressive inference by verifying multiple draft tokens in a single target forward pass. However, as the context grows, existing state-of-the-art drafters become increasingly expensive, eroding the very efficiency advantage they are designed to provide. We argue that this scaling is unnecessary. A standalone language model must grow with its prefix because it is solely responsible for every token it produces. A drafter, by contrast, only proposes candidates; the target catches and corrects every error before any token is committed. The drafter's decoding cost can therefore be made entirely independent of the prefix length. We introduce LongSpark, a block-diffusion drafter that achieves this by extracting fixed-size, multiscale views from the target's verification pass, thereby eliminating the need for a growing persistent state. Extensive evaluations demonstrate that LongSpark achieves state-of-the-art end-to-end efficiency across multiple model scales and realistic serving conditions. Notably, it delivers the lowest time-per-output-token on long-context tasks while reducing the drafter's context state by several orders of magnitude. --- ### Beyond Low-Rank Parameterization: Narrowing the Gap Between LoRA and Full Fine-Tuning via Gradient Decomposition Authors: Yihao Ouyang, Shiwei Li, Haozhao Wang, Xiandi Luo, Zhuoqi Hu, Jinglun Yu, Yichen Li, Ruixuan Li Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37027v1 Low-Rank Adaptation (LoRA) is a widely used approach to parameter-efficient fine-tuning (PEFT), yet a performance gap can remain relative to full fine-tuning (FFT). Many LoRA variants improve the initialization or optimization of low-rank factors. At each training step, however, their first-order weight-space directions are constrained by the current parameterization. We characterize the corresponding LoRA-accessible gradient space and show that it coincides with the tangent space induced by the current LoRA parameterization. This characterization yields an orthogonal decomposition of the full weight gradient at the current model parameters. We term the component orthogonal to this space the normal gradient. Based on this decomposition, we propose GDLoRA (Gradient-Decomposed Low-Rank Adaptation). GDLoRA reconstructs the full weight gradient from forward activations and backward signals, extracts its normal component, and directly updates the base weights with this component, while retaining standard AdamW optimization for the LoRA factors. GDLoRA incorporates complementary normal gradients without increasing standard LoRA's optimizer-state memory budget under matched adapter and optimizer configurations. Experiments on natural language understanding, mathematical reasoning, commonsense reasoning, and image classification show that GDLoRA consistently improves over LoRA and narrows the performance gap to FFT. The code is available at https://anonymous.4open.science/r/GDLoRA. --- ### AnyAct: Universal Action for Self-Evolving Agents Authors: Lingrui Xu, Yangqin Jiang, Jiachang Zhang, Xubin Ren, Chao Huang Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37025v1 As large language models (LLMs) advance, AI agents are increasingly deployed in open-world environments to tackle complex sequential tasks (e.g., document processing, cross-application collaboration), relying heavily on actions ranging from GUI operations to semantic APIs. However, three core challenges persist: the "scale dilemma" of massive tool ecosystems exceeding LLM context windows, the "non-stationarity" of tool quality due to updates or outages, and the "heterogeneity" of feedback formats (pixels, text, structured data) creating information silos. To address these, we propose AnyAct, a universal action layer that unifies available capabilities into a self-evolving action space, enabling agents to operate efficiently and reliably in large-scale, dynamic tool ecosystems. AnyAct's core design focuses on two objectives: constructing this action space via hierarchical progressive retrieval (filtering task-relevant actions) and test-time reliability evolution (pruning unreliable actions), and enabling reliability-aware action orchestration through a heterogeneous observation grounding module that unifies multi-modal feedback. Additionally, it defines a hybrid action space (primitive + semantic actions) and optimizes for a balance between task success rate and execution cost. Evaluations on LiveMCPBench and OSMCP (a new benchmark we developed for multi-granularity action collaboration) demonstrate state-of-the-art performance. AnyAct delivers substantial performance gains over baseline methods across various LLM base models on LiveMCPBench and improvements are particularly notable for models with constrained native capabilities. On OSMCP, it achieves 77.27% overall success with only 50 steps, which is half the steps required by most competitors. --- ### Language as the Interface: Foundation-Model Contrastive Learning Links Transcriptomes and Electrophysiology Authors: Junbo Shen, Jinying Gao, Bo Lei Published: 2026-09-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.37024v1 Integrating transcriptomic and electrophysiological data is essential for building multimodal foundation models for neuroscience. Patch-seq provides paired measurements of gene expression and intrinsic electrophysiology from the same neuron, establishing a basis for training cross-modal models. Here we introduce LangPatch, a foundation-model-based contrastive learning framework that uses paired Patch-seq data to align pretrained GenePT representations with electrophysiological phenotypes through a language-based interface. Gene descriptions and verbalized electrophysiological profiles are embedded by the same frozen text encoder. A context adapter and projection modules connect the modalities through paired contrastive learning. Across mouse visual, mouse motor, and human cortical cohorts, LangPatch achieves the highest mean transcriptome-to-electrophysiology prediction correlation among the evaluated foundation-model and representation-learning methods. It also improves held-out cross-modal alignment in the two mouse cohorts (FOSCTTM 0.107/0.135 vs. 0.208/0.222 for JAMIE, an existing cross-modal Patch-seq imputation method). It predicts transcriptomic family, type, cortical layer, and marker-gene expression from electrophysiology, exceeding other baselines on most endpoints. More importantly, the method transfers across brain areas and species: a model trained on mouse visual cortex predicts electrophysiology in motor cortex with approximately 70% correlation retention and in human cortex with 47% (58% on acute-slice recordings). Together, these results demonstrate alignment between molecular and functional representations of neurons, providing a building block for multimodal foundation models in neuroscience. --- ### Physics-Informed Multi-Agent Coordination for Hospital Patient Flow Optimization Authors: Guoqing Zhang, Rafik Hadfi, Takayuki Ito Published: 2026-09-29 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2609.37022v1 Efficient patient flow coordination across autonomous hospital departments is critical for mitigating overcrowding and balancing resource utilization. While classical queueing theory, specifically open Baskett--Chandy--Muntz--Palacios (BCMP) networks, provides an interpretable mathematical topology for healthcare operations, analytical models rely on stationary assumptions and fixed routing matrices that degrade under state-dependent real-world dynamics. Conversely, centralized reinforcement learning approaches struggle to accommodate the decentralized structure of hospital governance, where individual clinical departments function with localized observations, heterogeneous resources, and divergent operational objectives. In this paper, we present a Multi-Agent Systems (MAS) framework titled \emph{Physics-Informed Multi-Agent Coordination}, which embeds empirically calibrated BCMP queueing topologies as physical priors within a decentralized multi-agent reinforcement learning architecture. Formulated as a Decentralized Partially Observable Markov Decision Process (Dec-POMDP) under coupled resource constraints, our method enables autonomous departmental agents to cooperatively negotiate patient routing and dynamic service scaling. To mitigate environmental non-stationarity without inducing excessive communication overhead, agents exchange localized action fingerprints along network edges and optimize a spatially decomposed reward structure. Empirical evaluations driven by real-world MIMIC-IV patient trajectories indicate that this cooperative multi-agent approach substantially reduces cumulative system delay compared to static Markovian approximations, heuristic dispatching, and independent multi-agent baselines, while maintaining clinical safety constraints. --- ### LatCom: Cross-Agent Latent Compression for Efficient Multi-Agent Collaboration Authors: Shinan Zhang, Tao Zhang, Qihui Zhu, Mengjie Zhang, Dong Jin, Yunpeng Hou, Shuangwu Chen, Xiaobin Tan, Quan Zheng, Jian Yang Published: 2026-09-29 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.37017v1 LLM-based multi-agent systems (MAS) increasingly use latent collaboration to avoid the information loss and repeated encoding-decoding overhead of natural-language communication. However, directly forwarding all sender latents makes the receiver-side context scale with both the number of agents and the reasoning length, increasing computation, memory usage, and collaboration latency. A natural solution is latent compression. But we find that cross-agent redundancy remains unresolved in existing latent compression approaches, which typically compress each sender independently and then concatenate the results. We propose LatCom, a cross-agent latent compression framework for efficient multi-agent latent collaboration. LatCom maps multiple sender latents into a fixed number of receiver-readable and task-relevant slots. Rather than reconstructing all sender hidden states, it optimizes the compressed latents for receiver-side task utility. LatCom trains the compressor in two stages: single-sender readability learning first establishes a latent interface interpretable by the frozen receiver, and multi-sender fusion learning then trains the compressor to fuse complementary evidence and remove redundancy across agents. Experiments on multiple benchmarks with Qwen3-4B show that LatCom achieves an average 2.46x inference speed-up over LatentMAS and reduces output token usage by 70.3% while maintaining comparable average accuracy. --- ### Back2Struct: Making Structured Images Editable Again Authors: Pengyu Yan, Yixin Wu, Yunjie Tian, David Doermann Published: 2026-09-29 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.37016v1 Structured images, such as diagrams, charts, and flowcharts, are inherently symbolic and can be compactly represented in an editable format, yet in practice, they are often rendered as images, and therefore not graphically editable. This mismatch presents a significant challenge for researchers, engineers, and designers who wish to incorporate modified versions of existing graphic content into new materials without manually reconstructing it. In this study, we presentBack2Struct, which "makes structured images editable again" by directly recovering vector graphics code (SVG / XML) from image representations. Given an image of a structured graphic, Back2Struct predicts semantically object-level SVG / XML code that explicitly encodes text, shapes, topology, and layout, rather than performing low-level pixel vectorization. The generated code can be seamlessly imported into tools such as PowerPoint, allowing users to edit, refine, restyle, and reuse graphic content while preserving structural fidelity. Beyond supervised fine-tuning on ground-truth SVG token sequences, we further optimize Back2Struct with reward-based learning to better match deployment-time requirements: the output should be syntactically valid, properly concise, and visually faithful to the input diagram. Specifically, we design a composite reward that jointly encourages SVG / XML compilability, length consistency with the reference code, and structural or semantic similarity between the generated and ground-truth graphics. These complementary signals guide the model to produce SVGs that are not only closer to the training distribution, but also more complete, editable, and renderable in practice. Experiments show that Back2Struct improves accuracy, editability, validity, and user alignment over baselines. Dataset and code are available at: pengyu965.github.io/Back2Struct.github.io ---