Physical AI Brief
Daily cross-source signals for the Physical AI supply chain — silicon photonics, CPO, VLA models, humanoid hardware, embodied AI. Three streams, one page, zero filler.
742 items today · 678 arxiv · 2 SEC 8-K · 62 humanoid · 0 CN photonics
01 ARXIV · PHYSICAL AI PAPERS
678 items- arxiv:2610.26860 · cs.LGOpenVLHarness: Real World Visual Agentic Tool-Use with Multimodal MemoryYu Zhou, Cheng-Fu Yang, Rui Sun, Di Wu +7
Multimodal large language models (MLLMs) have made rapid progress on visual reasoning, yet many tasks require evidence that is difficult to obtain through internal reasoning alone: our controlled study on 22 benchmarks shows that simply increasing the inference budget of the base model yields only modest gains. Such tasks demand grounded perception, external knowledge, and numerical measurements obtained through heterogeneous tools, whose outputs must be made interpretable to the model and retained across reasoning steps. We introduce OpenVLHarness, a multimodal reasoning harness organized aro
memoryagentictool-usebenchmark - arxiv:2610.26861 · cs.CVCan It Actually Work? Synthesizing Functional CAD MechanismsWenqian Zhang, Yangyi Huang, Zigeng Xu, Weiyang Liu +1
While LLM agents can generate visually plausible CAD assemblies, these designs often fail to fulfill their intended physical functions. We study functional CAD design, a setting where agents interpret underspecified requests to construct and iteratively revise mechanisms that satisfy functional requirements in physical simulation. Verifying these requirements is nontrivial: different candidate mechanisms may realize the same function through entirely different components, controls, and operating sequences. To bridge this gap, we propose FunCAD-Harness, a framework that translates abstract requ
llm agentbenchmark - arxiv:2610.26856 · cs.CVReinforcement Learning with Visually Attributable On-Policy Self-DistillationShiao Xie, Siyu Chen, Jianwei Lv, Xiandong Li +1
Multimodal reinforcement learning with verifiable rewards (RLVR) provides reliable task-level supervision, yet offers limited guidance for localizing errors in visual reasoning. On-policy self-distillation (OPSD) complements this supervision with dense token-level corrections from a privileged teacher equipped with an evidence-centered visual view. However, the resulting corrections may conflate visual evidence with non-visual shifts induced by linguistic priors, response style, or the privileged view itself. We formalize this phenomenon as privileged-correction contamination: a token-level co
benchmark - arxiv:2610.26857 · cs.CVVideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine TokenSangbeom Lim, Seoung Wug Oh, Heeji Yoon, Seungryong Kim +1
Diffusion-based video matting has recently emerged as a leading approach, leveraging video diffusion priors from internet-scale pretraining to achieve strong zero-shot generalization from limited synthetic data. However, the heavy compute of these diffusion backbones limits inference to a short chunk of just 6 to 24 frames, so even ordinary videos span multiple chunks. Because each chunk is matted independently from coarse mask guidance, existing methods fail to maintain consistency across chunk boundaries, producing flickering edges and discontinuous transparency. We address this with a mask-
benchmark - arxiv:2610.26855 · cs.RORopedia W0: Scaling Loco-Manipulation World Models across Embodiments with Human ExperienceTianqi Liu, Hanfei Zhao, Qifeng Liang, Hui En Pang +11
World models enable embodied agents to anticipate action consequences, yet jointly modeling locomotion and manipulation across diverse embodiments remains challenging due to heterogeneous action spaces and limited multimodal supervision. We present Ropedia W0, the first egocentric loco-manipulation world foundation model spanning humans and robots with diverse embodiments. We curate Xp-Data, comprising 10k hours of egocentric human data and 3k hours of robot data with unified multimodal annotations and systematic quality control. A unified kinematic representation encodes locomotion through ca
embodiedmanipulationdexteroushumanoidteleoperationworld model - arxiv:2610.26854 · cs.ROPhysical Words: A Vocabulary for Structuring Latent Reasoning in Robot ManipulationJiaming Liu, Xinghao Zhu, Yiming Zhang, Shanghang Zhang +7
Multimodal reasoning provides robotic foundation models with predictive context needed for reliable action generation. However, language-based reasoning captures task semantics but struggles to express fine-grained physical dynamics, while latent reasoning models future state evolution but offers limited structure for organizing these dynamics into reusable reasoning primitives. To bridge this gap, we introduce PhysWord, a reasoning-before-acting foundation model built on Physical Words, which use atomic action verbs as shared semantic anchors for continuous representations of physical dynamic
manipulationdexteroustactileliberogripper - arxiv:2610.26852 · cs.ROCoopUQ: Robust Multi-Robot Social Navigation via Uncertainty-Aware Cooperative ForecastingZehao Wang, Jianpeng Yao, Yuping Wang, Yiping Li +8
Reinforcement learning enables mobile robots to make fast and scalable navigation decisions in constrained human-shared environments. However, directly extending single-robot approaches to multi-robot settings often fails to ensure safety and efficiency, especially under occlusions and out-of-distribution (OOD) scenarios, where individual robots may overlook potential risks or make inaccurate predictions. Leveraging cooperation effectively for both prediction and navigation remains a major challenge. We propose CoopUQ, a multi-robot social navigation framework that bridges uncertainty-aware co
multi-agent - arxiv:2610.26851 · cs.CVManga2Animate: An Agentic Framework for Faithful Manga Animation with Script and Storyboard PlanningSheng-Wei Zhang, Hsi-An Chen, Yan-Ting Liou, Chih-Hsuan Chen +2
Adapting existing manga into coherent animation that stays faithful to the source is a valuable yet challenging task. Existing methods either access the manga mainly through text, losing its characters and art style, or feed manga pages directly into video generation models, which misinterpret the manga's structure and produce incorrect storylines. In this paper, we propose Manga2Animate, a multi-agent framework that grounds every stage of generation in the source manga. We decompose the task into two intermediate representations, an animation script and a visual storyboard. Each representatio
agentmulti-agentagenticagent framework - arxiv:2610.26850 · cs.CVGenOlympiad: Benchmarking Frontier Image Models at Their Capability BoundariesXi Hu, Martin Bell, Liang Zheng
Text-to-image models have made tremendous progress and can produce high-quality images. However, a question which would be asked in any other discipline is where the frontier models fall short. We introduce GenOlympiad, a benchmark which provides the most challenging text prompts from nine capability axes for stress-testing frontier text-to-image models. The assessed capabilities include generating large quantities, performing deep reasoning, ensuring hard physical consistency, following complex rules and so on. The prompts are designed such that prompt alignment is visually verifiable (e.g.,
benchmarkevaluation frameworkevaluation protocol - arxiv:2610.26848 · cs.ROWhere is Physical RSI capability from? An Observational Scaling Law of LLM-as-policyZekai Li, Jiayi Yu, Jinlu Zhang, Luyang Niu +7
In this paper, we systematically investigate the mechanisms, scaling behavior, and generalization of physical Recursive self-improvement. We introduce PhysRSI-eval to quantify physical RSI through a five-axis metric: First-Attempt Performance, Efficiency, Success, Budget and Improvement from failure. Different VLMs are evaluated on 18 manipulation tasks under a fixed interaction budget. The analysis of experimental results motivates the Physical Barrel Hypothesis: different capabilities may constrain different parts of improvement. The five-axis metrics also distinguish systems with close publ
manipulationagentself-improvementbenchmark - arxiv:2610.26847 · cs.LGWeAct: Dynamic Weight Switching as Agent ActionsRui Shi, Xunjian Yin, Jiacheng Sang, Zichuan Liu +1
Long-horizon agents must learn heterogeneous capabilities from interleaved trajectories. Updating shared parameters to learn new capabilities can degrade previously acquired ones, resulting in catastrophic forgetting. As a trajectory unfolds, different stages may require different capabilities, yet conventional joint training and static adapter merging maintain a fixed parameter configuration throughout execution. We introduce WeAct (Weight-level Internal Actions), which trains independent capability-specific LoRA adapters over a frozen backbone and adds adapter selection to the agent's action
agentbenchmark - arxiv:2610.26846 · cs.ROGPT-6 Astra Is a Native Visuo-Tactile Policy for Contact-Rich ManipulationChen Yang, Linzhe Shi, Hang Zhang, Mu Xu +2
Recent advances in multimodal models have enabled GPT-6 Astra to perform robotic manipulation through in-context learning. Yet contact-rich control and tactile perception are often regarded as strengths of specialist policies, leaving open whether Astra can directly understand and use tactile observations. We provide, to our knowledge, the first demonstration that Astra directly interprets vision-based tactile images and uses their contact cues for manipulation. Tactile RGB images enter through its existing visual interface, enabling few-shot closed-loop control on nineteen simulated UniVTAC a
manipulationtactilegraspbenchmark - arxiv:2610.26845 · cs.AITrain An Imaginative AgentJiaxin Ge, Junyi Zhang, Tianning Chai, Hanjun Yoo +8
A seemingly reasonable push in Sokoban can make the puzzle unsolvable. To avoid such mistakes, an agent has to anticipate the consequences of its actions before execution, especially when feedback from the real environment is sparse or actions are irreversible. We introduce Agentic Imagination (Agima), a training framework that trains an agent to imagine before acting, i.e. to be imaginative. Agima trains a single omni model to propose actions, generate images of their outcomes, and then reason over these imagined futures to decide what to execute in the real environment. Rather than committin
world modelagentagentic - arxiv:2610.26844 · cs.AIAnyGoal: On World Model Generalization Beyond Expert RolloutsHeejeong Nam, Divyat Mahajan, Quentin Le Lidec, Randall Balestriero +1
The goal of a world model is to capture the dynamics of an environment. This has the great appeal that it allows agents to plan and simulate a course of actions with the help of a world model. For these uses to be reliable, however, the learned dynamics must behave well beyond the specific state-action combinations observed in expert rollouts. Despite this, the relationship between data coverage, dynamics generalization, and downstream planning performance remains underexplored. We study this relationship through the lens of compositional generalization: can a world model recombine familiar st
world modelbenchmark - arxiv:2610.26840 · cs.AILatComm: Reusable Context Compression for Latent Communication between Large Language ModelsHuawei Lin, Jinting Liu, Tan Yu, Weijie Zhao
Systems that combine several large language models (LLMs) often use one context for many questions. When the context is passed between models as text, the receiving model processes all of its tokens again for every question. Avoiding this cost requires a context representation that is compact, reusable across questions, and readable by a different LLM. Existing compression and latent communication methods each provide only some of these properties. We introduce LatComm, a Latent Communication interface that provides all three by separating encoding from reading: the context is encoded once wit
context compression - arxiv:2610.26836 · cs.AISame Simulation, Different Evidence: Where Retrieved Tool Output Costs LLM Agents AccuracyKe Zhang, Sahchit Chundur, Jerry Peprah Owusu, Mohammad Javad Abdolhosseini Qomi +1
LLM agents that compute the right answer can still answer wrongly. Agent harnesses increasingly keep long tool outputs out of the context window, saving them to files that the agent searches or reads on demand. We ask where this costs accuracy with a real scientific tool: PHREEQC, a widely used geochemistry simulator from the USGS, whose reports reach 846.8k characters and whose determinism lets every printed value be traced. Agents use it to answer 158 execution-verified multiple-choice items. When agents read the simulator's report through a section index rather than in context, GPT-5.1 and
agentllm agent - arxiv:2610.26834 · cs.LGKronecker Delta Rule for Long Context Linear AttentionAnamaria-Roberta Hartl, Levente Zólyomi, Michael List, Günter Klambauer +2
We propose KronLSTM, a linear attention architecture based on the Kronecker Delta Rule, a novel update rule for associative matrix memories. KronLSTM extends the xLSTM family with a memory-revision mechanism that corrects targeted associations while modifying the memory only along a single direction in matrix space. This minimally invasive memory revision reduces interference with previously stored information compared with delta-rule updates, making it particularly well suited for long-context language modeling. The resulting memory dynamics can be mathematically described as a Markov chain t
memorylong-contextlong contextbenchmark - arxiv:2610.26833 · cs.LGSupervised Learning with Quantum Hamiltonian EmbeddingsRiccardo Molteni, Minjun Jeon, Paolo Braccia, Abhinav Deshpande +7
Machine learning is widely regarded as a promising application in which quantum computing could offer a computational advantage. However, general quantum machine learning models with provable guarantees on performance, trainability, and efficient inference remain limited. In this work, we introduce quantum Hamiltonian embeddings, a supervised learning framework in which a classical training dataset is encoded into a single effective, parameterized Hamiltonian. Training optimizes the parameters of this data-dependent Hamiltonian, whose approximate ground state defines the classification rule. O
benchmark - arxiv:2610.26829 · cs.LGProteinGeoBench: Frontier Structural-Biology Agents Are Fluent at Lookup but Lost in GeometryMatt Turk, Brandon L. Neel, Diana M. Lopez, Zachary D. Smith +2
Frontier agents retrieve structural metadata well and compute geometry from coordinates less reliably. We ask how much of a benchmark score reflects that gap and how much reflects the evaluation configuration around the model. We study ProteinGeoBench, a suite of 131 protein-structure tasks, using 4,249 rollouts from four frontier models and 4,486 further trials across 26 models. Three of the four frontier models fall within 0.031 of one another (0.785 to 0.816) with no difference between them resolved, while the fourth scores 0.707 and all three gaps to it exclude zero. Three of the four do b
agentbenchmark - arxiv:2610.26828 · cs.LGAgent World Model in the Wild: Computer-Use Agent Environment SynthesisZhaoyang Wang, Qianhui Wu, Haorui Wang, Simon Yu +8
Agent learning is shifting from learning from human data toward learning through interaction with the world. However, scaling such learning directly in real-world computer systems is challenging and can introduce substantial security risks. We therefore propose Agent World Model in the Wild (WildAWM), a scalable pipeline for synthesizing controllable environments and tasks for training general computer-use agents (CUAs). Whereas existing approaches typically specialize in a single agent interaction interface, we construct executable environments that jointly support graphical user interfaces,
world modelagenttool usebenchmark - arxiv:2610.26827 · cs.CLKEEP-Bench: Evaluating Evidence Preservation in Context CompressionZiqing Man, Di Wu, Jia-Chen Gu, Junyi Zhang +2
Context compression improves long-context efficiency but may introduce faithfulness issues through an information bottleneck. Existing evaluations rely on QA metrics, but a reader can still answer correctly from an unfaithful compression, so these metrics underestimate such loss. To bridge this gap, we introduce KEEP-Bench, a diagnostic benchmark that directly assesses compression faithfulness with 4,102 cases across three challenging settings that require keeping evidence for plausible reference resolutions, explicit qualifiers, and minority perspectives. Experiments show that even with tailo
long-contextcontext compressionbenchmark - arxiv:2610.26826 · cs.CVRefCraft: Benchmarking Visual-Reference-Conditioned Agentic Generation through Web DevelopmentYuchi Wang, Rongyao Fang, Weihao Yu, Xilin Wei +6
Beyond basic tool use and coding skills, effectively incorporating visual references into agentic generation is equally essential for applying AI agents to real-world productivity tasks. We study this crucial capability through webpage development, a practical and readily evaluable setting in which existing benchmarks remain largely idealized: they typically focus on reconstructing visual targets or generating webpages from textual instructions alone, overlooking the more common workflow of developing from visual examples. Specifically, we introduce RefCraft, a benchmark comprising 300 realist
ai agentagentictool usebenchmark - arxiv:2610.26825 · cs.LGMemT-1.0: Towards Memory-Native EvolutionZhiyu Li, Zeyu Zhang, Linyun Liu, Yiwei Bu +70
Memory-native evolution requires making memory an intrinsic capability of foundation models, enabling them to continuously accumulate and utilize information. To this end, we introduce MemT-1.0, a new architecture of memory foundation models that internalize memory as part of model computation. Building on Metis, MemT-1.0 employs Gated DeltaNet-2 with a stable token-level memory update mechanism within each turn. We further introduce Mixture-of-Storage to scale native memory capacity. Following a deliberate backbone preparation, we develop a two-stage training pipeline consisting of continued
memoryretrieval-augmentedagenticpost-training - arxiv:2610.26819 · cs.CLASTOR: Benchmarking and Improving Autoformulation for Soft-Constraint Optimization with a Formal GrammarAmer Goel, Yeganeh Alimohammadi, Andrew Daw
Real-world decision problems combine two qualitatively different kinds of constraints: requirements that must be satisfied, called hard constraints, and preferences that can be violated, called soft constraints. Because these problems are often posed to decision-makers in natural language, large language models offer a natural tool for assisting with their solution. This has given rise to autoformulation, which focuses on using LLMs to translate natural-language descriptions of decision problems into executable mathematical optimization models. However, this research often neglects soft constr
multi-agentbenchmark - arxiv:2610.26817 · cs.LGNext Best Action: Causal Constrained Offline RL for Business Action TargetingSurya Ramachandiran, Yiming Li, Juan Mancilla-Caceres, Tao Ye +1
Next Best Action (NBA) systems must select among customer interventions (e.g., discounts, promotions, and lifecycle actions), maximize long-term economic value, satisfy complex operational constraints, and remain causally grounded. Existing approaches lack a rigorous framework for meeting these requirements jointly. We address this gap by formulating NBA as a Constrained Markov Decision Process (CMDP). We then integrate off-policy causal correction of the estimated customer dynamics directly into the Linear Programming (LP) formulation of the CMDP. The resulting policy selects actions based on
policy evaluation - arxiv:2610.26813 · cs.ROPICA-WAM: Visual Perturbation Invariance via Cross-Appearance Self-Alignment in World Action ModelsYude Zou, Yifei Yao, Guanbo Jiao, Jianming Ma +2
World Action Models (WAMs) jointly predict future video and generate actions, yet remain sensitive to task-irrelevant visual perturbations, i.e., changes in visual appearance that should not affect robot behavior. We identify a gap in conventional WAM training: video prediction encourages representations to preserve appearance information, while action generation does not explicitly enforce invariance to task-irrelevant appearance. Data augmentation exposes WAMs to such perturbations but does not require perturbed and nominal observations to produce consistent action representations, and align
liberorobotwin - arxiv:2610.26811 · cs.CVState World Models: Execute the Rules, Generate the PixelsChan Yan Ki, Yanran Zhang, Wenzhao Zheng, Xiaofeng Wang +4
Existing world models can generate action-conditioned video from a single image. Many focus on camera movement, allowing users to explore the generated scene. However, navigation alone does not ensure that an action follows persistent rules when its outcome depends on state that is invisible in the current frame. We present Stateful Harness, a framework that separates world evolution from video rendering to support stateful and rule-governed interaction. It maintains an evolving world state under specified rules, renders each resulting event with a frozen video model, and uses the generated cl
world modelaction-conditionedbenchmark - arxiv:2610.26809 · cs.CLSafety Mid-Training: Internalizing LLM Safety as a Foundational CapabilityZhengyue Zhao, Liwei Jiang, Yejin Choi, Chaowei Xiao1
Safety alignment for large language models (LLMs) is primarily performed during post-training, while recent work has also begun incorporating safety data into pre-training. Yet safety can degrade when later training no longer includes safety data, and gains in safety often come at the cost of general capability. This tension suggests that safety and general reasoning should not be treated as separate training goals. We instead view safety as an LLM capability that can be learned through a reasoning structure shared with general tasks. Both require the model to understand intent or task goals,
post-training - arxiv:2610.26805 · cs.CVWhat Language Adds, and at What Cost: Four Controlled Probes of Pathology Vision-Language ModelsWenhao Zhang, Zhongliang Zhou, Yi Ping, Zirui Wang +1
Vision--language models (VLMs) promise to make computational pathology queryable and able to draft reports, but how much language adds beyond the vision backbone is unclear, because almost every system is evaluated by its developers on test sets of their own choosing. We grade published evidence by how it was produced and test four claims with controlled probes that reuse released checkpoints and public data, varying one input at a time. Report text adds prognostic signal: adding a frozen embedding of the stage-redacted report to a slide embedding raises the pooled C-index by 0.019 for progres
agentic - arxiv:2610.26802 · cs.LGScaling Parameter-Efficient Fine-Tuning for Brain EncodingZhuofan Josh Ying, Guanglong Sun, Nolan Ho, Kieran Liu +5
Brain encoding models aim to predict neural responses to stimuli. The most widely used approach fits readouts from frozen pretrained backbones. Using neural data to train large models from scratch or fully fine-tune them after task-pretraining has proven difficult because neural data is costly and limited. We show that parameter-efficient fine-tuning (PEFT) enables us to effectively adapt very large pretrained backbones for brain encoding with a small number of trainable parameters. To make it computationally practical, we exploit the backbone's causal structure to predict multiple timepoints
leaderboard - arxiv:2610.26800 · cs.LGLearning to Refer: Inference-Time Collaboration in Decentralized Federated LearningYuheng Ma, Chendi Wang, Qiong Zhang
In decentralized federated learning, heterogeneous clients may develop complementary expertise, yet each query is typically evaluated only by the client where it arrives. To exploit this distributed expertise without global coordination, we formulate inference-time collaboration as budgeted local expert referral, a graph-constrained routing process in which a query can move only between neighboring clients and only visited clients contribute to its final prediction. EdgeDRM implements this framework by learning a density-based routing potential for each client: differences between neighboring
benchmark - arxiv:2610.26789 · cs.LGCooperative Multi-Agent Reinforcement Learning for Market Making on Prediction-Market Order BooksSkander Lejmi
A market maker provides liquidity by quoting prices to buy and sell an asset, earning the spread while managing inventory risk. We study market makers trained with multi-agent reinforcement learning on prediction markets, specifically Polymarket, whose thin order books pose distinct challenges. Three challenges underpin this research: adverse selection may cause a policy trained on real data to stop quoting; the consequences of a quote can arrive many steps after the action; and prediction-market L2 data does not identify individual orders needed for simulation. We address these challenges by
multi-agent - arxiv:2610.26787 · cs.AILG-Bench: A Graph-Structured Evaluation Benchmark for Life SciencesLu Sun, Xiangyi Zhang, Xiangyang Zhu, Kaiyuan Ji +5
Traditional evaluation benchmarks reduce inherently interconnected scientific knowledge in life sciences into flat lists of questions, disregarding the underlying topological structure of the knowledge. We introduce, the first graph-structured benchmark for life sciences, featuring over 10,000 high-quality multiple-choice questions across medicine, biology, and chemistry. Our approach constructs a weighted evaluation graph using bidirectional matching and semantic similarity algorithms, where nodes represent questions and edge weights capture their semantic relationships. Leveraging this graph
benchmark - arxiv:2610.26782 · cs.CLRESIN: Multi-Chain Reasoning with Entity-Structured IndexingXuanyu Meng, Jash Rajesh Parekh, Jiawei Han
Multi-chain reasoning is a challenging task not only because the multi-chain evidence is dispersed, but also because different parts of reasoning may require distinct, interacting chains of support. Flat retrieval can flood a model with loosely related text, whereas early selection of a single reasoning path can discard evidence that becomes useful only after other facts are considered. We introduce RESIN (Reasoning with Entity Structured Indexing), which uses structure to locate evidence while keeping the original source as the basis for reasoning. RESIN decomposes a question into focused one
benchmark - arxiv:2610.26781 · cs.AIGame of Phones: Harnessing Game Theory and LLMs in 'Bot Wars' to Counteract Phone ScamsNardine Basta, Robin Carpentier, Benjamin Zi Hao Zhao, Dali Kaafar
Phone scams impose substantial and growing costs on individuals and financial systems worldwide, yet existing countermeasures impose no resource cost on perpetrators. Strategic scam-baiting, in which automated agents engage scammers to deplete their operational capacity, offers a proactive alternative but requires dynamic strategy adaptation that static systems cannot provide. We introduce a game-theoretic framework enabling LLMs to learn effective counter-strategies through in-context experience without fine-tuning. The framework models scammer-baiter interactions as a two-player game with em
benchmark - arxiv:2610.26780 · cs.CVV-JEPA Hand: Simple, Accurate and Fast 3D Hand and Camera Trajectory Recovery from Egocentric VideosRuicheng Li, Qixiu Li, Yu Deng, Sicheng Xu +2
Egocentric video supports diverse applications, including robot learning, augmented reality, and human activity understanding. Recovering metric 3D hand trajectories from such video requires estimating both hand motion and camera egomotion to express the hands in a consistent reference frame. Many existing pipelines recover hand and camera motion in separate stages, combining hand detection, hand reconstruction, camera tracking, and trajectory refinement. We present V-JEPA Hand, a simple feed-forward video model that jointly reconstructs bilateral hand motion and camera motion from egocentric
v-jepa - arxiv:2610.26777 · cs.LGTopological Sheaves of Irreducibility: A Theoretical Framework for Multi-Agent Task ObstructionsSteven Motta, Davide Mambelli, Alessandro Rozza
Multi-agent systems can differ under identical agents and communication budgets when they support different forms of joint interaction. We introduce Topological Sheaves of Irreducibility (TSI), a sheaf-theoretic framework separating task-level compatibility obstructions from system capability. A task is represented by an interaction complex $K$ and a capacity sheaf $\mathcal{F}$; nonzero $H^n(K;\mathcal{F})$ defines a structural signature, while registered interventions test capability. For cocycle data, cochain realization is solvable exactly when the represented cohomology class vanishes. Ex
multi-agentagent system - arxiv:2610.26776 · cs.CLMonsoon: Scalable Spontaneous Speech Collection for Low-Resource ASR in 50 LanguagesShobhit Banga, Manas Dhir, Aaditya Pareek, Amritansh Walecha +7
For most languages, automatic speech recognition (ASR) is limited less by model architecture than by a lack of accurately transcribed spontaneous speech. We present Monsoon, a framework for collecting such data at scale. It elicits unscripted conversations between paired contributors, screens recordings with automated classifiers, corrects transcripts over five levels of human review, and selects segments for lexical and acoustic coverage. It has collected speech in 50 languages; we analyse seven 3,500-hour corpora. At equal volume, its quality stages reduce semantic word error rate (WER) by 1
benchmark - arxiv:2610.26770 · cs.LGHarnessPro: Recursive Harness Self-Improvement with Pareto Optimization for Performance-Cost Trade-offsZhuoshi Pan, ZhiYuan Feng, Yukuo Cen, Honglin Lin +5
Harness development plays a crucial role in building capable AI agents. Recent approaches improve harnesses through recursive self-improvement (RSI), but their performance gains often come with higher execution costs. We introduce HarnessPro, a label-free method for evolving harnesses to improve task performance while reducing execution cost. In this setting, candidate harnesses are evaluated using LLM-generated tests without access to ground-truth solutions or official tests. Specialized agents diagnose execution trajectories, plan updates with different performance-cost priorities, and imple
ai agentself-improvementbenchmark - arxiv:2610.26765 · cs.ROHandsOn: A Large-Scale Simulation Benchmark for Generalizable Dexterous ManipulationTianwei Luo, Yiming He, Hongjin Wu, Ting Pan +13
General-purpose manipulation requires reliable behavior across embodiments, objects, and task demands. We introduce HandsOn, a composable simulation platform and benchmark for cross-embodiment dexterous manipulation. It supports five robotic arms, six dexterous hands, and approximately 200 rigid, articulated, and deformable assets in randomized scenes. The benchmark pairs 18,000 curated human demonstrations with 66 human-verified tasks: Core-50 supports demonstration-based learning and execution tests, while Transfer-16 evaluates asset and relation transfer, skill composition, and deformable m
vision-language-actionmanipulationdexteroussim-to-realbenchmark - arxiv:2610.26762 · cs.RODexOAK: Object-Aware Kinematic RetargetingFederica Spinola, Robert Katzschmann, Cordelia Schmid
We introduce DexOAK, a learning-based object-aware kinematic retargeting framework for generating contact-consistent robot hand references from human motion-capture data. DexOAK first minimizes an interaction-graph Laplacian objective over a joint hand-object graph, so the robot reproduces the demonstrated object-relative interaction geometry rather than the human hand pose alone. These per-frame solutions form seed retargeted references, from which a feed-forward network conditioned on human keypoints and object point clouds learns the retargeting function. Object-aware finetuning then refine
dexterous - arxiv:2610.26759 · cs.ROTactile In-Hand Object Rotation: A Comparative Study of Hand Morphology in Simulation and HardwareWenhao He, Max Yang, Chenghua Lu, Yijiong Lin +2
Continuous in-hand rotation requires coordinated contact maintenance and renewal, linking manipulation performance to both hand geometry and tactile feedback. We investigate these relationships through rotational progress, finger participation, and fingertip loading on two versions of the low-cost LEAP Hand with anthropomorphic and symmetric finger arrangements. Both hands are integrated with TacTip vision-based tactile sensors and use a shared reinforcement learning framework. Simulation evaluates full tactile, binary touch, and proprioception-only observations on 12 unseen objects; hardware
manipulationtactiletactip - arxiv:2610.26757 · cs.ROSimulation Priors for Rapid Adaptation in Model-Based RL for Autonomous RacingAdrian Remonda, Jiajun Xi, Nicklas Hansen, Marcus Greiff +4
Reinforcement learning enables learning high-performance control policies, but its reliance on extensive interaction limits deployment on physical systems where data collection is costly, and failures are unsafe. We introduce Prior-Informed Adaptation (PIA), an offline-to-online model-based RL framework that uses structured simulation priors for rapid adaptation. PIA pretrains a world model on heterogeneous low- and high-fidelity simulated trajectories, then adapts it with limited target-domain interaction using uncertainty-regularized planning. Rather than requiring an accurate simulator for
world model - arxiv:2610.26754 · cs.AITokenGate: Adapting LLMs for Fast and Memory-Efficient Prefill with Learned Token PruningLiana Patel, Rishabh Ranjan, Ion Stoica, Matei Zaharia +1
Expert human readers process large inputs (e.g., code repositories and long documents) by skimming to closely read only what matters. On the other hand, language models process all input tokens through each model layer, resulting in significant processing costs for long-context inputs. We introduce TokenGate, a small, drop-in component that extends existing trained models to enable end-to-end learnable token pruning over long-context inputs for fast and memory-efficient prefill. To enable end-to-end learning despite the non-differentiable top-K operator for token pruning, we introduce token-ga
memorylong-context - arxiv:2610.26751 · cs.CVBeyond Omni: A Unified MLLM for Embodied Sensory UnderstandingShiyun Xiong, Dongming Wu, Peiwen Sun, Bokang Yang +4
Recent embodied systems increasingly build on multimodal large language models (MLLMs) for their pretrained understanding and reasoning capabilities, but their perception of the physical world remains largely limited to basic modalities such as vision and audio. However, broadening MLLMs to more diverse sensory modalities from the physical environment is challenging: their representations are often weakly connected to the MLLM and vary substantially in structure and statistics, while their interpretation still requires shared high-level understanding and reasoning. We therefore propose Unisens
embodiedtactilebenchmark - arxiv:2610.26749 · cs.LGTrajectory Replay: Continuous-Time Mixture Transport for Lifelong Generative ModelsAlessandro Palma, Charles Jones, Tristan Cinquin, Fabian J. Theis +1
Adapting pre-trained continuous-time generative models to sequentially incoming data without forgetting earlier tasks is a central challenge in lifelong learning. A leading strategy, generative replay, uses a generative model trained on historical tasks to augment its own continual update with synthetic samples from experience. However, it requires full trajectory simulations at every training step, which accumulate discretization errors and can degrade the replayed historical distribution. Here, we propose trajectory replay, which bypasses sample generation by training an updated velocity fie
lifelong learning - arxiv:2610.26740 · cs.CVCheckDistill: Adaptive Checklist Distillation for Efficient Distributional Reward Modeling in Image EditingYaqi Cai, Zuan Gao, Mengting Chen, Wei Li +1
Reinforcement learning for instruction-guided image editing requires reward models that provide reliable, accurate, and efficient feedback. Direct scalar prediction compresses multidimensional editing preferences, whereas generative evaluators can expose their judgment through explicit reasoning. However, broad scoring criteria do not specify all instance-dependent editing and preservation constraints, and generating long reasoning trajectories is costly for online optimization. We introduce CheckDistill, a pointwise reward modeling framework that internalizes instruction-adaptive checklists.
evaluator - arxiv:2610.26734 · cs.LGCAVA: Complementary-Label-Aware Data ValuationTan-Ha Mai, Hsuan-Tien Lin, Vu Nguyen
Complementary-label learning (CLL) trains a classifier from labels that name a class an instance does not belong to. Such labels are cheap to collect from crowdworkers or vision-language models (VLMs), making them attractive for large-scale annotation. However, both sources can mistakenly assign the true class as the complementary label. Such annotation errors are particularly harmful in CLL because the model is explicitly trained to suppress the correct answer. Optimal transport (OT) data valuation identifies harmful samples by comparing training data with a clean validation set, but transfer
benchmark - arxiv:2610.26730 · cs.CLEpiLens: Lightweight RAG for Epidemiological Knowledge AcquisitionVincenzo Perri, Ciro Cattuto, Daniela Paolotti
To prepare for future outbreaks, researchers need to know what studies have been done and what data they use. Mapping a large literature by hand is slow, while large language models alone can misstate or conflate sources. We present EpiLens, a lightweight, open-source retrieval-augmented generation package that applies a fixed retrieval template and output schema to every paper in a corpus and links outputs to source passages. We build a corpus of 2,297 papers from the sources of a multi-pathogen parameter study and evaluate EpiLens's paper-classification workflow against 214 manually labelled
retrieval-augmentedrag - arxiv:2610.26724 · cs.LGInjection of Space Radiation Errors on Low-power NVM using Generative Deep LearningEleonora Giudici, Antonio Miele, Luca Mottola
We develop an error model for a Non-Volatile Memory (NVM) device that represents the effects of physical faults experienced in a CubeSat mission in Low Earth Orbit (LEO). The CubeSat carries an off-the-shelf low-power Ferroelectric RAM (FRAM) chip and runs repeated read/write tests according to a predefined pattern. We compare every memory snapshot received from the satellite with a reference bitmap to identify bit errors. We visually analyze erroneous values distributions in the memory snapshots and we understand that they are frequently distributed according to regular spatial patterns. We d
memory - arxiv:2610.26717 · cs.RORobot Learning with Vision-Language-Action Models for Agricultural ManipulationHamid Hafezi, Javad Mohammadpour Velni, Mohammadreza Davoodi
Robotic manipulation in agricultural environments remains challenging due to variations in plant geometry, occlusion, and camera viewpoints. Pretrained vision-language-action (VLA) and generalist robot policies offer a promising approach to addressing these challenges by learning manipulation behaviors from multimodal observations and language instructions. However, adapting these policies to agricultural tasks requires domain-specific demonstrations, which are costly and time-consuming to collect on physical robots. In this work, we develop a simulation-to-real robot-learning framework for ag
vision-language-actionmanipulationteleoperationmanipulatorfrankagrasp - arxiv:2610.26714 · cs.CLHoloCosm: Beyond Single-Metric Benchmark Compression Across Multiple MetricsYupei Lin, Zhongzhan Huang, Junxin Li, Guoming Ling +4
Benchmark compression reduces evaluation cost by selecting a small set of samples. Existing methods typically guide selection using a single metric, although benchmarks often assess models along multiple dimensions. We find that model rankings are often weakly correlated across metrics, indicating that different metrics capture distinct aspects of model performance. Samples selected for one metric do not reliably preserve scores and rankings on others when used for compressed evaluation. At 80\% compression, each of the five single-metric baselines yields higher reconstruction error than equal
agentagent benchmarkbenchmark - arxiv:2610.26711 · cs.CVPVBench: Quantifying Physical Plausibility in Video Generation with Object-Level Physical MeasurementsKaiwei Chen, Moran Li, Zhenye Gan, Yabiao Wang +4
Recent video generation models produce visually compelling outputs, yet their physical plausibility remains poorly measured. Physical plausibility asks whether generated objects obey geometry, kinematics, dynamics, and conservation laws. Existing benchmarks rely on coarse visual-language model judgments or target embodied world models rather than video generation itself. We introduce PVBench, a physical video benchmark that quantitatively measures the plausibility of generated videos across 46 dimensions organized into seven domains, spanning: i) instruction and semantic fidelity, ii) behavior
embodiedhumanoidquadrupedworld modelmulti-agentembodied agent - arxiv:2610.26693 · cs.LGParallel Language Models via Variational DecodingFelix Draxler, Robin C. Geyer, Daniel Geyfman, Farrin Marouf Sofian +12
Autoregressive language models emit one token per forward pass. This causes high inference latency, in particular as models are scaled up for quality. In this paper, we introduce Parallel Language Models (PLMs), which generate mutually consistent tokens in one forward pass by routing dependencies through latent auxiliary variables. PLMs are trained using a variational bound that reduces to an autoregressive loss plus a parallel decoding loss. Empirically, PLMs are on a faster and better Pareto frontier than autoregressive and discrete diffusion baselines. Pretrained PLMs achieve near-autoregre
benchmark - arxiv:2610.26682 · cs.CVBrushOff: Stroke-Based Re-rendering as a Watermark Removal Attack on PaintingHao Chen, Junnan Xu
Invisible watermarks are used to mark who owns a painting. We show that a simple attack removes the most robust one we test. The attack redraws the watermarked painting with brush strokes, using a pretrained stroke-based renderer. We call this attack BrushOff. It needs no training, no knowledge of the watermark, and no generative model. We test four watermarks on 500 paintings from five styles. VINE, a recent watermark built to survive image editing, survives all 14 conventional distortions we test, with at least 97.8% detection. Repainting lowers its detection to 1.8%, while the output keeps
benchmark - arxiv:2610.26680 · cs.RORobotic Physical Comparison of Biologically Motivated and Counterfactual Sensory Arrangements using Biomimetic SkinAkihiro Miki, Shun Hasegawa, Yoshimoto Ribayashi, Yuta Sahara +2
Human skin contains diverse mechanoreceptors, including Merkel cells and Pacinian corpuscles, each localized in specific regions. Although their spatial organization is known to be shaped by molecular signaling and neural induction during development, experimentally examining its functional significance under controlled conditions in living systems remains challenging. In particular, physically comparing alternative receptor arrangements that do not occur biologically is difficult. In this study, we developed a layered biomimetic robotic skin designed to mimic key structural features of human
tactile - arxiv:2610.26679 · cs.AIStepBench: Diagnosing Real-World Procedural Understanding in AI AgentsHossein A. Rahmani, Bin Wu, Chaeeun Lee, Fanghua Ye +5
Evaluating an agent's planning ability remains challenging for open-domain tasks, where verifying procedures through execution is often impractical. Existing evaluations typically assess only the final plan and do not distinguish procedural errors such as missing, incorrect, or misordered actions. We introduce STEPBENCH, a benchmark of 6,885 verified procedures crawled from wikiHow across 81 open-domain categories. From each procedure, we derive 5 diagnostic tasks: four with deterministic answers involving withheld, incorrect, or shuffled steps, and one requiring next-step generation. We addit
ai agentbenchmark - arxiv:2610.26676 · cs.LGWhich Model for Which Role? Cost-Efficient Pareto Recommendation for LLM Agentic WorkflowsQian Xie, Wenyue Hua, Yueli He, Sripad Karne +6
LLM agentic workflows increasingly employ different models for specialized roles. Users may prefer different trade-offs between task quality and deployment cost, sometimes also considering latency. Finding good assignments incurs a separate search cost from evaluating multiple candidates. We introduce Cost-Coupled Gittins (CC-Gittins), a multi-objective extension of single-objective cost-aware Gittins search for Pareto recommendation. It exploits the coupling between evaluation and deployment costs: assignments that cost more to deploy also cost more to evaluate. The search alternates between
llm agentagenticbenchmark - arxiv:2610.26674 · physics.opticsTerabit-scale wavelength-multiplexed transmitter enabled by thin-film lithium tantalateShivaprasad U. Hulyal, Dengyang Fang, Jiachen Cai, Xinru Ji +4
Artificial-intelligence systems face a growing data-movement bottleneck and require higher interconnect bandwidth without a proportional increase in electrical power consumption or the number of laser sources and fibre connections. Moving optical interfaces closer to computing chips reduces electrical transmission loss, while demanding optical engines that combine high channel capacity, broadband modulation and simple packaging. Although wavelength multiplexing relieves pressure on the fibre budget, integrating shared multi-wavelength sources with dense wavelength routing and broadband modulat
co-packagedoptical interconnect - arxiv:2610.26671 · cs.LGPACT: Plan-Augmented Coordination via Trace-Graphs for Dynamic Multi-Agent Combinatorial OptimizationXiao Zhou, Alex Schutz, Victor-Alexandru Darvariu, Nick Hawes
Multi-agent approaches have been investigated recently as a means of improving the scalability of reinforcement learning for neural combinatorial optimization. Existing methods assume that all problem instance information is known upfront. However, real-world problems are often dynamic: relevant information is progressively revealed as the agents take actions (e.g., arriving at an edge in a routing problem only to discover it is untraversable). In such cases, existing methods fail to coordinate agents to explore and exploit the progressively revealed information. To address the arising coordin
multi-agent - arxiv:2610.26664 · cs.AIHoneypots for AI Agent Insider ThreatsJan Wehner, Jam Kraprayoon
AI agents deployed inside organizations can pose an insider threat if they make mistakes, are compromised by an attacker, or pursue misaligned objectives. We propose honeypots as an AI control method to improve the security of internal agent deployments. Rather than only using honeypots for pre-deployment evaluations, defenders should plant decoy opportunities for harmful actions throughout the production environment to detect misbehavior, incriminate malicious agents, and deter attacks. Such honeypots can elicit misbehavior under real deployment conditions, identify malicious agents even if c
agentai agent - arxiv:2610.26663 · cs.RORoboValue: A Fine-Grained Sim-and-Real Benchmark for Unified Evaluation of Robotic Value ModelsShengbang Liu, Zhengye Du, Zhilong Wan, Honghao Su +15
Robotic value models assess task execution from visual observations and instructions, providing feedback for policy learning and execution monitoring. However, existing evaluations cover only selected aspects of execution understanding, leaving the reliability of value feedback insufficiently tested. We introduce RoboValue, a fine-grained sim-and-real benchmark for unified evaluation of robotic value models. Its framework evaluates four complementary capability dimensions through shared interfaces: task-state understanding, temporal progress monitoring, failure and recovery reasoning, and valu
manipulationbenchmarkleaderboard - arxiv:2610.26662 · cs.AIAllPath: Layered Verification for a Deployed, Self-Revising LLM Trading Agent under Human-Gated and Autonomous RegimesTian Zhang, Meng Li
LLM trading agents are evaluated almost exclusively in backtests and autonomous benchmark harnesses. Deployment, where actions are irreversible, raises a different question: who verifies the agent's actions, and which failures can each verifier structurally catch? We report on AllPath, a personal LLM trading agent deployed against a live paper-brokerage account. The system watches positions through a deterministic rules engine, and each night it reviews its own day and proposes revisions to its own trading strategies. Verification is split across three layers. Code verifies what must not depen
agentbenchmark - arxiv:2610.26660 · cs.AITandem: Serving Hybrid Linear-Attention GUI Agents on Mobile NPUsZhuomin Chen, Jingchao Ni, Xu Zheng, Yong Chen +3
Smartphone GUI agents complete tasks in mobile apps by repeatedly reading the screen and asking a model for the next action. Mobile neural processing units (NPUs) enable local execution, keeping screen contents on the phone, but model latency accumulates with every action. Hybrid linear attention models reduce memory requirements, yet reusing their recurrent state requires checkpoints at the appropriate input positions. Agent requests retain instructions but rewrite recent history, so checkpoint policies designed for chat continuation can miss reusable prefixes. Retaining rollback states for s
memoryagent - arxiv:2610.26659 · cs.CVSpatialScout: Can VLM Agents Actively Explore and Map 3D Scenes?Zhaoxi Zhang, Hao Liang, Linxuan Wu, Yiming Wu +6
Spatial intelligence requires agents to actively acquire visual evidence and integrate observations into a coherent representation of the three-dimensional world. Yet spatial question answering offers limited evidence of whether models can construct accurate metric maps through autonomous exploration. We introduce SpatialScout, a benchmark for evaluating vision-language model (VLM) agents on active object-level 3D mapping. SpatialScout comprises 80 diverse indoor scenes containing 3,140 target objects. Given an object inventory, RGB observations, camera calibration, and camera poses, agents co
benchmark - arxiv:2610.26655 · cs.LGThe Development of Sycophancy in Language Models: Tracing Vulnerabilities from Pretraining to Post-TrainingHaonan Huang, Chengkun Cai
Language models can change a committed answer to accommodate a user's opinion or pushback without new evidence. However, when this behaviour develops during training and what a released model inherits from earlier stages remain unclear. We follow 741 checkpoints of 30 open model lineages, from early pretraining to released models. We plant initial verdicts on moral scenarios and measure how far a model follows a user's stated opinion or abandons its verdict under pressure. In most lineages the behaviour is present before post-training, and its average strength changes in either direction, at d
post-training - arxiv:2610.26650 · cs.AIFrom Outcomes to Components: A Counterfactual Evaluation Framework for LLM-Proposed RepairsHengyu Ye, Songhua Hu
An LLM-generated repair action specifies which records to change and by how much. A failed action may still contain a useful target set or numerical update, depending on how the two are paired. We evaluate this problem in traffic networks and databases, where aggregate observations obscure the records that need correction. We examine the pairings by executing each combination from the same initial state, using a reference action that restores the affected records. In 76 of 114 traffic actions, the gain from correcting both parts exceeds the sum of the gains from correcting each part alone. Six
evaluation framework - arxiv:2610.26645 · cs.AIReasoning-Aware Multicalibrated Reliability Predictors for Language ModelsEkaterina Fedorova, Alessandro Morosini, Chara Podimata
As people increasingly rely on large language models (LLMs) for information, model providers need to communicate meaningful estimates of whether individual responses are correct or, more generally, reliable. In the context of hallucinations, existing work uses multicalibration to make such reliability predictions meaningful across application-relevant groups, such as prompt topics. Topic groups alone, however, can conceal systematic errors: within the same topic and confidence level, responses with different chain-of-thought (CoT) properties may exhibit opposing errors that cancel on average.
benchmark - arxiv:2610.26644 · cs.CVDual-Scale Condition-Shift Editing for Robust Real-World Image DemoiréingShiyang Zhou, Haijin Zeng, Congpei Qiu, Zhengcen Li +5
Robust demoiréing of screen-recaptured images remains challenging: moiré varies in frequency, color, and spatial support across display-camera pairs and capture conditions. Dedicated restoration networks trained on finite paired data struggle with this diversity, especially when low-frequency color banding and high-frequency stripes co-occur. We present DemoiréEdit, a content-preserving editing framework that adapts the natural-image prior of a pretrained flow-matching editor to demoiréing. First, an endpoint-supervised training objective supplements velocity matching with pixel, perceptual, a
benchmark - arxiv:2610.26636 · cs.AIRevision-Based Learning from Interactive Experience and FeedbackJiawen Li, Chenming Ge, Alice Gao, Rishab Vempati +2
Code generation models are increasingly used to produce interactive artifacts, such as websites and user interfaces. Many properties that determine whether these interfaces are useful only become apparent through realistic interaction, making them difficult to incorporate into model training. In this paper, we introduce RELIEF (REvision-based Learning from Interactive Experience and Feedback), a method for improving code generation models using feedback on their generated interfaces. RELIEF uses a coding agent to revise these interfaces, then incorporates the improvements into their original g
agent - arxiv:2610.26632 · cs.LGWhen the harness decides success: state that cannot be re-observedRaimondas Lencevicius
Agent harnesses decide what a language model sees across turns: they drop old turns, summarize them, or replay the model's own earlier reasoning. We ask when these choices change whether an agent succeeds, not only what it costs, and report three findings. First, in a pre-registered synthetic probe, truncating the context window costs a low-cost reasoning model nothing in success when a needed fact can be looked up again (it looks the fact up instead), but drops its success to chance when the fact cannot be re-observed: the two penalties differ by 92.5 percentage points (90% CI [89.5, 95.5]).
agentbenchmark - arxiv:2610.26625 · cs.CVEgoXperience: Learning World-Grounded Egocentric Representations from Human ExperienceXinhao Ji, Hui En Pang, Hao Li, Kairui Hu +11
Reconstructing human activity from egocentric video requires more than recovering individual 3D components: the observer's motion, surrounding geometry, and hand-object interactions must fit together in a consistent world. Achieving this coherence requires both joint reconstruction and aligned supervision, yet existing methods and datasets offer limited support for either. We present EgoXperience, a full-stack paradigm that addresses these needs through data collection, automatic annotation, unified modeling, and benchmarking. Central to EgoXperience is an automatic annotation engine, EgoX-Eng
manipulationbenchmark - arxiv:2610.26595 · cs.LGConcerns in Benchmarking Jailbreaks: An Empirical Study of Attack Success RatesLeon Tu Nguyen, Rohan Menon, Rafael Tappe Maestro, Eric Aubinais +1
Jailbreak methods are commonly compared using Attack Success Rate (ASR), but these comparisons depend on how responses are generated and judged. In this paper, we examine eight attacks against three open-weight language models using three evaluation procedures. Low-resource language attacks achieve a mean ASR of 96.0% under keyword matching but 2.7% under a Qwen3 evaluator following StrongREJECT scoring instructions. Changing the number and arrangement of attempts reverses the ranking of DeepInception and rewriting. In a separate experiment, sampling 100 responses for each unchanged request at
benchmarkevaluator - arxiv:2610.26586 · cs.CVScaling Vision Foundation Models for Toxicologic PathologyJessica Loo, Theodoros Pissas, Byunghak Kang, Vincent A. Ella +6
Here, we present TOX-FM, a family of non-clinical PFMs trained on large-scale diverse histopathology datasets of over one million whole slide images consisting of over 60 organs across six animal species from over 14,000 toxicology studies across six decades. We introduce a comprehensive benchmark of 18 non-clinical downstream tasks, incorporate an established benchmark of 9 clinical downstream tasks, and evaluate TOX-FM against state-of-the-art clinical PFMs across both non-clinical and clinical domains. Notably, we explore open questions surrounding domain-specific training and cross-domain
benchmark - arxiv:2610.26584 · cs.ROHAF-VLA: Hierarchical Action Flow Framework for Humanoid Vision-Language-Action ControlLangzhe Gu, Chengkai Hou, Meng Li, Xinhua Wang +11
Humanoid robots hold great promise as general-purpose agents in human-centered environments, yet adapting generalist vision-language-action (VLA) policies to real-world humanoid loco-manipulation remains challenging. Unlike fixed-base or wheeled manipulators, humanoids must coordinate locomotion, posture, gaze, and dual-arm manipulation over a high-dimensional whole-body action space, where inaccurate base or waist motions can easily induce unstable and aggressive upper-body compensation. We propose HAF-VLA, a hierarchical action-flow framework for humanoid VLA control. Built upon a pretrained
vision-language-actionvlamanipulationhumanoidmanipulator - arxiv:2610.26583 · cs.CVDenseHap: Dense Hand Haptic Estimation from Monocular ImagesJunhan Xue, Rolandos Alexandros Potamias
As dexterous robotic manipulation advances, scalable annotations of human hand-object interactions are becoming increasingly important. Furthermore, previous hand contact estimation methods suffered from issues such as dorsal-side misclassifications and insufficient precision, while lacking a unified framework for integrating multiple hand to environment interaction modalities, including distance fields, contact thresholds, and pressure. To address that, we present DenseHap, a unified end-to-end approach to estimate hand distance field recognition, comprehensive hand contact detection, hand pr
manipulationdexterous - arxiv:2610.26582 · cs.LGWhatIf-Bench: Benchmarking Intervention-Aware Time Series ForecastingArjun Ashok, Mihir Parmar, Palash Goyal, Bhavana Dalvi Mishra +5
Recent time series foundation models (TSFMs) can consume covariates in-context, extending zero-shot forecasting to problems driven by external factors. This enables scenario analysis: forecasting how a target responds to events that set covariates to chosen future values. Whether they can do this reliably is an open question. We introduce WhatIf-Bench, a benchmark measuring how well forecasters use covariates to forecast a target's response to future events. WhatIf-Bench is semi-synthetic: time series come from mechanistic simulators of data centers, power grids, road traffic, and water networ
benchmark - arxiv:2610.26578 · physics.opticsSingle-Mode Laser Arrays with Size-Invariant Loss DiscriminationWenwen Liu, Shuang Zhang
In conventional coupled oscillator lattices and laser arrays, expanding the active volume inevitably causes spectral crowding and degrades spatial uniformity, posing a long-standing barrier to high-power single-mode operation. We uncover a dissipative selection mechanism in sparse star networks that decouples modal selectivity from lattice size constraints. By engineering dissipative auxiliary couplings between a central hub and N peripheral resonators, the system establishes an exact star-Laplacian symmetry. This enforces a target eigenstate that is immune to all lossy channels while maintain
microring - arxiv:2610.26564 · cs.CVPrismWAM: Decoupling State Representations in World Action ModelsTianyu Xin, Wanqi Zhong, Bowen Hu, Zhangchi Guo +7
World Action Models (WAMs) typically represent the world through latent embeddings from pretrained video models, which entangle heterogeneous aspects of the world, from how it looks (e.g., texture and color) to how it is physically configured (e.g., occupancy and pose). This raises an underexplored question: $\textbf{what should constitute the state of a WAM?}$ We argue that the world state should separate how the world appears from how the world evolves under action: the former captures semantic context, while the latter is governed by physical state. To this end, we propose $\textbf{PrismWAM
embodiedmanipulationtactileworld model - arxiv:2610.26554 · cs.AIBetaMem: Learning to Remember a Changing World for Embodied AgentsMai Liu, He Sun, Siru Zhong, Kerun Chen +4
Long-horizon embodied agents face temporal confusion and incomplete or distorted records, recurring errors from fixed memory rules, and growing storage and retrieval costs. We present BetaMem, a framework for building and evolving traceable long-term memory through six linked layers: working, spatial, entity, episodic, experiential, and procedural memory. A dual-level write contract constrains operation executability and source support for content: supervised fine-tuning learns compliant memory operations, while reinforcement learning combines repeated sampling with teacher verification of the
embodiedmemoryembodied agentself-improvementevaluation protocol - arxiv:2610.26550 · cs.ROLearning Wheel Motion Uncertainty for Degeneracy-Aware LiDAR-Inertial Odometry on Quadrupedal Wheel-Legged RobotsHairuo Wei, Yunfan Ren, Renda Yang, Qiangsheng Huang +1
Wheel kinematics can support odometry for quadrupedal wheel-legged robots when LiDAR geometry provides weak motion constraints. However, slip and contact loss make rolling predictions unreliable, and contact state alone cannot determine their reliability across motion directions. We propose a tightly coupled LiDAR-inertial-kinematic odometry framework that retains analytical rolling predictions and learns per-wheel, direction-dependent process covariances from proprioceptive history. The estimator maintains moving wheel-center states, allowing the learned uncertainty to regulate the influence
quadruped - arxiv:2610.26548 · cs.AIAgentic Orchestration of EDC Data Space Retrieval and Natural Language Driven Data ProcessingMarla Grunewald, Fin Gentzen, Iulisloi Zacarias, Admela Jukan
Participating in a data space still requires the Dataspace Protocol (DSP) negotiation flow to be hand-coded against each provider's access policy. We present a multi-agent architecture that automates this process using natural language requests. Built on LangGraph over an Eclipse Dataspace Components (EDC), the design separates concerns into cooperating agents. The orchestrator's role is to route and plan. The data extraction agent's role is to expose the DSP protocol's three composable tools (catalog discovery, contract negotiation and download). The agent also gates negotiations on the consu
agentmulti-agentagenticbenchmark - arxiv:2610.26545 · cs.ROReInsVLN: Revisiting Historical Evidence for Training-Free Vision-and-Language NavigationSiwei Cheng, Hao Wu, Zitong Qi, Xinming Zhang +6
Vision-and-language navigation (VLN) relies on retaining and using relevant observations. Yet growing histories introduce redundancy, and retained information may remain underused. We introduce ReInsBench, a diagnostic benchmark that examines history utilization through controlled continuation trials. Its results show that history-review reinstruction can recover navigation failures using the existing task context. Motivated by this finding, we propose ReInsVLN, a training-free framework supporting memory construction and utilization in frozen navigators. Coverage- and Instruction-Aware Compre
humanoidquadrupedmemorybenchmark - arxiv:2610.26531 · cs.CLReAnchor: Reusing Rejected Verifier States for Diffusion DraftingSanghyun Lee, Wonjin Nam, Seungho Jang, Jaeho Lee +3
Block diffusion drafters accelerate speculative decoding by generating multiple draft tokens in a single forward pass. However, each pass produces only a fixed-length block from a single target-generated anchor, limiting how far the draft can extend. Meanwhile, target verification computes hidden states even at positions beyond the first rejection, but these states are typically discarded rather than reused. We propose ReAnchor, a training-free method that reanchors a pretrained drafter at shifted positions using discarded verifier states to generate parallel draft blocks. These blocks improve
benchmark - arxiv:2610.26528 · cs.LGA Semantic Knowledge Graph Construction Pipeline for Vehicle Intention Prediction in nuScenesKhaled Nowara, Mohamed Manzour, Catherine Elias
Knowledge graphs (KGs) are increasingly used in autonomous driving (AD) to make traffic scene entities and relations explicit for prediction and reasoning. Existing traffic-scene KGs, however, primarily emphasize trajectory forecasting or general scene understanding. This paper presents the nuScenes Intention Knowledge Graph (NIKG), a focused Web Ontology Language (OWL)-based semantic representation and construction pipeline for ego-centric vehicle intention context. NIKG instantiates nuScenes scenes as a self-describing Turtle graph and promotes three intention-relevant relations to first-cla
knowledge graph - arxiv:2610.26525 · cs.LGA Fisher-Geometric View of Multi-Teacher On-Policy DistillationLiu Kaiyuan, Zhuang Ziyuan, Ma Haoxuan, Gu Yannian +3
Consolidating specialized foundation models into a unified generalist is a central goal of post-training. Two approaches are in common use: model merging, which is fast but treated as a heuristic, and multi-teacher on-policy distillation, which matches distributions but demands heavy rollout compute and suffers from cross-task gradient conflict. We show that the two are one objective seen at different orders. Within a shared basin of homologous specialists, the multi-teacher on-policy distillation objective is, to second order, exactly the Fisher-merging objective. Its minimizer is therefore t
post-training - arxiv:2610.26516 · cs.AIToward Structured Financial Foundation ModelsNatraj Raman, Saumitra Mishra, Vamsi Potluru, Sumitra Ganesh
Foundation models promise reusable capabilities and step-change performance gains across tasks. In finance, large-scale pretraining of domain-specialized foundation models can advance both goals. However, prevailing architectures and objectives do not by themselves address the combined demands of financial systems: partial observability, interacting entities, asynchronous events, changing regimes, action-dependent outcomes, and rare states with disproportionate consequences. Moreover, independently developed specialists may learn incompatible representations, while multi-domain monolithic alte
action-conditioned - arxiv:2610.26511 · cs.LGRouterWorld: Scaling Conversational AI with Latency-Aware LLM RoutingJinyang Li, Denis Emelin, Simone Tedeschi, Weiwei Cheng +5
Conversational AI systems built on Large Language Models (LLMs) require routing mechanisms that balance answer quality and response latency at scale. However, existing LLM routing benchmarks assume full observability of all candidate model outcomes, under-emphasize latency as a first-class metric, and fragment evaluation across heterogeneous protocols misaligned with real-world settings. We introduce RouterWorld, a comprehensive benchmark for LLM routing in conversational GenAI. RouterWorld unifies 371k single-turn responses and 76k multi-turn agentic trajectories, adopts natural user interact
agenticbenchmark - arxiv:2610.26507 · cs.ROLearn the Task, Not the Schedule: Reducing Spurious Cross-Arm Coupling in Bimanual Imitation LearningShuyuan Hu, Deng-Ke Wu, Yifan Jin, Qianhang Luo +3
A bimanual demonstration specifies both what the robot should do and one particular schedule for when each arm should act. Standard imitation learning does not distinguish task-required cross-arm constraints from incidental scheduling choices, and may therefore exploit opposite-arm progress as a shortcut cue, creating spurious cross-arm coupling even where relative timing is irrelevant for the current task phase. We introduce Constraint-preserving Temporal Recombination (CTR), a training-time data intervention that preserves recorded arm-local sensorimotor content while retiming the correspond
diffusion policyaction chunking - arxiv:2610.26503 · cs.AIScopeMem: Latent Memory That Learns Where to Revise Frozen GUI AgentsZhicheng Zheng, Cheng Zhang, Shengping Yu, Tianxiang Zhao
Latent memory is gaining ground as a lightweight history supplement for GUI agents. It supplies history to a frozen vision-language model via prefix vectors trained with step-wise cross-entropy (CE). While resulting gains are naturally attributed to encoded history, our controlled experiments show they primarily stem from action revision. The prefix learns to alter the model's default actions. However, this revision is unscoped because step-wise CE pushes outputs toward reference actions at every step without explicitly preserving correct decisions. Consequently, it repairs errors but also bre
memory - arxiv:2610.26496 · cs.AIMitigating Reward Hacking in Preference Reward Models with Evolving PenaltiesAfonso Baldo, António Farinhas, Marcos Treviso, Ricardo Rei +1
Reinforcement learning for language models has moved from preference reward models to verifiable rewards, which sustain much longer optimization. Preference models are prone to reward hacking: the policy raises their score without improving response quality, or even while degrading it. Yet for open-ended tasks, where quality can neither be verified nor fully written down as rubrics, they remain valuable. We introduce Dynalize, which mitigates reward hacking online, during policy training. Periodically and asynchronously, a language model analyzes training rollouts and writes penalties for unde
benchmark - arxiv:2610.26487 · cs.ROSPAR: Structural Pruning and Action Recovery for Vision-Language-Action ModelsShenli Zheng, Zhou Fang, Yuan Du, Yijiang Liu +1
The limited onboard computing resources of robots make the inference demands of vision-language-action (VLA) models costly to accommodate. Reducing these demands through aggressive compression, however, can degrade manipulation performance beyond what supervised fine-tuning alone recovers. We present SPAR (Structural Pruning and Action Recovery), which simultaneously prunes the hidden and feed-forward intermediate dimensions of the large language model (LLM) backbone through pruning-aware tuning, then applies reinforcement learning (RL) to recover policy performance. Applied to $π_{0.5}$, SPAR
vision-language-actionmanipulationlibero - arxiv:2610.26475 · cs.CV3D Diffusion Models for Medical ImagingFrank-Peter Schilling, Daniel Barco, Philipp Denzel, Marc Stadelmann
Generative deep learning methods such as Variational Autoencoders, Generative Adversarial Networks or Diffusion Models have achieved remarkable results in image synthesis. In particular, Diffusion Models have demonstrated superior image fidelity and sample diversity compared to other approaches on standard benchmarks, while also offering more flexible training dynamics as evidenced both in natural image domains and in medical applications. In medical imaging including Computed Tomography, Magnetic Resonance Imaging or Positron Emission Tomography, such models have various important use-cases i
benchmark - arxiv:2610.26473 · cs.LGA Simple and Unified Method for Gradient-Variation RegretHang Yu, Peng Zhao
Gradient-variation online learning has attracted increasing attention for its connections to game theory and optimization. It has been studied in both non-stationary online learning, which aims to minimize dynamic regret against time-varying comparator sequences, and universal online learning, which seeks to adapt to unknown loss curvature. Prior work obtained gradient-variation bounds using different techniques tailored to each setting. We propose a simple and unified method based on surrogate losses with a quadratic penalty on decision movement, yielding gradient-variation bounds in both set
online learning - arxiv:2610.26469 · cs.LGVeriCUA: Learning Proactive Verification for Reliable GUI Agent RewardsZhizhou Wang, Yaxi Lu, Zhong Zhang, Zikun Hu +3
Reinforcement learning for computer-use agents increasingly relies on proactive reward agents that interact with the post-execution GUI environment to verify what recorded screenshots cannot reveal. The reliability of such verification hinges on two coupled decisions: what to check and when to stop. Existing methods leave these decisions to prompted foundation models. However, neither scaling up the model nor providing few-shot examples reliably improves verification: prompted Judges ranging from 4B to GPT-5.4 all plateau at 68%-70% macro-average agreement. This is because the value of a check
agentbenchmark - arxiv:2610.26458 · cs.LGResidual-Based Adaptive Refinement in Hybrid Variational-Collocation PINNs for Singular Elliptic PDEs with Mixed Dirichlet-Neumann Boundary ConditionsSomnath Giri, Muslim Malik, Rishikesh Yadav
Solutions of second-order elliptic PDE problems on nonconvex or cracked domains with mixed Dirichlet-Neumann conditions typically lie outside $H^2(Ω)$. Near re-entrant corners, crack tips and points where the boundary condition changes type, the gradient grows like $r^{λ-1}$ with $λ<1$. We study residual-based adaptive refinement in hybrid variational-collocation PINNs. A multiscale Fourier-feature network is trained on a composite loss. The loss combines a variational residual tested against Legendre bubble functions, a strong-form collocation term and penalties for the Dirichlet and Neumann
benchmark - arxiv:2610.26457 · cs.CVProgressive multi-scale tumor-state memory for incomplete multimodal brain tumor segmentationMilad Rostamian, Mohammadreza Mohammadi, Mohsen Soryani
Brain tumor segmentation uses complementary magnetic resonance imaging (MRI) sequences, but some sequences may be unavailable in practice. Their absence changes the information available for segmenting different tumor regions. Fusion methods can support different input subsets. However, fusion at several scales does not by itself specify how tumor evidence is retained and updated to guide fusion at later stages. We propose a progressive multi-scale tumor-state memory network (PMTM) for segmentation from any non-empty subset of T1ce, T1, FLAIR, and T2. PMTM maintains a compact state for the cur
memory - arxiv:2610.26456 · cs.LGRegime-Aware Interpretable Ranking for Sector RotationKelvin J. L. Koa, Ziling Liu, Xinyang Li, Shuowei Feng +6
Sector rotation is a common decision in benchmark-relative portfolio management, where portfolio managers rank sectors for overweight and underweight positions. We formulate this task as ranking sector actions by expected future risk-adjusted active performance, measured with forward sector-benchmark Sharpe ratios. Such rankings should preserve temporal information, express model outputs through economically familiar signal characteristics, account for changes in signal relevance across macroeconomic environments, and capture dependencies among sectors. We propose RegimeRank, an interpretable
benchmark - arxiv:2610.26455 · cs.ROALIGN: Aligning Condition Estimation and Expression in Advantage-Conditioned Behavior CloningZechen Wang, Zhenqi Qiu, Houlong Xiong, Suohang Zhang +7
Human-collected robot manipulation data often mix successful behavior with retries, pauses, recoveries, and failures. Advantage-Conditioned Behavior Cloning (AC-BC) can exploit such mixed-quality data, but only if the conditioning signal is both accurately constructed from offline trajectories and effectively incorporated into the policy's action-vector-field prediction. Existing pipelines can weaken this signal by thresholding estimator-derived advantage scores into noisy conditions, and by routing conditions through prompts that weakly affect the action vector field. We introduce ALIGN, a co
vision-language-actionmanipulationfranka - arxiv:2610.26453 · cs.AIRemembering Has a Price: Learning Memory Worth under Resource ConstraintsGuannan Lai, Zi-Yang Bai, Jun-Peng Jiang, Han-Jia Ye
Lifelong agents continually accumulate and reuse experience to improve after deployment, yet in practice they often operate under tight generation, interaction, and memory budgets. Under such stringent joint constraints, we find that existing lifelong-memory methods can become ineffective, with several sophisticated approaches offering little advantage over simple retrieval or even no-memory agents. This reveals a more fundamental problem: resource scarcity makes each memory decision more consequential. When only a small amount of past experience can be retained or reused and opportunities for
memorylifelong agent - arxiv:2610.26448 · cs.CLTeleASR-Amharic: A Benchmark for Robust Amharic ASR in Realistic Telecommunication ConditionsDagnachew Mekonnen Marilign, Enrique Molero, Abrham Haile, Louis Powell +1
The real-world usability of ASR models remains challenging, particularly for low-resource languages, where existing speech resources are often collected in clean studio environments. In this work, we introduce TeleASR-Amharic, a benchmark for evaluating Amharic ASR under various acoustic and telecommunication conditions. We analyze the major degradations observed in real telecommunications recordings and develop a benchmark that covers multiple channel conditions, environmental disturbances, and telecom-domain variations. Our results show that telecommunication degradation can substantially re
benchmark - arxiv:2610.26445 · cs.ROA Human-in-the-Loop Comparison of Proportional Control and Adaptive Gravity Compensation in Elbow EMG-Driven ExoskeletonsFrancesco Pascucci, Eldison Dimo, Davide Costanzi, Mario Sedilesu +1
Electromyography (EMG)-driven control enables wearable robots to assist users based on their neuromuscular activity. Upper-limb exoskeletons typically rely on proportional myoelectric control (PMC), which amplifies user effort but may compromise stability at high assistance levels. EMG-driven adaptive gravity compensation (AGC) offers a promising alternative, counterbalancing gravitational torques while continuously adapting to neuromuscular activity. However, these approaches have not been systematically compared in a human-in-the-loop setting. This study evaluates five control algorithms: tw
human-in-the-loop - arxiv:2610.26438 · cs.AIFake News, Real Crashes: Information Injection Attacks on LLM Financial Trading AgentsMengxiao Wang, Nitesh Saxena
Large language model (LLM) agents now trade real capital using news as evidence, yet they cannot reliably tell fabricated headlines from real news. Because they read the same feeds under near-identical policies, a single fabricated post can push them into the same trade at the same moment. We present a systematic attack benchmark for information injection against LLM trading agents. Unlike prompt injection, which targets the model's instructions, information injection targets the evidence the agent trusts. The benchmark spans five injection vectors and seven event categories, evaluated over ~1
agentbenchmark - arxiv:2610.26433 · cs.CLThe Geometry of Agent Memory: Forecasting Retrieval-Bound Memory Failures, and Characterizing Its LimitsIliya Zhechev, Radoslav Ralev, Chris Coleman, Srijith Rajamohan +3
We study whether geometric relationships between conversation turns, memory texts, and queries can predict answer correctness before generation. We evaluate four memory implementations with dense or hybrid retrieval and one observation-summary comparator. We embed these texts in a shared space. A logistic classifier uses 25 features describing query proximity, memory-conversation correspondence, and conversation structure. On LongMemEval-small, pooled AUROC is 0.640 over 2,367 question-system pairs, below a question-difficulty baseline of 0.655. Performance varies by question type: geometry re
memoryagent memoryagent - arxiv:2610.26431 · cs.CVLong-form Video-to-Text Summarization using Recursive Language ModelsAlessia Intini, Lorenzo Vaiani, Enrico Chen, Luca Cagliero
While VideoLLMs have made significant progress in understanding short video benchmarks, they often struggle to summarize long-form ones. This is mainly due to performance degradation caused by long-context prompts, the inherent difficulty of preserving temporal continuity and a lack of semantically rich concepts in event descriptions. To tackle these issues, this paper introduces the first recursive Video Language Summarization Model (RVLSM) designed for video processing and tailored to long-form video-to-text summarization. RVLSM recursively decomposes videos into semantically coherent segmen
long-contexttool callingbenchmark - arxiv:2610.26429 · physics.opticsParameter-efficient neural computing with programmable photonic meshesArtem Popov, Stanislav Straupe, Mikhail Saygin
Memory capacity and bandwidth are becoming major constraints in neural-network computing, motivating parameter-efficient representations that reduce model storage and data movement. Tensor-network methods already enable substantial compression of neural networks, but their structured operations are often poorly matched to conventional processors optimized for dense linear algebra. Such representations can be implemented in photonic hardware, although this typically requires optical architectures tailored to a particular factorization. Here we propose a different approach based on programmable
memorypersistent memory - arxiv:2610.26426 · cs.CLHarnessRL: Co-Evolving Policies with a Just-in-Time Training Harness for Open-Ended RLXiaoqiang Wang, Ting-En Lin, Yuchuan Wu, Pinyi Zhang +6
Rubric-based reinforcement learning (RL) trains policies for open-ended generation with natural-language criteria as rewards, but these rewards can become ineffective as training progresses. We identify two failure modes. Under non-discriminative rewards, scores collapse to uniformly low or high values due to insufficient exploration or a mismatch between task difficulty and policy capability. Under spurious rewards, higher scores do not correspond to better responses. Both reflect a training configuration whose task sampling, exploration, and evaluation stay fixed as the policy evolves. We in
memory - arxiv:2610.26425 · cs.CLSNO-PI: Label-Scarce SNOMED CT Entity Linking with Pre-trained Encoders and Curated Terminology KnowledgePiyawoot Songsiritat
We present SNO-PI, the top-performing submission on SNOMED International's SNOMED CT Entity Linking Benchmark as of 18 October 2026. SNO-PI achieves a macro character IoU of 0.5387, outperforming the organisers' reference system (0.4582; based on the 2024 challenge winner's approach) by 18% relative. Linking clinical mentions to SNOMED CT is a zero- and few-shot problem over a very large label space: the 272 annotated MIMIC-IV discharge summaries cover only 6,595 of the roughly 219,000 eligible concepts, and a third of those occur only once. This motivates a design built on transfer from pre-t
benchmark - arxiv:2610.26421 · cs.CLSelf-Consistency of Entity Sets and Citations in a Deep Research System under Repeated PromptingLauri Pykälä, Kimmo K. Kaski
Deep research agents (DRA) are commonly used for open-ended synthesis tasks where closed-world QA benchmarks are poorly suited to characterize relevant aspects of performance. We study the self-consistency of one commercial deep research system under repeated identical prompting, representing each report with the set of named entities it mentions, categorised by a custom schema, and the set of citations it retrieves. Across repeated runs of a single open-ended cybersecurity related prompt, we quantify pairwise overlap to characterize how stable the retrieved and synthesized output structure is
benchmark - arxiv:2610.26416 · cs.LGCo-Evolving Policy and Reward for Code Generation: A Mechanistic Study of Single-Adapter DriftAmarsaikhan Batjargal, Noriko Tomuro
Reinforcement learning with verifiable rewards (RLVR) is particularly well suited to code generation because generated programs can be objectively evaluated through execution. Execution also provides behavioral information that can be reused as additional supervision. The Synergistic Policy and Reward Co-Evolving Framework (SPARK), originally proposed for reasoning tasks, exploits this idea by recycling training rollouts and their evaluation results to train a model not only to generate solutions but also to evaluate and improve them. We investigate whether this policy-reward co-evolution appr
self-improvement - arxiv:2610.26414 · cs.RORoboMMT: Learning Stable Tracking of Modulated Expressive and Interactive Robot MotionTill Hielscher, Kai O. Arras
Expressive robot motion can improve naturalness, effectiveness, and perceived likability in human-robot interaction. When deployed on dynamically unstable platforms, however, reactive execution of expressive body motion may bring the robot to its stability limits, especially when expressivity is decoupled from physical control. In this paper, we present RoboMMT (Robot Modulated Motion Tracking), a modular control architecture that couples parameterized expressive motion synthesis, physics-based character control, and reactive adaptation to spatial interaction goals. Our approach trains a refer
humanoidquadrupedsim-to-real - arxiv:2610.26412 · cs.CLWhere Does Rhetorical Distortion Live? A Multi-Platform, Multi-Language Measurement of Influencer ContentXucheng Yu, Jiangping Chen, Haohan Wang
Much of the manipulation in online influencer content is not false but disproportionate: an accurate fact is framed as historic, urgent, or catastrophic. Fact-checking and clickbait detection are not designed to capture such framing, and little is known about where it originates: with the creator, the topic, the platform, or the audience that replies. We operationalize rhetorical distortion through five literature-grounded dimensions (Significance Inflation, Anxiety Manufacturing, Novelty Claims, Loaded Language, and Temporal Distortion) and measure it with a three-tier pipeline in which bilin
manipulation - arxiv:2610.26410 · cs.CVPolicy-Adaptive Credit Assignment for Spatial Reasoning with Verifiable RewardsHailiang Yang, Shengchao Chen
Image-based spatial reasoning requires perception, decomposition, and inference, but the stage that limits accuracy can change during training. Answer-level rewards neither localize failures to individual stages nor provide discriminative group-relative feedback when all rollouts receive the same reward. Process supervision provides intermediate feedback, but fixed weights cannot adapt reward allocation to these changing bottlenecks. To address this gap, we propose CALIPER, an adaptive process-reward framework supported by SpatialTrace, a corpus of stage-decomposed spatial reasoning traces. CA
benchmark - arxiv:2610.26401 · cs.CVOne Fact, Many Places: Redundant Evidence in Long Visual Documents Breaks Single-Gold Evaluation of Retrieval, Citation and AttributionXiaokang Yuan, Binhui Yao, Luwei Yang, Ruimin Chen +1
Document understanding is increasingly judged on where an answer comes from, scored against annotated gold locations as if each fact lived in one place. In long visual documents, a fact often lives in several places: a chart value is restated in the text, a key figure reappears in the summary. We ask how often such copies occur in standard benchmarks and what they do to evaluation. We annotate 420 questions from MMLongBench-Doc and MMDocRAG with the locations that independently establish each required fact, checked by two independent reviewers. Copies are common and often cross-modal: 19.5% of
benchmark - arxiv:2610.26399 · cs.CLHow Benchmark Graph Construction Shapes Multi-Hop KG Navigation: A Controlled Audit of PathQuestion and PathQuestion-LargeEduin E. Hernandez, Sergio A. Diaz, Luis F. Garcia, Nurassyl Askar +1
Multi-hop knowledge-graph question answering (KGQA) benchmarks are often treated as fixed language-and-reasoning tasks, but for navigation-based methods, difficulty also depends on the graph in which those tasks are instantiated. This makes graph construction a consequential but often entangled part of the navigation task. To isolate this effect, we audit PathQuestion (PQ) and PathQuestion-Large (PQL) as graph-navigation benchmarks under a controlled intervention: we hold questions, topic entities, released answers, released path annotations, and family-level splits fixed while replacing each
benchmark - arxiv:2610.26395 · cs.ROForagerSim: Evaluating Embodied Map-Based Reasoning in Large-Scale Dynamic WorldsTomáš Musil, Martin Kotík, Michal Vavrečka, Karla Štěpánová +2
Animals build cognitive maps of vast environments, use them to perform map-based navigation, and might even be reusing the same circuitry used in spatial reasoning for creating and using maps of arbitrary task-spaces. In contrast, current embodied AI systems either rely on hand-engineered mapping pipelines, which break in the presence of unmodeled dynamics and are restricted to metric space, or on learned policies that rarely form map-like representations at all. Developing architectures that could learn to build and use map-like representations as flexibly as animals is, however, difficult be
embodiedbenchmark - arxiv:2610.26394 · cs.AIWhen Coding Agents Outpace Review: Evaluating PRD-Driven Development in Mature Enterprise RepositoriesTianqi Liu, Yandong Guan, Songze Li, Wenkai Wang +2
Coding agents increasingly undertake development tasks in mature enterprise repositories, yet existing benchmarks offer limited insight into whether they fulfill complete Product Requirement Documents (PRDs) and where delivery breaks down. Issue resolution and development from scratch leave incremental PRD fulfillment underexplored, while isolated tests provide incomplete evidence of business and integration requirements. We introduce PACE, a benchmark of 43 enterprise tasks with 750 expert-audited Code Points. It combines requirement reconstruction, criterion construction, and evidence-based
agentbenchmark - arxiv:2610.26383 · cs.AIBOSS: Token-Budgeted Online Sampling and Selection for LLM ReasoningYupeng Chang, Yahan Li, Chenlu Guo, Diana Inkpen +2
Test-time scaling can improve LLM reasoning, but existing approaches typically adapt generation or apply answer selection after a predetermined generation schedule. We study within-instance allocation: under a hard per-instance prompt-plus-completion token cap, when should inference stop, generate another candidate, or spend residual tokens adjudicating candidates already observed? BOSS is a training-free policy for this three-way decision, with worst-case budget checks before every generation and judge call. Its key contribution is online generation-adjudication allocation under one shared to
benchmark - arxiv:2610.26380 · cs.LGSelecting the Best Few with Human Evaluators: Adaptively Combining Ratings and Pairwise ComparisonsMotti Goldberger, Nils Rudi
This paper considers a decision-maker tasked with identifying the $k$ best items from a large pool of candidates, such as job applicants, LLMs, award nominees, or new product ideas. The decision-maker learns about the items' quality by soliciting human evaluations. Each evaluation can be costly, so the decision-maker aims to use the limited evaluation budget efficiently by allocating evaluations adaptively. A key consideration is the form of evaluation to solicit, which should be something evaluators are comfortable providing. Two popular forms are ratings and pairwise comparisons, and we allo
benchmarkevaluator - arxiv:2610.26379 · cs.LGCTLite: Lightweight Irregular Multivariate Time Series Forecasting with Reusable Continuous-Time SummariesShusen Ma, Xudong Zuo, Yangchun Xie, Yun-Bo Zhao
Irregular multivariate time series forecasting (IMTSF) requires capturing temporal dynamics and cross-variable dependencies from sparse, asynchronous observations. Existing methods often rely on learned temporal encoders, coupling historical representation construction with model optimisation and repeatedly processing the same observations. We propose CTLite, a lightweight forecasting framework that decouples historical summarisation from learned prediction. Specifically, a parameter-free continuous-time moment encoder summarises local levels and trends at multiple temporal scales while suppre
benchmark - arxiv:2610.26372 · physics.opticsSelf-Oscillating Photonic Frequency Comb for Sub-THz Local Oscillator Generation in Earth Observation SystemsDimitrios Kastritsis, Ioannis Stratakis, Stavros Iezekiel
Earth observation missions rely on microwave and sub-terahertz radiometry for retrieving atmospheric parameters such as temperature, humidity, and cloud properties, with the 183- and 229-GHz spectral regions being particularly relevant. Local-oscillator (LO) generation at these frequencies requires low phase noise and high frequency stability while meeting stringent size, weight, and power (SWaP) constraints. Conventional architectures employ dielectric resonator oscillators followed by cascaded frequency-multiplication stages, requiring dedicated generation chains for different frequency band
mach-zehnder - arxiv:2610.26369 · cs.AIAutonomous intelligence for mining productivity: a multi-agent AI approach to operational value creationN Guerin, A Dehoux, M Nakhla
Integrated Remote Operation Centres (IROCs) are emerging as the architectural frontier of Industry 4.0 for capital-intensive sectors such as mining, oil and gas, and energy. Yet two intertwined gaps constrain their evolution towards full autonomy. First, no integrated agentic architecture has been proposed for mining IROCs that orchestrates Supervisory Control and Data Acquisition systems, digital twins, predictive analytics and human-machine interfaces under safety-critical constraints. Second, no evaluation framework jointly addresses the real-time, Health-Safety-Environment, regulatory-audi
multi-agentagenticbenchmarkevaluation framework - arxiv:2610.26362 · cs.AIGAIDA: Generative AI for Human-in-the-Loop Document AnnotationIslam Mesabah, Ali Elhosary, Priyabanta Sandulu, Darko Obradovic +1
Document annotation is a fundamental yet labor-intensive step in building document information extraction systems. While Optical Character Recognition (OCR) and Large Language Models (LLMs) can automate parts of this process, we argue that annotation systems should be judged not by how completely they remove human effort, but by how effectively they redirect it toward review and correction. We propose a modular human-in-the-loop architecture that integrates OCR-based extraction, LLM-based semantic pre-annotation, and human validation in a single editable workflow, and present \textbf{GAIDA}, a
human-in-the-loopbenchmark - arxiv:2610.26359 · cs.MADiscovery-Probability Graph Planning for Embodied Agents in Tasks with Hidden Cooperation RequirementsZihuang Cai, Changjian Chen, Zhuo Tang, Chang Tang +1
Embodied agents are increasingly capable of perceiving, planning, and acting autonomously in dynamic environments, demonstrating strong performance on complex cooperative tasks. Recent advances typically assume that cooperative tasks are known in advance. In real-world applications, however, hidden-cooperation tasks are more common: the need for cooperation only becomes apparent after an agent observes and interacts with task-related targets. In this paper, we first present HiddenCoop, a benchmark for effectively evaluating agent cooperation in hidden-cooperation tasks. We further propose Disc
embodiedagentembodied agentbenchmark - arxiv:2610.26342 · cs.AIVibeTraining-Bench: Data-Trainer-Harness Co-Design for Iterative System ImprovementRuotian Peng, Yuanyi Ren, Shuaiqi Liu, Xin Guo +6
Recursive self-improvement (RSI) requires agents to diagnose system bottlenecks and coordinate improvements across training and inference. We refer to the agents' capability to jointly optimize training data, the trainer, and the inference harness as cross-component co-design. We introduce VibeTraining-Bench, a benchmark for evaluating this capability across tasks spanning terminal operation, software engineering, multi-turn tool use, and mathematical reasoning. Within each task, research agents start from an identical initial agent system under a fixed resource budget. Guided by validation fe
agentagent systemtool useself-improvementbenchmark - arxiv:2610.26339 · cs.AILarge-Scale Biomedical RAG: Corpus Scaling, Document Representation, and Efficiency Trade-offsAdrián Sánchez-Freire, Gonzalo Pérez-Gómez, Aitziber Atutxa, Ander Barrena
We present a controlled, large-scale evaluation of how corpus scale, document representation, and reranking affect the effectiveness and efficiency of biomedical Retrieval-Augmented Generation (RAG) systems. Retrieval collections were progressively expanded while retaining the relevant source documents, enabling consistent comparisons across corpus sizes. We compared abstract, flat full-text, and hierarchical full-text representations across first-stage retrieval, reranking, and end-to-end multiple-choice question answering. Configurations selected on development data were evaluated on five ex
retrieval-augmentedragbenchmark - arxiv:2610.26337 · cs.LGCIS-Transolver: A Cross-Layer Incremental-State Neural Operator for Fine-Scale Automotive Aerodynamic Flow PredictionJianglong Liu, Zhicheng Zhu, Jia Hao, Liangyue Jia
Neural operators offer an efficient data-driven approach to rapid prediction of complex automotive aerodynamic flows. However, spectral bias and other factors limit accurate reconstruction of fine-scale flow variations near complex geometries. Existing studies improve the representation and fitting of fine-scale information mainly through information extraction and training objectives. Further study is needed to determine how extracted local details can be used continually in subsequent operator computations and contribute effectively to the overall flow-field representation. We propose CIS-Tr
benchmark - arxiv:2610.26324 · cs.LGThe Control Layer Behind Rule-Following Agents Trading Across Market ConditionsSaad Naja, Nasser Odtallah, Jake Van-Praag, Sarah May
Trading discipline often fails people. Our study tested whether four AI trading agents followed their strategies and risk rules. Offering traders the evidence and an agent evaluation toolkit to separate luck from following the rules, before giving agents meaningful autonomy or capital. Our study analysed 2,627 closed positions from 12 February to 27 August 2026 across the same asset universe and execution conditions. Most positions were simulated against live prices with a 0.09% round-trip fee; 18% involved real Hyperliquid orders. Each position declared a strategy, stop and target, with a lim
agent - arxiv:2610.26323 · cs.LGTriScore: Post-Hoc Out-of-Distribution Detection with Energy, Boundary Probes, and Transform ConsistencyEge Gursoy, Nicolas Mansard
Post-hoc out-of-distribution detection usually relies on a single signal extracted from a frozen classifier, such as confidence, energy, features, or gradients. This dependence becomes fragile under classifier-level adversarial shift, where the shift can preserve the detector's chosen cue while changing the model's decision behavior. We investigate whether this failure mode can be mitigated by combining complementary detection criteria instead of selecting a single score. We introduce TriScore, a post-hoc OOD detector that combines three ID-calibrated signals from the same frozen backbone: an
benchmark - arxiv:2610.26313 · cs.AIStrucChunk: Structure-Aware Retrieval for Legal Document RAGSurya Deep Singh, Ashutosh Tripathi, Gaurav Deep Singh
Retrieval-Augmented Generation (RAG) can ground LLM outputs in domain-specific knowledge, but legal documents are hard: standard chunking fragments hierarchical argument structure, and generic embeddings struggle with both precise citation matching and semantic understanding. We propose StrucChunk (Structure Aware Chunking) with hybrid dense-sparse retrieval for hierarchically structured legal documents. It treats documents as directed acyclic graphs of provisions, preserves clause boundaries, resolves cross-references before embedding, and combines BM25 with BGE-M3 via Reciprocal Rank Fusion.
retrieval-augmentedrag - arxiv:2610.26312 · cs.CVCARNet: End-to-End Coupled Association-Recognition Network for Fine-Grained Multi-Drone Collaborative DetectionZhao Zhang, Ruigang Fu, Yu Zhang, Jiajun Xiang +2
Multi-drone collaborative detection aims to enhance target recognition by leveraging complementary observations from overlapping aerial viewpoints. Existing approaches typically follow a two-stage pipeline that first establishes cross-view correspondences and then performs fusion-based recognition. However, this decomposition suffers from inconsistent optimization objectives, redundant pairwise modeling, and irreversible error propagation, where incorrect associations inevitably undermine subsequent recognition performance. To overcome these limitations, we propose the Coupled Association-Reco
benchmark - arxiv:2610.26311 · cs.AISelective Transfer through Sequence Distillation: From Benchmark Knowledge to Backdoor BehaviourJonibek Mansurov, Sholpan Bolatzhanova, Filbert Aurelian Tjiaranta, Vallerie Alexandara Putra +1
Knowledge distillation is widely used to transfer capabilities from teacher models to student models, but recent work shows that it can also transfer unintended teacher properties. We study whether different hidden teacher properties survive the same distillation channels. Specifically, we compare two properties: benchmark-associated knowledge and refusal backdoors, and show that distillation treats them very differently. Teachers fine-tuned on benchmark data successfully pass that knowledge to students through all four distillation channels we test, including when students are trained only on
benchmark - arxiv:2610.26310 · cs.AIThe Scheduler Is a Hidden Retriever: How Asynchrony Shapes Evidence Acquisition in Search AgentsRuiyang Ren, Yuhao Wang, Yinan Zhang, Chunyan Miao
Language-model agents increasingly solve complex tasks through parallel reasoning and asynchronous information gathering. Controlled replay across two model families and three benchmarks shows that delivery timing redirects search decisions made before all results arrive, despite identical initial evidence. Acquisition disagreement reaches up to 24 percentage points above same-order repeat variation. This dependence persists after all initial evidence arrives and through a second search, turning temporary differences in what agents have seen into lasting differences in what they acquire. An ac
benchmark - arxiv:2610.26297 · eess.SYMobile Embodied Agent Network (MEAN): Integrated Framework, Enabling Technologies, and Task-Driven DeploymentZhaoyang Zhang, Lingxiao Sun, Weijie Zhou, Chunhui Feng +7
Mobile networks are evolving from human-centered, data-oriented communication infrastructures toward machine- and/or agent-dominant, task-oriented platforms that are to serve the rapidly increasing networked machines and embodied agents. In order to fulfill the newly arose complex task objectives, the framework of future mobile network is to be innovated to reflect and facilitate the intrinsic interplay and integration of the underlying processes of perception, communication, computation, decision, action, and feedback, etc. This survey establishes the mobile embodied agent network (MEAN) as a
embodiedagentembodied agent - arxiv:2610.26287 · cs.RORobotic Photography System based on Vision-Language-Action ModelSong-Li Wu, Chubin Zhang, Wenbo Ding, Yansong Tang
Modern robotic photography requires understanding photographic intent, planning camera motions, and producing aesthetically pleasing images in complex environments. We present RoboPhoto-VLA, a unified Vision-Language-Action framework that jointly models perception, reasoning, and camera control for autonomous robotic photography. RoboPhoto-VLA incorporates PhotoCoT for structured action planning and Aesthetic Direct Preference Optimization (ADPO) to align camera behaviors with human aesthetic preferences. To facilitate training and evaluation, we construct OpenPhotography, a robotic photograph
vision-language-action - arxiv:2610.26286 · cs.LGToward Online Robust Zero-Sum Markov Games with Function ApproximationChenhao Zhou, Jiyu Wei
Online robust zero-sum Markov games pose a coupled operator-learning and game-solving problem: from nominal interaction data, the learner must recover a worst-case minimax Bellman operator while maintaining strategic coverage and controlling stage-game approximation error. Existing robust reinforcement-learning theory does not yet provide such an operator-learning perspective beyond specialized multi-agent settings. We develop an optimistic dual robust Bellman framework that combines dual robust Bellman fitting, optimistic minimax planning, backward-validated confidence sets, and exploiter-mix
multi-agent - arxiv:2610.26275 · cs.CVFrom Cause to Effect: Generating Intermediate Event Chains in VideosRidouane Ghermi, Vicky Kalogeiton, Ivan Laptev
Video-Language Models (VLMs) excel at visual recognition but struggle with causal reasoning. Current evaluations are flawed: multiple-choice QA allows shortcuts, and open-ended QA lacks robust automatic evaluation. We introduce Causal Chain Generation (CCG), a novel task where models have to generate the sequence of intermediate events connecting a cause and an effect. This structured setup demands genuine reasoning and enables reliable automatic scoring. We accompany this task with two complementary metrics: SODA-Embedding, which measures the semantic matching of a generated chain against a r
benchmark - arxiv:2610.26272 · cs.MABidirectionality is Necessary in Cellular Machines for Exact-Arrival Multi-Agent Path FindingMasashi Aono
On ranked directed acyclic graphs coupled by vertex conflicts we study exact-arrival multi-agent path finding (departure and arrival pinned in place and time) posed as a distributed output task: every cell must emit its own bit of a valid joint path-support mask. The machines are cellular: one automaton per cell, state factored over cells, communication only along a fixed local fabric, synchronous rounds, a query-independent output round, no central decoder, and a per-topology transition code. "Necessary" here means: on an explicit hard topology, over all its feasible queries, no universally c
multi-agentbenchmark - arxiv:2610.26267 · cs.LGA Pitfall in the Evaluation of Gene Perturbation ModelsGeorge Panagopoulos, Saman Omidi, Tsz Pan Tong, Jun Pang
Gene perturbation models aim to predict transcriptional responses to unseen genetic interventions, but comparisons rely on heterogeneous evaluation metrics and differentially-expressed (DE) gene subsets. We investigate the consequences of these choices using eight methods across six single-gene perturbation datasets under a unified evaluation protocol. Varying the number of top DE genes reveals substantial instability in Precision-based rankings, with the winning method changing $86.7\%$ of the time, while rankings based on the rest of the metrics remain comparatively stable. Investigating thi
benchmarkevaluation protocol - arxiv:2610.26262 · cs.CVDynamics-Aware Sparse Attention for Efficient Autoregressive Video DiffusionYuanyu He, Zhuokun Chen, Yefei He, Zhiwei Tang +4
Autoregressive Diffusion Transformers have emerged as a powerful paradigm for long video generation. However, they are often constrained by prohibitive computational overhead and substantial memory footprints. This inefficiency arises from dense computation mechanisms that allocate uniform processing resources across all spatiotemporal regions, failing to leverage the inherent temporal redundancy of video where motion is typically confined to sparse areas. In this paper, we propose Dynamics-Aware Sparse Attention (DASA), a unified framework that shifts from dense computation to sparse, motion-
memory - arxiv:2610.26261 · cs.AIREFLEX: RNA Ensemble Generation via Flexibility-Calibrated Stochastic BridgeWanli Ma, Tianmeng Hu, Ke Li
RNA molecules often function through multiple experimentally observed conformational states. A central challenge is to capture ensemble diversity without sacrificing structural fidelity: samples may collapse toward averaged representative structures or drift into unsupported conformational regions. We introduce REFLEX (RNA Ensemble generation calibrated by FLEXibility), a sequence-conditioned generator for RNA backbone-frame ensembles based on a heteroscedastic stochastic bridge on SE(3). During training, experimental B-factors calibrate residue-wise bridge widths, exposing each observed confo
benchmark - arxiv:2610.26255 · cs.CVXGLASS: X-Ray, Gaze and Language Assisted Slice Sequencing for CT VolumesMichael Glöckler, Timo Ropinski
Vision-language models (VLMs) achieve remarkable performance on 2D medical images, such as X-rays or pathology images, while 3D VLMs that directly process CT/MRI volumes lag behind due to limited 3D training data and the predominance of 2D foundation architectures. To leverage powerful 2D VLMs for analyzing medical volumes, which contain hundreds of slices, it is therefore necessary to focus only on those few slices, which are diagnostically relevant. To enable such a prioritization, we propose XGLASS, a method that transfers medical expertise from 2D X-rays to 3D CT for task-aware slice prior
memory - arxiv:2610.26251 · cs.AIBeyond Static Leaderboards: Capability-Focused Evaluation with Structural Diversity for Time Series Foundation ModelsHongyin Zhang, Mingyi Xiang, Rongzhao Chen, Yankai Wang +6
Static public benchmarks for time series foundation models (TSFMs) typically have fixed and narrow test distributions. Their repeated reuse encourages benchmark-specific tuning, making reported performance reflect adaptation to the fixed test set rather than genuine forecasting generalization. In addition, a single aggregate score cannot fully characterize model capabilities across distinct temporal structures and input relationships. We propose CaFE (Capability-Focused Evaluation), a refreshable evaluation extension grounded in original benchmark data that enriches static benchmarks with stru
benchmarkleaderboard - arxiv:2610.26239 · cs.AICoRS: Fast Neural Flexible Job-Shop Scheduling via Composite Interval State ReuseJiang Sutong, Zhao Peng, Wu Xuan, Wang Yan +5
Neural solvers for the Flexible Job-Shop Scheduling Problem (FJSP) typically refresh graph representations and update feasibility after every operation--machine decision. However, only feasibility must remain exact at every step, whereas graph encoding is substantially more expensive and can potentially be refreshed less frequently. We present CoRS, which decouples them: an interval-level graph representation is reused across multiple sequential decisions, while a stepwise feasibility state is updated after every action. This two-timescale design amortizes graph encoding without parallelizing
benchmark - arxiv:2610.26233 · cs.CLHeartbeat Thinking: Speculative Reasoning and Tool Execution While Listening in Realtime Voice AgentsZhenyu Zhang, Kewei Fu, Yongqi Tong, Chenzhi Zheng +3
Tool-using voice agents often leave users waiting after speech ends, even when actionable information arrives earlier in the request. Speculative approaches begin work during speech, but chunk-bounded or interrupt-on-update generation can truncate a tool call before its arguments are complete. This couples computation time to speech updates, limiting useful work completed in advance. We introduce Heartbeat Thinking, which handles updated input through new Thinker invocations while allowing ongoing invocations and dispatched tools to finish. Periodic invocations read the latest partial transcri
benchmark - arxiv:2610.26225 · cs.CLRethinking BPE: Merge Selection as Model SelectionVijini Liyanage, François Yvon
Subword tokenization methods such as Byte-Pair Encoding (BPE) are a core component of modern language models; yet, tokenizer capacity remains typically determined by a fixed merge budget, chosen heuristically. Since larger vocabularies increase memory usage and computational cost, selecting compact and effective token inventories is essential for efficient language modeling. We study the optimization of a BPE merge budget as a statistical model selection problem and investigate information-theoretic criteria for \emph{computing optimal merge sets} directly from corpus statistics. To further im
memory - arxiv:2610.26216 · cs.CVRoAR-VAE: Improving Latent Diffusability through Robust PredictionYang Yue, Fangyun Wei, Yitong Wang, Jinjing Zhao +3
An effective visual tokenizer must preserve image fidelity while producing latents that downstream diffusion models can learn efficiently. Prior work improves latent diffusability through semantic alignment or joint tokenizer--diffusion training. Revisiting REPA-E, a representative approach, we find that removing its diffusion objective and dedicated generative blocks while retaining semantic prediction from corrupted latents yields comparable downstream generation performance under a matched setting. This observation motivates a general principle of robust prediction: generation-relevant info
benchmark - arxiv:2610.26209 · cs.LGOceanTensorCube and ST-MMT: A Multimodal Data and Transformer System for Seven-Day Forecasting of Pyropia BleachingGwangjin Kim, Cheolyoung Jang, Yosep Jo
Bleaching of cultured Pyropia (laver) causes losses in Korean seaweed farming. We present a system that forecasts bleaching seven days ahead for farm clusters. OceanTensorCube aligns 32 channels from public marine, meteorological, ocean-model and satellite sources on a 30 m grid, and ST-MMT, a 9.2-million-parameter spatiotemporal transformer with hub-style multi-head cross-attention between variable groups, maps a 14-day input window to a seven-day-ahead bleaching probability. On the validation set, a six-seed ensemble detects every reported positive at a false-alarm rate of at most 3.5%, an u
evaluation protocol - arxiv:2610.26203 · cs.CVFocused Omni-Modal GUI AgentsFuyao Zhang, Wenwen He, Jiaming Zhang, Yurong Hao +5
Graphical User Interface (GUI) agents increasingly act over streaming omni-modal observations spanning screenshots, audio, video, interaction history, and device states. However, exposing this complete context uniformly at every step ignores that relevance is transient and sparse: the signal determining one action can become clutter at the next. Because executed actions feed back into the context, these errors compound into Sequential Distraction, driving the agent into irreversible trajectory drift. We formulate the underlying requirement as Focused Perception: a representation that is step-a
agent - arxiv:2610.26201 · cs.ROK-AoI: A Spatio-Temporal Synchronization Metric for Digital Twin-based Teleoperation in Dynamic Network EnvironmentsLu Tian, Zhenhui Yuan, Zexin Deng, Wenjuan Zhang
Synchronization between a digital twin (DT) and its physical robot is critical for effective teleoperation, as mismatches directly degrade operator awareness and task performance. Existing synchronization metrics are predominantly time-based and rarely capture the spatial consequences of delayed information during motion. In this paper, we introduce Kinematic Age of Information (K-AoI), a spatio-temporal metric that weights AoI by instantaneous Cartesian velocity, translating temporal staleness into physically meaningful spatial divergence. We implement and evaluate K-AoI on a DT teleoperation
teleoperationmanipulator - arxiv:2610.26190 · cs.AIVEILED-SH: Vulnerability Expression through Indirect Language Evaluation Dataset (Self-Harm)Leanne Tan, Keith Chua, Shaun Khoo, Onno P. Kampman +6
AI systems may miss indirect, culturally situated distress or provide support mismatched to the application context. Existing benchmarks do not jointly address cultural variation, application context, and practitioner disagreement. We introduce VEILED-SH, a benchmark and construction methodology for evaluating single-turn responses to indirect expressions of suicide, self-harm, and distress. Its 10,003 prompts span counselor chat, forum posts, platform search, and triage, drawing on Singapore-grounded sources and public datasets. An expert-validated multi-label taxonomy distinguishes intent, c
benchmark - arxiv:2610.26185 · cs.AIMoveBench: A Rubric-Based Benchmark for Evaluating Large Language Models in Sports and Health CoachingYitong Li, Jiang Bian, Jingling Wang, Pan Ding +1
Despite rapid advances in large language models (LLMs), their deployment in the sports and health domain remains high-risk due to the absence of domain-specific evaluation benchmarks. Existing general-purpose benchmarks (e.g., MMLU and HELM) primarily assess broad knowledge and reasoning capabilities, but fail to capture critical domain requirements such as exercise safety, physiological appropriateness, personalization, and actionable coaching quality. This gap is particularly concerning in fitness and health applications, where incorrect or unsafe recommendations can lead to physical harm. T
benchmark - arxiv:2610.26179 · cs.CVLearning Goal-Conditioned Subgoals through Frozen Hierarchical World ModelsXiang Kaibo, Liu Qingfeng
Hierarchical world models plan over long horizons by passing predicted states from a coarse model to a lower-level controller. Choosing these states can be expensive: the upper-level planner repeatedly differentiates through candidate macro-action trajectories. We replace this search with a lightweight goal-conditioned network trained through the frozen high-level predictor. The network supplies a conditioning vector, while the predictor generates the next latent waypoint. Training uses hindsight goals and state-prediction losses on offline trajectories, without search-solution targets or macr
world modelevaluation protocol - arxiv:2610.26174 · cs.ROPhantomTac: Hijacking Robotic Tactile Systems with Deploy-and-forget TagsBinbin Zhang, Yifeng Zhang, Xiaomeng Chen, Dongyao Chen
Tactile perception is crucial for precise robotic manipulation, providing an important foundation for dexterous, closed-loop force control. Among various tactile sensing techniques, magnetic tactile sensors stand out thanks to their fine-grained, three-dimensional force perception capability and strong robustness against visual occlusion. In this study, we discover a physical loophole called PhantomTac in magnetic tactile sensors. The attacker can alter the magnetic field, thus spoofing the force interpreted by magnetic tactile sensors. The attacker can stealthily deploy such an attack by simp
manipulationdexteroustactilegrasp - arxiv:2610.26170 · cs.ROData-Efficient Learning from Demonstration for Humanoid Robots via Feature-Level Perception and Prompt-Based Object DetectionKeyhan Rayati, AliReza Beigy, Anoosheh Saadati, Pourya Shahverdi +3
Learning from Demonstration (LfD) allows robots to learn new skills from human examples; however, many vision-based methods require large datasets and resource-intensive end-to-end training. This paper introduces a data-efficient LfD framework for humanoid robots that learns manipulation skills from compact, task-specific features instead of raw images. The proposed approach integrates layered visual preprocessing, Kinect-based human joint estimation, prompt-based object detection, and depth information to extract human motion features and three-dimensional object locations from demonstration
manipulationhumanoid - arxiv:2610.26169 · cs.ROCompositional Humanoid Manipulation through Isomorphic Teleoperation and Independent Skill RoutingYiming Wang, Peixian Yang, Lifeng Zhuo, Lai Wei +4
Humanoid manipulation requires reusable arm skills, yet demonstrating every possible skill combination is impractical. We present Humanoid Semantic Mixture-of-Experts Diffusion Policy (HS-MoDP), which composes separately demonstrated arm skills through independent semantic routing over a shared expert pool. This design preserves arm-specific skill control while sharing action-generation parameters across arms. A semantic scheduler integrates the learned policy with an existing locomotion controller through alternating manipulation and repositioning stages, extending the effective workspace and
manipulationhumanoidteleoperationdiffusion policy - arxiv:2610.26168 · cs.CLVISAGENT: An Agent-Based End-to-End Voice Interaction SystemXiang Wang, Zhaoxiang Liu, Xiang Chen, Liqiang He +2
Voice-based public services have evolved from manual switchboards to intelligent platforms to meet modern user needs. However, current systems remain constrained by rigid menu-driven interaction, fragmented vertical services, and limited extensibility. To address these issues, we present VISAGENT, an agent-based end-to-end voice interaction system for large-scale public services that integrates end-to-end speech interaction, hierarchical agent routing, a scalable voice-agent marketplace, and user memory support. The system replaces menu-driven workflows with natural speech and dynamically disp
memoryagentagent systemhierarchical agent - arxiv:2610.26155 · cs.CLDeepSEA: A Multilingual Evaluation Benchmark for Southeast AsiaJiuzhou Han, Abhishek Bijay Mishra, Trevor Cohn, Shubham Mittal
While language models have demonstrated remarkable capabilities, their performance remains disproportionately skewed towards high-resource languages. This disparity is particularly acute for the 700 million inhabitants of Southeast Asia, where vast linguistic diversity remains under-represented in existing benchmarks, often suffering from limited language coverage or narrow task scopes. To address this, we introduce DeepSEA, a comprehensive benchmark constructed from established datasets through professional human translation into 12 Southeast Asian languages, spanning both high-resource (e.g.
benchmark - arxiv:2610.26143 · cs.LGFLMNet: Self-Ensemble Distillation from Masks to Seeds for Single-Tower CTR PredictionYifan Wang, Yifei Du
Click-through rate (CTR) prediction underpins recommender systems and online advertising, where high-dimensional categorical features interact in combinatorially many ways. Several recent CTR models, such as FinalNet and QNN, report higher accuracy when a network also learns from its own averaged predictions, for example by running the same network under different dropout patterns and training each run toward the average. Why this self-distillation helps, and whether its random sampling is needed, remains unclear. This work shows that as the number of dropout patterns grows without bound, the
benchmark - arxiv:2610.26114 · cs.LGFrom Retrieval to Recognition:Knowledge-Enhanced Post-training of Time Series Foundation Models for ClassificationZhen Liu, Yucheng Wang, Yingpeng Du, Tianjun Wei +4
Time series foundation models (TSFMs) have achieved notable progress in classification tasks through large-scale pretraining. However, their post-training practice mainly focuses on implicit parameter fine-tuning, where task-relevant information is entangled with abundant irrelevant knowledge in vast parameter spaces. This may hinder TSFMs from precisely activating useful knowledge for downstream classification. To this end, we propose Retrieval-to-Recognition (R2R), a plug-and-play post-training paradigm that adapts pretrained TSFMs to time series classification. Rather than pretraining a new
post-training - arxiv:2610.26106 · cs.ROA GPU-Resident Pipeline for Multi-Robot 3D Inspection with Limited PerceptionKryštof Pšenička, Stefan Edelkamp
We present a complete, GPU-resident pipeline for multi-robot 3D inspection of underwater structures. Given a triangle mesh and a fleet of $K$ robots, each carrying a short-range, limited field-of-view camera, the pipeline computes 6-DOF trajectories that jointly cover a target fraction of the surface while minimizing a linearly interpolated makespan/total-cost objective. The joint problem is NP-hard, so we decompose it into five core stages - candidate viewpoint generation, visibility computation, set cover, min-max vehicle routing solved as a mixed integer program (MIP) with lifted Miller-Tuc
multi-agent - arxiv:2610.26101 · cs.AIReliable Before Profitable: QuantHarness for Governing Financial Strategy Development with LLM AgentsJiasheng Mao, Jiacheng Liu, Hengwei Ye, Zheng Tian
LLM agents can turn natural-language financial research requests into executable and backtested strategies, but a successful run may still implement the wrong rule, use information unavailable at decision time, or obscure how repairs and candidate selection changed the strategy. Because these failures arise throughout development, final code and performance metrics alone are insufficient to establish either fidelity to the request or the validity of the development process. We therefore address the reliability of financial strategy development at the trajectory level: the system must control h
agentllm agentbenchmark - arxiv:2610.26098 · cs.AIFrom Code to Playable Systems: CEGame-Bench for Long-Horizon Game Creation AgentsZixuan Dong, Yanghai Wang, Hanyan Bian, Jiaxi Tao +7
A game ships when a player can launch it, finish it, and return to a saved world, none of which a successful build guarantees, yet game-creation agents are mostly tested on one engine or on one-prompt browser games. We present CEGame-Bench, the first long-horizon benchmark for complete-game creation across five production engines: Defold, Godot, Roblox Studio, Unity, and Unreal Engine. Agents build each game from an empty project and a design document in runs averaging up to 295 steps and 65M tokens; the document doubles as a contract, and a hidden evaluator accepts a game only by real-input p
agentbenchmarkevaluator - arxiv:2610.26096 · cs.CVPnP-Transformer: A Neural Pose Solver for 6D Object Pose EstimationEmirkan Burak Yilmaz, Yakup Genc
We present PnP-Transformer, a neural pose solver designed as an iterative-free, differentiable alternative to classical PnP-RANSAC algorithms. PnP-Transformer regresses the 6D object pose by processing predicted semi-dense 2D-3D correspondence fields alongside confidence scores, while mapping geometric inputs into a continuous latent space. To evaluate its practical utility, we integrate PnP-Transformer into a zero-shot 6D pose estimation pipeline, where the correspondence module and pose solver are trained separately. PnP-Transformer achieves a 41.2% mean AR score with low latency (0.46 s/obj
benchmark - arxiv:2610.26091 · cs.CVGeoNaviDiff: Transferring Disentangled Semantic Controls to Diffusion Features via Local Riemannian NavigationTao Sun, Huanting Guo, Baao Xie, Zhou Fang +3
Disentangled real-image manipulation requires factor-specific semantic control while preserving image fidelity and non-target content. StyleSpace $S$ provides interpretable controls but is limited by GAN inversion, whereas diffusion models offer stronger image realization without an explicit factorized control basis. We introduce GeoNaviDiff, an unsupervised cross-space framework that transfers disentangled StyleSpace interventions to local Riemannian trajectories in diffusion $h$-space. GeoNaviDiff constructs reconstruction-decoupled supervision by extracting a spherical Log-map tangent from
manipulation - arxiv:2610.26090 · cs.CVRGB--Event Object Detection Across Operating Frequencies via Reliability-Supervised Routing and Asymmetric-Resolution FusionPeijun Shi, Jiannan Zhao, Xiujin Zhu, Zhenguo Lao +2
Autonomous vehicles must detect objects faster than a frame camera delivers images. Event cameras record brightness changes with microsecond resolution, so RGB--event detectors can decide between frames, at an operating frequency, the rate of decisions, far above the frame rate. As this rate rises, each decision receives a sparser event window, and the reliability of the events relative to the frame varies over the scene and over time. Existing detectors weight the sensors per sample and encode the frame at the reduced event resolution. Reliability-supervised routing assigns a fusion weight to
event camera - arxiv:2610.26088 · cs.LGLeakBench: Auditing Reward Hacking through Prediction-Time Leakage in Clinical Modeling AgentsXiaoyuan Bao, Jingsi Jiang, Jiao Shan, Yulong Cao +3
Autonomous clinical-modeling agents can optimize predictive performance while violating conditions that make a model usable. We introduce LeakBench, an executable, guideline-derived audit suite that tests agent decisions for prediction-time leakage, calibration, site-level evaluation, support awareness, and decision-curve net benefit. LeakBench combines a synthetic task with known failure structure and two real-data settings: MIMIC-IV (430k admissions) and governed multi-hospital inpatient record front-page (MHFP) data. Our primary analysis uses reward hacking descriptively, without attributin
agentbenchmark - arxiv:2610.26081 · cs.CVDeOverlay: A Benchmark and Source-Anchored Flow for Video Overlay RemovalDeshui Miao, Yuhang Ma, Shaoyong Jia, Wenzhao Gao +3
Video overlay removal aims to recover scenes obscured by subtitles, titles, and logos while preserving surrounding content. Existing methods use temporal propagation or source-conditioned generation, but can leave residual overlays and distort scene structure when useful source information is obscured. We propose DeOverlay, a source-anchored flow model that reconstructs the clean video from the source latent with fixed source guidance. Specifically, we train a shared velocity network along paired source-to-clean paths using region-weighted flow matching, enabling successive cor- rections from
benchmarkevaluation protocol - arxiv:2610.26079 · cs.CVToolkit for Facial DeidentificationEsteban Leiva-Montenegro, Matthieu Pillot, Manfred Gonzalez-Hernandez, Vitomir Struc +2
This work presents a novel open-source toolkit designed for the evaluation of face deidentification techniques with available evaluation models. The framework provides a complete evaluation pipeline, including preprocessing, deidentification, and performance assessment, based on data utility, image quality, and identity protection metrics. It supports a variety of datasets, enabling researchers to easily benchmark new deidentification techniques. Results show the effectiveness of some techniques on maintaining privacy while preserving image utility. The toolkit is aimed at unifying future rese
benchmarkevaluation framework - arxiv:2610.26071 · cs.AICUA-Lite: An Open Platform for Computer Use AgentsZhanhui Zhou, Weichen Zhang, Haoran Liu, Lingjie Chen +5
Open resources for computer use agents (CUAs) -- environments, datasets, and models -- remain scattered: without a standard, these resources cannot be integrated and scaled; additionally, evaluation, RL, and SFT cannot share one infrastructure. We present CUA-Lite, an open platform for developing CUAs, built around three standardized abstractions: (1) Unified environment interface, under which we integrate 12 benchmarks across desktop, browser, and mobile, plus lightweight VM-free sandboxes that replicate OSWorld's desktop at much lower cost, providing over 30,000 verifiable tasks. (2) Unified
benchmark - arxiv:2610.26069 · cs.CVRisk Decomposition for Hallucination-Aware Video SamplingChenyan Zheng, Fan Wu, Ruibo Li, Jiaxing Li +3
Despite achieving high visual fidelity, video models still suffer from physical hallucinations, limiting their reliability as world simulators. Existing methods leverage model-internal or external signals to improve physical plausibility, but their roles and limitations require a unified explanation. We propose a two-dimensional physical-hallucination risk framework that defines low-density sampling risk and high-density drifting risk. The first risk captures generations assigned low likelihood by the video model, for example, a ball suddenly teleporting despite the model favoring continuous m
benchmark - arxiv:2610.26066 · cs.AIFinSafeGuard: A Quality-Driven Synthetic Dataset for Financial AI SafetyJoseph Thomas Thacil, Reetu Raj Harsh, Krati Saxena, Bhaskarjit Sarmah +1
Generating high-quality synthetic datasets at scale remains a fundamental challenge for training specialized AI systems in data-scarce domains. We present a comprehensive methodology that produces 14.3 million quality-filtered synthetic training examples from a structured taxonomy of 632,000 scenario specifications. Our four-stage pipeline ad- dresses critical challenges in synthetic data quality, diversity, and realism through: (1) systematic pre-generation research establishing optimal generation parameters via diminishing returns analysis, (2) conditional auto evolution mechanisms that reme
llm-as-judge - arxiv:2610.26062 · cs.CVProcedural Exposure Bias in Skill-Augmented Language Model AgentsQing Yang, Yuqi Xiong, Pengcheng Huang, Zhipeng Xu +4
External skills provide language-model agents with reusable procedural knowledge, yet how models use such knowledge after it is supplied remains poorly understood. We identify Procedural Exposure Bias (PEB), the tendency of language models to increasingly favor an action as its associated procedural pattern is repeatedly exposed in a skill, even when the interaction state and action relevance remain unchanged. We find that PEB strongly depends on how procedural knowledge is represented: presenting the same content visually substantially attenuates exposure-induced shifts in action preference r
agent - arxiv:2610.26059 · cs.AIHoneyed Words, Hidden Daggers: Benchmarking and Training LLMs to See Through Mock PolitenessCaixin Kang, Sitong Gong, Tianyu Yan, Kaipeng Zhang +2
The Chinese idiom "kou mi fu jian" ("honeyed words, hidden daggers") names speech that is polite in form but hostile in function. We present ZhenhuanBench, a 1,592-item benchmark built from expert commentary on Empresses in the Palace, covering speaker intent, emotion-language mismatch, and per-listener uptake. Text-only evaluation covers 34 LLMs. GPT-6 Astra leads the leaderboard at 73.48% overall, while its listener-reaction score is 58.37%. The annotations also supply supervised traces and string-verifiable training rewards. With 338 gold traces, an 8B model outperforms teacher distillation
benchmarkleaderboard - arxiv:2610.26046 · cs.CVAVTime: Reinforcing Dense Audio-Visual Captions with Bidirectional AlignmentMingyang Wu, Kaituo Feng, Bohao Li, Kaixiong Gong +2
Dense temporal audio-visual captioning is important for multimodal video understanding and generation. However, prior work is constrained by three main limitations: (1) the lack of a dataset combining high temporal resolution, longer videos, and bidirectional alignment between time and events; (2) the reliance of most existing RL methods on checklists; and (3) the lack of a benchmark that bidirectionally evaluates event semantic correctness and temporal accuracy. To address these challenges, we propose: (1) AVTime-50K, a high-quality 50K-scale dataset with temporally precise, semantically accu
benchmark - arxiv:2610.26043 · cs.ROOmniInteract: Towards Unified Whole-Body Interaction Control for Dual HumanoidsYinghao Li, Tianyu Zhang, Zhuohao Li, Yi-Lin Wei +3
Physical interaction is a key capability for humanoid collaboration, requiring coordinated whole-body motion and precise contact. However, learning diverse dual-humanoid interactions faces two major challenges: (1) human-robot morphological differences disrupt contact interfaces and spatial relationships during retargeting, and (2) jointly learning motor skills and partner coordination makes it difficult to retain control fidelity as interaction diversity grows. In this paper, we propose OmniInteract to address these challenges through surface-aware modeling and scalable residual policy learni
humanoid - arxiv:2610.26040 · cs.CLBalance-Aware Pivot Selection for Parallel Diffusion LLM DecodingYejin Lee, Hyundong Jin, Yo-Sub Han
Diffusion language models enable parallel token generation, offering a promising alternative to sequential autoregressive decoding. Existing acceleration methods use predictive confidence to enable parallel token commitment while preserving generation quality. However, confidence-based selection prioritizes reliable token predictions without explicitly considering how token positions shape the remaining decoding schedule. Moreover, such ordering does not always reflect inter-token dependencies, suggesting that confidence need not be the sole criterion for token selection. Motivated by this lim
benchmark - arxiv:2610.26036 · cs.LGFinHopper: A Progressive Multi-Hop Benchmark for Complex Financial Agents with Step-wise EvaluationYuxuan Wu, Miaopei Lin, Qiufang Ying, Lin Dan +8
Financial agents are expected to answer factual investment queries through open search, but financial multi-hop retrieval differs from general multi-hop QA in a crucial way: each intermediate answer is an exact typed constraint, such as an entity, date, event, or numerical indicator, that determines the next query. This strict answer-conditioned dependency makes small intermediate errors cascade into plausible but wrong search trajectories. We introduce FinHopper, a benchmark for complex financial agents with controlled reasoning depth from 1 to 6 hops, covering stocks, funds, and macroeconomi
benchmark - arxiv:2610.26035 · cs.AIDecision4Rec: A Jev-Inspired Framework of Decision Models on Knowledge and Active Graphs for Multi-Agent Recommender SystemsMarco Polignano, Het Darshan Mehta, Pasquale Lops, Ernesto William De Luca +1
Decision models, such as the commercial Jev, are language models that return one probability for each option of a question instead of generating text. We present Decision4Rec, an open and modular framework that makes decision models useful for recommendation. A knowledge graph writes the logs of each recommendation scenario as short statements on a defined scale, such as "sold: top 10%". A pool of agents, proposes candidates, and a residual ranker keeps the best 30. A decision model, trained once on real choices of three scenarios, reads them through one interface of typed questions, and a fus
knowledge graphmulti-agent - arxiv:2610.26031 · cs.AIAuDQ: Auditable Data Quality through Agentic Rule DiscoveryJacopo Biggiogera, Nicolas Colombo, Krati Saxena, Bhaskarjit Sarmah +1
High-stakes tabular pipelines rely on scattered, static quality checks that are hard to evolve and audit against policy intent. Regulated domains particularly suffer as the LLM-era automation raises expectations for traceability and accountability. We propose a vision for an ontology-grounded, repository-native data quality layer in which rules are first-class artifacts: versioned, provenance-linked to governance texts, and re-playable through immutable execution traces. As a first step toward validating this vision, we present a concrete implementation of its agentic rule-discovery core with
agentichuman-in-the-loopevaluation framework - arxiv:2610.26026 · cs.CVAfter the Federation: Benchmarking Knowledge Portability to Heterogeneous NewcomersXinyuan Zhao
Once historical clients become unavailable, a federation must serve newcomers through the artifact it leaves behind. How much of its knowledge remains useful in an architecture chosen afterward? We introduce FEDNBO, a seven-dataset benchmark that separates knowledge supply, policy identification, and private-model compilation under disjoint data roles. Controlled studies expose inflated oracle gains, architecture-dependent transfer reversals, and a gap between teacher accuracy and private-model utility. Ordinary public-response distillation provides a competitive reference without public task
benchmark - arxiv:2610.26022 · cs.CVFrameGRPO: Incentivizing Video Reasoning via Multi-Frame Rollouts and Temporal CreditJiaxuan Sun, Biao Wang, Shuai Li
Reinforcement learning with verifiable rewards, particularly Group Relative Policy Optimization (GRPO), has improved video reasoning in multimodal large language models. However, standard video GRPO varies only linguistic trajectories while conditioning every rollout in a group on the same sampled frames, which can produce homogeneous rewards and weaken group-relative learning when the shared visual evidence is insufficient or distracting. We propose \textbf{FrameGRPO}, a simple and effective reinforcement-learning framework that requires neither auxiliary networks nor changes to the model arc
benchmark - arxiv:2610.26020 · cs.AIUnderstanding When Reflective Prompt and Skill Optimization Helps on Locally Served Reasoning Models: A Two-Model Study of GEPA and SkillOptAndrea Giummarra, Carmelo Baglieri, Antonio Di Maria, Andrea Farruggio +2
Reflective text-space optimizers such as GEPA and SkillOpt have reported large gains on frontier and mid-size models. We test replication across two reasoning-native target-serving configurations: DeepSeek V4 Flash through DwarfStar and Qwen3.8-27B through LM Studio. For each target, we evaluate the same held-out tasks and four conditions - minimal baseline, GEPA prompt, SkillOpt skill, and manually engineered prompt - on HotpotQA (n=500 per condition) and MATH-500 (n=250). In the primary Qwen v2 rerun under the same frozen protocol, GEPA selects the HotpotQA seed and exactly reproduces baseli
benchmarkevaluator - arxiv:2610.26009 · cs.LGProtGlycanDock: Towards Accurate Protein-Glycan Docking with Tailored Dataset, Benchmark and ModelMinghao Xu, Jiaze Song, Jinyu Wang, Wentao Zhang
Performing accurate protein-glycan docking is critical to understanding protein-glycan interactions which govern many biological processes. However, this important task is hampered by the lack of tailored datasets, benchmarks and models for protein-glycan docking. To fill this blank, we first curate a standard benchmark ProtGlycanDock for training and evaluating protein-glycan docking models under both protein and glycan generalization settings. On this benchmark, we evaluate ten models in three categories, physics-based docking methods, machine-learning-based docking models and molecular stru
benchmark - arxiv:2610.26006 · cs.CLScaling Synthetic Pre-training Data and Evaluation for Indic LLMsAbhinav Khattar, Aishwarya Chhabra, Ameya Velingker, Anusha Kamath +23
Building capable language models for India requires scalable multilingual data and robust evaluation. We introduce Nemotron-Pretraining-Indic-v1, a synthetic pretraining corpus of approximately 2.83 trillion tokens covering all 22 scheduled Indian languages in native script and twelve in romanized form. It is the largest openly released Indic pretraining corpus. Using our GPU-accelerated pipeline, we translate curated English documents, assess translation quality, and repair low-quality outputs while preserving technical content and document structure and maintaining source traceability. Throu
human-in-the-loopbenchmark - arxiv:2610.25999 · cs.LGTempoMoE: Identifying and Steering Fast-Slow Thinking Experts for Efficient Multimodal ReasoningXiyue Zhang, Keru Zhou, Ziyi Wang, Bing Hu +1
Multimodal Mixture-of-Experts (MoE) reasoning models rely on long chain-of-thought (CoT) for complex tasks, yet often overthink easy inputs, increasing token cost and latency, and sometimes degrading accuracy. Existing methods optimize reasoning length during training or impose external control at inference, but leave the underlying reasoning mechanism a black box. We find that fast and slow thinking exhibit distinct expert-routing patterns: holding the input fixed and varying only the prompted reasoning tempo produces stable routing shifts. Suppressing slow-thinking experts shortens the reaso
benchmark - arxiv:2610.25996 · cs.AIProcedural Memory for Language Agents: A Neurosymbolic Perspective on Explicit Goal StructureJonathan Leung, Zhiqi Shen, Deheng Ye
Large language models remain unreliable on long-horizon tasks requiring procedural reasoning. We argue that a key cause is the absence of explicit procedural memory: accessible, modifiable knowledge of how tasks are accomplished, distinct from semantic and episodic memory. Goal-Oriented Graphs offer such a representation, encoding goals and prerequisite dependencies as inspectable symbolic structure. We examine their construction, the challenges they pose, and principles for integrating them with neural reasoning.
memoryepisodic memory - arxiv:2610.25993 · cs.LGGraph-to-Table Interfaces Control Relational GeneralisationAlessandro Cagiano, Lorenzo Molfetta, Gianluca Moro, Matteo Zignani +1
Graph-to-table systems combine three design choices: an encoder that maps relational evidence to feature rows, a context constructor that selects labelled examples, and a tabular decoder. We investigate whether these components can be evaluated independently through matched comparisons that change one component while holding the rest of the pipeline and the evaluation constant. Across logistic regression, CATBOOST, and TABPFN-3 on CLUTRR, 57 GraphLog worlds, and 57 knowledge graphs, we show that conclusions about one component can change when the others change. On CLUTRR, absolute positional e
knowledge graph - arxiv:2610.25991 · cs.AIMap-Distribution Mixture-of-Experts for Multi-Agent PathfindingFumimaro Suzuki, Keisuke Okumura
Designing capable neural policies for multi-agent pathfinding (MAPF), which map local observations to action distributions, represents an important challenge in multi-agent learning, with practical relevance to scalable collective automation problems such as warehouse logistics. In contrast to prior studies focusing primarily on individual policy design, this study explores a meta-level Mixture-of-Experts (MoE) approach to learning-based MAPF. Specifically, given a collection of compact yet high-performing experts specialised for different map distributions, our architecture learns a router th
multi-agent - arxiv:2610.25988 · cs.ROSelf-Supervised Inertial OdometryYuheng Qiu, Yutian Chen, Carlton Cort, Weison Huang +9
Learning inertial odometry for a new platform usually requires ground-truth trajectories. We show that an inertial odometry model can instead be trained from scratch by a myopic teacher that integrates the IMU over each short window independently, a scheme we call integration self-supervision.Interestingly, although supervised only by short-window velocity changes, the student learns to directly predict velocity, and does so more accurately than its integration teacher. Pose graph distillation then refines the model with sequence-consistent targets from a pose graph over many windows. Training
humanoidquadruped - arxiv:2610.25985 · cs.LGNIOCIS: Typed-Decision Next-Hop Routing with Node-Conditioned Option Spaces for Tool-Use AgentsJin Han, Mingfeng Xin, Feng Liang, Jingtao Yu
Tool-use agents pay per decision: at each hop a language model re-reads the state and thinks for seconds, though its actions are finite and decisions recur. We present NIOCIS, a routing layer that folds execution logs into a loop-collapsed action graph (forward transitions plus recovery back-edges) answering one typed question: which declared option at a state, plus stop, does the fleet take next? A small model reads a calibrated distribution over them in one pass, emitting no text. On 1,615 production security-operations episodes we measure what the layer replaces, returns and is bounded by.
tool-use - arxiv:2610.25981 · cs.ROFR-VLA: Implicit Force Reasoning VLA for Contact-Rich ManipulationJiaju He, Zhiyuan Yang, Ruyang Han, Jiahong Tang +2
Integrating force awareness into Vision-Language-Action (VLA) models involves not merely encoding wrench measurements as an additional modality, but determining when and how strongly physical interaction should influence robot action generation. Existing VLAs that use force typically inject it through a fixed pathway, without explicitly determining when it is relevant to the current action. This calls for an intermediate force relevance mechanism inside the VLA, in which visual-language context estimates the relevance of physical feedback to the current task before force enters the action path
vision-language-actionvlamanipulation - arxiv:2610.25980 · cs.AIClueFlow: Learning Long-Horizon Search through Attributed Research-State TransitionsYicheng Wang, Wentao Tan, Yidan Huang, Yifei Liu +1
Long-horizon search requires agents to acquire relevant evidence and integrate it into an evolving research state. Existing methods maintain research states to guide exploration or assign turn-level progress signals, yet grounding these signals in explicit, attributable state changes remains underexplored. We introduce ClueFlow, which learns long-horizon search from attributed research-state transitions. ClueFlow constructs Source-State-Final attribution chains that connect evidence-acquiring browser actions, Evidence and Candidate updates, and the final answer supported by those updates. Duri
benchmark - arxiv:2610.25973 · cs.LGCorrectness-Dependent Asymmetric Target Distributions for Self-Distillation Policy OptimizationSuwan Yoon, Seongryong Jung, Changhee Lee
Self-distillation policy optimization provides dense token-level supervision for LLM post-training through feedback-conditioned self-teachers. However, directly matching the self-teacher can disrupt successful behaviors on correct rollouts while providing insufficient correction on incorrect ones. We propose Asymmetric Policy Optimization via Self-Distillation (APSD), which constructs correctness-dependent targets by mixing student and self-teacher distributions: interpolation preserves successful behaviors on correct rollouts, while extrapolation beyond the self-teacher strengthens correction
post-training - arxiv:2610.25970 · cs.AIGLOBE: Geometry-Led Online Behavioral Early Exit via Unit-Sphere TrajectoriesXinyue Peng, Zhenlin Liang, Meng Cai, Hua Wang
Chain-of-thought (CoT) reasoning can improve problem solving, but models often keep generating after the correct answer is already reachable; the extra steps raise serving costs and may replace a correct intermediate result with an incorrect one. We propose GLOBE (Geometry-Led Online Behavioral Early Exit), an online, training-free algorithm that terminates the model's CoT process early. GLOBE embeds each completed reasoning segment as a point on a unit sphere and reads the shape of the resulting semantic trajectory. GLOBE first applies a trial phase that generates short trial answers in a sep
agent memoryagenttool callingbenchmark - arxiv:2610.25967 · cs.ROReconstruction and Discrete-Event Simulation Assessment of an AnalyticalMohsen Reda Mahmoud Moursi, Ahmed Stohy
This article presents an analytical-DES benchmark for intra-aisle robotic picking in automated warehousing. The investigated cycle combines horizontal shuttle travel, tote handling and item picking at storage locations. Simulation results are compared with the analytical formulation of Lerher et al. under the same rack geometry, shuttle-motion, handling and picking conditions, and the same definitions of cycle time and derived item throughput. Reproducing the published analytical tables requires the distance interpretation dES = 1.5 m and dTB = 0 m. The analytical-DES benchmark comprises 180 s
benchmark - arxiv:2610.25964 · cs.CVToken-Set Choice Confounds POPE: A Systematic Audit of Yes/No Extraction in VLM Hallucination EvaluationKesav Kumar Jayakumar, Karthigeyan Thilak
We set out to reproduce a recent hallucination-mitigation result on the POPE benchmark. The LLaVA-1.5-7B baseline we measured, 82.21 F1 on the adversarial split, sat 1.21 to 8.67 points above the adversarial baselines we planned to compare against (81.00, 79.30, and 73.54). We looked at the evaluation code and found a readout choice that moves F1 on its own. A two-token logit readout decides yes versus no by comparing two vocabulary IDs, 3582 ("yes") and 1217 ("no"). Across 9,000 greedy-decode questions spanning all three splits, neither ID is generated even once. The model emits " Yes" (token
benchmark - arxiv:2610.25962 · cs.ROSkateNav: Learning Energy-Efficient Navigation through Adaptive Skating on Passive-Wheel Quadrupedal RobotsJeil Jeong, Taeyun Kim, Sangjin Kim, Jinyoung Choi +1
Passive-wheel quadrupedal robots offer energy-efficient locomotion through leg-driven propulsion and momentum-driven gliding. However, navigation requires coordinating steering with intermittent propulsion and adapting push-glide timing to task demands. Velocity-tracking objectives may suppress the speed variations inherent in skating, discouraging efficient gliding. To address these challenges, we propose SkateNav, a unified skating-navigation framework based on reinforcement learning. A time-budgeted goal-reaching objective allows the policy to jointly learn steering, propulsion, and their t
quadruped - arxiv:2610.25961 · cs.ROPaceVLA: Adaptive Temporal Action Quantization in Vision-Language-Action ModelsHojin Jeon, Jimin Lee, Taekwan Kim, Nayoung Oh +5
Fast and reliable manipulation is crucial for practical robot deployment. Although Vision-Language-Action (VLA) models have recently shown remarkable progress as generalist robot policies, their execution often remains temporally inefficient. A key source of this inefficiency is that policies learn fine-grained temporal patterns from teleoperated demonstrations, requiring many low-level action steps. Temporal action quantization offers a simple way to accelerate execution by aggregating consecutive low-level actions into temporally coarser actions. However, applying action quantization without
vision-language-actionmanipulationdexterousbenchmark - arxiv:2610.25959 · cs.AIEmbodiedWorld-200K: Toward Embodied World Agents via Embodied-Planning Dataset and Interactive World ModelsX. Feng, J. Tang, Z. Lin, L. Chen +3
Building embodied agents that can perceive, reason, and act across diverse 3D worlds is a central research direction toward artificial general intelligence. Yet most existing work remains confined to specific tasks in a single scene and transfers poorly to new environments, leaving general embodied agents for open-world settings under-explored. We refer to such agents as Embodied World Agents (EWAs), whose defining property is the ability to understand diverse visual environments and execute general actions in open-world conditions. To advance EWA research, we study open-world embodied navigat
embodiedworld modelembodied agentpost-training - arxiv:2610.25957 · cs.LGProductive State-Space Exploration for Multi-turn RL in LLM AgentsMohammed Haroon Dupty, Yong Liang Goh, Song Guo, Wee Sun Lee
Reinforcement Learning (RL) for Large Language Model (LLM) agents in multi-turn, long-horizon environments remains challenging due to inefficient action selection and repetitive behavioral patterns. In particular, agents often become trapped in high-confidence yet unproductive behaviors, repeatedly executing invalid actions or entering action loops that hinder task completion. We observe that these frequently violate environment preconditions, revealing a mismatch between the agent's representation of the internal state and the dynamics of the underlying environment. In addition, repetitive hi
llm agent - arxiv:2610.25954 · cs.AICXBench: A Benchmark for Process-Compliant LLM Agents in Regulated Customer-Experience WorkflowsBoxi Yu, Yuzhong Zhang, Zhenfeng Su, Linfu Xi +6
Tool-using language agents increasingly perform regulated customer-service workflows, where a correct final state is insufficient: agents must also provide disclosures, obtain consent, and follow required procedures. Existing tool-agent benchmarks largely score final business states, accepting trajectories that reach correct outcomes through non-compliant processes. We introduce cxbench, a benchmark that evaluates outcome and process compliance separately across six regulated domains. Each task combines a customer request, a tool-based environment, and explicit process requirements. Rule cards
llm agentagent benchmarkbenchmark - arxiv:2610.25950 · cs.AIAdaExploreBench: Benchmarking LLM Agents for Proactive Exploration and Test-time Adaptation in Evolving EnvironmentsRui Wang, Zishen Lai, Jiayu Liu, Cheng Qian +14
Large Language Model (LLM) agents are increasingly deployed in unfamiliar and evolving environments, requiring them to proactively explore, understand, and adapt to environmental dynamics. However, existing benchmarks rarely test whether agents can proactively discover how an environment works and adapt as it evolves. To address this gap, we introduce AdaExploreBench, a benchmark for proactive exploration and test-time adaptation in evolving environments. By instantiating environmental dynamics as hidden rules, AdaExploreBench requires agents to actively acquire evidence and infer the rules. U
llm agentbenchmark - arxiv:2610.25947 · cs.AIClaim Laundering: When Agent Decision Models Trust Records Over RealityTina Khezresmaeilzadeh, Seyedarmin Azizi, Erfan Baghaei Potraghloo, Mehdi Kamal +1
Agent decision models select actions in a single forward pass from inputs that combine environmental observations with the agent's own record of previous actions. When an action fails silently, these sources can conflict: the record reports success, but the environment shows that the step did not take effect. We ask which source determines the next decision. Across Android screen traces, ALFWorld, and WebShop, we construct paired states that independently vary the observation, the logged action, and an explicit success claim. In all 12 configurations of four decision models and three domains,
agent - arxiv:2610.25946 · cs.ROPolicy Search in Code Space for Multi-Agent PathfindingMatthias Eder, Pantelis Emmanouil, Vivek Farias, Fabian Fischer +6
Developing coordination policies for complex multi-agent systems relies on either labor-intensive algorithm design or opaque task-specific reinforcement learning. With recent advances in autonomous coding agents, we show that policy search directly in code space offers a compelling third path that combines automated optimization with explicit interpretability. We investigate a human-directed workflow in which coding agents iteratively propose, implement, evaluate, diagnose, and eventually refine algorithms guided by simulation feedback and explicit correctness verification. We conducted two ca
multi-agentagent system - arxiv:2610.25932 · cs.LGKeep the Lineages Alive: Diversity-Preserving SMC for Diffusion LLM DecodingMinoo Ahmadi, Seyedarmin Azizi, Mehdi Kamal, Massoud Pedram
Inference-time sequential Monte Carlo (SMC) decoding can improve the reasoning accuracy of masked diffusion language models (diffusion LLMs) without post-training by resampling a population of partial generations according to a reward. Its effectiveness therefore depends on the resampling step, which must concentrate computation on promising candidates while retaining enough diversity to explore alternative reasoning paths. The reward can be internal, the decoder's own confidence on the committed prefix, or external, a process reward model scoring that prefix. Under either reward, multinomial
post-trainingbenchmark - arxiv:2610.25931 · cs.LGEvoSkill v2: Fully Agentic Skill Evolution from Benchmark Adaptation to a Stateful EvolverDastin Huang, Baran Nama, Salaheddin Alzubi
EvoSkill v2 is a three-agent system that edits the instructions (skills) an agent is given, leaving its model and harness intact. The user starts a run by describing a benchmark and supplying a grader, which a benchmark adaptor turns into a benchmark-specific plugin. A solver attempts the tasks, and an evolver reads the solver's training rollouts and rewrites the skills at each epoch. We implement the evolver as one stateful coding agent with a bash tool for the whole run, unlike the two most common designs among the skill optimizers we surveyed: a sequence of single model calls, as in SkillOp
agentagenticagent systembenchmark - arxiv:2610.25930 · cs.AITESS: LLM-guided Translation, Exploration, Simulation, and Search for Embodied Task PlanningPeiran Xu, Yixiang Yang, Jinghan Li, Fei Li +3
In this work we focus on the problem of embodied task planning, which requires an agent to accomplish a complex task by separating it into atomic actions. Recent studies have developed LLM-based methods to address this problem. However, the executability and correctness of the generated plans remain suboptimal in complex environments and tasks. On the other hand, classical search algorithms have the potential to solve the planning problems perfectly, yet they require fully observable environments, known state transitions and goal state, which are often unattainable in general task domains. The
embodiedagent - arxiv:2610.25929 · cs.CVOmniGround: Strengthening Omni-Modal Temporal Grounding via Attention-Consistency TrainingZixuan Chen, Depeng Wang, Hao Lin, Ke Xu +4
Omni-modal large language models can jointly understand video, audio, and text, yet they remain weak at temporal grounding, a capability essential for real-world audio-visual applications. To understand this limitation, we look beyond final predictions and examine the model's internal temporal evidence. We identify a temporal generation gap: the internal attention of the model localizes the target moment more accurately than the timestamps it generates. To close this gap, we propose TACG (Temporal Attention-Consistency Grounding), a post-training method that treats internal attention as a reli
post-training - arxiv:2610.25927 · cs.ROCondenseVLA: Learnable History Condensation for Efficient Multi-Frame VLAFeiyang Hong, Yujie Wei, Bo Zhao, Xiu Su +2
Robot manipulation often depends on short-term cues - such as brief contacts, temporary occlusions, and intermediate object states - that are ambiguous in a single frame and can only be resolved by integrating observations over time. However, injecting historical information solely by enlarging the history window is computationally costly and frequently yields limited gains due to substantial redundancy and task-agnostic, inflexible sampling of observations. We propose CondenseVLA, a multi-frame VLA policy with a learnable decoder-side history condensation module for using short-term visual co
vlavla policymanipulationlibero - arxiv:2610.25926 · cs.CVMIRA: A Multimodal Illustrated Deep Research Agent for Controllable ReportsHongrui Wu, Kaiwen Tuo, Congcong Wang, Gen Li +6
Deep research agents are increasingly capable of searching the open web and synthesizing evidence, yet they remain poorly suited to controllable report generation in multimodal deep research. This task requires agents to handle interleaved textual and visual inputs, gather multimodal evidence through multi-step web search, and deliver comprehensive illustrated reports under user-specified requirements. Existing systems face two limitations. During reasoning, they often underuse images discovered along the search trajectory; during generation, they tend to produce fixed-form answers with limite
agentagent benchmarkbenchmark - arxiv:2610.25924 · cs.LGUtilityBench: Measuring Language Model Utility When Errors Have Different CostsQifeng Lei, Zhiyong Yang, Zitai Wang, Shilong Bao +3
As language models are increasingly deployed in real-world scenarios with important outcomes, the impact of their mistakes often extends far beyond what correctness can capture. In domains like medicine, a single error can lead to serious consequences and put patient safety at risk. However, mainstream evaluation protocols still treat all errors as equal, overlooking the spectrum of real-world costs. To address this, we propose a utility-based evaluation paradigm that explicitly incorporates the varying consequences of errors. Guided by this principle, we develop UtilityBench, a large-scale be
benchmarkevaluation protocol - arxiv:2610.25920 · cs.LGMaestro: Agentic Formal Proof Development by CoordinationHaoyu Li, Zeren Shui, Yuhao Zhang, Ferhat Erata +3
Lean is increasingly used to verify mathematical proofs and software produced by LLM agents, but writing formal proofs remains hard. Existing prover systems take a divide-and-conquer approach that recursively decomposes hard proof obligations into easier lemmas and splits again any lemma that resists proof. However, a lemma can resist proof for different reasons, including insufficient guidance, a false statement, or a flaw in the decomposition that produced it. Existing systems cannot tell these apart as a failed attempt reports little beyond its failure, and responses are confined to the fai
agentllm agentagenticbenchmark - arxiv:2610.25919 · cs.CLLEAPS: Long-context Evaluation of Adaptive Pattern Learning and State TrackingHoward Yen, Nitya Mani, Alex Renda, Christopher S. Hardin
Long-context language models (LCLMs) power agents that act over long trajectories, from improving a codebase to running research experiments to forecasters that condition on lengthy histories. These applications draw on three distinct long-context capabilities: recall, where the answer is a fact present in the context; state tracking, where the outcome can be computed by performing a series of transformations to a specified state; and pattern learning, where the goal is to make a prediction by sampling from a distribution inferred from many observations. Existing benchmarks primarily focus on
long-contextagenticbenchmark - arxiv:2610.25916 · cs.AISecure Agentic Coding through Counterexample-Grounded FeedbackTue Le, Anh Nguyen Hoang, Gabriel Sasseville, Doseok Jang +2
As coding models and agent harnesses become more capable, reliable security verification is becoming a central bottleneck. Improving secure code generation requires a signal whose reported progress remains aligned with actual program behavior under optimization; otherwise, an optimizer can improve the signal's verdict without removing the underlying vulnerability. Existing model-based approaches require models to either infer program behavior from source code or specify correct outcomes for adversarial tests, placing substantial reasoning demands on the security signal. We introduce \falcon{},
agentagenticbenchmark - arxiv:2610.25914 · cs.CLFisherEdit: Fisher Information Guided Knowledge Editing for Large Language ModelsYizhou Yang, Jiahong Fu, Zhiwu Wang, Kailong Fan +5
Sequential knowledge editing in Large Language Models (LLMs) often leads to a degradation of general capabilities due to cumulative parameter interference. In this paper, we revisit prominent projection-based methods like AlphaEdit and reveal a surprising mismatch: their robustness is primarily driven by update regularization rather than the intended null-space projection. Based on this insight, we propose FisherEdit, a theoretically grounded method built on Fisher geometry, which replaces coarse activation-space projections with importance-aware regularization in parameter space. It penalizes
benchmark - arxiv:2610.25908 · cs.CLCALM: Consistency-Aware Logit Masking in Decoding Space for Mitigating Overthinking of Large Reasoning ModelsShengjing Liu, Feiyang Li, Yuxuan Yang, Yile Wang +1
Large reasoning models (LRMs) have achieved strong performance on challenging mathematical and scientific tasks through long chain-of-thought reasoning, but they often suffer from overthinking with unnecessary reasoning steps. Existing truncation-based methods typically detect consecutive trial-answer agreement and force early exit, yet such local stopping signals do not necessarily indicate final-answer stability and can disrupt useful subsequent reflection or verification, potentially leading to performance degradation. We propose consistency-aware logit masking (CALM), a training-free decod
benchmark - arxiv:2610.25906 · cs.LGAn Analysis of Semantic and Spatio-Temporal Co-evolution for Edge Classification on Dynamic Text-Attributed GraphsTyler Bonnet, Marek Rei
Dynamic edge classification on text-attributed Web graphs, such as review networks, community forums, and news-derived knowledge graphs, requires high quality, adaptive semantic representations in the context of network evolution. Existing dynamic graph architectures typically initialize text feature vectors with pretrained language models and hold them static during training to preserve their semantic structure, forgoing the potential benefits of allowing them to co-evolve with network topology. In this work, we investigate the effects of enabling such co-evolution with a unified text-graph a
knowledge graphbenchmark - arxiv:2610.25899 · cs.CLSWE-Residual: Evaluating Starting-State Robustness of Coding AgentsJinze Li, Weijie Liu, Jinfeng Xu, Shuo Yang +5
Software engineering (SWE) benchmarks are the primary way to measure the capabilities of coding agents: an agent is placed in a code repository or machine environment, given a request, and judged by hidden tests. These benchmarks almost always start the agent from a clean state that contains no earlier work on the task. In practice, however, agents often continue work that a developer or another agent has already done and reported as complete, with no indication of what remains incorrect or missing. We introduce SWE-Residual, a benchmark for Starting-State Robustness: the ability to maintain p
agentbenchmark - arxiv:2610.25896 · cs.AIWhen Should AI Chime In? Benchmarking Timely Contribution of Proactive Agents in Multi-Party ChatRuotong Wang, Brian S. Yu, Xinyi Zhou, Joseph Chee Chang +4
Proactive agents can support group collaboration by surfacing useful information without waiting to be asked, but even relevant contributions can be disruptive if poorly timed. In this work, we isolate the timing question: given a fixed candidate contribution such as a clarification or relevant fact, when should an AI agent chime in? We introduce PITCH (Proactive Intervention Timing in CHat), to our knowledge the first benchmark with large-scale human judgments of proactive agent timing collected from participants in the same unscripted multi-party conversations they took part in. To collect h
agentai agentbenchmark - arxiv:2610.25881 · cs.LGLearn the Interface, Not the Implementation: Output Representation Dominates Training Regime in LLM-Based RTL GenerationSiddharth Patel, Vikram Singh, Siddharth Savyasachi Malu, Rohit Singh
Large-language-model register-transfer-level (RTL) generation has concentrated on how to train the model: fine-tuning, continued pre-training, and reinforcement learning from a verifier. We argue this is the second-order question. On a bounded, formally specifiable task of generating single-error-correcting, double-error-detecting (SECDED) ECC memory wrappers, we hold the task, data, and correctness oracle fixed and vary two axes independently: the training regime (seven settings) and the output representation (four levels, from fully expanded flat Verilog to an eight-field configuration expan
memory - arxiv:2610.25880 · cs.LGAsynchronous On-Policy Self-DistillationXidong Wu, Qi Qi, Yuqiong Ji, Shangqian Gao
Unlike on-policy distillation (OPD), which depends on a separate teacher model, on-policy self-distillation (OPSD) leverages a single language model acting simultaneously as the student and the teacher, with the teacher conditioned on privileged information. Despite this architectural simplicity, synchronous rollout generation dominates overall training time, creating a severe systems bottleneck when scaling OPSD. To address this, we introduce Asynchronous On-Policy Self-Distillation (AsyncOPSD), a framework that decouples rollout generation from learner updates to improve training throughput.
benchmark - arxiv:2610.25877 · cs.LGDeltaTrace: Efficient Signed Attribution for Reasoning Language ModelsChen Liang, Daoping Wu, Donghua Yang, Zhiyu Liang +3
We study how words in the context support or oppose a language model's response and present DeltaTrace, a novel attribution framework that explains the effect of input tokens on a fixed model response. DeltaTrace measures how the response's log-probability changes when the context tokens being explained are masked. To distribute this difference among source tokens, it constructs operator-specific local rules for finite changes and composes them backward through the computational graph using effective secant slopes. Each token is assigned a signed attribution as the inner product of its embeddi
memory - arxiv:2610.25876 · cs.LGLearning to Reason Efficiently with Interventional Step CreditGuanda Li, Xuefeng Bai, Yulong Chen, Zhiyong Wang +2
Large reasoning models often produce lengthy chain-of-thought (CoT) traces mixing useful intermediate steps with unnecessary ones. Recent methods explore step-level signals such as attention mass and step entropy to identify and suppress unnecessary steps. Despite their effectiveness, such signals reflect the model's behavior at a step rather than that step's contribution to the answer. To address this, we propose the interventional step credit (ISC), which takes a causal-intervention view and assesses step importance to the reasoning through targeted parameter-level interventions. Building on
benchmark - arxiv:2610.25874 · cs.ROVision-Language-Delta-Action: Learning Compact Deltas as Actionable Transition Interfaces for Robotic ManipulationLingsen Zhang, Wei Li, Jie He, Henghua Zhang +4
Robotic manipulation is not merely observation-to-action generation, but the realization of task-specified state changes in the physical world. Existing vision-language-action (VLA) methods directly learn observation-to-action mappings, imposing a substantial learning burden due to the gap between semantic understanding and action execution. Although world-action models (WAMs) improve control through future prediction, modeling complete future states wastes capacity on static appearance and action-irrelevant changes. We propose Vision-Language-Delta-Action (VLDA), a paradigm that reformulates
vision-language-actionembodiedmanipulation - arxiv:2610.25871 · cs.LGCODA: Cohort- and Drift-aware Foundation Model for Multimodal Clinical ReasoningChangshuo Liu, Jiaqi Zhu, Wenqiao Zhang, Xiaokui Xiao +1
Recent advances in multimodal foundation models have demonstrated strong performance across diverse clinical tasks. However, existing approaches predominantly operate at the individual sample level and fail to model the interplay between population-level cohort patterns and patient-specific variation that underlie real-world clinical reasoning. In practice, clinicians routinely reason through patient cohorts sharing similar demographic, physiological, or pathological characteristics, while accounting for patient-specific deviations from the group prototype. Motivated by this gap, we propose CO
benchmark - arxiv:2610.25869 · cs.ROAeroIntent: A Synchronized Multimodal Benchmark for UAV Mission Intent RecognitionYutong Chen, Jingxin Mao, Zhiwei Wei, Bing Li +2
Reliable recognition of mission-level UAV intent from multimodal observations is essential for low-altitude airspace surveillance and security. Existing multimodal UAV datasets primarily support detection, tracking, and localization, leaving limited support for evaluating mission-intent recognition across held-out trajectory programs and incomplete radar histories. We introduce AeroIntent, a simulation-based benchmark comprising 990 synchronized multimodal sequences generated by 22 trajectory programs across five mission intents. The benchmark provides radar, RGB, infrared, and depth observati
benchmarkevaluation protocol - arxiv:2610.25865 · cs.AIMinimum-Cost Robust Evidence Certificates for Horn Reasoning under Correlated Source FailuresYukun Yang
A small justification for a rule-based conclusion may disappear when a source is withdrawn, and multiple derivations need not provide independent evidence. We study minimum-cost certificates that preserve a Horn conclusion under every withdrawal of at most k source domains. Evidence records remain distinct from asserted facts, supporting replicated attestations and overlapping failure dependencies. Transversals of minimal-support footprints characterize robustness and explain why disjoint-proof requirements can reject valid certificates. An exact procedure alternates weighted evidence selectio
benchmark - arxiv:2610.25861 · cs.CVActionRoPE: action as coordinates for spatially consistent control in planar-camera game world modelsZhiyang Deng, Yuzhe Wu, Zijun Lin, Yeying Jin
Game world models have made rapid progress toward real-time interactive generation. Yet action control is still predominantly implemented as feature conditioning, leaving the model to learn from training data how each action should move the visual world. This formulation is poorly matched to a broad class of planar-camera worlds, such as top-down and isometric games. In these worlds, player motion induces a geometrically well-defined planar camera displacement. We observe that the key to spatially consistent control in such worlds is to represent the effect of each action directly in world coo
world model - arxiv:2610.25858 · cs.LGAre Self-Evolving Agents Continual Learners?Xiaoyin Chen, Yizheng Huang, Yinglun Xu, Ziyun Zhao +3
Large language model agents are often deployed in environments where new knowledge, tools, and workflows keep arriving. Such an agent is therefore expected to adapt continually, learning new tasks as they arrive without forgetting earlier ones. This resembles the continual learning problem, but at a pace that conventional gradient updates cannot sustain. Nonparametric self-evolving agents are a natural alternative: they adapt by rewriting their own harness, such as prompts, memory, and skills, from task feedback while the model weights stay frozen. We study whether such agents are effective co
memoryagentself-evolving - arxiv:2610.25854 · cs.CLArabic Financial NLPMo El-Haj
Financial communication connects numerical accounts of economic activity with explanations of performance, responsibility and future expectations. Arabic financial natural language processing offers methods for studying these connections across corporate reports, market disclosures, journalism and banking interactions. Its contribution to digital humanities depends on retaining the relationship between a computational representation and the document from which it was made. This chapter follows that relationship from source collection and Arabic text processing to the analysis of vocabulary, ev
benchmark - arxiv:2610.25852 · cs.AIMASIF: Implicit Failures for Process Observability in Multi-Agent SystemsYanwei Huang, Arpit Narechania, Atanu Sinha
Multi-agent systems (MAS) can complete a task despite local execution errors that are corrected, bypassed, or carried forward without individually determining its outcome. We call these implicit failures (IFs) and study them as signals for process observability beyond final task outcome. Across 1,828 traces from nine MAS-benchmark configurations, IFs occur in 54.3% of successful and 75.2% of failed traces. We develop taxonomies of error types and subsequent absorption mechanisms, supported by human and cross-model annotation audits. Counterfactual replay further probes how repairing an IF chan
multi-agentagent systembenchmark - arxiv:2610.25851 · cs.MALearning Who Talks to Whom: Role-Aware Communication Topology Scheduling for Efficient LLM Multi-Agent CollaborationHaibin Li, Wenya Yu, Xuefeng Jiang, Jia Li +3
Multi-agent systems driven by large language models (LLMs) rely on structured communication topologies to facilitate information exchange and collaborative reasoning. However, effectively scheduling these topologies to dynamically determine who talks to whom and when to communicate remains a fundamental challenge in balancing collective reasoning performance against exponential token overhead. Existing topology scheduling approaches either enforce static communication graphs that cannot adapt to dynamic problem-solving states, or overlook agent role complementarity while failing to explicitly
agentmulti-agentagent systembenchmark - arxiv:2610.25847 · cs.LGACE-Trader: Adversarial Co-Evolution for Generator-Critic Trading SystemsChang Zhou, Jiarun Zhou, Xingtong Yu, Hong Cheng +2
Generator--critic trading systems pair an LLM-based trader that makes trading decisions with a red team agent that evaluates these decisions and provides feedback for revision. However, improving individual trading decisions does not ensure that both agents learn from experience and improve together over time. This raises an important question: how can the trader and red-Team select useful experience and learn from each other's evolving behavior? On one hand, experience that benefits decisions under one market condition may become ineffective or harmful under another. On the other hand, tradin
memoryagentbenchmark - arxiv:2610.25843 · cs.AIA Cue's Hidden Price: Auditing Seller Influence on AI ShoppersFelix Buecheler, Gerardo Berbeglia, Mingyu Guo
Large language models are increasingly asked to shop on a user's behalf, which lets seller-controlled content such as badges, endorsements, and framing steer what they recommend. That poses the question at the core of generative engine optimization (GEO): not whether seller content shifts an agent's choice, but how much. The usual gauge of that influence, how often a cue wins the choice, has a blind spot: once a cue is strong enough to be picked almost every time, the choice rate can no longer show how much stronger one cue is than another. We introduce price-equivalent (PE) auditing, which re
benchmark - arxiv:2610.25840 · cs.ROOV-PEAR: Observation-Verified Pose-Scale Estimation and Reconstruction for Unknown Industrial Manipulation ObjectsGuoqin Tang, Qingxuan Jia, Yichen Tan, Zihang Lv +3
Fine industrial robotic manipulation requires 6D pose, scale, and an operation-ready 3D model. Existing industrial pose estimation relies on CAD models, object-specific training data, or stable workcells, limiting use with unseen objects; zero-shot methods suffer from single-view occlusion, generative bias, and scale uncertainty. We present OV-PEAR, an observation-verified framework for 3D reconstruction and pose-scale estimation of unknown industrial objects. Its three-stage object-centric pipeline: (1) segments RGB-D instances and generates object-level 3D hypotheses; (2) anchors them to the
manipulationgrasp - arxiv:2610.25839 · physics.opticsHigh-Efficiency Silicon-Lithium Niobate Grating Coupler enabled by Fully-Etched Subwavelength StructuresJing Wang, Yang Liu, Shitong Su, Weiwen Zou
A new approach of silicon-lithium niobate (Si-LN) grating couplers (GCs) with a fully-etched structure is proposed. The coupling strength is engineered through subwavelength structures, allowing improved directionality. The adjoint-based inverse design method is simultaneously adopted to achieve efficient coupling between photonic circuits and fibers. Considering the birefringence of LN, the proposed Si-LN GCs are optimized and fabricated for TE mode along the crystallographic Z and Y directions. The experimentally measured coupling efficiencies are -3.3 dB and -3.9 dB along the crystallograph
photonic integrated circuitgrating coupler - arxiv:2610.25834 · cs.AIFrontierRefactor: Benchmarking Multi-Scale Performance-Oriented Codebase RefactoringShijian Wang, Fanqing Meng, Jiajun Chen, Qiguang Chen +7
Autonomous codebase refactoring requires coding agents to restructure working software while strictly preserving externally observable behavior and prioritizing execution performance and compute efficiency over stylistic preference. Existing benchmarks: refactoring benchmarks assess functional correctness while overlooking runtime regressions, whereas code performance benchmarks target isolated micro-kernels that ignore codebase-scale complexity. We observe that refactoring scale is governed by the scope of the observable contract, the caller-visible interface fixing behavioral invariants whil
benchmark - arxiv:2610.25832 · cs.CVHihaze: One-Step Diffusion for High-Resolution Real-World Image DehazingZheng Chen, Yuchen Huang, Xiaoyang Liu, Jue Gong +2
Diffusion models have recently achieved encouraging results in real-world image dehazing. However, restoring high-resolution images remains challenging due to spatially non-uniform haze and inconsistent predictions across local patches. To address these issues, we propose Hihaze, a one-step diffusion framework for high-resolution real-world dehazing. Hihaze adopts two haze-aware modulation modules: attentive skip compensation (ASC) and a spatial degradation adapter (SDA). ASC selectively transfers reliable encoder information to the variational autoencoder (VAE) decoder, preserving structural
benchmark - arxiv:2610.25827 · cs.AIWhen Retrieval Partially Conflicts with Skills: Understanding Procedural Hallucination in Language ModelsGuanqun Jiang, Pengfei Cao, Kang Liu, Jun Zhao
Language models increasingly use skills that provide reusable instructions and procedures for task execution. After a relevant skill has been selected, the model may need to apply its governing procedure together with retrieved task context. When the retrieved context is useful but contains cues that violate the skill, the model must retain compatible information while rejecting the conflicting cues. We study failures in which the model adopts such cues and produces a retrieval-supported but skill-invalid answer, which we call procedural hallucination. We introduce SourceClash, a matched bench
benchmark - arxiv:2610.25819 · cs.AIWhen to Stop Asking: Factor Coverage Bias and LLM Determinability Judgments with Underspecified Information in MedicineYusuke Watanabe, Yohei Kobashi, Taisei Tosaki, Takeshi Kojima +3
Clinical decision-making often begins with underspecified information and requires not only gathering missing information but also knowing when to stop. This calls for judging determinability (whether the available information is sufficient for a decision) by considering whether missing information could change the conclusion, not simply the amount of available decision-relevant information. Reliance on this amount as a cue for determinability, which we term \textit{factor coverage bias}, could undermine the trustworthiness of large language models (LLMs) through premature conclusions or exces
benchmarkevaluation framework - arxiv:2610.25812 · cs.LGIs it RL or Data? A Controlled Comparison of GRPO and Rejection-Sampling Fine-TuningLakshya Chaudhry
Bauer et al. (2026) showed that in low-data reinforcement learning with verifiable rewards, training GRPO on mixed-difficulty prompts gives up to 5x the sample efficiency of easy-only prompts. They did not test whether this benefit is specific to GRPO, or whether another post-training method given the same data would obtain it too. We test this on their procedural counting task with Qwen3-4B-Base and LoRA, comparing GRPO to rejection-sampling fine-tuning (RFT) on the same prompts, verifier and token cap. Six primary arms cross the two methods with three prompt sets (Easy, Mixed, and a Curated
post-training - arxiv:2610.25806 · cs.LGSector-Ring Linear Reservoirs: Sector-Wise Decay-Timescale Allocation With State Reuse Across Candidate SettingsYudai Ebato, Sou Nobukawa, Yusuke Sakemi, Takashi Kanamaru +2
In reservoir computing, performance depends on whether the fixed dynamics of the reservoir suit the task, and the appropriate memory length and frequency response differ across tasks. The reservoir settings are found by a search on each task, and driving the reservoir with the input and collecting its states for every candidate setting dominates the cost of the search. In a reservoir with a linear recurrence, the dynamics decompose into modes, one per eigenvalue, and the eigenvalue distribution becomes a design variable. We propose the Sector-Ring Linear Reservoir (SRLR), in which the eigenval
memorybenchmark - arxiv:2610.25790 · cs.CLCoMem: Learning to Anticipate Failures from Retrieved ExperienceYun Shen, Ming Shen, Tianjun Yao
Reflection lets LLM agents improve from their own attempts, but the memory it builds is local to a single task, so an agent reflecting on its own tends to revisit the same explanation for failure. Cross-task reflective memory addresses this in two forms: retrieved memory keeps past experience explicit and looks up related cases, which is specific but bounded by the bank, and parametric memory internalizes experience into parameters, which generalizes but is grounded in nothing beyond the task description. When both are used, existing agents combine them additively, feeding their outputs to the
memoryagentllm agentbenchmark - arxiv:2610.25788 · cs.AICan LLM User-Proxies Stand In for Humans in Reliance Measurement? Replicating Rel-A.I. with Simulated UsersShuai Wang, Rui Hu
Can language-model user proxies reproduce human reliance on an AI answer across contexts? Published simulated-user studies do not establish validity for this partial-answer decision, and Rel-A.I.'s released human record omits trial-level choices. We audit its 90 printed questions and 15 domain-by-marker reliance rates using sampled decisions and counterbalanced A/B probabilities from open models through 72B parameters. A matched crossover varies the printed agent specialty within each question, separating question content, label averages, and their interaction. Mistral Small's strong-marker ma
agent - arxiv:2610.25785 · cs.ROLong-Horizon Multi-Object Rearrangement via Distilled Tree Search over Continuous ActionsKallol Saha, Yining Hong, Amber Li, Yejin Choi +2
Despite recent advances in policy learning, long-horizon manipulation remains difficult because learned policies must avoid compounding errors while preserving future feasibility. While search-based planning can explicitly reason over such future consequences, it quickly becomes expensive in high-dimensional continuous action spaces. This raises a central question: how can long-horizon reasoning from search be transferred into policies in continuous action spaces? To this end, we propose DISCO: Distilled Tree Search over Continuous Actions. DISCO uses progressive-widening MCTS in physics simul
manipulation - arxiv:2610.25781 · cs.CVSee the Future, Classify the Present: Stochastic Latent World Modeling for Hyperspectral ImagesMuhammad Ahmad, Gemine Vivone, Ronny Hänsch, Ahmed Sohaib +2
Hyperspectral image classification (HSIC) relies on static spatial-spectral representations learned from labeled samples, limiting its ability to model structured dependencies in the latent representation space. To address this limitation, this work proposes Stochastic Latent World Modeling (LWM), a trajectory-based representation-learning framework that enriches classification through stochastic latent-state prediction. Given a hyperspectral patch, LWM tokenizes and encodes spectral and spatial information separately using dedicated transformer branches, then fuses their outputs into a unifie
world modelbenchmark - arxiv:2610.25777 · cs.AIStanding on the Shoulders of Giants: Enhancing LLM Agents for Automated Fact-Checking with Human Expert SkillsXinwen Chen, Haorui He, Yupeng Li
Automated fact-checking (AFC) systems powered by large language model (LLM) agents still face challenges in evidence retrieval and claim verification. Existing approaches improve their capabilities through computationally costly post-training or human-in-the-loop feedback that is difficult to scale. To transfer professional fact-checking expertise to LLM agents efficiently, we distill publicly available guidelines published by Agence France-Presse (AFP) Fact Check into a reusable, modular Professional Fact-Checking Skill (PFC-Skill). We then present SkillCheck, the first AFC system that gives
llm agentpost-traininghuman-in-the-loopbenchmark - arxiv:2610.25776 · cs.AIFrom Trace to Fix: Residual Task State in Long-Horizon AgentsZile Huang, Ser-nam Lim
Recovering from failures in long-horizon, stateful language agents is challenging because the action introducing a defect, the symptom exposing it, and the persistent state requiring repair frequently diverge. Existing methods largely focus on retrospective failure attribution, leaving sequential diagnosis and state repair poorly integrated under constrained budgets. We introduce TRACE-FIX, a unified recovery framework centered on versioned residual task state. Trace transforms execution logs into a structured dependency-linked ledger, prioritizing unresolved task requirements rather than sali
persistent state - arxiv:2610.25770 · cs.CVVHA: VQA-based Hallucination Assessment for Generative Image Super-ResolutionDonghun Ryou, Inju Ha, Jaeyoo Park, Bohyung Han
Generative image super-resolution produces visually realistic high-resolution images, but may also introduce hallucinations by adding unsupported content or altering existing details. Because existing super-resolution metrics fail to measure whether generated details remain grounded in the reference image, we present a hallucination assessment technique based on bidirectional visual question answering (VQA), which we term VQA-based Hallucination Assessment (VHA). The proposed approach uses a vision-language model (VLM) to determine whether the semantic content and low-level visual evidence in
benchmark - arxiv:2610.25768 · cs.CLTruncated Reasoning Under a Budget: Evaluating System-1 Models as Decision-makersZhongyin Wang, Zhaoqian Xue, Junting Mu, Haozheng Liu +3
Large language models (LLMs) often generate more reasoning tokens than needed. Limiting the reasoning budget reduces cost, but forcing an answer from an incomplete trace can hurt accuracy. System-1 decision models select among candidate options without producing a reasoning response. We evaluate Jev as a decision reader of truncated reasoning across seven LLM reasoners and five multiple-choice benchmarks (35 pairs). Jev achieves higher mean accuracy than forced answering at tested budgets from 500 to 16,000 tokens, including the estimated costs of reasoning and answering. At a 1,500-token budg
benchmark - arxiv:2610.25767 · cs.AIEvalRx: Agentic Diagnosis and Verified Repair of Foundation ModelsJiaqi Liu, Ruinan Jin, Yitong Tang, Beidi Zhao +3
Model evaluation reveals where models fail, but rarely why or whether a proposed fix genuinely works. We introduce EvalRx, an open-source agentic framework for statistically grounded model diagnosis and repair. EvalRx discovers systematic failure modes, explores candidate mechanisms, formulates falsifiable hypotheses, and verifies them by executing frozen, pre-specified tests before using supported diagnoses to guide tiered repair. Candidate repairs are validated against the unchanged model and promoted only when they improve the target behavior without unacceptable regression. Across language
agentic - arxiv:2610.25764 · cs.AIActTime: Action-Conditioned World Modeling for Time SeriesXinxing Zhou, Xiaojie Yuan, Yanlong Wen, Ming Jin
Time-series models increasingly condition on known future inputs. In controlled systems, some of these inputs are actions that can be changed, compared, and optimized before execution. A model must therefore do more than predict the realized future: it must predict how outcomes change under different actions from the same observed history and support reliable comparison of candidate plans. We introduce ActTime, an action-conditioned world model for time series. Over the observed history, ActTime updates variable-specific latent states through an action-conditioned transition and corrects them
world modelaction-conditioned - arxiv:2610.25763 · cs.ROTrajWAM: An Efficient Trajectory-Guided World Action Model for Dynamic Robotic ManipulationHaohan Geng, Sha Zhang, Weishuo Sun, Guanting Ye +6
Dynamic object manipulation requires robots to anticipate target motion while replanning fast enough to avoid executing outdated actions. Existing approaches either rely on fast reactive control without an explicit future target state or encode anticipated motion only implicitly. World Action Models (WAMs) offer a promising alternative by coupling future visual prediction with action generation, but scene-level video objectives may underrepresent target dynamics that are critical for manipulation, while iterative video-action denoising introduces substantial latency. We present TrajWAM, a traj
manipulation - arxiv:2610.25760 · cs.LGComputation, Not Memory: Chain-of-Thought and State Tracking in LLM AgentsPedram Mouseli, Massieh Moayedi
Language-model agents often reason before they act, and the value of monitoring that reasoning depends on how it relates to their actions. We ask how an agent's actions depend on its reasoning when the correct action depends on a changing state. In a small file-system task, background events announce which file's lock has been toggled but not its new value. The agent therefore has to reconstruct the current locks from its history before it can move files safely. With Qwen3-8B and Qwen3.5-9B, we separate two roles of reasoning: generating it before each action, and retaining earlier reasoning i
agentllm agent - arxiv:2610.25758 · cs.LGRemember Less, Adapt Longer: Bounded-State In-Context Reinforcement Learning with Recurrent Summary MemoryJiawen Wang, Cheng Xue, Peng Zhang, Jochen Renz
In-context reinforcement learning (ICRL) enables agents to adapt to new tasks without parameter updates, relying on memory to retain what they learn from experience. For agents operating continually, existing approaches using memory trade off bounded cost against access to past experience. Approaches using growing memories, such as full-history attention and accumulating summaries, incur storage and per-decision costs that increase with task length. Fixed-size memories bound storage but restrict access: sliding windows discard older steps, while recurrent networks compress even recent interact
memory - arxiv:2610.25749 · cs.LGCognitively Decoupled Meta-Induction for Out-of-Distribution GeneralizationYuqian Wang, Yuanzheng Wang, Chaojun Xiao, Ke Xu +6
In-context learning (ICL) tackles downstream tasks conditioned on demonstrations. This paradigm relies heavily on neighbor-based matching between queries and demonstrations, limiting out-of-distribution (OOD) performance. For better generalization, rule-based inductive reasoning derives task-level regularities, which we term task-level inductive representations (TIR), before directly answering. However, existing methods lack generalizable inductive ability across diverse tasks. Inspired by meta-learning, we propose Cognitive-Decoupling meta-Induction (CoDI) for TIR-based inductive reasoning on
benchmark - arxiv:2610.25748 · cs.CVMAPS: A Multi-Agent Personalized Script-Driven Video Advertising System Enhanced by Reinforcement LearningZhongyu Wang, Chen Zuo, Qingxuan Sun, Jiayang Qiao +3
With the rapid growth of online video consumption, video advertising has become a dominant component of digital advertising. Most personalized video advertising systems follow a retrieval-based paradigm, which matches ad content with user preferences in a feature space and recommends the most similar candidates. However, since the candidate creatives are pre-produced, this paradigm limits the granularity of personalization and leads to performance disparities across diverse user groups. To address this limitation, we propose a multi-agent personalized script-driven video advertising system (MA
agentmulti-agentiterative refinement - arxiv:2610.25747 · cs.LGBayesian Model Discovery with Language Models: Closing the Loop between Hypotheses and ExperimentsJooKyung Yuh, Byungmin Park, Eunseo Choi, Jinheon Baek +5
Scientific model discovery requires proposing candidate models of an unknown system and designing experiments that discriminate among them. However, LLM-based approaches typically rely on in-context learning to predict outcomes and select experiments, lacking explicit model evaluation to systematically guide hypothesis search, while Bayesian approaches enable principled model comparison but are confined to a predefined candidate set which results in limited search spaces. To this end, we introduce BMD, which closes the loop between LLM-driven hypothesis generation and Bayesian experimental des
benchmark - arxiv:2610.25742 · cs.CVFunctional Cross-layer Experts Routing for One-step Diffusion DistillationZile Huang, Ser-nam Lim
One-step distillation compresses diffusion trajectories into a single forward evaluation, yet how internal capacity should be structured within that remaining pass remains largely unexplored. In this setting, identical parameters must simultaneously preserve general generative priors and resolve prompt-specific details without iterative refinement. We formalize this challenge as functional allocation. Our analysis reveals that decoupling representations is beneficial only when expert interference outweighs shared transfer, favoring an always-active shared backbone augmented with conditionally
iterative refinement - arxiv:2610.25736 · cs.ROLearning from Fleet Experience: A Collective Learning Paradigm for RobotsChao Yu, Shu'ang Yu, Hao Lin, Hongzhi Zang +38
Experience generated during deployment offers an opportunity for robots to improve policy reliability through learning from their own executions. However, learning through human corrections requires teleoperators to monitor execution and intervene when policies fail, a demand that is difficult to meet as robot fleets grow. We introduce collective experience learning (CEL), a learning paradigm in which deployed robots share their execution experience to improve a common policy. We realize CEL through failure-guided experience retrieval fine-tuning (FERT), which uses failure contexts from robot
manipulation - arxiv:2610.25734 · cs.ROThink Densely, Act Sparsely: Latent Expert Cognitive Chains for Vision-Language-Action Autonomous DrivingJie Wang, Guang Li, Zhijian Huang, Jinlong Li +4
Autonomous driving fundamentally demands dense understanding of visual, semantic, and geometric information, yet existing Vision-Language-Action (VLA) models are typically trained with sparse supervision such as language instructions or trajectory signals. This direct mapping from dense visual observations to sparse representations forces the model to compress or even discard critical scene information essential for safe driving. We argue that this limitation does not stem from insufficient model capacity, but from the lack of explicit mechanisms that encourage the formation of dense world rep
vision-language-actionvlavla modelworld modelcurriculum learningbenchmark - arxiv:2610.25733 · cs.AITimelineAgent: An LLM-based Framework for Query-Centric Event Timeline Extraction Across Multiple DocumentsQingkai Min, Yunxin Bao, Yifeng Lu, Zhenyu Feng +3
Query-centric cross-document event timeline extraction aims to construct event timelines from multiple documents under diverse query focuses. The task is very challenging due to the heterogeneous event modeling and reasoning requirements arising from both the intrinsic complexity of cross-document timeline construction and the diversity of query focuses. In this paper, we propose TimelineAgent, an LLM-based multi-agent framework featuring a multi-stage reasoning workflow, where heterogeneous event features are organized into reusable skills and orchestrated through a goal-oriented protocol. To
multi-agentagent frameworkbenchmark - arxiv:2610.25728 · cs.AINo Matter Who You Are: Addressivity and Sycophancy in Large Language ModelsSeungjong Sun, Hyunmin Jun, Minsik Yoon, Eungu Lee +3
Sycophancy, where an LLM becomes overly agreeable, deferential, or flattering toward the user, has been identified as a major failure mode. While post-training has been implicated as a cause, how it internally gives rise to sycophantic behavior remains insufficiently understood. Post-training transforms the model's output from text completion into utterances directed toward an addressee, thereby giving the output \textit{addressivity}. We hypothesize that the accompanying pressure to follow the addressee's instructions and align with their preferences links this addressivity to sycophancy. To
post-training - arxiv:2610.25725 · cs.LGAnatomy of Off-Policy Policy Gradient: Importance Sampling, KL Regularization, and BaselinesHaoqun Cao, Yurun Yuan, Tengyang Xie
Reinforcement learning has become a central tool for large language model (LLM) post-training, where policy gradient methods are routinely deployed off-policy, even though vanilla policy gradient assumes on-policy sampling. We study when such off-policy deployment is justified, and what roles its core ingredients, including importance sampling, KL regularization, and baselines, play in handling the resulting distribution shift. We show that for a general class of off-policy policy gradient updates, consistency, in the sense that the optimal policy remains a stationary point, requires one of tw
post-training - arxiv:2610.25723 · cs.LGLearning to Couple: Context-Adaptive Joint Denoising for Diffusion Language ModelsZilei Shao, Ian Li, Heng Zhao, Kewei Lian +3
Diffusion language models accelerate generation by predicting multiple tokens in parallel, but factorized denoisers treat simultaneous predictions as conditionally independent. Decoding heuristics reduce the resulting errors by controlling which tokens are committed, but leave the underlying distribution factorized. Joint modeling addresses the source of the problem, yet existing methods attach separately learned layers to frozen denoisers, leaving the backbone unadapted to the joint objective and the coupling parameters fixed across contexts. We propose Codpula, a context-parameterized denois
benchmark - arxiv:2610.25717 · cs.AIFine-Grained Units, Flat Decisions: Hierarchical Micro-Expert Pruning for Mixture-of-Experts ModelsZukang Xu, Xinpei Gao, Xing Hu, Zhe Jiang +1
Micro-expert pruning reduces the capacity of mixture-of-experts (MoE) models by removing coupled channels within routed experts. However, finer pruning granularity alone does not guarantee the preservation of downstream capabilities, as similar language-modeling perplexity can accompany substantially different task performance. We revisit micro-expert mask construction as two distinct decisions: allocating retained capacity across experts and selecting channels within that capacity. We propose UC-RAMP, a two-stage post-training pruning framework that explicitly separates these decisions. CAPE
post-training - arxiv:2610.25716 · cs.CVLEVI: Benchmarking Visual Identity Fidelity with Verifiable Low-Entropy ConstructsXiao Ma, Zhonghao Zhan, Hamed Haddadi, Qi Chen
Modern generative models can produce visually plausible images while violating exact identity constraints. Evaluating such failures requires distinguishing violations of the requested identity from permissible variation in appearance. We introduce Low-Entropy Visual Identity (LEVI), which characterizes visual identities determined by compact facts and explicit rules. The low-entropy requirement concerns identity uncertainty, while appearance may remain diverse. We formulate identity fidelity in terms of the probability that a conditional image distribution assigns to the identities permitted b
benchmark - arxiv:2610.25711 · cs.ROSWAP: Test-Time Steerable Humanoid Whole-Body Collision AvoidanceZeji Yi, Guanqi He, Xiaofeng Guo, Guannan Qu +4
Humanoid robots operating in human-centric environments must coordinate the entire body to traverse tight, cluttered spaces while maintaining balance. Existing methods often struggle to jointly consider dynamic stability and collision avoidance from high-dimensional perceptual inputs. Moreover, for humanoids, safe behavior is inherently multimodal: the same obstacle may admit multiple valid traversals, which is difficult to represent with a single deterministic policy. To address these challenges, we propose SWAP (Steerable Whole-body Avoidance Policy), a hierarchical framework that separates
humanoiddiffusion policy - arxiv:2610.25710 · cs.LGEvidence Sufficiency and Model Comparisons in MexRegQA: An Item-Level Audit of Mexican-Spanish Regulatory QAFred Hu, Yeqi Wang, Xing Zhou
Similarity-gated coverage does not establish that retrieved evidence contains the facts needed to answer a question. We audit Mexican-Spanish financial-regulatory QA and find that 35 of 135 covered items lack some or all load-bearing facts. On the post-hoc full-evidence slice (n=98), eleven prompt-matched models score 95-98/98 and none differs significantly from the anchor; this does not establish equivalence. Removing evidence-defective items changes the nominal significance status of 14 of 55 pairwise comparisons. On 56 zero-evidence items, accuracy spans 33.9%-89.3%. Removing evidence from
benchmark - arxiv:2610.25707 · cs.AIWhose Credential Is It? How Owner Attribution in Agent MemoryShapes Cross-User Account MisuseHau-Sing So, Daixian Liu, Carl Hou, Yuhaohang He +2
Shared long-term memory lets an LLM agent serving one user inadvertently retrieve another user's credentials, but it is unclear when such exposure leads to actual harm. We introduce a three-stage audit that separates three events: a foreign credential reaches the prompt (exposure), the agent acts with it (behavior), and the other user's account changes (state). In AppWorld, a store that mixes 763 past logins from all users retrieves another user's credential for 84.7 percent of required apps, yet agents rarely act on it. Misuse appears in a stress test where memory lacks the user's own credent
memoryagent memoryagentllm agent - arxiv:2610.25704 · cs.LGPredictive Bandit Curriculum Reinforcement Learning via Gradient-Based Sample Group DiscoveryLiang Xu, Shufan Shen, Zhongni Hou, Chao Bi +4
Curriculum learning, which organizes training samples according to their learning priority, has been widely adopted in reinforcement learning to improve training efficiency. However, existing methods commonly construct curriculum units using predefined data properties and estimate their utility by simply aggregating a single proxy signal from recent training history.Such designs may group samples with different optimization effects and fail to anticipate how their utility evolves in relation to the policy. We propose Predictive Gradient-based Bandit (PGB) Curriculum, which first discovers opti
curriculum learningbenchmark - arxiv:2610.25701 · cs.ROValidation of Human Motor Development Scales with Humanoid RobotsFanxing Meng, Jing Xiao
We investigate how humanoid robots can learn a series of motor skills defined by human infant motor development scales, and we evaluate which scales are comprehensive enough to serve as self-contained guides for developing gross motor skills up to standing and walking. We represent each motor development scale as a developmental graph that guides the training of a humanoid robot. We then apply this graph within a reinforcement learning framework so that the robot learns skills following the natural phases of human motor development described by the scale. Our work demonstrates that graphs deri
humanoid - arxiv:2610.25697 · cs.AICVaR-RMD: CVaR Selection Pressure for Robust Metaheuristic DiscoveryRuodai Cui, Li Niu
Large language models have recently been used to discover heuristics through iterative program search, offering a promising approach to automated algorithm design. However, candidates selected by average performance may degrade across instance families, random seeds, and problem scales because mean-based objectives can obscure rare but severe failures. We propose CVaR-RMD, a risk-sensitive evolutionary framework that uses an LLM to mutate executable heuristic programs and ranks candidates by empirical conditional value-at-risk over heterogeneous evaluation conditions. Across three combinatoria
benchmark - arxiv:2610.25693 · cs.AIToolOPD: Credit-Aware On-Policy Distillation for Tool-Use AgentsChu Zhao, Lei Tang, Yanming Wang, Jianzhe Zhao +4
On-policy distillation (OPD) enhances the capabilities of small language agents by providing dense, token-level supervision from a teacher model on student-generated trajectories. However, standard OPD implicitly assumes that credit is uniformly distributed across long-horizon trajectories, assigning the same strength of teacher supervision to different actions. Through credit-distribution analysis and outcome-reversal counterfactual experiments, we find that trajectory credit is concentrated on a small number of critical actions. Moreover, even when the trajectory is held fixed, reversing onl
tool usetool-usebenchmark - arxiv:2610.25692 · cs.ROProgress-Guided Preference Generation for Aligning Generative Control PoliciesTung M. Luu, Tri Ton, Duc M. Nguyen, Xuan Thanh Nguyen +7
Generative policies have achieved strong performance in imitation learning tasks, but their success is often constrained by the limited availability of expert demonstrations. Although preference-based fine-tuning can leverage suboptimal rollouts to improve performance, existing approaches heavily rely on expensive human annotations. In this work, we introduce Progress-Guided Preference Generation (PGPG), an unsupervised preference alignment framework that eliminates the need for human-in-the-loop supervision. Our method leverages ordered coverage alignment to measure sequential progress relati
manipulationhuman-in-the-loop - arxiv:2610.25691 · cs.ROReDOA: Real-World Distillation via Observation Alignment for Image-Based Manipulation Policy LearningKotaro Kajitsuka, Masaki Murooka, Fumio Kanehiro
Data-driven learning for general-purpose robotic manipulation presents a fundamental trade-off between data-collection cost and adaptability to real-world environments. Imitation learning (IL) requires costly human demonstrations, whereas simulation-based learning is vulnerable to sim-to-real gaps in dynamics and sensor observations. We propose ReDOA (Real-World Distillation via Observation Alignment), a framework that decouples control learning from perception learning to learn image-based policies in the real world. A state-based teacher policy trained in simulation is executed on a real rob
manipulationsim-to-real - arxiv:2610.25690 · cs.ROScaleNav: Scalable Internet Video Pretraining for Language-Conditioned Local NavigationYihui Yang, Yuanhong Yu, Jiawei Wang, Zhongrui Yu +6
We study language-conditioned local navigation, where a robot must navigate to a text-specified target visible in its initial observation and stop at the described location. End-to-end navigation policies require diverse training data to generalize across scenes, yet real-robot data is costly to collect and simulated data leaves a sim-to-real gap. Internet egocentric videos offer a scalable source of real-world navigation experience. Our key insight is to learn goal-directed visual dynamics from Internet egocentric videos, and use these dynamics to support robot action generation. To this end,
quadrupedsim-to-realbenchmark - arxiv:2610.25687 · cs.RORevealing and Bridging the Physical Executability Gap in Vision-Language NavigationGang He, Zhile Cheng, Kepeng Xu, Zhenyang Liu +1
Continuous vision-and-language navigation (VLN) is typically evaluated by goal-reaching success, assuming it reflects an agent's ability to execute language-grounded behavior. We show that this assumption can fail: permissive contact models in current simulators may mask physically infeasible actions, creating a physical executability gap between semantic competence and embodied realization. To bridge this gap, we introduce Visual-Semantic Traversability Alignment (VeSTA), a framework that aligns VLN policies with physically executable futures. We define visual-semantic traversability as the a
embodied - arxiv:2610.25679 · cs.LGLOOP: Recovery-Aligned Iterative Layer Pruning via On-Policy DistillationLiang Xu, Shufan Shen, Zhongni Hou, Yufei Zhang +3
Layer-level pruning is attractive for large language model (LLM) deployment as it reduces the model size while preserving a dense architecture that can be efficiently executed on standard hardware. Current methods typically follow a one-shot paradigm, which first removes selected layers to reach a target model size and then performs additional training to recover model capabilities. However, pruning directly to the target size can substantially degrade reasoning capabilities, making subsequent recovery more difficult under a limited training budget. Moreover, this paradigm separates layer sele
benchmark - arxiv:2610.25678 · cs.AICross-Pollination Retrieval: Finding Transferable Ideas Across Fields in Scientific LiteratureSeokwon Jung, Yerang Kim, Yejin Kim, Alexander Rubinstein +4
Many scientific advances come from cross-pollination, carrying an idea from one field over to a problem in another. Yet retrieving papers across fields remains underexplored, with existing benchmarks confined to a narrow range of fields. We study cross-pollination retrieval, the problem of identifying transferable ideas across research fields, and introduce CPR-Dataset, a large-scale dataset constructed by mining cross-topic citations. Its evaluation split CPR-eval has 1,300 queries in 26 fields, and its training split CPR-train has 10,573 pairs. On CPR-eval, the best of eleven existing retrie
benchmark - arxiv:2610.25675 · cs.AILIRQuant: An Efficient Locally Isolated Rotation Quantization Method for Large Language ModelsDesheng Xu, Hanglin Li, Tianchen Song, Zhiyong Zhao +1
Post-Training Quantization (PTQ) enables efficient deployment of Large Language Models (LLMs), yet activation outliers remain a key challenge for low-bit quantization. Existing global transformation methods overlook heterogeneous channel sensitivity to quantization noise, leading to suboptimal accuracy-efficiency trade-offs. To address this, we propose \textbf{Locally Isolated Rotation Quantization (LIRQuant)}, a sensitivity-aware framework that selectively applies transformations. LIRQuant identifies sensitive channels using a noise-weighted sensitivity metric, applying expressive learnable r
memorypost-training - arxiv:2610.25673 · cs.AIHyperTaxa: Multiscale Genome Representations for Generalized Microbial Taxonomic InferenceAkash Rao, Sushim Gupta, Vishalini Ramnath, Akhilesh Ramachandran +1
Reliable microbial taxonomic inference is challenged by the rapid growth of newly sequenced organisms and incomplete reference coverage. While alignment- and database-driven tools remain essential for reference-backed identification, they are not designed for learned inference over species absent from genome-level training. We introduce HyperTaxa, a taxonomy-aligned representation learning framework for open-set microbial taxonomic inference. HyperTaxa encodes whole genomes as multiscale hyperspectral k-mer transition fields and aligns these representations to a taxonomy-derived semantic space
benchmark - arxiv:2610.25666 · cs.CVFrechet Forcing: Training Long-Horizon Streaming World Models from Short ClipsYian Zhao, Yu Yan, Hongcan Guo, Ruochong Zheng +2
Streaming video world models generate the future chunk by chunk, conditioned on their own past and a stream of actions. Trained on short clips, they drift once a rollout exceeds the training horizon, as prediction errors re-enter the context and compound. A common remedy is to train the model on its own rollouts via distribution matching distillation (DMD). For action-conditioned world models, the teacher becomes a bottleneck: an external video model is agnostic to the actions, whereas the model's own pretrained network drifts in the same manner as the student. We observe that a long rollout n
manipulationworld modelaction-conditioned - arxiv:2610.25665 · cs.ROSpaRV: Sparse Radiance Voxels for Scalable Geometrically Consistent LiDAR-Visual Mapping and RenderingJianheng Liu, Yunfei Wan, Fangming Cheng, Xiangyu Hong +6
Robotic simulation, navigation, and inspection need photorealistic and geometrically accurate maps. Radiance fields achieve photorealism, but building them from robot data raises three challenges. A robot's trajectory serves its task, not reconstruction, so images from few viewpoints leave the geometry under-constrained. Residual errors in camera poses and intrinsics blur the map. The map's GPU memory grows with scene size, which limits its level of detail in large scenes. We present \textbf{SpaRV}, a \textbf{Spa}rse \textbf{R}adiance \textbf{V}oxel map rendered by differentiable ray marching,
memory - arxiv:2610.25660 · cs.CVFluidGen: Physics-Aware Fluid Video Generation via Motion-Guided Post-TrainingYang Zheng, Zizhang Li, Boyang Deng, Jan Ackermann +5
Video generation models produce visually compelling motion but systematically fail on fluid phenomena, including smoke, fire, and explosion debris, where outputs lack coherent vortical structure and collapse to bland dynamics that violate the physics of real fluids. Classical solvers achieve high physical fidelity but require detailed 3D scenes and carefully tuned parameters, making them impractical for in-the-wild imagery and video generation. We address this gap by injecting a data-driven fluid prior into a pre-trained video model, trained on a mixture of high-fidelity Houdini simulation dat
post-training - arxiv:2610.25659 · cs.AIBeyond Leaderboards: A Method for Application-Specific Capability Profiling of Large Language ModelsOleg Grynets, Vasyl Lyashkevych, Andrii Salyk, Dmytro Koziuberda +10
Public leaderboards rank large language models (LLMs) in general, but an engineer has to choose a model for a specific application, architecture and cost. This paper presents a method for application-specific capability profiling of LLMs and its realization as a containerized testbed. The method treats each application as a unit of extension: an Oracle-to-PostgreSQL migration agent, a documentation-grounded chatbot or an enterprise-support automation agent runs in its own container behind a small contract, brings its own dataset, evaluator and owned attributes, and is onboarded together with i
agentevaluatorleaderboard - arxiv:2610.25651 · cs.LGOPDArena: Understanding Learning and Forgetting in On-Policy DistillationYuanze Xu, Tianzhu Ye, Hui Sun, Wenhui Wang +2
On-policy distillation (OPD) is increasingly used for language model post-training, yet how teacher supervision shapes learning and forgetting remains unclear. We present OPDArena, a framework comparing supervised fine-tuning (SFT), reinforcement learning with verifiable rewards (RLVR), and OPD across five pretrained models and diverse teacher sources. It jointly evaluates mathematical learning, general capability retention, and predictive distribution changes. Teacher choice strongly shapes retention: students of RLVR-trained teachers improve on the target task while retaining prior capabilit
post-training - arxiv:2610.25649 · cs.ROInteractAgent: Agentic Human Motion Interaction with Memory-Augmented ReflectionTao Zhang, Senhe Zhang, Zeyu Zhang, Yujia Zhang +1
Long-horizon human-scene interaction requires coordinated spatial reasoning, physical feasibility, and the ability to recover from compounding execution errors. Text-driven motion models are largely scene-agnostic and lack closed-loop correction, while geometry-conditioned approaches depend on paired scene-motion training or explicit 3D preprocessing that limits generalization to novel layouts and open-vocabulary goals. We present InteractAgent, an agentic multimodal LLM system for long-horizon human-scene interaction that plans, executes, observes, and reflects in a closed loop, using a singl
agenticevaluator - arxiv:2610.25641 · cs.CVVisual Memory and Textual Alignment for Cross-Modal Few-Shot SegmentationTianqi Wei, Zhi Chen, Xin Yu, Scott Chapman +1
Few-shot segmentation (FSS) aims to segment target objects with limited annotated examples. Recent studies have begun to explore large foundation models to improve generalization for this task. In particular, SAM 3 provides strong visual representations and supports text-guided segmentation. However, its performance remains limited in specialized domains, where semantic prompts and visual representations are often poorly aligned. To address this challenge, we propose ViTA-SAM, a framework built upon SAM 3 that integrates visual memory and textual features through cross-modal alignment. Specifi
memorymemory modulebenchmark - arxiv:2610.25640 · cs.AIThink, Compress, Answer: Reinforcement Learning for Multimodal Grounding via Reasoning CompressionJinbae Seo, Hyeongjun Kwon, Jiyoung Lee, Hyunsung Jang +3
Extended reasoning has substantially improved multimodal large language models (MLLMs) on tasks requiring compositional and relational inference. However, multimodal grounding imposes a distinct requirement: reasoning must ultimately converge on a target-specific compact representation that supports precise localization. As reasoning trajectories grow longer, they often expand into broad scene narratives and peripheral relational details, which can dilute grounded perceptual evidence and lead to ambiguous or incorrect predictions. Motivated by the information bottleneck principle, we propose T
context compressionbenchmark - arxiv:2610.25638 · cs.CVACorD: Adaptive Correspondence Discovery for Medical Image RegistrationArghavan Rezvani, Kun Han, Xiaohui Xie
Medical image registration is a fundamental task for aligning anatomical structures across images. Exhaustively searching for point-wise correspondences is prohibitively expensive in both computation and memory, particularly for high-resolution 3D medical images. Most recent learning-based methods directly predict correspondences in a coarse-to-fine manner, but often propagate intermediate estimates across scales without explicitly weighting correspondence evidence, which can allow coarse-scale errors to persist during refinement. To address these challenges, we propose ACorD, an adaptive corr
benchmark - arxiv:2610.25632 · cs.ROSteerAct: Unlocking Action Models through Few-Shot Temporal SteeringKaichen Zhou, Zhiyan Li, Chongkai Gao, Yuhan Wang +10
Improving pretrained robot action models at test time from only a few rollouts remains challenging. Activation steering offers a lightweight approach by modifying feature activations while keeping model weights fixed. However, effective steering requires identifying which features to modify and determining the timing and strength of modification. We introduce SteerAct, a few-shot temporal steering framework that uses a shared temporal descriptor of sparse-feature dynamics for offline feature selection and online control. A pretrained sparse autoencoder provides candidate features, and our temp
manipulationrobotwin - arxiv:2610.25630 · cs.AIThe Forward/Inverse Duality of Controlled Emergence:A Computational Design Program from Explanation to EngineeringHao Li, Minghui Ren, Shaowen Lu
The macroscopic capabilities of intelligent systems arise from the local interactions of many simple units. When such systems are deployed in the real world, their behaviour must nevertheless be constrained, leaving designers caught between releasing adaptability and preventing loss of control, yet without formal tools to resolve the dilemma. This paper proposes a unified research program for controlled emergence that restates this tension as a pair of dual problems: the forward problem explains macroscopic behaviour from local rules, and the inverse problem solves for controllable rules and r
multi-agent - arxiv:2610.25629 · cs.AISelecting and Composing Agent Design Patterns: Industrial Case Evidence and Controlled Reference-System ExperimentsJia Huang, Bingsheng Ru, Bo Liang, Yuke Xiong +2
Agent pattern catalogs describe architectural alternatives, but engineers also need evidence for selecting and composing them. We extend a cognitive-function--execution-topology framework with an evidence-driven selection procedure and explicit composition contracts. The study combines two industrial case accounts, a contributor's collaboration design, seven practitioner workshops, and executable fault-injection experiments. Three local reference systems use SQLite state, shared artifact files, and a file-backed map-publication surrogate. A 2 x 2 ablation separates function-specific contracts
agentevaluator - arxiv:2610.25628 · cs.ROHumanoid SRL: Augmenting Humanoids with Supernumerary Limbs for Coordinated Loco-ManipulationFukang Liu, Feiyang Wu, Paul Barsa, Jaehwi Jang +11
Humanoid robots are envisioned to perform diverse human-level tasks through their anthropomorphic morphology. Their dual-arm configuration enables a broad range of everyday actions. Yet, as in humans, this morphology can constrain loco-manipulation when both arms are occupied or additional contacts are required beyond their reach. Supernumerary robotic limbs (SRLs) offer a way to extend these capabilities, but their integration requires the added limb to be coordinated with whole-body motion and balance. In this study, we develop and integrate a supernumerary arm into a humanoid platform to in
manipulationhumanoid - arxiv:2610.25626 · cs.AIMapSteal: Map-based Autonomous Knowledge Stealing Attack Against Large Language ModelsZhengao Li, Chuxu Zhang, Yushun Dong
Knowledge acquisition from black-box large language models (LLMs) offers a practical approach to adapting smaller student models. However, many existing methods rely on predefined queries or student feedback to guide acquisition, leaving autonomous expansion of knowledge targets an important challenge. We propose \textbf{MapSteal}, a student-independent framework for structured recursive knowledge acquisition. Starting from a small set of initial roots, MapSteal represents acquisition targets using structured states and recursively expands them through teacher-proposed directions. Question rea
benchmark - arxiv:2610.25625 · cs.LGZephyrus-R1: Agentic Reinforcement Learning for Weather ReportingFengfei Yu, Marshall Fisher, Isadora White, Sumanth Varambally +12
Reinforcement learning (RL) has emerged as a key method for training AI models in scientific applications, but RL remains difficult when outputs cannot easily be verified against a known ground-truth answer. Natural-language weather forecast reports are one such domain: they translate atmospheric data into assessments of weather evolution, hazards, and uncertainty for both the general public and key decision-makers. Unlike numerical weather-analysis questions, they admit many valid outputs, which limits deterministic verification. To address this, we develop Zephyrus-R1, a practical framework
agentictool usecurriculum learning - arxiv:2610.25622 · cs.CVRecurrent Vision Transformers for Event-Based Spacecraft Pose Estimation Under Slow Relative MotionsSofia McLeod, Mohsi Jawaid, Carl Seubert, Tat-Jun Chin
Vision-based object pose estimation is vital for conducting spacecraft rendezvous and proximity operations (RPO). Recent research has demonstrated the robustness of event cameras against the harsh lighting conditions that often occur in orbit, thanks to the extremely high dynamic range of event sensors. However, event-based spacecraft pose estimation (EBSPE) suffers when the relative motion between the chaser and target is slow, since the event rate drops significantly, leading to data with low signal-to-noise ratio. Unfortunately, for safety reasons, real RPO trajectories often involve slow r
memoryevent camera - arxiv:2610.25621 · cs.LGWeighted State Souping for Efficient Long-Context RetrievalWilliam L. Tong, Aryo Lotfi, Emmanuel Abbe, Josh Susskind +2
State space models (SSMs) combine fast inference and fixed state size with competitive language-modeling performance. Recently, state souping has been proposed as a procedure whereby an SSM independently encodes chunks of tokens like documents, tool definitions, or memory records. The resulting state encodings may be cached and recombined to cheaply construct a context for downstream queries. However, both state souping and vanilla SSMs deteriorate on long, crowded contexts, limiting their utility for long-horizon domains. To address this issue, we propose weighted state souping, a simple meth
memorylong-contextagentictool use - arxiv:2610.25612 · cs.LGCurriculum Learning Teaches LLMs Long-Horizon Implicit Reasoning, Up To A LimitJimson Huang, Maxwell Nickell, Ruowang Zhang, Abulhair Saparov
LLMs "reason" by externalizing intermediate steps as tokens through chain-of-thought (CoT), yet even frontier reasoning models hit abrupt performance cliffs on long-horizon tasks. Implicit reasoning (reasoning within a single forward pass) is useful during CoT generation, as the model can "search ahead" to decide the correct next move without explicit search. Models with weak implicit reasoning are more likely to commit to wrong paths and must attempt to recover via backtracking, while models with strong implicit reasoning identify the correct proof strategy directly, shortening the CoT. Howev
curriculum learningbenchmark - arxiv:2610.25611 · cs.LGSafety Offset Learning: When Benign Fine-Tuning Misaligns LLMsJesson Wang, Zhanhao Hu, Minghao Lin, Yuchen Wu +2
Safety risks from fine-tuning are commonly characterized by the harmfulness of the training data:harmful fine-tuning is considered dangerous, while benign fine-tuning is often assumed to preservealignment. We argue that this view misses a more fundamental factor-the behavioral transition inducedby supervision. Harmful fine-tuning and safety-degrading benign fine-tuning can both be understoodas learning the same refusal-to-compliance transition, but on different data distributions. We studywhether such a transition learned on benign inputs can generalize to harmful requests. We constructbenign
benchmark - arxiv:2610.25604 · cs.LGAdversarial Post-Training of Foundation Models for High-Frequency TradingAlexandre Bismuth, Kang Li, Aramis Marti-Shahandeh, Valentin Mohl +2
While autoregressive foundation models trained on limit order book (LOB) data show state-of-the-art performance for market simulation, pre-training with cross-entropy neither penalizes the compounding of sampling errors over free-running rollouts nor targets desired market properties. We introduce a post-training method for LOB foundation models using a Wasserstein generative adversarial network (GAN) objective. A critic, trained online on rollouts through the true, non-differentiable matching engine, guides updates from evolution strategies over rank-4 LoRA perturbations of the pre-trained ge
post-training - arxiv:2610.25600 · cs.ROVector-Quantized World Model: Latent Feedback Motion Planning for Robot ManipulationXuezhou Xu, Chung-Pang Wang, Zhenzhen Li, Benjamin Riviere
Visual world models have shown strong potential for robot planning, yet long-horizon manipulation remains challenging due to compounding prediction errors, large search complexity, and latent plans that the low-level controller cannot reliably reach. In this work, we introduce Vector-Quantized World Model (VQ-WM), a latent feedback motion planning framework that integrates discrete long-horizon search with continuous visual feedback control: a discrete latent predictor models transitions between vector-quantized latent states, allowing the planner to look far ahead and find a shortest path to
manipulationworld modelbenchmark - arxiv:2610.25599 · cs.LGText Dominance in Paralinguistic Conflict: Tracing Suppression Through Audio LLM CircuitsAnkur Banerjee, Vishnu Guntakala, Danity Pike, Advay Roongta +2
Audio LLMs deployed in settings such as crisis hotlines and customer support are only useful beyond speech-to-text pipelines if they use acoustic cues rather than the transcript alone. We test whether they prioritise transcript content over paralinguistic cues (emotion, speaker identity, sarcasm) and trace the mechanism when they do. Across six audio LLMs under LISTEN-style conflict conditions, the mean text-dominance rate on mismatched text+audio items is 0.78 (0.86 excluding sarcasm). In the most text-dominant screened model, probing shows speaker identity remains linearly decodable at ceili
benchmark - arxiv:2610.25585 · cs.ROAdaptive Bendpoint Formation in Octopus-Inspired Soft Continuum ArmsJessica Gumowski, Julian Finn, Krishna Manaswi Digumarti, David Howard
Octopuses exhibit a powerful form of embodied intelligence from a highly adaptable soft body. One key mechanism is the ability to instantiate bending points at any location along the arm, which enables performant and dexterous behaviours. Taking inspiration from the octopus for its dexterity, this study focuses on changing the curvature of a tendon-actuated arm by adaptively changing the anchoring end-point along the arm. An array of inflatable shells is integrated into the arm, where inflating a shell causes an increase in friction at the location, jamming the tendon and creating a temporary
embodieddexterousmanipulator - arxiv:2610.25584 · cs.LGCVTA: Cross-Variable Temporal Attention for Multivariate Irregular Time Series PredictionAnkitkumar Joshi, Milos Hauskrecht
Irregularly sampled multivariate time series arise in many real-world domains such as healthcare, climate science, and large-scale monitoring systems. Characterized by asynchronous observations and heterogeneity across variables, such time series pose significant challenges for forecasting. In this work, we introduce Cross-Variable Temporal Attention (CVTA), a short-term, continuous-time forecasting framework that learns prediction functions from its compact, interpretable state summarizing recent temporal dynamics across variables. CVTA uses Variable-Conditioned Embeddings to project heteroge
benchmark - arxiv:2610.25575 · cs.LGBenchmark for Structured Retrieval-Augmented Generation in Perovskite PhotovoltaicsQiwei Han, Chi Zhou, Gaoxiang Yuan
While retrieval-augmented language models increasingly serve as literature assistants in the physical sciences, existing benchmarks evaluate them primarily on qualitative text overlap or through model-based scoring rather than against verified experimental ground truth. In experimental materials disciplines such as perovskite photovoltaics, however, the most critical inquiries are inherently quantitative and aggregate measurements across dozens or hundreds of papers. Here, we present a quantitative question-answering benchmark where all 755 ground-truth answers are computed directly from 45,44
retrieval-augmentedbenchmark - arxiv:2610.25568 · cs.AIDesign and Cross-Domain Evaluation of an Agent Execution HarnessSaurav Joshi, Manas Gupta, Ashwin Venkatesh Raman, Akshay Budhkar
Completing complex tasks with language-model agents requires coordinating planning, execution, and review so the final result addresses the user's original objective. We present an agent execution harness that keeps the original objective separate from a revisable plan and an execution record. A planner assigns scoped tasks, workers produce outputs and artifacts, and a reviewer examines results against the original request before submitting or requesting additional work. Reusable domain-specific instructions guide these roles and are refined during construction using task inputs and execution
agent - arxiv:2610.25565 · cs.LGConcurrent two-scale topology optimization of printable bone scaffolds via rectified-flow generation and exact latent-space refinementReza T Batley, Sourav Saha
Bone scaffolds must reconcile a macroscale mechanical response with microscale geometric feasibility, yet conventional approaches to their design restrict attainable constitutive properties. This work presents a concurrent two-scale topology optimization framework; a conditional rectified flow generates periodic lattice microstructures, which are then refined in latent space by exact finite-element homogenization sensitivities. The resulting unit-cell geometries and their spatial assignments are jointly optimized under mechanical and feature-size constraints. In a distal rabbit femur benchmark
benchmark - arxiv:2610.25563 · cs.AILearning Which Device to Call: Online Tool Grounding for LLM Agents in IoTYichen Liu, Xiaochong Jiang, Haoran Yu, Lifei Liu
The Internet of Things (IoT) gives large language model (LLM) agents access to physical observations, but one requested capability is usually offered by several devices, and the agent's plan does not say which one to call. Existing work retrieves relevant tools and checks plan feasibility, without learning which eligible device returns a useful observation. Unlike previous works, we study online tool grounding: binding an already specified tool call to one physical endpoint under service cost and device participation constraints. We propose a framework that learns which physical device should
llm agent - arxiv:2610.25558 · cs.LGFTFM: Factorized Tabular Foundation ModelsJosif Grabocka
Tabular foundation models (TFMs) are becoming a powerful alternative to conventional tabular learning, adapting to new datasets in context without task-specific training. Current architectures, however, face a trade-off: models that preserve cell-level representations are expensive because they repeatedly attend across rows and columns, while faster models collapse each row into a fixed embedding and lose this fine-grained state. We introduce Factorized Tabular Foundation Models (FTFM), which preserve cell representations while factorizing their computation through one contextual latent per ro
benchmark - arxiv:2610.25554 · cs.CVTricking Synthetic Image Detectors with Amortized Generative AttacksMohamed Awad, Khaled Elhady, Yasser Rohaim, Samuele Poppi +1
Rapid advances in image synthesis have increased the need for reliable synthetic-image detectors, yet their robustness to transferable attacks remains incomplete and understudied. Existing evaluations generally examine whether an attack constructed against one detector transfers to another, but rarely isolate how its effectiveness changes when the attacked images originate from different generators. We conduct a systematic evaluation of three representative attacks: two iterative perturbation-based methods, MIG and MuMoDIG, and the amortized generative attack Tricker. Across four detectors and
evaluation protocol - arxiv:2610.25552 · cs.AISymbolic Constraint Agents: Mining Formal Tool-Call Constraints to Prevent Agent ErrorsJinu Lee, Shraddha Barke
Symbolic constraints (programmatic, deterministic checks for tool calls) enable deterministic verification of LLM agents' tool calls, but existing approaches primarily enforce predefined safety policies rather than tool dependencies in complex tool environments. We explore whether symbolic constraints can be mined from environments and observations beyond predefined policies, and whether they can improve end-to-end task performance. We present Symbolic Constraint Agents (SCA), a framework that mines symbolic constraints from API documentation, refines them against execution traces, and checks
agentllm agent - arxiv:2610.25550 · cs.AIEditing as Language: Teaching LLM Agents a Grammar for Native Context EditingYunlong Deng, Yan Li, Xing Hu, Yida Zhu +3
A synopsis in novels folds a story into a few lines; a function in code gathers many steps with a single execution. Context editing is native to language itself. In this paper, we propose the Native Context Editing (NCE) framework, which gives language model agents a grammar for natively editing their own context. By encoding context-editing operations in a grammar of dedicated special tokens, we teach the agents to invoke edits through the language itself. We train agents on iteratively constructed trajectory segments, with each teacher edit conditioned on the evolving context. This establish
llm agentagentictool-usebenchmark - arxiv:2610.25544 · cs.LGReinforcement learning on the EPRI AI-ccelerating Unit Commitment benchmark: diagnosing over-commitment and guiding MILP with learned policiesMaharani Devira Pramita, Aidan O'Sullivan
Reinforcement learning (RL) is increasingly proposed as a fast solver for unit commitment (UC), but standalone agents remain far more expensive than mixed-integer linear programming (MILP). In the first study of RL on the 51-generator, 72-hour benchmark of the Electric Power Research Institute (EPRI) AI-ccelerating Unit Commitment competition, we trace this gap to the agents' own exploration, close most of it, and show that the resulting policy can accelerate an exact solver. In every configuration tested, Proximal Policy Optimisation (PPO) and an action-branching Deep Q-Network (DQN) trained
benchmark - arxiv:2610.25540 · cs.AIAuditing Tool Traces in Agentic LLMs: Source Localization and Counterfactual Dependence TestingJunhao Liu, Tianxiao Li, Yongxian Wu, Ziheng Duan +3
Agentic LLMs increasingly support scientific prediction by grounding decisions in external evidence. In biology, they promise accurate, interpretable predictions for important problems by integrating complex experimental and computational evidence. However, serialized tool results can contain hidden label shortcuts, such as dataset identifiers or template artifacts, that create apparent success without biological reasoning. To address this risk, we introduce a two-stage audit. Stage 1, Per-Tool TF-IDF Decomposition (PT2D), localizes label-predictive patterns to individual tool results while av
agenticbenchmark - arxiv:2610.25535 · cs.CVResidential Construction Image Dataset Using CSI Divisions for Deep Learning-based Object DetectionFahad Iqbal, Shayan Mirzabeigi
Accurate recognition of construction site elements is essential for the application of deep learning (DL) in the construction industry. Despite rapid advances in computer vision, the adoption of DL in construction has been limited by the lack of structured, publicly available datasets aligned with established industry standards. To address this gap, we introduce the Residential Construction Dataset Using CSI Divisions (ResCD-CSI) image dataset, a manually labeled image benchmark organized according to the divisional structure defined by the Construction Specifications Institute (CSI). The data
benchmark - arxiv:2610.25534 · cs.ROAQUAACT: A Joystick-Level Underwater Teleoperation Dataset and Closed-Loop BenchmarkZhenqi Wu, Trung Tien Dong, Aditya Penumarti, Mohamed A. Fouad +4
Underwater robots support a growing range of work, from inspecting offshore structures and pipelines to surveying reefs and accompanying divers. Most of these tasks, however, are still flown by hand by a pilot, which raises the question of whether they can be carried out autonomously instead. On land, vision-language-action (VLA) models already let robots carry out tasks autonomously from language instructions, which makes them a natural candidate for taking the pilot's place. Underwater, however, there is little data to train such models. We take inspiration from how these tasks are flown tod
vision-language-actionvlavla modelteleoperationbenchmark - arxiv:2610.25522 · cs.AIWhen Empathy Misses the Goal: A Benchmark for Goal Displacement in LLM AdviceDean Ariel, Guy Laban
Conversational large language models can respond with warmth and reassurance while failing to advance the user's stated goal. We conceptualize this failure as $\textbf{goal displacement}$, the substitution of persona-maintenance content for goal-relevant substance. To measure it, we introduce $\textbf{GDB}$ (Goal Displacement Benchmark), comprising 230 scenarios across nine domains and $4{,}134$ human-scored responses from 18 models. A $2 \times 2$ taxonomy crosses hard constraints with emotional distress, distinguishing failures to confront infeasible goals or combine empathy with practical g
post-trainingbenchmarkarena - arxiv:2610.25517 · cs.CVReDT: Resolution-Decoupled Visual Geometry TransformerAbteen Arab, Guile Wu, Dongfeng Bai
We introduce ReDT, a feed-forward visual geometry transformer designed to scale multi-view reconstruction to high-definition (HD) inputs. Existing visual geometry transformers are typically trained at fixed, standard definition (SD) resolutions and exhibit degraded reconstruction quality and poor computational scaling to high input resolutions. While recent efficient architectures reduce the complexity of inter-frame attention, expensive intra-frame and encoder attention remain largely unaddressed, making HD reconstruction impractical with existing architectures. ReDT addresses both limitation
memory - arxiv:2610.25516 · cs.AIBud ContextCompact 2.0: Context Compaction for Agent Success and Accuracy with Small Language Models and Constrained Context WindowsJithin VG, Ditto PS
Agents that work for many turns outgrow the context window of the model that serves them, and on small language models (SLMs) and constrained inference systems that window is small. Context compaction replaces older turns with a summary so a session can continue, but it is lossy, costs a model call, and can help or hurt depending on when it runs. We describe Bud ContextCompact 2.0, the compaction built into the Bud agent runtime, and evaluate it in two settings. On recall questions over recorded agent sessions, compaction to a 16K budget raised a 1.2B-parameter model's accuracy to 0.335, from
agent - arxiv:2610.25511 · cs.LGDynamics-Aligned Representation Learning under Induced Partial ObservabilitySM Mazharul Islam, Manfred Huber
Model-based reinforcement learning (MBRL) has enabled sample-efficient learning of continuous control policies from high-dimensional observations through learned latent dynamics models. These models are typically trained with observation reconstruction objectives, which provide strong supervision but can bias representations toward visual fidelity leading to an undertrained prior and unreliable long-horizon predictions. We propose a stochastic gating mechanism that induces controlled partial observability during training by sampling a spatial binary mask from a Beta--Bernoulli random field and
latent dynamics - arxiv:2610.25510 · cs.LGKPI-Aware Multimodal Anomaly Detection for 5G and Open RAN TelemetryChiman Salavati, Liang Wu, Kelly Wan, Mayank Darbari +1
Reliable anomaly detection is a deployment bottleneck for 5G/Open RAN observability and emerging self-optimizing 6G systems, where hundreds of heterogeneous KPIs, severe class imbalance, and strict alert budgets make purely numerical scores brittle. We present KAC, a deployment-oriented KPI-aware multimodal anomaly-detection framework that uses KPI-level residual-text alignment to turn compact operator-style summaries into a learnable feedback signal over full-KPI temporal evidence. This gives the detector broad numerical coverage like modern time-series models, while retaining the sparse KPI-
benchmark - arxiv:2610.25508 · cs.MAHow Far Can an Agent Push? Trust-Safety Margins in Live Language-Based Economic GamesDima Galat, Marian-Andrei Rizoiu
The previous version contained Unicode em dashes. Paste this ASCII-only version: Game theory identifies when a strategic action should be accepted, but not how long a live, language-based counterpart will continue to accept it. We entered the GLEE competition in its final week with one protected agent, Milton, and four instrument agents that ran deterministic, per-game randomized trials of policy changes; Milton received changes only through pre-specified gates or registered, rollback-guarded production rules. In negotiation, learned pricing improves adjusted percentile by 0.071 (n=21,168); in
agentbenchmark - arxiv:2610.25497 · cs.CLSelf-Distillation for Supervised Fine-TuningWenxuan Zhu, Bang An, Shuming Liu, Wenqing Cui +5
Supervised fine-tuning treats the expert demonstration or gold response as a one-hot target, suppressing valid alternative continuations that the model would otherwise assign probability to. We propose Gold-Conditioned Distillation (GCD), a method that turns expert demonstrations into model-native supervision. GCD uses one model as both teacher and student: the teacher conditions on a prompt containing the gold response in context, the student conditions on the original prompt as at inference, and the student is trained to minimize the forward KL divergence to the teacher's next-token distribu
benchmark - arxiv:2610.25493 · cs.CVGroundedVDR: Grounding Video-to-Web Evidence Dependencies in Video Deep ResearchXiyuan Liang, Yu Wang, Zuhao Yang, Ziting Wang +6
Video large multimodal models have improved at identifying cross-frame visual evidence and conducting open-web retrieval, as reflected in recent video deep research (VDR) benchmarks. However, existing VDR benchmarks primarily evaluate tool-use trajectories and final responses without verifying whether a model actually uses the temporal video information or the retrieved web evidence. To fill this gap, we introduce \method{}, an evidence-grounded benchmark with three features. First, it combines human review with restricted-access probing across multiple models to exclude questions answerable f
agentagent systemtool-usebenchmarkevaluation protocol - arxiv:2610.25486 · cs.CVPCXR-Copilot: Device-aware pediatric chest radiographic analysis with segmentation, suppression, and edema-related cuesXueyang Li, Alec Girten, Peng Zeng, Mingze Jiang +5
Patients admitted to pediatric ICU and cardiac ICU often require intensive perioperative support and frequent chest radiography to monitor support devices, extravascular lung water, and body-wall edema. Radiographic interpretation and analysis is often difficult due to superimposed anatomic and iatrogenic structures that obscure clinically relevant signals. While deep learning can provide decision support, translation to pediatric radiographs is still limited by high device burden and scarcity of pediatric datasets with clinically grounded device annotations. To address this gap, we first intr
benchmark - arxiv:2610.25480 · cs.LGConsistent LLM Call Handling Across Java Frameworks: A Reference Implementation and BenchmarkOleksandr Klymenko
Java applications reach large language models (LLMs) through an integration framework, and rely on it to handle a failing provider: to retry what can be retried and nothing else, to wait as long as the provider asks, never to present a truncated answer as complete, and to record each failure in the application's metrics. We state these expectations as a six-clause reliability contract and measure Spring AI 2.0.1 and LangChain4j 1.21.0 against it with a scripted OpenAI-compatible provider: fifteen failure scenarios, synchronous and streaming, three repetitions. LangChain4j meets it in 47 of 77
benchmark - arxiv:2610.25476 · cs.ROCusADi2: Latency-Optimized Batch Computation of Robot Kinematics, Jacobians, and Hessians on the GPUJohn Viljoen, Lasse Peters, Masayoshi Tomizuka, Negar Mehr
Trajectory optimization methods, from sampling-based zero-order to gradient-based first- and second-order approaches, require many evaluations of dynamical system functions, i.e., functions derived from a robot's dynamical model, such as forward kinematics and their Jacobians and Hessians. These evaluations are largely independent, e.g., across sampled trajectories or time steps, and can thus be evaluated in parallel as a batch. When used to make decisions online, these methods require low latency, i.e., each batch must complete in as little wall-clock time as possible. Existing frameworks uti
humanoidquadrupedbenchmark - arxiv:2610.25463 · cs.AIDiaLong: Benchmarking Reminiscence in Long-Context DialoguesOwen Bianchi, Samer Aslan, Ryunosuke Saito
How well can transformer-based Large Language Models (LLMs) retain and recall memories in long contexts of dialogue? LLMs are known to forget long-range factual information in their context, yet the field lacks a comprehensive benchmark to quantitatively assess these shortcomings. We present DiaLong, a long-context dialogue dataset and accompanying memory task that standardizes evaluation in this space. It consists of 100 unique conversations created by humans and then augmented with LLMs, extending beyond lengths found in existing datasets. Our selected memory task is fact recall: classify 40
memorylong-contextlong contextbenchmark - arxiv:2610.25460 · cs.LGReliability-Preserving Compression for Deployable Medical Image SegmentationSagar Kumar, Shailja Sharma, Tirthajit Baruah, Punit Rathore
Post-training compression of pretrained medical segmentation models commonly uses quantization, pruning, or distillation to reduce model size and computational cost. Existing medical segmentation quantization studies mainly evaluate Dice, model size, memory, and latency, but this is incomplete for safety-critical use because compressed models should also preserve uncertainty signals that indicate unreliable predictions. We study this problem in compressed MRI/CT segmentation and formalize why Dice can remain stable while uncertainty changes. Motivated by this failure mode, we propose score-gui
post-trainingbenchmark - arxiv:2610.25459 · cs.ROBiDex-100: A Comprehensive Benchmark for Bimanual Dexterous Manipulation Across 100+ Simulation TasksYong Wang, Kecheng Zheng, Xing Zhu, Chao Xu +2
Bimanual manipulation with dexterous hands has attracted growing attention in research on robotic action models. However, few studies have developed benchmarks specifically tailored for dexterous manipulation, leaving the dexterous manipulation capabilities of existing models insufficiently characterized. In this paper, we present BiDex-100, a simulation benchmark for bimanual dexterous manipulation that combines broad task coverage, multi-dimensional evaluation, and high-DoF hand embodiments. Concretely, the benchmark comprises 106 tasks organized along five capability axes: generalization, i
vision-language-actionmanipulationdexterousbenchmark - arxiv:2610.25458 · cs.CLQuality-Controlled Reply and Summary Data for Post-Training Small Language Models in AI Call-Screening AgentsJiaqi Gan, Tsang Ng, Ankit Raj, Tommy Duong +4
Small language models have improved to the point where a phone can plausibly screen its own calls, answering on an absent owner's behalf without sending call content off the device, a privacy-respecting alternative to cloud call handling. Realizing that promise takes a model that can do two things well: reply within the owner's authority during the call, and afterward tell the owner what happened and what needs attention. Released small models do neither reliably - they raise unwarranted alarms about ordinary callers and write summaries that drop material conditions or report a caller's claims
post-trainingevaluator - arxiv:2610.25456 · cs.AISystem X: An LLM Decision Model That Replies from Its Own KV CacheSyed Usama Khalid Bukhari, Syed Safwan Khalid, Muhammad Aamir
Decision models return probabilities over supplied options, but a deployed agent may also need to reply in words. The present study introduces System X, a large language model (LLM) decision model that adds a text-reply option to the LLM-as-Jev readout. When this option wins, or confidence is low, the same model continues from the decision's key-value (KV) cache, avoiding a repeated prompt prefill. Qwen3-0.6B has been fine-tuned with LoRA on Reason-9k and Bitext reply rows. JevBench accuracy shows no statistically significant change (55.4% to 54.5%, p=0.87), while expected calibration error fa
agent - arxiv:2610.25448 · cs.CVUBD-Bench: Benchmarking and Analyzing Unimodal Branch Degradation in Multimodal Perception for Autonomous DrivingAng Liu, Xiang Gu, Ruikai Li, Qiyue Wang +9
Multimodal perception improves autonomous-driving performance by combining complementary sensor information. However, stronger multimodal predictions do not necessarily imply stronger representations in each modality-specific branch. Such branch-level weaknesses can remain hidden in conventional output-level evaluation. We introduce UBD-Bench, a systematic benchmark for studying unimodal branch degradation (UBD) in multimodal perception. UBD-Bench uses a Paired Frozen-Branch Probe to compare the task-relevant representation quality of the same modality after independent unimodal training and m
benchmark - arxiv:2610.25447 · cs.LGEnhanced Masked Diffusion UnlearningTomer Garber, Tom Tirer
Masked diffusion language models (MDLMs) generate text by iteratively unmasking tokens with bidirectional context. As these models reach real-world deployment, machine unlearning becomes essential. This process aims to remove the influence of a designated "forget set" without retraining from scratch, so that sensitive, copyrighted, or restricted content can no longer be elicited while preserving retained capabilities. Masked Diffusion Unlearning (MDU) is the first publicly available unlearning method customized for these models: it tunes the model $p_θ$ to match a frozen prompt-masked anchor $
benchmark - arxiv:2610.25440 · cs.CLDesign Considerations in Low-Resource Archival OCR: A Case Study in LakotaTéa Wright, Laura X. Ma, Claudia Iron Hawk, Hannah J. Haynie +4
Building and expanding access to digital archives supports language reclamation and creates lasting resources for community revitalization. Designing effective tools to support these efforts requires taking into account the community-, language-, and document-specific context. We investigate how small ($\leq$3B) vision-language models (VLMs) can be adapted as optical character recognition (OCR) tools for Lakota, a North American Indigenous language. In addition to benchmarking modeling approaches, we inspect four practical facets of low-resource archival OCR that arise in previous work and con
benchmark - arxiv:2610.25436 · cs.ROIntent-Informed Model Predictive Control for Shared Autonomy in Remote Manipulation TasksLorena Maria Genua, Camilla Tabasso, Mya Schroder, Joshua Migdal +2
Shared autonomy is a control paradigm that combines human cognitive abilities with robot precision to enable effective remote manipulation in complex environments. A key challenge is how to adapt the level of autonomy using information extracted from the human, task, and/or environment. In this paper, we propose a novel arbitration strategy that dynamically adjusts the level of autonomy in the system using the entropy of a belief distribution derived from a multimodal intent inference. Human intent is modeled with a Dynamic Bayesian Network (DBN), which updates beliefs over goals given real-ti
manipulationteleoperation - arxiv:2610.25434 · cs.LGRARE: Reinforcement Learning with Adaptive Rubric Evolution for Open-Ended GenerationXueyun Tian, Bingbing Xu, Chuyue Li, Jiaming Liu +8
Verifiable rewards have advanced reinforcement learning in mathematics and code. However, many valuable open-ended tasks share three properties: no automatic verifier of overall quality, multiple requirements to satisfy jointly, and multiple valid responses to each query. Training therefore needs feedback on requirement satisfaction across different valid responses. Rubric-based reinforcement learning provides this feedback by aggregating scores from explicit criteria. As learning progresses unevenly, some criteria become fully satisfied and lose discrimination while others remain informative.
benchmark - arxiv:2610.25431 · cs.LGMOSAIC: Multi-Objective Search for Allocating Inference Cache Across LLM Decoder LayersSravanth Kodavanti, Manjunath Arveti
The Key-Value (KV) cache is the main memory bottleneck of long-context LLM inference. Token eviction methods such as SnapKV, H2O, AdaKV and L2Norm shrink it, but they give every transformer layer the same budget, although layers differ in which tokens they need. We present MOSAIC, a multi-objective neural architecture search that learns per-layer KV-cache budgets for any eviction method. One search maps the trade-off between average budget and task score; its best allocation transfers to every target budget without further search, and an optional budget-constrained search refines it. Across fo
memorylong-context - arxiv:2610.25425 · cs.CLMixture-of-Memories: Low-Overhead Language Modeling with Modular Knowledge InjectionYifan Lu, Rubin Wei, Ziyun Zou, Junhong Liang +7
Large language models store substantial factual knowledge in their parameters, yet this knowledge is difficult to control, update, and reliably elicit at inference time. Existing memory-augmented approaches typically rely on a single auxiliary mechanism, such as retrieval, parameter updates, or neural memory, and therefore apply the same form of correction to tokens with very different knowledge demands. We propose Mixture-of-Memories (MoM), a memory-augmented language modeling framework that separates routed neural memory from writable factual memory. MoM combines a frozen base LM with two ne
memorymemory module - arxiv:2610.25414 · cs.CVUniState: Unifying Memory and Intervention for Controllable World SimulationJaeseok Jeong, Kyungmook Choi, Sohyun Chung, Youngsik Yun +1
World models offer increasingly rich environments for exploration, yet intervening in a world while observing the consequences from freely chosen viewpoints remains challenging. We tackle these challenges by introducing UniState, a framework for controllable world simulation built on a unified world-state representation. This representation unifies spatial memory and user interventions. Users specify object motion directly within spatial memory, while independently choosing the camera trajectory from which the resulting world state is observed. This unification makes the relationship between t
world modelmemory - arxiv:2610.25410 · cs.ROT(R,O) Grasp++: Cross-Embodiment Dexterous Grasping in Cluttered Environments at ScaleXin Fei, Zhixuan Xu, Huaicong Fang, Fangjun Chen +4
Dexterous grasping in cluttered scenes remains a central challenge in robotic manipulation, as it requires high-dimensional hand coordination under complex object-environment constraints. To address this problem, we introduce one of the largest simulation benchmarks to date, comprising around 0.5B grasp annotations across eight dexterous hands and approximately 50K objects. Furthermore, we propose T(R,O) Grasp++, a scalable diffusion-based framework for cross-embodiment dexterous grasp synthesis in cluttered environments. Extensive experiments demonstrate that our model significantly outperfor
manipulationdexterousgraspbenchmark - arxiv:2610.25407 · cs.ROAeroDASH: Asynchronous Vision-Language-Action Control for Continuous UAV FlightZhe Liu, Yuxuan Zheng, Jinzhe Qin, Ziqi Gao +1
Onboard inference latency creates a mismatch between the trajectories predicted by vision-language-action (VLA) policies and the ongoing motion of unmanned aerial vehicles (UAVs). We present AeroDASH, an asynchronous VLA framework for continuous language-conditioned flight. A B-spline policy is initialized through supervised fine-tuning on expert demonstrations, providing continuous trajectory references in position and yaw. During execution, trajectory fusion removes the expired prediction prefix and connects the remaining segment to the active reference with derivative continuity while limit
vision-language-actionvlaopenvla - arxiv:2610.25406 · cs.CLTowards Figurative Language Understanding in Moroccan Darija: An Evaluation of Large Language ModelsJihad R'baiti, Youssef Hmamouche, Emmanuele Chersoni, Chu-Ren Huang +1
Figurative language remains a major challenge for Machine Translation (MT) systems, as the intended meaning of such expressions often goes beyond literal translation. Although recent advances in Large Language Models (LLMs) improved the translation of Arabic dialects, benchmark resources for evaluating figurative language translation in Moroccan Darija remain scarce. To this end, we introduce the Moroccan Darija Figurative Language Corpus (MDFC), a benchmark dataset for MT and multiple-choice question (MCQ) selection covering idioms, metaphors, and proverbs. Using MDFC, we evaluate 19 open- an
benchmark - arxiv:2610.25401 · cs.ROEnergy-Score Stabilized Attentive Mixture-Density Policy: A Lightweight Approach to Multimodal Robot MotionSachiya Fujita, Hideyuki Ichiwara, Shigeki Sugano, Tetsuya Ogata
For practical motion generation, it is essential to handle multimodality (multiple actions can be correct for the same observation) while maintaining low inference cost and strong generalization from limited demonstrations. We introduce an attentive mixture-density policy trained with the Energy Score that represents future visuomotor trajectories as explicit probability distributions with a single forward pass. Concretely, we output Gaussian mixture parameters for joint commands (and optionally attention points) and optimize them with the Energy Score (computed as a sum of 1D CRPS under a dia
diffusion policymanipulatorgrasp - arxiv:2610.25399 · cs.CLFrom Unverifiable Domain Text to Cross-Lingual RLVR via Code-Switched Task SynthesisCheonbok Park, Jihwan Kim, Shinyoung Joo, Joosung Lee +3
Reinforcement Learning with Verifiable Rewards (RLVR) can improve reasoning in large language models, but specialized non-English domains expose a data mismatch. Raw English domain text is often available, while target-language RLVR tasks with verifiable rewards are scarce. We introduce code-switched task synthesis for cross-lingual RLVR, a data-centric approach that derives target-language and code-switched RLVR task variants from otherwise unverifiable English domain text. Using fill-in-the-middle multiple-choice QA as a seed task format, we synthesize automatically verifiable tasks from dom
benchmark - arxiv:2610.25396 · cs.CVPolarVis: A Dataset and Diffusion Framework for RGB-to-Polarization Perception in Adverse Driving ScenesRuilin Huang, Jue Wang, Zhen Tian, Xia Hua +1
Reflections, glare, transparent structures, and material-dependent appearance variations can introduce ambiguity into RGB-based visual perception in vehicle-relevant environments. Polarization provides complementary reflection- and material-sensitive information, but its acquisition requires dedicated sensing hardware. This paper presents RGB-P Wild, a real-world dataset containing 10,000 paired RGB and four-angle linear-polarization image sets, together with PolarVis, a diffusion-based framework for recovering polarization representations from a single RGB image. PolarVis combines hierarchica
evaluation protocol - arxiv:2610.25387 · cs.CVOnly Means Only: what a harmonization benchmark score averages overImad Buljić, Delaida Muminović
Image harmonization methods adjust the colour and lighting of an object pasted into a photograph so that it looks native to its new scene, and they are ranked by average error on one benchmark whose test images are real photographs with one object recoloured. One photograph can yield several such test images, and the benchmark promises that no photograph appears in both the training and the test split. We find that 1,399 photographs do, and the composites built on them are a quarter of the test set. Duplicate checks over file names or bytes miss this, because every composite is a different fil
benchmark - arxiv:2610.25384 · cs.LGOutcome-Anchored Credit Reallocation for Agentic Knowledge Graph Question AnsweringMingyuan Li, Yucheng Li, Xiaxia Wang, Gong Cheng
Agentic knowledge graph question answering (KGQA) requires large language models (LLMs) to solve complex questions through multi-turn interactions with a knowledge graph (KG). Outcome-supervised reinforcement learning derives verifiable outcome rewards from final answers but uniformly assigns each rollout's outcome-level advantage to all policy-generated tokens, creating a trajectory-to-operation mismatch since KG operations within a rollout can have varying effects on the search process and final answer. To address this, we propose Outcome-Anchored Credit Reallocation (OACR), which employs a
knowledge graphagenticbenchmark - arxiv:2610.25382 · physics.opticsDesign and synthesis of laser inscribed flexible Bragg gratings containing silicon oxynitride-doped silicon thin film for authentication applications with the insights from X-Ray Absorption Spectroscopy and Quantum Mechanical CalculationsAli Karatutlu, Timuçin Emre Tabaru, Umut Taylan, Zehra Gizem Mutlay +11
Silicon-on-insulator is mostly the main driving material system for the field of silicon photonics, including PICs, due to high index contrast and compatibility with current semiconductor manufacturing processes [1]. A recent example is silicon oxynitride-doped silicon [2] discovered recently by our group, offering to set a new position for the fields of silicon photonics with a precise control of refractive index from 2.1 to 2.5 together with the photosensitivity property at different wavelengths. In this study, we demonstrate that the synthesis of silicon oxynitride-doped silicon can be perf
silicon photonicsilicon photonics - arxiv:2610.25376 · cs.ROTILDE: Teacher-in-the-Loop Distillation from Embodied Foundation ModelsYuxiang Lu, Yantong Liu, Jinghua Hou, Xueli Sun +3
Large-scale action pretraining equips Embodied Foundation Models (EFMs) with strong manipulation capabilities, with particular strengths in generalization and closed-loop robustness, but reproducing such pretraining requires substantial robot data and computation. To transfer these capabilities from pretrained EFMs to policies without action pretraining, we introduce TILDE (Teacher-In-the-Loop Distillation from EFMs), which synergizes on-policy distillation at student-induced states with learning from teacher-executed recovery trajectories. We further propose AutoDAgger, an autonomous data agg
embodiedmanipulationhuman-in-the-loop - arxiv:2610.25374 · cs.LGMining the Self-Rewriting System: A Vision for Data Mining on AI-Native InfrastructureMayank Darbari, Chiman Salavati, Liang Wu, Dawei Wang +4
Classical data mining treats data as generated outside the learner. Emerging AI-native infrastructure invalidates this assumption: as agents observe system state, act, update retrieval stores and models, and promote new artifacts into the serving path, their own decisions reshape the future logs, traces, corpora, and evaluation records used for later training and release. We call this new research object a self-rewriting system and the resulting system-induced distribution shift endogenous drift. Our central, falsifiable claim is that cross-surface causes are not identifiable in general from m
retrieval-augmentedagentbenchmark - arxiv:2610.25373 · cs.AIMemDerive: A Diagnostic Benchmark for Agent Memory over Evolving DataSen Gao, Yongan Li, Han Chen, Bingsheng He +4
Long-running agents rely on memory of their past interactions for task state, such as what has been completed and which conditions still hold. Such state is usually a derived state. No record states it, and it must be computed by rules from multiple records of an evolving history. Existing memory benchmarks score only the final answer, with no reference for the intermediate states behind it. A wrong answer therefore cannot be traced to the state that was missed, and a state that silently became outdated goes undetected. We present MemDerive, a diagnostic benchmark for agent memory over evolvin
memoryagent memoryagentbenchmark - arxiv:2610.25371 · cs.LGWhistle and Wrong Lyrics: Synthetic Queries Are Not Enough for Real-World Query-by-MelodyLuigi Rivolta
People look up a song by humming it, but also by whistling it or singing it with the wrong words. Published query-by-humming systems and benchmarks focus on hums and sung queries, and no public dataset pairs whistles or wrong-lyrics singing with song recordings. We study both query types with a melody-only retrieval system: a key-normalized pitch track embedded by a 5.2M-parameter contour encoder, with a spectral-peak tracker for whistles. First, a physics-grounded generator renders synthetic whistles and nonsense-syllable singing from the vocal melodies of 11,083 songs. Training on them raise
benchmark - arxiv:2610.25369 · cs.CVPainting Change Map as guidance for generating painting processesMarkus Pobitzer, Chang Liu, Niccolò Biondi, Nicu Sebe
Digital art generation has undergone a massive paradigm shift, moving from static image synthesis to dynamic video generation that can mimic the human painting process. While existing models can synthesize visually stunning time-lapses of paintings being created or "unpainted," they largely treat the creative process as a black box. Current conditioning mechanisms are either too rigid, relying on static geometric cues like Canny edges or depth maps, or too holistic, utilizing simple scalar progress variables that offer no localized control. Consequently, users lack the agency to steer how a pa
benchmark - arxiv:2610.25366 · cs.LGGreenNIDS: A Benchmark Workflow and Analysis Dataset for NetFlow Intrusion Detection ExperimentsLilia-Eliana Popescu-Apreutesei, Sabina-Cristiana Necula
Comparing the predictive performance and computational cost of flow-based intrusion detection experiments requires both quantities to be available for each run. GreenNIDS connects benchmark execution on public NetFlow datasets to a reusable analysis dataset. Model implementations write a common set of outputs, which are ingested into a run-level table with companion tables for model configuration, feature importance, and training history. The run-level table combines predictive metrics with timing, throughput, memory use, and energy estimates based on thermal design power. A CPU-only demonstra
memorybenchmark - arxiv:2610.25362 · cs.AIAligning Cultural Behaviors in Arabic Language Models via PreferencesJunhong Liang, Daniil Orel, Saeed Almheiri, Bilal Elbouardi +10
Arabic LLMs have improved rapidly, but aligning them with country-specific cultural behavioral preferences remains underexplored. Previous work has focused primarily on instruction-tuning for factual knowledge, overlooking nuanced cultural behaviors. We propose ArabPref, which to our knowledge is the first large-scale pan-Arab preference data set for the alignment of cultural behavior conditioned on the country. Our contribution includes two resources in English and Modern Standard Arabic: a culturally grounded preference dataset for training and evaluation, and a multiple-choice benchmark des
post-trainingbenchmark - arxiv:2610.25360 · cs.LGOperator-Mismatch Contrastive Learning for World-Relative Trajectory CoherenceOne Octadion, Firmansyah Sundana, Novanto Yudistira
World models are increasingly judged by how well they rank plausible futures, yet plausibility alone is the wrong target. The same scene can evolve under many latent rules, and a future that is smooth and physically valid under one rule can be incoherent under the rule actually in force. We formalize the resulting failure, operator-blindness, and show that it is systematically hidden by standard evaluation: a scorer can attain near-perfect generic ranking while separating current-world from other-world futures no better than chance. We introduce Operator-Mismatch Contrastive (OMC) learning, wh
world modelbenchmark - arxiv:2610.25359 · cs.LGCD-OPD: Contextual Digest for On-Policy DistillationMingyuan Li, Yucheng Li, Xiaxia Wang, Gong Cheng
Teacher takeover in on-policy distillation (OPD) can produce incorrect responses even when the student and teacher independently answer the same problem correctly, a failure we quantify using Collaboration Failure Rate (CFR). To address it, we propose Contextual Digest for On-Policy Distillation (CD-OPD). CD-OPD structures the earlier context of teacher guidance through a teacher-generated dependency graph, curating the required problem conditions and preceding reasoning as node summaries alongside the unchanged guidance in a Contextual Digest before student continuation. We further develop on
benchmark - arxiv:2610.25358 · cs.ROLateMatch: Breaking the Interaction Bottleneck via Cacheable Matching for Efficient Visual LocalizationMohammad Omama, Aeyan Ashraf, Sandeep Chinchali
Visual localization is critical for embodied agents, but deploying state-of-the-art deep feature matchers on edge devices is fundamentally limited by their massive computational cost. Standard deep matchers tightly interleave intra-image context building and inter-image matching at every layer, creating a severe interaction bottleneck. Localizing a single query against K database images forces the system to execute the full network K independent times with no capacity for caching. To break this bottleneck, we advocate for a paradigm shift toward cacheable image matching. We realize this paradi
embodiedembodied agent - arxiv:2610.25333 · cs.CVCoRe-WM: Conditioned Response-State World Modeling for Longitudinal Treatment-Response Prediction in Breast DCE-MRILin Zhao, Bowen Chen, Qiaohui Gao, Shaowen Wan +3
Pathologic complete response (pCR) after neoadjuvant chemotherapy is a key prognostic indicator in aggressive breast cancer, yet existing prediction approaches either rely on pretreatment imaging alone, missing the patient's evolving response, or optimize multi-visit classifiers under direct outcome supervision, coupling representation learning with the clinical endpoint. We formulate longitudinal pCR prediction as a world-modeling problem and propose Conditioned Response-state World Modeling (CoRe-WM), which learns a latent state-transition model that predicts the next imaging state from the
world model - arxiv:2610.25330 · cs.LGFollow the Mixed Groups: Ridge-Tracking Gaussian Sampling for Zero-Variance Web RolloutsTomas Duncun, Jonah Ellison, Owen Pell, Mateo Ruiz +3
Group relative policy optimization learns from a question only when the sampled responses disagree. A group that is entirely correct or entirely incorrect has zero reward variance and a zero advantage. On the Web, that group has already spent its searches and page fetches. Gaussian curricula aim the next batch at pass rates where a finite group can still mix. A Gaussian on the latest empirical rate treats every all-fail history as the same point, and a center stepped by the average signed change of those rates is pushed by prompts already stuck at 0 or 1. RIDGE keeps the Gaussian curriculum an
retrieval-augmented - arxiv:2610.25325 · cs.LGWhere Groups Split: A Probe Prior for Nonzero Within-Group Variance in RL Post-TrainingJonah Ellison, Mateo Ruiz, Owen Pell, Tomas Duncun +3
Group relative policy optimization learns from a prompt only when the rewards in the sampled group disagree. With a binary verifier, a group that is entirely correct or entirely incorrect has within-group variance zero and a zero advantage. Uniform sampling continues to spend rollouts on prompts that are unanimous now and stay unanimous as training goes on. Pass rate is a weak forecast of that future. Two prompts at the same pass rate can move apart: one into the range where a finite group splits, and one toward a ceiling where the group becomes unanimous again. Aperture estimates the later pa
post-training - arxiv:2610.25323 · cs.LGWhere and Why Search-Augmented LLMs Fail When Reasoning over Rapidly Evolving KnowledgeTianyi Hu, Bhavyajeet Singh, Niket Tandon, Akhil Arora
As the world changes, new information is published while older records remain online. This complicates reasoning for large language models (LLMs), which must retrieve evidence and resolve conflicts to determine which information remains valid. Existing evaluations treat information seeking as a black box, revealing little about where and why LLMs fail. To address this gap, we introduce TREK, a suite of three components: TREKTRACE is a plug-and-play framework that attributes failures to specific stages of retrieval and reasoning. We apply it to TREKBENCH-WILD, a meta-benchmark of real-world que
benchmark - arxiv:2610.25317 · cs.CLAtoms, Not Knowledge: What Production-Scale Addressed-Memory Writes DepositWei Gao, Cheng Tian, Duan Yang, Jianqing Yin
Deterministic hash-addressed conditional memory permits local post-hoc knowledge injection. Extending arXiv:2610.21553's bounded-feasibility audit rejecting under-specified generality claims with repaired probes and training-dose/capacity axes, we give, to our knowledge, the first production-scale write-content verdict. We test 124M-7B self-trained/publicly pretrained beds, Qwen3.8-Flash-Next-FP8 (176.95B; 320,001,536 FP8 rows), and DeepSeek-V4.1-Flash. Write coverage requires a sufficiently trained backbone prior, full-width readout, trainable gates; undertrained-bed intervention confirms a t
memory - arxiv:2610.25309 · cs.CVSENRI: Understanding the Space-Event Nexus in Real-Life VideosKehan Lan, Kaining Ying, Henghui Ding, Kaipeng Zhang
Human memories are organized in a way that connects events with their corresponding spatial contexts. An AI assistant should also be able to memorize and recall the nexus of events and spaces. Thus we introduce SENRI, a task that studies this space--event connection in long, continuous real-life egocentric video. It contains six task types in two directions: Event$\rightarrow$Space, which recovers spatial knowledge from event cues, and Space$\rightarrow$Event, which retrieves events from regions and trajectories. To evaluate this, we build SENRI-Bench, containing 500 manually annotated multipl
memorypersistent memory - arxiv:2610.25301 · cs.AIShared-Private Cross-Modal Discrepancy Learning for Imbalanced Multi-Source Audio-Visual Deepfake DetectionQiwu Wen, Jahangir Alam
Audio-visual deepfake detectors are often trained under a single dataset or manipulation regime, but multi-source training collections may contain heterogeneous attacks and uneven source distributions. Training separate detectors for each source is inefficient, while a single shared detector may be biased toward dominant source-specific patterns. We study known-source imbalanced multi-source audio-visual deepfake detection using FakeAVCeleb and LAV-DF, where source identity is supplied during training and inference. To address this setting, we build on the Multi-Scale Cross-Modal Transformer E
manipulation - arxiv:2610.25298 · cs.AIAligning Language Models with Indigenous Community Cosmovision through Constitutional AISilvia Fernández-Sabido, Jorge Reyes-Magaña, Jesús Adrián Chay Tun, Jean Carlos Buenfil +2
Large language models (LLMs) are predominantly trained on corpora that reflect Western, anglophone worldviews. This asymmetry is not neutral: when a model answers questions about territory, community authority, or the relationship with nature, it does so from assumptions that may be alien to the communities using the system. This paper presents a system for aligning language models with the ethical principles of Indigenous communities in Latin America. Drawing on interviews with five Indigenous leaders and defenders, we built an operational cultural constitution grounded in an Indigenous cosmo
retrieval-augmented - arxiv:2610.25296 · cs.LGWorld Model Interpretability: From Decoding Representations to Understanding Latent WorldsJize Wang, Yongcheng Jing, Xinyi Le, Cailian Chen +2
Emerging Joint-Embedding Predictive Architectures (JEPAs) are moving beyond representation learning towards latent world models that support prediction, planning, and control. Yet their latent worlds naturally lack a human-interpretable interface. In this Perspective, we argue that as JEPA latents become internal states for prediction and action, interpretability must move beyond what they encode to how latent variables interact, evolve, and causally drive prediction. To this end, we envision a four-stage research agenda, involving Ground, Structure, Verify, and Act, to guide this shift toward
world model - arxiv:2610.25290 · cs.CVMotionShift: Environment-Compatible Motion Transfer via Planner-Evaluator Synergistic AdaptationYanchen Liu, Junyao Gao, Yanan Sun, Wensheng Chen +2
Existing motion transfer methods excel at exact, rigid motion replication from reference videos to target scenes, whereas they often fail to account for the compatibility between the transferred motion and the target scene environment in accordance with physical laws. In this work, we propose MotionShift, a novel motion transfer framework that achieves high-fidelity motion transfer while ensuring physical compatibility with the target scene. In particular, we design a planner--evaluator iterative motion adaptation mechanism, in which a Large Language Model (LLM) acts as the planner to generate
evaluator - arxiv:2610.25289 · cs.LGThe Explainability Illusion: Must ViTs Exhibit Patch-Decoding Interpretability?Ronen Schaffer, Hila Manor, Tomer Michaeli
Transformers are widely assumed to exhibit token-decoding explainability. In particular, in vision transformer (ViT) classifiers, decoding individual patch tokens has become a common diagnostic tool due to its tendency to yield human-interpretable outputs. In light of the increasing popularity of this approach, an important question is whether token-decoding explainability is an inherent property of transformers that should be expected to hold across architectures, scales, and training objectives, or merely an artifact of some implicit bias. In this paper we systematically study token-decoding
world model - arxiv:2610.25288 · cs.ROVision as Proprioception: Estimating Humanoid Robot Pose from VideoJiaxi Jiang, Matthias Heyrman, René Zurbrügg, Marco Hutter +1
Estimating a humanoid robot's internal state from video has a wide range of real-world applications but remains unexplored. During interactions with a robot, its onboard state information may be unavailable due to privacy, ownership, or communication constraints. Meanwhile, online and AI-generated robot videos provide rich sources of data but typically lack corresponding joint configurations, limiting their direct use in robot learning. To address these issues, we present HumanoidPose, a new framework that treats vision as proprioception to recover full-body joint configurations in world space
humanoid - arxiv:2610.25287 · cs.ROThe Bitter Lesson Comes to RoboticsChristopher Leet, Achyutha Menon, Sravanthi Machcha, Sven Koenig +2
General purpose multimodal language models (LMs) are increasingly capable of directly controlling robotic arms (e.g., by commanding end effector poses) to manipulate their environment. Tracking the development of these capabilities is important for understanding their potential benefits, such as improving industrial automation, and risks, such as enabling malicious users to produce dangerous chemical or biological agents. To this end, we investigate the ability of frontier multimodal LMs to directly control physical robotic arms. We evaluate seven models at multiple reasoning-effort settings o
embodiedmanipulationbenchmark - arxiv:2610.25281 · cs.LGBridging Concepts: Multi-Agent Reasoning over Scientific Concept Graphs for High-Impact Research PredictionDonggyu Yoon, Minsung Hwang, Juneha Baek, Donghyuk Shin +1
Within a research community, new topics often emerge when previously unconnected concepts are first studied together. We construct scientific concept graphs (SCGs)--domain knowledge graphs whose nodes are fine-grained research concepts extracted from paper text by LLMs and whose edges capture their co-occurrence in papers--and study the task of forecasting high-impact concept pairs, i.e., discovering concept pairs that are currently unconnected but will have high impact when connected. To this end, we introduce new benchmark datasets constructed from CIKM, KDD, and SIGIR conference papers, wit
knowledge graphagentllm agentmulti-agentbenchmark - arxiv:2610.25273 · cs.CVGeoART: Geometry-Aware Refinement via Test-Time Scaling for Camera-Conditioned Video DiffusionLiudi Yang, Ziyuan Liu
Camera-conditioned video diffusion models often suffer from inaccurate camera control and geometric inconsisten- cies because camera constraints and estimated geometry are only weakly enforced during denoising. Existing reme- dies often require costly geometry-aware fine-tuning or ad- ditional 3D supervision. We introduce GeoART, a training- free self-refinement framework that improves camera con- trollability through geometry-aware test-time scaling. At selected denoising steps, GeoART generates multiple can- didates within the Predict-and-Perturb cycle and eval- uates them using a geometry p
self-refinement - arxiv:2610.25269 · cs.LGEG-GRPO: Evidence-Guided Step-level Credit Assignment for Deep Search AgentsSijia Luo, Mengjie Gao, Yuxiao He, Mengshu Sun +3
LLM-based search agents demonstrate strong capabilities in long-horizon information seeking and reasoning, typically trained via reinforcement learning. However, existing approaches predominantly rely on trajectory-level outcome rewards, providing limited supervision for complex search and reasoning processes. Although step-level credit assignment offers a promising solution, obtaining reliable and task-grounded supervision for open-ended search remains challenging. In this paper, we propose Evidence-Guided Group Relative Policy Optimization (EG-GRPO), a framework that derives step-level super
benchmark - arxiv:2610.25265 · cs.CLSPARC: Spoken Profile Authorship Recognition CorpusEfthymios Tsaprazlis, Bowen Yi, Tiantian Feng, Junghwan Kim +2
Speech anonymization has focused on the acoustic signal, leaving open how much speaker information the transcribed words themselves retain. In this work, we study pairwise speaker and demographic-attribute verification from transcribed speech. We introduce SPARC (Spoken Profile Authorship Recognition Corpus), a benchmark built from two podcast corpora comprising over 12,500 speakers, 1.5M quality-filtered utterances, and speaker-level inferred or source-provided labels for sex, age, and accent, and evaluate five attribution models and three LLMs on identity- and attribute-verification trials w
benchmark - arxiv:2610.25261 · cs.ROGOFusion: Real-Time Gaussian-Occupancy Submap Fusion for Fisher-Informed Autonomous Robot NavigationBoxin Bu, Simon Boche, Haoming Zhang, Jaehyung Jung +5
In this letter, we present GOFusion, a real-time visual mapping system for autonomous robot navigation that anchors Gaussian primitives to local occupancy maps, providing both spatial awareness for path planning and high-fidelity appearance that enables information-driven robot navigation. GOFusion builds 3D Gaussian maps from pose estimates and local occupancy maps obtained via visual-inertial SLAM and enables pose corrections to maintain both geometric and visual consistency. Furthermore, it restricts the 3D Gaussian Splatting optimization to active submaps to bound GPU memory usage. Buildin
memory - arxiv:2610.25254 · cs.LGBatchParScale: Sparse Batch-Mixing for Sequence ModelsHans Peter Lyngsøe Raaschou-jensen, M. Onurcan Kaya, Desmond Elliott
Recent test-time scaling methods improve model performance by increasing inference-time compute. One practical form of this is parallel scaling, where a model samples multiple independent reasoning streams and aggregates their answers using methods like voting, reranking, or verification. While effective, this approach leaves the streams uncoordinated, and a discovery made in one rollout cannot influence any other. Multi-Sequence Verifiers jointly score candidate answers after generation, but the reasoning streams remain independent during inference. ParScale takes a step toward model-internal
multi-agentbenchmark - arxiv:2610.25250 · cs.LGImproving the Expressive Power of Hybrid Sequence Models via Multi-Hop RetrievalYam Eitan, Gilad Yehudai, Haggai Maron
Hybrid architectures offer a promising approach to efficient language modeling by combining a small number of full-attention layers with a larger number of efficient sequence layers, such as sparse attention and recurrent layers. While full-attention layers maintain a KV cache that grows linearly with sequence length, efficient layers typically rely on separate, fixed-size states rather than accessing this richer existing memory. Recent work has begun exploring ways to allow efficient layers access to the KV cache, but the theoretical benefits of such access, and the extent to which existing a
memory - arxiv:2610.25246 · cs.LGEstimating Causal Bounds from Observational Data With Neural Generative ModelsMax Zhu, Srivathsan Amruth, Sina Akbari, Qingyuan Zhao
Identifying causal bounds from discrete observational data is a fundamental problem in causal inference. While exact analytical bounds exist for certain graphs, this is not generally the case. In this paper, we introduce a framework for estimating such bounds using neural generative models together with a new extrapolation procedure to estimate these bounds more accurately. Across multiple causal inference settings, it produces bounds that are consistently closer to the true bounds than those obtained with existing general neural approaches. We validate the framework on a range of benchmark pr
benchmark - arxiv:2610.25241 · cs.LGAssessing the Empirical Evidence for Loss of Control from Agentic General-Purpose AIRisto Uuk, Santeri Koivula, Lorenzo Pacchiardi, Rokas Gipiškis +2
Theoretical literature warns that advanced generalist agentic systems might pursue instrumental goals that are misaligned with human preferences, thus posing risks of irreversible loss of human control. The EU's Code of Practice for General-Purpose AI lists loss of control as a systemic risk that must be assessed, yet no structured assessment of the quality of the empirical evidence for this risk exists. This study fills that gap. We evaluate 19 recent empirical studies across eight properties identified from threat models (such as strategic deception, situational awareness, and autonomous rep
agentic - arxiv:2610.25223 · cs.CVMap2Vid: Embodied Navigation Video Generation with Human-Friendly Motion ControlYulin Dai, Yuqian Fu, Tianwen Qian, Zhen Xing +2
Controllable video generation has recently achieved impressive visual quality, yet existing control paradigms remain poorly suited to embodied navigation. Text prompts are spatially ambiguous, while camera-trajectory control relies on low-level motion specifications that are unintuitive for human users and difficult to manipulate. In this work, we introduce Map2Vid, a novel task that reformulates embodied navigation video generation as a BEV-map-to-video synthesis problem, together with a large-scale benchmark for systematic evaluation. By using BEV maps, which provide a visual representation
embodiedbenchmark - arxiv:2610.25211 · cs.LGNeuroAct: Advancing EEG Analysis and Decoding via Dual-Loop Autonomous ResearchWei Xiong, Jiangtong Li, Han Zhao, Linxiao Yang +2
EEG analysis and decoding underpin brain-computer interfaces, cognitive assessment, and sleep monitoring. Fixed pipelines and reward-driven search do not explicitly determine how signal structure, representations, predictor behavior, and training dynamics contribute to performance or generalization gaps, leaving intervention choice underdetermined. We introduce NeuroAct, an autonomous research agent framework for EEG analysis and decoding. NeuroAct implements dual-loop autonomous research across EEG foundation models, task-specific predictors, and model combinations. In an inner loop, research
agentagent framework - arxiv:2610.25208 · cs.AIBrandEval: A Two-Track Multi-Agent Benchmark for Risk-Sensitive LLM Crisis CommunicationHaotian Gan, Weidong Tang, Hanbin Sun, Yi Luo +4
Evaluating LLMs for crisis communication requires more than judging isolated responses, because public feedback and competitor actions can reshape trust over time. Existing benchmarks remain largely static and single-turn, offering limited evidence about long-horizon risk-sensitive communication. We introduce \textbf{BrandEval}, a two-track benchmark for product-crisis communication, focusing on public brand responses to product-related safety, quality, service, and trust incidents. BrandEval-Static evaluates stakeholder-aware strategic analysis, while BrandEval-Dynamic stress-tests response p
agentmulti-agentagent benchmarkbenchmark - arxiv:2610.25206 · cs.CLA Thousand Faces: A Persuasion Benchmark with Experience-Grounded Persuadees Informed by Psychology and PragmaticsZiqing Zhuang, Dingyi Zhang, Deyu Zhou
People can express the same beliefs and desires yet need different reasons to change their minds. What makes an argument relevant depends on how those views arose: experiences shape what a person finds credible, what matters to them, and what could prompt reconsideration. These personal grounds may also be expressed indirectly, through contextual clues in conversation. Benchmarks that specify beliefs and desires without these connections can reward arguments that match a description while overlooking whether they address the person's reasons. We introduce A Thousand Faces (ATF-Persu), with 1,0
benchmark - arxiv:2610.25202 · cs.LGViCoR: Estimating Visual Necessity via Counterfactual Residuals for Multimodal Medical Data SelectionHaotian Gan, Yudong Li, Xintian Li, Jiarui Han +1
Medical vision-language models often exhibit hidden stratification, where high overall performance masks failures on subtle visual findings and overlapping clinical conditions. We trace this failure mode to modality imbalance during fine-tuning, where strong linguistic priors can suppress visually grounded learning and induce plausible but weakly grounded rationales. Existing data selection methods usually score image-text pairs as holistic units, without separating visually necessary samples from shortcut-solvable ones. To address this problem, we propose ViCoR, a counterfactual-residual-base
benchmark - arxiv:2610.25201 · cs.LGRAIL-5G: A Challenging Real-World Dataset and Benchmark for AI-Based 5G Indoor LocalizationJonas Pirkl, Andreas Eidloth, Maximilian Kasparek, Tobias Feigl +1
Reliable radio-based indoor localization is essential for industrial automation, yet existing methods are often evaluated in environments that do not fully reflect the severe multipath and non-line-of-sight (NLOS) conditions encountered in real deployments. This limits our understanding of how AI, i.e., modern geometric, unsupervised, and semi-supervised localization methods generalize to industrial 5G settings. We present RAIL-5G, a public real-world 5G NR channel impulse response (CIR) benchmark recorded in an operational railway maintenance facility. RAIL-5G covers an 811m$^2$ indoor area w
benchmark - arxiv:2610.25192 · cs.CLAcharya 2.0: An AI Tutor for Self-Learning Subjects in HindiShantipriya Parida, Debasish Dhal, Sonal Khosla, Sambit Sekhar +1
Text processing is crucial for low-resource languages as it enables the development of tools and technologies that bridge the digital divide and provide equitable access to information and services. By empowering low-resource languages with robust text processing, we can preserve linguistic diversity, enhance education, improve communi- cation, and foster inclusivity in a rapidly digitizing world. Due to the lack of sufficient annotated datasets and lin- guistic resources, it is challenging to develop accurate and reliable language technologies for these low-resource languages. Large Language
retrieval-augmentedrag - arxiv:2610.25185 · cs.AIReranking Answers of Large Language Models with Knowledge GraphsMikhail Salnikov, Hai Le, Olga Tsymboi, Ivan Lazichny +5
Answering natural language questions over knowledge graph data is challenging due to the vast number of facts, which can be difficult to process and navigate. One potential solution for this issue is to use mined subgraphs related to the query, although this process still requires extracting these subgraphs. This research presents a solution for extracting subgraphs related to entity candidates from a question-and-answer set, which can be obtained by inferring a large language model by calculating the shortest paths between entities. The proposed approaches detail various features that can be
knowledge graph - arxiv:2610.25182 · cs.LGDialogue Generation for Conversational Recommendation Using User Search Interactions and Knowledge GraphsMinsung Hwang, Donggyu Yoon, Gowoon Hwang, Kwanghee Jang +1
Large-scale conversational recommendation datasets typically require manual dialogue construction or annotation. We propose SearCl, which automatically generates conversational recommendation dialogues grounded in user search and click behaviors. SearCl segments user interactions into query sessions and derives each recommendation target from content selected after a search query. A configurable probabilistic state machine controls the dialogue flow, while two LLM agents acting as the seeker and recommender generate utterances according to turn-level constraints. To support recommendation, the
knowledge graphllm agentbenchmark - arxiv:2610.25179 · cs.AIMemory, Know Your PLACE: Governing Lifecycle State and Query-Time Eligibility in Persistent-Memory AgentsChen-Fang Hu, Hong-Han Shuai, Woan-Shiuan Chien
Persistent memory turns prior interactions into durable agent state, but existing systems provide limited control over whether that state remains valid and whether it should influence a later query. We argue that reliable memory requires governance over both state formation and query-time influence. We introduce PLACE, a factorized governance layer that adds write-side state certification and read-side authorization to a memory backend. A factorial evaluation on Mem0 with Qwen2.5-3B and Qwen2.5-7B isolates the contribution of each mechanism. Results show that write-side governance substantiall
memorypersistent memoryagent - arxiv:2610.25170 · cs.LGWorld Models, Trees and Foundation Models for Retail Forecasting: Evidence from Four Public Data SetsKarthik Iyer, Madan Dabbeeru
A retailer makes different decisions at different grains of the product, store and time hierarchy: store orders for an item in a store, supplier funding for an item across stores, category performance by store, and the financial plan for the chain. Which forecasting model to use depends on the grain. This paper compares models at each grain on four public data sets (M5, six Dominick's categories, Breakfast at the Frat, and Corporación Favorita, the last kept out of all development), over rolling forecast origins with paired tests and model confidence sets. It covers price and promotion through
world model - arxiv:2610.25161 · cs.LGIndistinguishable Agents: Training Agent Privacy as Twin-World IndistinguishabilityJiaming He, Guanyu Hou, Yi Yu, Xudong Jiang
LLM agents search, shop and send messages for their users, and every action is seen by some party. Existing privacy training rewards agents for not mentioning protected items, yet an observer can infer a private fact from what the agent does. Since an observer learns a fact only if its view changes with the fact, we propose \methodfull (\method), which trains on twin worlds that differ only in the private fact. A frozen distinguisher scores each observer's view by its log-likelihood ratio for the true world, and paired GRPO trains the agent on task success minus this ratio. We prove that this
agentllm agent - arxiv:2610.25155 · cs.CLMorpheus: a fully local, defense-in-depth de-identification engine for Greek clinical text - a benchmark evaluationFilippos-Paschalis Rorris, Nikoleta Stanitsa, Tatiana Rorri, Foteini Rorri +1
Background and Objective: Clinical free text cannot leave the hospital without de-identification, yet no de-identification system or benchmark exists for Greek clinical text. We evaluated Morpheus, a fully local de-identification pipeline for Greek clinical documents, on 500 synthetic discharge summaries. Methods: The notes (300 cardiac surgery, 100 cardiology, 100 nephrology) were generated by a frontier large language model (Gemini 3.1 Pro) constrained to authentic Greek hospital document layouts. Notes were pre-annotated at generation; two clinicians then independently reviewed and correcte
benchmark - arxiv:2610.25150 · cs.LGLearning without Overwriting: Structural Knowledge Fusion for Continual LearningKanishk Bakshi, Amitansu Pattanaik, Kathiravan Srinivasan
Catastrophic forgetting in sequential deep learning arises from a fixed-capacity architectural constraint: parameter updates encoding new tasks perturb weight subspaces supporting prior representations. We introduce the Knowledge Fusion Network (KFN), which mitigates the stability-plasticity trade-off in class-incremental learning through novelty-gated structural expansion rather than parameter negotiation within a shared capacity budget. KFN decouples task acquisition from knowledge consolidation: each incoming task is learned in an isolated lightweight specialist network (2.8M parameters) an
memory - arxiv:2610.25146 · cs.CVPLANE: Plug-and-Play Fine-Grained Visual Experts for Hallucination Mitigation in Large Vision-Language ModelsJianghao Yin, Lang Yu, Qianxi Yu, Qin Chen +3
Large Vision-Language Models (LVLMs) commonly suffer from hallucinations, which appear as fabricating non-existent objects or outputting incorrect attributes and spatial relationships in images. Previous works empirically indicate that CLIP-style vision encoders tend to focus on primary objects while ignoring image details, which is one of the causes of hallucinations. To address this issue, we propose plug-and-play fine-grained visual experts (PLANE) comprising two components: LoRA visual experts operating on CLIP for multi-aspect image encoding, and an instruction-aware gate network for adap
benchmark - arxiv:2610.25142 · cs.CVImage Dehazing as In-Context Tabular InferenceLe-Anh Tran
Image dehazing is traditionally addressed through handcrafted physical priors or task-specific deep learning models, often requiring carefully designed architectures or extensive model training. Recent advances in tabular foundation models (TFMs), however, offer an alternative learning paradigm in which continuous relationships can be inferred directly from contextual examples without task-specific model training. Motivated by this capability, this paper investigates whether single-image dehazing can be effectively reformulated as an in-context tabular regression problem and introduces DehazeT
benchmark - arxiv:2610.25140 · cs.LGSpectral Adaptive Graph Transformer for Efficient and Scalable Traffic ForecastingAmirhossein Ghaffari, Susanna Pirttikangas, Ekaterina Gilman
Traffic forecasting models based on spectral graph convolution offer an interpretable account of how information is mixed between sensors, but they rely on an eigendecomposition of the graph Laplacian. It costs cubic time, is fixed by the road network, and provides a symmetric operator that cannot represent the directed propagation of traffic congestion. Attention-based spatio-temporal transformers remove these limitations but generate a dense sensor-by-sensor attention map at each layer and time step, so memory, rather than accuracy, becomes the binding constraint on large networks. We introd
memorybenchmark - arxiv:2610.25134 · cs.LGTIRE: Top-Down Formula Instantiation with Reliable Execution for Physics Problem SynthesisXiaoyu Wu, Zhaoxin Yu, Xinglin Xiao, Nan Xu
Advancing physics reasoning in large language models requires scalable training data that connect physical laws, concrete problem scenarios, and verifiable numerical answers. Expert-designed problems offer top-down control over physical laws, conditions, and scenarios, but require costly human effort. LLM-based synthesis from texts scales efficiently, but model-generated answers and assessments leave supervision vulnerable to hallucinations. We introduce TIRE, a framework for Top-Down Formula Instantiation with Reliable Execution that brings knowledge-driven problem construction to large-scale
benchmark - arxiv:2610.25130 · cs.ROStructure over Reasoning: Unified Zero-Shot Navigation with a Small Vision-Language ModelZeqi Li, Xunyi Zhao, Zerui Li, Sihao Lin +2
Vision-language navigation is moving into photorealistic, physically simulated scenes, where the leading zero-shot agents put a vision-language model in the control loop to choose every move and stop. Robots favour small models, whose abilities are uneven: Qwen3-VL-4B can judge whether a single view looks like the goal, but its route reasoning is at chance, its distance estimates nearly constant, its stop votes unreliable. Our agent therefore asks it two perceptual questions per step, which direction looks like the goal and whether the view justifies stopping, plus one gated frontier query on
quadrupedagent - arxiv:2610.25127 · cs.CLMemBubble: Benchmarking How Agent Memory Shapes LLM Response DiversityJiajun Tan, Fei Sun, Xinyu Ma, Shixuan Zhang +5
Memory serves as the core component of LLM-based agents, enabling personalized responses via extracting and using user preferences from past interactions. However, its use may reduce response diversity, thereby narrowing the range of information available to users. Existing evaluations rely largely on synthetic interactions and cover few memory systems, leaving the effects of memory on response diversity insufficiently evaluated in real-world scenarios. We introduce MemBubble, a benchmark of 1000 instances built on preferences extracted from real user-assistant conversations. To measure the ex
memoryagent memoryagentbenchmark - arxiv:2610.25126 · cs.LGTrackable Skill Library Lifecycle Management for Agentic Reinforcement LearningXi Zhu, Yuanlin Duan, Changyi Yang, Kai Mei +3
Skill-augmented agentic RL has emerged as an effective approach to improving LLM agents on long-horizon tasks. Despite this progress, individual skills may follow distinct trajectories as the policy evolves, with some becoming redundant after the model internalizes their capabilities, others remaining valuable as external scaffolding, and still others turning harmful. However, existing skill library systems fail to identify each skill's true contribution due to free-riding among co-retrieved skills, difficulty confounding across tasks, and the absence of skill-specific counterfactual evidence,
llm agentagentic - arxiv:2610.25123 · cs.CLMemories That Outlive Their Reasons: Dependency-Tethered Memory for Tool-Using LLM AgentsYiqi Wang, Jinqian Ju, Jiaqi Zhang, Zhangkai Wu +4
Agents now act on the Web, publishing packages, models and datasets through registry APIs, and remembering past verifications spares them the work. But a verdict rests on premises. When a signing key is revoked, everything built on it stops passing, yet the advisory names only the key, and the old verdicts of its dependents may never share a retrieval window with it. This non-locality of invalidation turns memory into a liability: in long simulated chats, an agent reading a memory of atomic facts publishes 62% of the objects whose failure lies upstream, each backed by a record that was true wh
memoryagentllm agent - arxiv:2610.25122 · cs.CLBehave Like Experts: Fusing Data with Semantic Knowledge for Effective and Efficient Semantic Map ConstructionZhu Liu, Zhen Hu, Yuxuan Liu, Ying Liu
Semantic map models (SMMs) construct a graph that reflects conceptual similarity by comparing cross-lingual form-meaning mappings. Experts build such models not only by obeying the data constraint, i.e., the connectivity hypothesis, which states that conceptual nodes belonging to the same form must be connected, but also by incorporating their prior semantic knowledge. Current automatic methods ignore this knowledge, which harms both accuracy and efficiency by enlarging the search space of candidate graphs. In this paper, we fuse the distributional semantic knowledge hidden in the internal rep
benchmark - arxiv:2610.25119 · cs.LGMemory as Belief: Bayesian Evolution of Memory for Self-Evolving LLM AgentsYiming Xu, Qian Yang, Zihan Chen, Yuefei Wu +3
Large language model (LLM) agents increasingly rely on external memory to accumulate past experiences for continual adaptation. However, effectively reusing memory requires accurately assessing its value, as retrieved experiences may become unreliable when transferred across task contexts. Despite its importance, memory valuation remains largely underexplored, with existing methods relying on historical retrieval success rates that overlook the context dependency of memory usefulness and the varying reliability of estimates under different amounts of evidence. Inspired by the Bayesian brain hy
memoryexternal memoryllm agentagenticself-evolvingbenchmark - arxiv:2610.25114 · cs.ROCompile Once, Run Cheap: Frozen, Portable Skills for Embodied Agents, and What Recursive Self-Improvement AddsJingwen Bai, Guochao Bai, Canbo Zhang, Shuangqi Li +3
Embodied AI lacks the data-rich pretraining that gave language models their agents and skills, so new manipulation capability costs demonstrations per task or a frontier model in every episode. We study a third route: a strong general model acts only as an offline author that writes, runs, diagnoses and repairs a manipulation skill in simulation until it passes an exam on layouts drawn after freezing; the frozen skill then runs as code alone or under a cheap caller that reads only its description. We call this skill-level recursive self-improvement Skill-RSI and specify a compile protocol (iso
embodiedmanipulationrobotwinembodied agentself-improvement - arxiv:2610.25113 · cs.LGSelf-organizing reinforcement learning: a swarm approaches the explore-exploit optimumGalen J. Wilkerson
The explore/exploit trade-off is the core problem of reinforcement learning (RL). We show that a minimal foraging swarm approaches its optimum - the ideal-free / Nash allocation of a multi-agent resource-sharing game (an equilibrium, not a return maximum) - with no learner and no reward signal. Agents explore by heavy-tailed (Levy) flights, recruit neighbours to a discovery through a threshold cascade on their proximity network, and disperse as a patch crowds; this cascade cycle balances exploration and exploitation and drives foragers toward the ideal free distribution. The swarm attains the
embodiedmulti-agent - arxiv:2610.25111 · cs.AIExplore or Confirm? LLM Agents Misallocate Measurements Under NoiseAbdul Karim, Matt Keon
As model reasoning gets cheaper, the physical experiment becomes the slow and costly step in discovery. Each experiment returns a noisy reading: a high value can mean a good candidate or a lucky draw. With a fixed budget, an AI agent must keep choosing between exploring a new candidate and confirming a promising one by re-testing it. We measure that choice in Wellocate, a simulated assay campaign with known true values. It is built on a generated landscape and on two real protein datasets, GB1 and PhoQ, scrambled so no model can answer from memory. We test five language models as minimal agent
memoryagentai agentllm agentbenchmark - arxiv:2610.25100 · cs.CLDeep Fusion Memory: A Chunk-Level Parametric Memory for Large Language ModelsRubin Wei, Siyuan Sun, Hao Sun, Yuliang Liu +4
Large language models need effective ways to acquire and use external knowledge beyond what they retain from pretraining. Continued pretraining risks catastrophic forgetting, while retrieval at inference time can add substantial overhead. External parametric memory avoids online retrieval, but constructing its training data requires extensive offline search, and output-level interpolation provides only shallow fusion. We introduce Deep Fusion Memory (DFM), a parametric memory architecture that reduces this construction workload and incorporates learned memory into intermediate hidden states. A
memorymemory architecture - arxiv:2610.25098 · cs.AIPre-Query Retention with Bounded Bit and Word RetrievalPranav Singh, Prashant Singh
An agent can retain information before receiving a future question and later retrieve evidence from an archive. Evaluating this arrangement requires accounting for both retained information and the width of each retrieval response. Decision-tree capacity bounds describe individual sets of distinguishable records, but a memory writer must encode every possible record before the query is known. We study fixed-length retention beside an immutable binary archive, with a worst-case limit on raw bit or word reads. We characterize feasible one-bit-read memory cells and construct encoders that partiti
memoryagent - arxiv:2610.25096 · cs.AIToward Agentic Retrieval-Augmented Generation for Enterprise Retail Workflows over Heterogeneous Data SourcesJeonghwan Choi, Jibin Hwang, Hayoon Park, Hyeonjae Cheon +1
Enterprise retail workflows require agents to chain evidence across heterogeneous sources, from structured product and order records to unstructured inquiries and reviews, yet existing agentic RAG benchmarks rarely capture this challenge. To enable controlled evaluation of such workflows, we introduce RetailBridge, a compositional retail benchmark built from persona-driven simulations. Its questions are constructed by linking entities and intermediate answers across heterogeneous sources, requiring agents to identify the relevant sources, carry evidence between them, and produce a grounded ans
retrieval-augmentedragllm agentagenticbenchmark - arxiv:2610.25092 · cs.LGConformal Prediction-Enabled Sequence-level Policy Optimization for Reasoning Language ModelsHui Yang, Wan Tian, Tao Ren, Zhongyi Li +1
Reinforcement learning improves the reasoning capability of large language models through verifiable rewards for mathematical and coding tasks. In token-level methods such as GRPO, clipping acts on individual token ratios while the reward and advantage apply to a complete response. Recent sequence-level methods align these units but still use fixed clipping bands, whose tail probabilities change as the score distribution evolves. We propose Conformal Prediction-Enabled Sequence-level Policy Optimization (CSPO), which adaptively calibrates sequence-level clipping bounds using recent training st
benchmark - arxiv:2610.25090 · cs.CVThink Visually Across Space and Time: Grounded Reasoning for Video SegmentationJunwei Wen, Deshui Miao, Guangming Lu, Ming-Hsuan Yang +2
Video Reasoning Segmentation (VRS) requires a model to identify an implicitly described target and predict its pixel-level mask trajectory, while resolving the spatial relations and temporal events that determine the answer. Text-only Chain-of-Thought (CoT) can explain a target choice without making the underlying visual references explicit, leaving a \emph{video reference gap} between linguistic inference and the evidence distributed across frames. We introduce \textbf{Temporal Chain of Video Primitives (TCoVP)}, a spatiotemporally grounded form of Video CoT that organizes query-relevant obse
benchmark - arxiv:2610.25089 · cs.AIDAGLoop: Aligning Method Design with Experimental Feedback in Automated AI ResearchDengyuxing Wang, Yonggang Zhang, Qianshu Cai, Huajiang Zheng +3
Automated AI research agents aim to improve machine learning methods by iteratively proposing, implementing, and evaluating candidate methods. Existing work uses the outcome of each evaluation to determine how the next candidate should be revised for performance improvement. However, a candidate often combines multiple design decisions, and its overall outcome alone does not reveal which decisions produced their expected effects, whether their required conditions were satisfied, or which decisions should be retained or revised. Inspired by ablation studies in the machine learning community, we
memoryagent - arxiv:2610.25087 · cs.CLBeyond Factual Accuracy: A Benchmark for Evaluating Perspective Alignment in LLMsSunghoon Kim, Sangyeop Kim
Large Language Models often generate factually correct responses that fail to maintain user-requested perspectives, blending multiple perspectives and forcing users to filter through unwanted information---a perspective control failure that factual accuracy metrics cannot detect. We introduce PerspectiveBench, a benchmark quantifying perspective alignment by constructing scenarios where multiple valid reasoning paths coexist, and propose the Perspective Alignment Score (PAS) to measure adherence to specified perspectives. Experiments reveal that state-of-the-art models successfully include tar
ragbenchmark - arxiv:2610.25084 · cs.CVSufficient Frame Subsets for Video UnderstandingLahiru Wijayasingha
Understanding when during a video a model makes its critical observations is essential for interpreting video-understanding systems. However, existing attribution methods often spread evidence diffusely across near-duplicate frames. We introduce Sufficient Frame Subsets: compact sets of frames in which all other frames are masked, yet each subset independently preserves the model's prediction. We evaluate our method on two complementary benchmarks, UCF101 and Something-Something v2, using both CNN- and transformer-based architectures. Our results show that the proposed algorithm identifies hig
benchmark - arxiv:2610.25079 · physics.opticsDynamic Hotspot Migration and Self-Induced Resonances in Ultrafast Laser-Irradiated Silicon Thin FilmsGeorge D. Tsibidis, Yi Wang, Emmanuel Stratakis
The spatial distribution of energy in silicon thin films under femtosecond laser irradiation is demonstrated to be influenced by the interaction between carrier-induced refractive index modulation and self-induced Fabry-Pérot resonances. We resolve three characteristic excitation regions such as interface-localized, binodal and multi-nodal regimes considering the analysis of silicon-on-insulator (SOI) configurations across thicknesses of 100, 230, and 350 nm. We report a nonlinear hotspot migration effect where the transient reduction of the refractive index stretches the internal standing wav
photonic fabric - arxiv:2610.25078 · cs.LGCoEvo-RSI: When Harnesses Evolve, Which Models Should Follow?Jiaqi Su, Guibin Zhang, Jie-Jing Shao, Cong Pang +4
Model--harness co-evolution has emerged as a prominent paradigm for recursive self-improving (RSI) agents, jointly evolving their execution scaffolds and intrinsic capabilities. We study the persistent yet often overlooked brain--body mismatch problem in this paradigm, where an evolving harness (body) demands behaviors its model (brain) has not learned or restricts capabilities it already possesses. We introduce CoEvo-RSI, a RSI framework that advances model--harness co-evolution through agent-driven, bidirectional historical inheritance. Drawing on a shared evolutionary archive and execution
agentmulti-agentself-improving - arxiv:2610.25076 · cs.CLP2N: Reusing Deep Representations for Greater Effective DepthYiqin Hao, Boyi Zeng, Zitong Wang, He Li +4
Mathematical problem solving and autonomous agents are increasingly important applications of language models, demanding stronger reasoning capabilities. Computational depth is a key resource for such reasoning, but adding model layers increases both training and inference costs, while test-time scaling incurs substantial additional inference computation. We introduce P2N (Previous to Next), which reuses the previous token's deep representations, already available during generation, to increase effective computational depth with negligible decoding overhead. Specifically, P2N adds the previous
autonomous agentbenchmark - arxiv:2610.25074 · cs.CVMoWE: Spatially Adaptive Mixture-of-Wavelet-Experts Pooling for Ultrasound Lesion SegmentationReaz Hassan Joarder, Sidratul Tanzila Tasmi, Alif Arshad Bakshi, Md. Bakhtiar Hasan +1
We present Mixture-of-Wavelet-Experts (MoWE), a spatially adaptive routing mechanism that replaces max-pooling and strided convolutions in U-Net architectures for ultrasound lesion segmentation. Standard pooling operations suppress high-frequency spatial details, including potential boundary cues in speckled regions. Recent frequency-domain methods preserve some of these details, but a single wavelet family applies the same spectral bias across the image. MoWE instead uses a context-aware router to assign local regions to fixed wavelet families according to local texture and global feature con
benchmark - arxiv:2610.25069 · cs.LGAuto-MD-bench: Evaluating Autoresearch Agents for Real-World Medical TasksShenqi Zhou, Qiran Zou, Wenhao Zhao, Zhenyu Liao +1
Autonomous agents are increasingly capable of developing machine learning (ML) models end to end. In medical practice, however, strong performance on the target metric does not fully characterize how good the models are in real-world use cases. For example, medical ML models should also provide well-calibrated risk estimates and remain robust across patient populations. Yet autonomous model development is typically guided by task objectives that emphasize simple predictive discrimination (e.g. AUROC). This raises a key question: are these objectives sufficient to guide agents toward models tha
autonomous agentbenchmark - arxiv:2610.25068 · cs.ROLocalization and Neural Splatting on the Edge: An FPGA Scene Memory for RobotsQue Haohua, Mingkai Liu, Handong Yao
Resource-constrained robots and autonomous vehicles need scene memories that localize a camera and render nearby views without a desktop GPU. Scene-coordinate regression (SCR) and 3D Gaussian splatting (3DGS) fit because they avoid an SfM stage, but ACEsplat-style pipelines are GPU-centered, and moving them to an edge FPGA is hard because SCR is pose-sensitive and 3DGS rendering is irregular and capacity-sensitive. We present LENS, an edge scene-memory acceleration system on the AMD KR260 that brings an ACEsplat-style workload to a low-power FPGA/ARM platform. Rather than proposing a new scene
memory - arxiv:2610.25064 · cs.AIOne Skill Takes Over: Clashing Agent Skills and a One-File FixParsa Razmara, Seyedarmin Azizi, Ardalan Aryashad
Agent Skills are instruction files that tell a coding agent how to do one job. Users install skills from many authors, so two skills for the same job can be active at the same time. We ask what a model does when both skill files are in its context: does it combine their compatible rules, or follow one skill and drop the other? We build SkillClash, 16 pairs of same-job skills in which code checks every rule and a reference output shows the best result that can be reached. On seven models from five families, from 4B open models to Qwen3-235B and GPT-4.1, a skill's rules are followed 93 to 99% of
agent - arxiv:2610.25060 · cs.LGPolicy Is Not Capability: Some Safety Benchmarks Measure Which Definition of "Unsafe" You Trained OnVahid Jebraeeli, Jonathon Romero, Yijin Wang, Adrian Chan +1
A guardrail classifier is reported as if unsafe were a property of text. It is not: it is a property of a policy, and the policy enters the model through its labels. We exploit an accident of an industrial deployment in which the same ~53k prompts were labelled twice, under two operational definitions ("is this a request for harmful content?" and "would a moderator block this?"), and used to train ten guardrail arms sharing architecture, codebase and optimiser family, at two model scales. First, the guardrail's own training corpus already contained the benchmarks: JailbreakBench-benign is 88%
benchmark - arxiv:2610.25059 · cs.LGHCFM: Helmholtz Component Flow Matching for Time Series Anomaly DetectionShyam Sundar Saravanan, Tie Luo, Sanjay Madria
Flow Matching (FM) is an efficient simulation-free approach for training flow-based generative models, making it a promising foundation for scalable time-series anomaly detection (TSAD). However, standard FM learns a monolithic vector field in which nominal transport and local density-changing behavior are entangled. This creates a central challenge for TSAD; local expansion and contraction carry crucial anomaly-sensitive information, yet isolating this signal requires expensive Jacobian-trace estimation over the entire field. We propose \emph{Helmholtz Component Flow Matching} (HCFM), an unsu
benchmark - arxiv:2610.25054 · cs.CVPRET is a few-shot system for pan-cancer recognition without example trainingYi Li, Ziyu Ning, Tianqi Xiang, Qixiang Zhang +16
Pathological examination stands as the cornerstone in cancer diagnosis, impacting millions worldwide annually. With the shortage of pathologists globally, artificial intelligence (AI) has emerged rapidly to automate the diagnostics process. However, conventional AI models require substantial labeled data for each disease, posing huge challenges in scalability and practicality. Therefore, we introduce PRET (pan-cancer recognition without examples training), a few-shot system to achieve flexible, scalable, and effective cancer recognition across diverse organs, hospitals and tasks without traini
benchmark - arxiv:2610.25050 · eess.SYDynamic Skill-Based Routing for Call Centers with Transfers: A Value-Based FrameworkBaris Ata, Ebru Kasikaralar
We consider a multi-class, multi-pool queueing model of a telephone call center in which a service completion does not always end the call. The caller may be transferred to another agent for a second service, so the assignment of an agent to a call affects not only the caller's waiting time but also whether the call is resolved and how much further work it creates. We model these transfers as Markovian routing from original to rework classes, and the system manager dynamically assigns agent pools to classes so as to maximize the value generated by service, net of transfer penalties and waiting
agentbenchmark - arxiv:2610.25048 · cs.LGPACT: NVFP4 Data-Flow LoRA Fine-Tuning on Consumer GPUsLei Gao, Hossein Entezari Zarch, Chaoyi Jiang, Murali Annavaram
Fine-tuning language models on consumer GPUs is constrained by computation and memory. Although LoRA reduces trainable parameters and optimizer state, frozen projections remain expensive, and activations retained for backward can dominate memory usage. We present PACT, a LoRA fine-tuning system that accelerates frozen projections with native NVFP4 GEMMs while reducing quantization and dequantization overhead and limiting quantization error. To reduce activation memory, PACT reuses quantized RMSNorm outputs and saves gate/up activations in NVFP4 for backward. To limit numerical error, it preser
memory - arxiv:2610.25043 · cs.ROVLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action ReuseSicheng He, Yi Pan, Qiuting Yu, Jiaming Huang +9
Flow-matching vision-language-action (VLA) models generate an action chunk by iteratively denoising Gaussian noise, but typically execute only a short prefix before replanning with updated observations. Under this receding-horizon paradigm, the unexecuted remainder---the action tail---is discarded, and the next chunk is regenerated from fresh noise. We show that this discarded tail retains transferable denoising progress: action-tail initialization can match or outperform a native 10-step Gaussian baseline using fewer denoising steps. However, the optimal denoising budget varies substantially
vision-language-actionvlaliberorobotwinbenchmark - arxiv:2610.25034 · cs.LGBeyond the Teacher: Understanding and Preserving Distillation Overshoot in Agentic On-Policy DistillationJike Zhong, Ming Li, Yuxiang Lai, Xi Xiao +8
On-policy distillation (OPD) trains agents on their own trajectories using supervision from a stronger teacher. We uncover a counterintuitive phenomenon in multi-turn agentic tasks: a smaller, initially weaker student can substantially outperform its teacher during training, only to regress as distillation continues. We term this phenomenon distillation overshoot. Through decision- and task-level analyses, we trace overshoot to two competing effects of OPD: repairing weak student decisions and eroding successful student choices that the teacher disfavors. Early in training, repair outpaces ero
agenticbenchmark - arxiv:2610.25026 · cs.AIOne Head Away from Language Confusion: Single Attention Heads Across Model FamiliesJiwon Jeong, Anjelo Jann Laroza, Chaewoon Yoon, Arjun Pillai +3
Large language models sometimes reply in an unexpected language, a failure known as language confusion. We ablate individual attention heads in 12 model checkpoints spanning seven model series, using FLORES-200 continuations in five languages and Language Confusion Benchmark (LCB) chat replies. In ten models, zeroing one head causes substantially more departures from the expected language than zeroing any other head in its layer; its LCB pass-rate drop persists, though reduced, under mean ablation. Seven heads move 8.0 to 51.5 percent of all FLORES continuations out of the prompt's language, v
benchmark - arxiv:2610.25021 · cs.LGSynergy-Aware Residual Flow Matching for Single-Cell Perturbation PredictionJingrui Wu, Wangyun Dan
Predicting single-cell responses to genetic and chemical perturbations is important for understanding regulatory mechanisms and prioritizing therapeutic interventions. Existing methods have made progress by modeling perturbation-conditioned expression shifts with neural networks, graph priors, foundation-model representations, or distributional generative models. However, combinatorial perturbation prediction remains challenging because much of the response can be explained by additive single-perturbation effects, while the biologically informative signal often lies in a smaller non-additive r
benchmark - arxiv:2610.25019 · cs.LGLLM Agents as Suppliers of Physical System Models: Calibration and Acceptance in ModelicaJörg Frochte, Herbert Schmidt
To procure a model of a device, a client hands a supplier a description, sometimes with recorded data or a target response, and accepts the model against this handover. We let LLM coding agents act as the supplier on handovers with data, building Modelica models from library components for four simulated systems and a real switching magnet, and ask what acceptance certifies. Several agents reach the noise floor, but a handover admits a set of models: the description leaves parts of the structure open, and the data narrow it only to the agreed accuracy. We therefore accept by membership in this
llm agent - arxiv:2610.25015 · cs.CVIECLA: Interpretable Emotion Contribution Learning for Affective Image EditingYijie Guo, Dexiang Hong, Weidong Chen, Xinyan Liu +1
Affective image editing (AIE) aims to evoke a target emotion while preserving the content and structure of a source image. Existing methods improve emotional expression or visual fidelity, but the target emotion alone does not reveal which source regions should be changed and which should remain intact, especially in complex scenes. We introduce Interpretable Emotion Contribution Learning for Affective Image Editing (IECLA), a framework that separates source-emotion attribution from target-aware editing. IECLA constructs signed object-level supervision through counterfactual interventions and
agentevaluator - arxiv:2610.25013 · cs.ROLatent World Models Should Plan with Distances Along the Latent ManifoldZunzhi You, Linwei Tao, Siyu Xu, Yunke Wang +1
Latent world models such as LeWM plan by searching for an action sequence whose predicted terminal latent is closest to the goal latent under an ambient norm, such as the Euclidean distance. Planning with such a cost assumes that proximity in the ambient latent space reflects proximity to the goal in the environment. Studying LeWM, we find that this assumption holds only in a small neighborhood around the goal, whereas farther away the L2 cost forms spurious low-cost basins that trap the planner. Yet neighborhood-based embeddings show that the latents retain the underlying physical state. We t
world modelbenchmark - arxiv:2610.25012 · cs.LGConjugate Compilation: Self-Improving Agentic Co-Compilation for Speculative DecodingAnnabelle Sujun Tang, Rohan Mahapatra, Christopher Priebe, Hanyang Xu +3
As LLM adoption surges, scaling model serving requires unlocking stranded GPU capacity alongside deploying more GPUs. Speculative decoding brings two models into the serving process: a small draft model proposes tokens that a larger target model verifies. Yet compiling these models individually leaves their shared execution outside the compiler's optimization boundary. We introduce CONJUGATE COMPILATION, which reframes speculative decoding as a cross-model compilation problem and jointly optimizes the draft and target models within their shared execution rather than treating each in isolation.
agenticself-improving - arxiv:2610.25006 · cs.ROScalable, Intuitive, Mixed Reality-Based Semi-Autonomous Robotic Telemanipulation for the Execution of a Fine-Grained Bimanual TaskShifei Duan, Zhe Wang, Francesco De Pace, Minas Liarokapis
Nowadays, robotic telemanipulation enables humans to control robots so as to execute tasks in a manner similar to how they perform them, and serves as an essential approach for collecting on-robot data for robot learning by human demonstrations. The quality of collected demonstrations directly impacts the performance of the learned policy. However, existing telemanipulation systems rely heavily on human operation, particularly for large-scale data collection. Collecting dozens or hundreds of demonstrations is a demanding process, as operators tend to fatigue quickly, especially during the exec
manipulationdexteroushuman-in-the-loop - arxiv:2610.25005 · cs.CVReTrace: Visual Backtracking for Efficient Multimodal Large Language ModelsGuangzong Si, Dong Wang, Yifan Yu, Miao Liu +2
Multimodal Large Language Models (MLLMs) encode images into dense visual token sequences, creating substantial attention and KV-cache overhead. Existing visual token compression methods reduce this cost by pruning low-saliency tokens in shallow layers, but this practice couples temporary computational deactivation with irreversible information deletion. We show that this coupling breaks an intrinsic inference pattern in MLLMs. Through layer-wise attention analysis, we identify \textbf{Visual Backtracking}: visual pivots remain spatially stable, yet in middle layers they actively retrieve conte
benchmark - arxiv:2610.25004 · cs.LGUpLoop: Upcycling Large Language Models with Gated LoopTieyuan Chen, Huabin Liu, Yingtao Zhang, Shijie Li +3
Looped large language models trade parameter storage for recurrent computation, but most are trained from scratch. In practice, pretraining with looped depth is expensive, making post-training upcycling of a pretrained non-looped LLM an attractive alternative. We introduce \textbf{UpLoop}, a post-training recipe based on \emph{layer-wise} recurrence. Instead of looping a whole model or a middle block, UpLoop loops each selected layer locally, preserving the pretrained depth hierarchy and avoiding feature-space mismatch. To make loop execution useful, we add hidden-wise gates on the attention a
post-trainingbenchmark - arxiv:2610.24999 · cs.LGSoK: PIAtlas: Decomposing the Unbounded Prompt Injection Attack SpaceStavros Damianakis, Konstantinos Psounis
Prompt injection is the dominant security risk for large-language-model (LLM) applications and agents, yet the field still lacks a coherent map of how these attacks are constructed. More than twenty public datasets describe prompt injection using incompatible vocabularies, often organized around delivery vector, attacker goal, or generation method. In this Systematization of Knowledge, we present PIAtlas, a taxonomy organized instead around how the injected instruction is constructed to subvert the underlying model. Its six classes, one per subversion mechanism, form the roots of a hierarchy o
agentagent benchmarkbenchmark - arxiv:2610.24993 · cs.CLWhen Does Better Acceptance Pay? A Controlled Study of Speculative DecodingYongchao Long, Chunrui Zheng, Yuxuan Zhou, Guangkun Nie +9
The throughput gains of speculative decoding depend on the trade-off between acceptance length and execution latency. We ask when additional acceptance is worth the cost of obtaining it. We establish a comparison protocol spanning training, inference, and evaluation, remeasuring released checkpoints and retraining six drafter recipes on shared target-generated data across ten benchmarks and two target models. Alignment changes the picture: gaps that appear large in published comparisons shrink substantially under shared training and inference conditions. This narrowing motivates two deeper que
benchmark - arxiv:2610.24991 · cs.LGCertified Sparse Attention via Perturbed RetrievalRunguo Li
Sparse attention speeds up long-context decoding by reading few keys of the KV cache, but existing methods do not control the error of a query, and on real KV caches their errors have heavy tails. We show that perturbing the score of every key with an independent Exp(1) variable turns threshold retrieval into sampling keys with probability proportional to their attention weights. Lowering the threshold along a geometric grid gives nested samples, a Horvitz--Thompson estimate of the output at every level, and confidence bounds that hold simultaneously over all levels, so each head stops at the
long-context - arxiv:2610.24990 · cs.CLThe World as Viewed by 136 Countries: An Open, Curated, Annotated, and Multilingual Dataset of 118 Million State Media ArticlesBryan Li, Samar Haider, Grace Kim, Fiona Luo +5
State media outlets often project their governments' narratives to target both domestic and foreign audiences. Because this content is plentiful and free to access, it appears widely in the web corpora that large language models (LLMs) are trained on and that retrieval systems draw from. However, the prevalence of state media in these corpora and its potential influence on LLM outputs remain largely unexplored. In this paper, we introduce WorldStateMedia, an open dataset of 118 million state media articles from 136 countries in 114 languages. We extract these articles from existing pretraining
retrieval-augmented - arxiv:2610.24981 · cs.RODream What I Can: Sparse Visual Planning for Long-Horizon Robot ManipulationChongkai Gao, Wenlong Huang, Haonan Chen, Lin Shao +4
Video models provide rich priors for generalizable robotic manipulation, but plausible visual futures are not always executable. We introduce DreamCan, a hierarchical policy that grounds video models through sparse visual planning: a high-level planner adapted from a pre-trained video model and a low-level visuomotor policy. Rather than predicting dense trajectories, the high-level planner generates sparse visual subgoals that balance reachability and task utility. From offline, action-free robot videos, we estimate these quantities and use dynamic programming to select subgoal sequences witho
manipulationlibero - arxiv:2610.24977 · cs.AICorrectRTL: Boost RTL Generation Quality Through Automated Self-Correction with Instruction DecompositionJing Wang, Shang Liu, Hangan Zhou, Wenji Fang +3
Large language models (LLMs) have shown strong promise in specification-to-RTL generation. Among the explorations, the self-correction paradigm has significantly advanced LLM-based RTL generation. However, two challenges limit further progress. Existing benchmarks are saturated by advanced LLMs, leaving little room to evaluate the agentic self-correction framework. The monolithic self-correction paradigm rewrites the entire design at once, where LLMs are susceptible to error confusion and corruption of originally correct code. To address the benchmark challenge, we contribute ScaleBench, a mor
agentllm agentagenticself-correctionbenchmark - arxiv:2610.24972 · cs.CLA Measurement Perspective on Benchmarking LLM-Generated Text Detectors using Item Response TheoryYanbin Zhao, Zhipeng Chen, Guangsheng Bao, Guandong Zhang +1
Benchmarking large language model (LLM)-generated text detectors commonly relies on aggregate detector-level metrics such as accuracy, F1 score, and AUC. Although useful for summarizing overall performance, these metrics provide limited insight into how benchmark composition shapes detector comparisons or which texts distinguish detectors with different response profiles. We propose a measurement-based benchmarking framework for jointly characterizing benchmark heterogeneity and detector behavior using multidimensional item response theory (IRT). By treating benchmark texts as measurement item
benchmark - arxiv:2610.24971 · cs.CLRubrics Carry the Standard: Shapley Attribution for Checklist-Based LLM JudgesSaeid Asgari, Rui Ying, Sara Malvar, Rakshanda Agarwal +4
Checklist rubrics applied by large language model (LLM) judges have become a standard instrument for evaluating open-ended outputs and for rewarding them during training. Two assumptions surround this practice: that leniency or strictness is a property of the judge model, and that a rubric can be improved by asking a model to review it. We test both in a controlled grading task with planted flaws and on two human-labeled datasets, SummEdits and RAGTruth. Holding the judge and the outputs fixed, rubric content moves the pass rate of correct outputs from 2% to 100%, whereas changing the judge mo
judge model - arxiv:2610.24963 · cs.CLDeGMatch: Target-Aware Reasoning Data Selection via Deconfounded Gradient MatchingChangho Lee, Janghoon Han, Rajdeep Mukherjee, Yeda Song +3
Reasoning post-training on long trajectories has substantially improved the reasoning capabilities of large language models, but scaling such training to large datasets is increasingly expensive. We study target-aware reasoning data selection, where a small reference set specifies a target capability and guides the selection of training trajectories for specialization. We find that existing target-aware methods can exhibit limited target discrimination on long reasoning trajectories, while dominant population-level structure in candidate gradients can obscure finer target-dependent variation d
post-training - arxiv:2610.24960 · cs.AICoworkEnv: Synthesizing Training Environments for Long-Horizon Cowork AgentsZhongwei Xie, Yiming Du, Jiaxin Bai, Baixuan Xu +9
AI coworkers must coordinate tools, communicate across turns, and act on changing workspace state while respecting workflow constraints. Constructing training experience for these capabilities requires coherent requests, service records, tool interfaces, and grading checks. We present CoworkEnv, a capability-guided framework for synthesizing executable training environments for AI coworkers. Workflow requirements guide task construction, and programmatic validation with feedback repair checks generated specifications and grading behavior. Teacher executions yield curated demonstrations for sup
benchmark - arxiv:2610.24957 · cs.LGRevealing what benchmark scores hide: Jointly measuring the breadth and reliability of LLM capabilities using accessibility profilesKazem Faghih, Yize Cheng, Shoumik Saha, Mobina Pournemat +2
Language model benchmarks assign each question a single outcome from one fixed wording, even though the same underlying request can be expressed in many equivalent ways. We argue that the quantity of interest should therefore not be correctness on one wording, but the accessibility of an inquiry, which we define as the probability that a model succeeds across meaning-preserving formulations. The distribution of accessibility across inquiries reveals a structure that accuracy and existing best-, worst-, or average-case prompt scores collapse: how broadly a model's capabilities extend across inq
benchmark - arxiv:2610.24955 · cs.AIMemPrice: Online Memory Allocation via Adaptive Pricing for Long-Horizon AgentsYiming Zong, Shiyuan Piao, Jiashuo Jiang, Fugee Tsung
Long-horizon LLM agents face a fundamental allocation problem as interaction histories grow: which information to retain, how much context to devote to it, and at what granularity. Existing memory mechanisms largely focus on relevance, retrieval, or compression, leaving this allocation trade-off implicit. We introduce MemPrice, which treats active context as a shared resource and dynamically allocates it across memories and representation levels. MemPrice turns accumulated memory pressure into an adaptive price on context, encouraging detail only when its estimated retention value justifies th
memoryagentllm agentbenchmark - arxiv:2610.24952 · cs.LGCasePlay: Self-Play Reinforcement Learning from Case Reports for Medical ReasoningHao Wang, Zihan Wang, Kai Wu, Jian Yao +5
Reinforcement learning (RL) is a promising way to improve the clinical reasoning abilities of large language models (LLMs), but current pipelines often rely on expert-labeled questions or costly synthetic-data generation. Meanwhile, the medical literature offers a wealth of high-quality unlabeled documents, including clinical case reports that trace the full clinical course from initial presentation to final treatment decisions. We study whether document-grounded self-play can convert these reports into effective RL training signals. We identify two challenges that limit existing self-play met
self-playbenchmark - arxiv:2610.24951 · cs.LGControlJEPA: Principled Trajectory Regularization via Lyapunov Tube LossThai-Khanh Nguyen, Nhi Y. Pham, Uyen N. B. Vo, Thieu N. Vo +2
Next-token prediction leaves hidden-state trajectories largely unconstrained, while existing geometric regularizers primarily control the direction rather than the magnitude of trajectory deviations. We introduce \emph{ControlJEPA}, a trajectory regularization framework based on a discrete-time Lyapunov contraction condition from practical Input-to-State Stability theory. Its Lyapunov Tube Loss penalizes violations of this condition on a normalized, scale-invariant measure of transversal deviation. When the contraction condition holds at every step, the hidden-state trajectory remains within a
benchmark - arxiv:2610.24944 · cs.AIEpi: Towards a Persistent, Self-Improving Agent for Autonomous ResearchYue Huang, Zhangchen Xu, Xiaonan Luo, Zheyuan Liu +15
We present Épi, an autonomous research agent that sustains investigations across model sessions and evolves its methods through accumulated experience. Épi combines persistent research history, hierarchical memory, branching and independent review to maintain research continuity, learn from experiments, and explore and revisit alternative approaches. Its experiments run across session boundaries, and a fixed verifier evaluates their outputs. Épi produced 120 record-setting results, comprising 20 numerical bound records across 12 families and 100 construction records across eight families. Thes
agentself-improvingbenchmark - arxiv:2610.24937 · cs.LGMLFont: Machine Learning-Driven Font Subsetting and Range-Compact Unicode-Range Slicing for Low-Latency Web TypographyQuan Wang, Yiran Mao
CSS3 @font-face rules with unicode-range descriptors partition web fonts into downloadable slices. While Latin fonts need ~20 KB in WOFF2, Pan-European (LGC) and East Asian (CJK) fonts contain 2,297-30,759 codepoints (158 KB-5.9 MB in WOFF2), where heuristic partitioning causes unused-glyph downloads, high HTTP RTT, CSS bloat, or missing-glyph fallbacks. We present MLFont, a machine learning and optimization framework that learns font slicing strategies from corpora via three stages: (1) a Zipfian High-Frequency Tier (most) isolating top-frequency codepoints; (2) a Latent Co-Occurrence Topic T
benchmark - arxiv:2610.24930 · cs.LGTrustAgent-RFF: Agentic Diagnostic Planning for Reliable RF Fingerprinting under Domain ShiftChuan Liu, Tianya Zhao, Xuyu Wang
Radio frequency fingerprinting (RFF) models can achieve accuracy under source domain conditions but become unreliable when the receiver, distance, collection day, or environment changes. Incorrect predictions in unseen domains may receive high softmax confidence, making confidence thresholding insufficient for trustworthy deployment. We propose TrustAgent-RFF, a prediction level reliability framework for RFF that operates with a frozen in-phase and quadrature (I/Q) classifier without target domain labels. For each device prediction, the framework acquires radio frequency evidence: counterfactu
agentic - arxiv:2610.24929 · cs.AIWhat Can't Jev Decide? Probing the Limits of One-Pass Decision ModelsWang Yang, Xinpeng Li, Yanyan Zhang, Chuang Ma +7
Jev and its open replicas choose among the options a caller supplies with a single forward pass, fast enough to decide inside an agent loop. We ask what such a model cannot decide, and find a simple limit: Jev cannot make a decision that an LLM makes only by writing out intermediate steps. On a benchmark that varies one factor at a time, Jev finds and selects records without error up to 20k tokens and 64 options, but once the answer depends on two or more updates of one user, accuracy falls, to 53-80% at sixteen updates, because similar records interfere. Given a stale summary of the state, Je
agentbenchmark - arxiv:2610.24921 · cs.LGDF-PALM: Dynamic Fusion Prompt Learning for Audio--Language ModelsYulin Sun, Shenghua Li, Yong Dou, Xinwang Liu +2
Few-shot audio classification requires audio--language models to distinguish variable acoustic events from limited labeled data. PALM's main formulation adapts encoded class representations without changing pretrained encoders, but its class representations remain input-independent and classification uses audio features without class-conditioned refinement. We introduce DF-PALM, a dynamic feature-space prompting framework that combines audio-conditioned class representations with asymmetric cross-attention for audio refinement. Class representations are conditioned after text encoding, held fi
benchmark - arxiv:2610.24916 · cs.AISame Data, Different Bytes: Serialization-Invariant Integrity Checks for Agent-Published DatasetsJin Gao, Maura Clark
Agent networks require reliable ways for requesting agents to verify where data is located, how it can be accessed, and whether the data received matches what the data owner made available. Checksum mismatches can reflect either altered content or serialization differences that preserve data equivalence under specified rules. Data identification mechanisms for agents, such as DataFacts, describe where data can be accessed and provide checksums for checking received data against published references. We propose a serialization-invariant identification method that assigns the same identifier to
agent - arxiv:2610.24914 · cs.LGArctic-SWE: Revisiting the Long Horizon Agent Training by Modeling-System Co-DesignBoyi Liu, Yite Wang, Xinyu Lian, Fan Shu +5
Long-horizon agent training, including software engineering, is advancing rapidly, but the recipe and infrastructure needed for a strong reinforcement-learning baseline remain difficult to reproduce. This makes gains from new algorithms or data hard to distinguish from improvements to an under-tuned baseline. We present Arctic-SWE, a fully asynchronous RL framework that establishes such a baseline through model-system co-design, using overfitting on a small task pool as the testbed. This testbed exposes failures that early reward gains conceal and guides the integration of established componen
agent - arxiv:2610.24913 · cs.AIA Cross-Domain Evaluation of Prompting Techniques for Large Language ModelsKevin Krazius, Tim Schlippe, Paul Libbrecht
Prompt engineering has become a central mechanism for guiding the behavior of Large Language Models (LLMs) across a wide range of reasoning and language understanding tasks [1], [2]. However, existing studies frequently evaluate prompting techniques using different LLMs, datasets, or task settings, making direct comparisons across methods and domains difficult [3]. Consequently, in this paper we present a controlled cross-domain evaluation of Zero-Shot prompting, Few-Shot prompting, Retrieval-Augmented Generation (RAG), and automatic prompt optimization based on a PE2-inspired optimization pip
retrieval-augmentedragbenchmark - arxiv:2610.24899 · cs.AIHybridResearcher: Bridging the Structured-Unstructured Boundary for Deep Research AgentsSoyoung Yoon, Peiran Xu, Han Wang, Ruofan Wu +7
While LLM agents are increasingly capable of gathering and synthesizing evidence from a single homogeneous source, such as structured tables or unstructured data, real-world analytical questions frequently span heterogeneous sources. For example, answering one enterprise question may require both a result computed over an internal database and evidence found on the web. Recent evaluations on benchmarks like HybridDeepResearch show that even frontier agents struggle when evidence must cross this boundary. While these benchmarks have exposed the limitation, methodologies to explicitly teach cros
agentllm agentbenchmark - arxiv:2610.24896 · cs.CLPeer Pressure: Execution-Verified Co-Evolution of Attacker and Defender Populations for Code SecurityKhushbu Pahwa, Yubin Ge, Daniele Bonadiman, Xiaofei Ma +1
LLM code agents increasingly ship code with little oversight, yet defenders trained on static vulnerability datasets fall behind adapting adversaries. We co-train attacker and defender instead, and let execution decide every reward: an attacker (Red) edits a real repository and earns nothing unless its patch applies, builds, passes the tests, completes its development task and reproduces a real exploit in an isolated container, whose path also labels the entry point, the critical operation and the trace between them. Peer-Relative Policy Optimization (PRPO) makes both sides populations of LoRA
agent - arxiv:2610.24895 · cs.ROAnticipatory Gait Switching for Quadrupeds via Look-Ahead MDP PlanningKelvin Aladum, Justin Peel, Nisar Ahmed
Quadruped robots often switch gaits to traverse mixed terrain safely. Many practical supervisors enforce hardware constraints (e.g., dwell or guard intervals) using a transition handler. However, the switch decision itself is often driven by near-field thresholds, which can delay commitment when rough terrain is only detected shortly before contact in the local height map. We propose a lightweight height-map-only gait supervisor that unifies look-ahead and locomotion constraints in a single decision policy. A robot-centered height map is partitioned into near-, mid-, and far-field bands, each
quadruped - arxiv:2610.24892 · cs.LGScalable and Adaptive Distribution Regression via Random FeaturesYinsong Wang, Xiao Liu, Yu Ding
Distribution regression seeks to learn mappings from probability distributions to response variables and has emerged as a unifying framework for kernel methods, multi-instance learning, and set-valued learning. Despite its success, kernel-based distribution regression remains computationally expensive and sensitive to kernel choice, motivating scalable and adaptive representations. Building on random feature approximations, we develop a finite-sample analysis of Random Feature Distribution Regression (RaDR) under two-stage sampling. Our perturbation analysis separates the effects of distributi
benchmark - arxiv:2610.24888 · cs.LGMeasuring Data Diversity with Embeddings: A Taxonomy and Benchmark for Diversity MeasuresCantao Su, Anna Wegmann, Esther Ploeger, Dong Nguyen
Diversity is a critical property of text corpora, with direct implications for training data curation, test set design, and model generalization. However, its measurement remains fragmented and poorly understood. Existing embedding-based diversity measures vary widely in how they are computed and can yield substantially different diversity scores on the same data, yet no systematic comparison of their empirical behaviour exists. We make two contributions toward addressing this gap. First, we present a taxonomy of 21 embedding-based diversity measures, organizing them into four families based o
benchmark - arxiv:2610.24885 · cs.CVGoal Images Are Control Targets: Spatially-Grounded Goal-Image Synthesis for Robotic ManipulationRiddhi Chatterjee, Ritish Shrirao, Kushagra Mehrotra, Viswanath Gopalakrishnan
Goal-conditioned manipulation policies are steered by a goal image, a picture of the desired post-manipulation scene synthesized from the current observation and a language instruction. The prevailing assumption is that better image-editing quality yields better goals, so goal generators are built and compared using general image-editing metrics. We show this assumption is false. Across two independent goal-conditioned policies on CALVIN, the image editors that top aggregate editing-quality leaderboards (Gemini 3 Pro Image and GPT Image 1.5) produce worse control targets than our proposed far
manipulationbenchmarkevaluatorleaderboard - arxiv:2610.24875 · cs.LGRAFT: Redundancy-Aware Input Filtering for Temporal Graph Neural Network TrainingYue Dai, Sarah Zhou, Youtao Zhang
Temporal graph neural networks (TGNNs) have shown strong performance on dynamic graph learning tasks by continuously updating node memories from evolving interactions (i.e., events) and leveraging these memories for downstream predictions. As graphs grow rapidly in modern applications, training requires processing an increasing number of events; meanwhile, each event can update node memories used by later events, creating dependencies that constrain parallel execution and limit the benefits of additional computing resources. These costs motivate an upstream question: Are all events, and the de
memory - arxiv:2610.24873 · cs.AIThe One-Pixel of Reasoning: One Truthful Relation Is Enough to Mislead LLM AgentsJinwei Hu, Youcheng Sun, Yi Dong, Xiaowei Huang
Can one sentence manipulate an agent's belief by targeting a single implicit warrant linking evidence to a conclusion? In this paper, we introduce the one-relation attack (ORA), a novel cognitive attack on this inferential relation. ORA localizes a critical implicit warrant and performs a one-relation intervention using a single source-grounded sentence that qualifies this warrant and steers the agent toward an incorrect belief. Across eighteen models on biomedical inference and vulnerability detection, best-of-$10 attacks overturn initially correct answers at mean rates of 54.9% for proprieta
agentllm agentmulti-agent - arxiv:2610.24871 · cs.LGBellman Language Models: Solving the Recurring Decisions of LLM AgentsAdarsh Vatsa, William Eiers
A general observation in agentic workflows, especially bounded ones, is that the decision space of the LLM, when we separate a decision from generation, starts approaching saturation: the agent meets the same states again and again, and after enough conversations it stops meeting new ones. LLMs generally don't treat these decisions as an optimization problem and instead re-derive them from their context/prompt on every turn. We show that treating decision making within reasonable bounds as an optimization problem, solving it once, and reusing the solved decisions can often make an agent faster
agentllm agentagentic - arxiv:2610.24854 · cs.CLLearning to Program Retrievals: Composing Search Primitives via Reinforcement LearningKyuyoung Kim, Zilin Xiao, Zhiyuan Wang, Avinash Atreya +3
Retrieval-augmented language models increasingly use multi-step search to solve knowledge-intensive tasks, yet existing systems typically expose retrieval through fixed pipelines or collections of atomic tools. We introduce Retrieval Program Learning (RPL), a programmable search framework that allows a policy to compose search primitives into executable programs with local state and control flow. This enables adaptive search that can retrieve broadly, process intermediate results, and selectively construct model context without tying retrieval breadth to context consumption. However, the resul
retrieval-augmentedbenchmark - arxiv:2610.24847 · cs.RODiPLO: Self-distilled Diffusion Planning for Loco-manipulationShafeef Omar, Sebastian Sanokowski, Elham Daneshmand, Dian Yu +3
Humanoid loco-manipulation is particularly challenging because it requires coordinating locomotion, balance, and object interaction, making it difficult to learn from limited demonstrations or from scratch through Reinforcement Learning (RL). We introduce DiPLO (self-distilled Diffusion Planning for Loco-manipulation), a framework that combines a self-distilled diffusion planner with a low-level whole-body tracking policy. Our planner is pretrained via supervised learning on motions derived from human-to-humanoid retargeting, while the tracker is trained via RL to execute them. DiPLO centers a
manipulationhumanoiddiffusion policypost-training - arxiv:2610.24845 · cs.MAAgentSoV: Virtual Sensing for LLM Multi-Agent Systems via Stream of VariationCongjing Zhang, Siyuan Xu, Yunfei Bai, Ke Ren +2
Large language model (LLM) multi-agent systems (MASs) solve complex tasks through interacting agents, yet their reliability is difficult to analyze because failures depend on both individual agents and how information and errors propagate across the system. Existing approaches largely treat this as post-hoc failure attribution, but do not provide a common model across heterogeneous MAS architectures or support broader analysis of information influence and observability. We introduce AgentSoV, a virtual-sensing framework that models MAS execution as a stochastic dynamic system through Stream of
multi-agentagenticagent system - arxiv:2610.24841 · cs.CVPCHMR: Empowering and Benchmarking Human Mesh Recovery in Privacy-Constrained Real-World SettingsZeyu Cao, Qingxuan Wu, Zhiyang Dou, Rui Xu +5
Fine-tuning directly on user-side data is an effective way to improve the performance of widely adopted data-driven human mesh recovery (HMR) models. However, doing so for existing HMR methods often assumes aggregating user-side data from real-world HMR-deployed devices to a central training server, posing a significant privacy risk as sensitive human images are transmitted. How can human mesh recovery be evaluated and improved in privacy-constrained real-world settings? We study this question through a benchmark in which state-of-the-art HMR models are trained with federated learning and a se
benchmark - arxiv:2610.24835 · cs.AIPattern-Driven Evaluation of LLM Code Optimization: Isolating Compiler-Transcendent Semantic ReasoningWesley Lu, Runzhou Liu, Hongru Du, Xinyao Yi
Existing evaluations of LLM-based code optimization primarily measure end-to-end speedup, obscuring whether a model actually performs the intended optimization. For compiled languages, this is especially problematic: observed gains may reflect transformations already recoverable by modern optimizing compilers rather than optimization behavior attributable to the model itself. We present PatternBench, a benchmark for studying compiler-transcendent optimization, i.e., performance improvements that remain after aggressive compiler optimization. PatternBench targets residual optimization opportuni
benchmark - arxiv:2610.24833 · cs.MAMeasuring Consensus in AI Agent NetworksJin Gao
As AI agents become increasingly connected through networks and take on tasks that require negotiation and joint decision-making, a systematic approach is needed to assess how their views converge and what information supports their decisions. We present an evaluation framework that treats consensus formation as the evolution of publicly stated positions toward a shared region, using accumulated communication history as context. The framework tracks position trajectories, group dispersion, and information expressed in public messages. Two case studies examine staffing decisions based on distri
agentai agentevaluation framework - arxiv:2610.24832 · cs.CVVTF900: A 900-Subject Visible-Thermal Face Recognition Dataset and BenchmarkJacob Abaare, Bianca Ianetta, Cedric Nimpa Fondje, Shuowen Hu +1
Despite advancements in visible-thermal face recognition, where thermal probe (query) face images are matched against visible enrollment templates, existing benchmarks are currently limited by small gallery sizes (N $\approx$ 100 subjects). This limitation inhibits the ability to accurately evaluate visible-thermal facial recognition at scale and to report significant recognition performance metrics, which have better operational relevance. Therefore, we introduce VTF900, a large-scale visible-thermal face recognition benchmark with 900 subjects and paired imagery that includes expression, pos
benchmark - arxiv:2610.24830 · eess.SYHEL: A Heterogeneous Extensible Layer for GPU-Resident Runtime OrchestrationAgnieszka M. Łupińska, Serban D. Porumbescu, John D. Owens
We describe a GPU-resident runtime system, HEL, built on stateful services that persist throughout the lifespan of a user application. A HEL programmer expresses a workload as an application that spawns coroutine-like tasks, which are then scheduled and run by the HEL runtime. HEL supports synchronous and asynchronous tasks, tasks with runtime dependencies, batched task submission, and automatic retries. Applications and tasks run alongside each other on the spatially partitioned GPU. Using producer-consumer and host-RPC benchmarks together with a PFPL compression case study, we show that HEL
benchmark - arxiv:2610.24826 · cs.LGCompose the Horizon: Parameter-Shared Blockwise Forecasting under a Sequential-Compute Trade-offZeyan Li, Weiheng Li, Shengda Zhuo, Libing Chen +2
Most long-horizon forecasters emit the complete future in one pass. We ask whether a compact operator can instead be reused to emit contiguous future blocks, trading stored parameters for sequential computation. Recursive Functional Decomposition (RFD) factorizes the horizon mapping into three parameter-shared block predictions. The blocks occupy disjoint future intervals and are concatenated, not added as trend, seasonal, or residual components, and each emitted block joins the rolling context that conditions the next application. In controlled three-seed experiments, removing forward feedbac
long-context - arxiv:2610.24824 · cs.AITeaching Code Agents to Ask Without Leaking Private CodeWenjia Jiang, Jieke Shi, Zhensu Sun, Chengran Yang +8
Local coding agents can extend their reach by consulting a stronger cloud model. Such consultations, however, often expose the repository's private identifiers, file paths, and source code. Host-side masking hides recognized identifiers but leaves the surrounding code and the agent's phrasing unchanged, while training for less disclosure risks suppressing consultation or producing questions too vague to answer. We propose utility-verified abstraction (UVA), which teaches the agent to describe private entities by their role and type while retaining the technical details an expert needs. From so
agentbenchmark - arxiv:2610.24823 · cs.LGPair-Aware Optimization: Coordinating Exploration and Refinement for Jailbreaking LLM AgentsJuanyang Xu, Zheng Wang, Xingyu Zhao, Siddartha Khastgir +1
Large Language Model (LLM) agents are vulnerable to adversarial inputs that induce attacker-specified malicious behaviors, commonly referred to as jailbreaking. Existing optimization-based jailbreak attacks primarily employ first-order (FO) methods that leverage gradient information or zeroth-order (ZO) methods that rely on function-value feedback. However, their complementarity and optimization ordering remain underexplored. We systematically investigate four strategies: pure ZO, pure FO, ZO->FO, and FO->ZO, and develop a ZO-FO framework that explores adversarial prefixes through ZO optimizat
agentllm agentbenchmark - arxiv:2610.24818 · cs.LGEfficiency Measured in English: Tokeniser Coverage Bounds Small Language Models on Indic ScriptsAnna Pramod
Small Language Models with quantisation and parameter-efficient adaptation are widely offered as the route to multilingual AI under compute constraints, and their efficiency is almost always reported as a single memory figure. We show that this figure is measured in English and does not transfer. Across a 131k-sentence corpus spanning eleven Indic scripts, TinyLlama-1.1B's tokeniser emits 1.34 subword tokens per Latin-script word and 22.60 per Kannada word. This is not inefficiency; it is absence. Its vocabulary contains 0% of Odia's codepoints, 1% of Telugu's and 2% of Kannada's, and for five
memory - arxiv:2610.24816 · cs.LGFirst-Order LoRA Training with Phone NPUsXu Zheng, Wei Cheng, Zhuomin Chen, Dongkuan Xu +3
On-device fine-tuning lets a large language model (LLM) learn from its user's private data without uploading it. Low-rank adaptation (LoRA) reduces the trainable parameters, but each first-order training step still multiplies by the frozen base-model weights during the forward pass, activation recomputation, and input-gradient computation. A phone's neural processing unit (NPU) already holds these weights for inference; using it for training also requires computing input gradients, which its inference interface does not support. We present a system that runs all three frozen-weight products on
benchmark - arxiv:2610.24814 · cs.CVSL-Ground: Sentence-Level Grounding for Fine-Grained Open-Vocabulary Object DetectionYang-Syu Li, Cheng-Huan Yang, Chia-Yung Jui, Peggy Joy Lu
Grounding-based open-vocabulary detectors provide a widely used framework for text-conditioned localization, with a phrase-grounding formulation that aligns object queries to individual text tokens and recovers class scores through token-to-class aggregation. We observe that this token-level scoring design becomes a bottleneck for fine-grained open-vocabulary object detection, where the target category is defined by a complete compositional description rather than by isolated tokens. Based on this observation, we propose SL-Ground, a sentence-level grounding reformulation for grounding-based o
benchmark - arxiv:2610.24807 · cs.AIMedical Fine-Tuning Reshapes, Not Reduces, Patient-Facing AI Risk: A Rubric-Based, LLM-Judge Evaluation FrameworkSavan Doshi
Patient-facing medical question answering with large language models (LLMs) is moving from research demonstrations to production deployments, yet evaluation remains anchored on factual correctness. Factuality is necessary but not sufficient: a response can be factually grounded and still issue a directive that, if acted on, would harm the patient. We introduce a risk-sensitive evaluation framework that scores responses by the presence and severity of six categories of risk-bearing language (treatment directives, contraindications, dosage expressions, urgency-triage cues, high-alert medication
benchmarkevaluation framework - arxiv:2610.24804 · cs.ROSARAM: Sparse Adapter Allocation with Reuse and Merging for Lifelong Robot LearningSeongJun Jeong, Chungwoo Lee, Gi-Cheon Kang, Byoung-Tak Zhang
Lifelong robot learning requires acquiring new tasks sequentially without forgetting earlier ones. With large pre-trained vision-language-action (VLA) backbones, adapter-based methods effectively mitigate this challenge. However, their task-specific adapters and selection modules accumulate with the task stream, creating a scalability challenge in on-disk storage and deployment cost under fixed onboard resource budgets. We propose SARAM, a lifelong learning framework that limits storage growth while keeping GPU memory and per-step inference cost independent of the task stream length. For each
vision-language-actionliberomemorylifelong learning - arxiv:2610.24801 · cs.ROWhen Better Policies Leave Harder Failures: Coverage Shift in VLM Success MonitoringIsabel Guan, Dingyuan Zhang, Shipeng Lyu, I-Ming Chen
Can a better robot policy become harder to monitor? We study a fixed vision-language model (VLM) that judges success while a policy improves. In a controlled 20-seed tray-placement experiment, training selects only simulator-verified successes. Success rises from 66.8% to 97.0%, and accepted hover failures fall from 5.0% to 0.9% of executions. Yet the judge accepts 50.0% of remaining hovers, versus 31.1% under untrained sampling. These residual failures more often lie below the tray rim, which can conceal the gap beneath a grasped cube. A separate boundary probe accepts 148/160 below-rim hover
robot policygrasp - arxiv:2610.24793 · cs.AICooperation Without Shared Rewards via Reciprocal Q-AugmentationFrancesco Morri, Adel Abusitta, Quentin Cappart, Margarida Carvalho
General-sum games are more challenging than fully cooperative or adversarial settings because agents may deceive or exploit one another. Achieving high social welfare requires sustained cooperation, while a single defection can drive agents toward a welfare suboptimal Nash equilibrium. Moreover, assuming access to other agents' rewards or internal states is often unrealistic and implicitly introduces additional cooperation into the setting. To address this issue, we introduce \textit{Reciprocal Q-Augmentation} (RQA), a method that uses a latent-variable opponent model to augment the Q-function
agent - arxiv:2610.24791 · cs.CVUniAVC: Unified Audio-Video Creation via Progressive Training and Asymmetric SchedulingKien T. Pham, Yuhao Liu, Chong Hou Choi, Zhenqi He +5
Unified audio-video creation requires a single model capable of text-conditioned joint synthesis, bidirectional cross-modal generation, and instruction-guided editing across both modalities. Existing methods rely heavily on paired corpora, which are scarce for joint generation and even rarer for editing. Consequently, they are spent learning single-branch generative and editing capabilities that abundant unimodal data could readily teach, rather than on the cross-modal coupling. Furthermore, prevailing systems apply a single shared noise schedule to both modalities, neglecting two critical lim
self-refinement - arxiv:2610.24790 · cs.AIWhen LLM-Generated Refactorings Ripple: An Empirical Study of Repository Adoption ReadinessMuhammad Waseem, Aakash Ahmad, Pekka Abrahamsson
Large Language Models (LLMs) are increasingly used to refactor source files, but file-local evaluations do not reveal changes to contracts, dependencies, or caller assumptions affecting repository adoption. To address this gap, we conduct 1,500 generation attempts across 150 production files from 15 open-source Java repositories, five LLM configurations, and two dependency-aware generation conditions: single-prompt refactoring and LangGraph-based multi-agent refactoring. Of these attempts, 1,455 yield complete candidates eligible for analysis. We evaluate these candidates using local changes,
multi-agent - arxiv:2610.24789 · cs.AIIndustrialCAD-1400: Benchmarking Detailed Industrial Assemblies from TextHebao Zhu, Kang An, Zhuoke Liao, Faqiang Qian +3
Industrial assemblies require more than a recognizable shape: functional parts, interfaces, dimensions and service features must satisfy a detailed request together. We introduce IndustrialCAD-1400, a text-to-CAD benchmark for this industrial setting. Its 1,400 requests combine 400 equipment categories across 40 domains with 250 four-case design families. Family requests specify a median of 17 named component or configuration roles and six explicit numeric targets; controlled variants test whether dimensions change while shared structure is retained. Task construction draws on documented manuf
benchmark - arxiv:2610.24783 · cs.CVATLAS: Learning to Adapt LVLM Attention for Reasoning Video SegmentationHamidreza Dastmalchi, Ali Cheraghian, Aijun An, Abdelwahed Khamis +1
Reasoning Video Object Segmentation (RVOS) aims to localize and segment objects referred to by natural-language expressions throughout a video. Although large vision-language models (LVLMs) implicitly encode localization cues in their attention maps, these attentions are often noisy, diffuse, and scattered across layers, heads, and modalities, limiting their direct use for segmentation. Existing methods typically rely on handcrafted attention-selection heuristics or computationally expensive attention refinement. We propose ATLAS, an attention adaptation framework that learns to transform raw
benchmark - arxiv:2610.24779 · cs.ROLearning Structured Locomotion References for Optimization-Based Whole-Body ControlÖmer Tarık Banuş, Uluç Saranlı, M. Mert Ankaralı
Quadruped locomotion requires coordinating body motion,contact forces, foothold placement, and gait timing. We present a hierarchical controller in which reinforcement learning adapts these structured locomotion references while a gait scheduler, reactive footstep planner, and whole-body impulse controller retain responsibility for high-frequency execution. The model-based backbone can sustain a nominal trot independently, allowing the policy to focus on when and how to modify an already functional locomotion controller. We train the complete hierarchy in a vectorized controller-in-the-loop en
quadrupedwhole-body controlevaluation protocol - arxiv:2610.24771 · cs.CVDesign Anything, Design at ScaleXuchao Ren, Siyuan Feng, Jing Huang, Yanhua Yao +7
As multimodal resources are transformed into media outputs, agentic AI is expected to stay aligned with design intent over long horizons. Designers follow briefs specifying goals, constraints, and deliverables, with later work building on earlier decisions. Recent work reuses design experience across tasks, yet alignment among artifacts within one project remains challenging. Agents therefore need to preserve and apply this intent across domains and as projects grow. We introduce Design Anything, a graph-based harness that preserves design intent outside the model and binds deliverables to the
agentic - arxiv:2610.24767 · cs.AIDialogMem: Evaluating Dialogue-Grounded Embodied Memory in 3D EnvironmentsSeongJun Jeong, Junghyun Kim, Gi-Cheon Kang, Byoung-Tak Zhang
Embodied agents in human-centered environments require memory that integrates what they observe with what humans tell them over time. However, existing embodied memories are built primarily from visual experience, leaving open how dialogue with humans can be grounded in the objects encountered throughout that experience. We introduce DialogMem, a two-phase benchmark in which agents build memory from time-aligned streams of posed RGB-D observations and dialogue, and later use it for instance-level object navigation and question answering (QA). DialogMem provides 142 episodes across 27 real-worl
embodiedmemoryembodied agentbenchmark - arxiv:2610.24764 · cs.CLBanglaBBQ: A Bias Benchmark for Question Answering in the Bangladeshi ContextSumit Anthony Gomes, Yasin Rahman, Taeeba Tasnia, Faiza Nooren Suhita
This paper introduces BanglaBBQ, a question-answering bias benchmark designed to capture stereotypes grounded in the Bangladeshi context, in both Bangla and English. Following KoBBQ and PakBBQ, BanglaBBQ adapts the original Bias Benchmark for Question Answering (BBQ) by classifying its templates into three groups: directly translated (DT), modified to align target groups or contexts with Bangladesh (TM), and removed when deemed unsuitable for the Bangladeshi context (SR). This process retains 140 templates across six categories, comprising 76 DT and 64 TM templates. Five newly authored (NA) ca
benchmark - arxiv:2610.24757 · cs.CVCorrCLIPv2: Harnessing Unlabeled Images for Open-Vocabulary Semantic SegmentationDengke Zhang, Fagui Liu, Quan Tang, C. L. Philip Chen
Open-vocabulary semantic segmentation aims to classify pixels without being restricted to a fixed set of categories. Although Contrastive Language-Image Pre-training (CLIP) demonstrates remarkable potential for open-vocabulary recognition, its segmentation capability is impaired by noisy inter-class patch interactions. To this end, this work proposes a unified training-free object-centric framework, named CorrCLIPv2, from three complementary aspects of patch correlation, feature representation, and output regularization. Specifically, for patch correlation, we leverage class-agnostic mask gene
memorybenchmark - arxiv:2610.24747 · cs.ROEndoVLA-OOV: Language-Conditioned Out-of-View Target Acquisition in Robotic EndoscopyChi Kit Ng, Yidong Zhang, Tianchun Wu, Zhiqing Tang +8
The scan-then-revisit workflow of upper gastrointestinal endoscopy requires bringing a previously observed suspicious region back into view for closer inspection. Acquiring such targets requires translating anatomical and appearance descriptions into camera motions from partial observations of a confined, deformable environment. We present EndoVLA-OOV, a language-conditioned vision-language-action policy for out-of-view target acquisition in robotic endoscopy. The policy adapts a pretrained backbone to continuous two-degree-of-freedom distal bending commands and combines initial, intermediate,
vision-language-actionvla - arxiv:2610.24744 · cs.AIReTRACE: Feedback-Guided LLM Agent in a Chemistry Harness for Retrosynthesis PlanningJinxiang Xie, Yijian Zhang, Qingyuan Yi, Jianyi Wang +1
Computer-aided retrosynthesis has advanced through single-step prediction, tree search and route scoring, but complex targets still require judgement over the whole route, where stereochemistry, functional-group compatibility, cost and greenness interact. Here we present \textbf{\mbox{ReTRACE}} (\textbf{RE}trosynthesis with \textbf{T}ool-orchestrated \textbf{R}easoning \textbf{A}nd \textbf{C}ontext-aware \textbf{E}vidence), an agent that places a large language model (LLM) in a chemistry harness that integrates molecular tools, complementary single-step prediction engines, a purchasable starti
memoryagentllm agentbenchmarkevaluator - arxiv:2610.24740 · cs.AIAre We Evaluating VLM Token Pruning Right? Looking Beyond Accuracy and Token CountGeng Li, Guohao Chen, Shilin Shan, Jianfei Yang
Visual token pruning aims to reduce inference cost while preserving the capabilities of the original VLM. Yet existing evaluations rely on aggregate task accuracy and retained token count as proxies for pruning quality and actual efficiency. We reveal two limitations: 1) Aggregate accuracy can obscure whether a pruning method truly identifies critical visual tokens; many examples remain solvable under arbitrary token removal and models can exploit textual shortcuts when visual evidence is degraded. 2) Retained token count is an unreliable proxy for efficiency, as identical token budgets can in
benchmarkevaluation framework - arxiv:2610.24739 · cs.AIToward Collective Model Intelligence: Characterizing and Coordinating ModelsSinapisAI Team, Jian-Dong Liu, Qun Dai, Zhendong Du +26
Independently developed AI models accumulate diverse and complementary capabilities, but a growing model collection does not automatically realize Collective AI. Addressing this gap requires understanding model capabilities with limited evaluation, organizing complementary models to collaborate on diverse tasks, and executing model calls efficiently under resource constraints. We present a systematic approach to pursuing Collective Model Intelligence through three connected components. First, Model Eval constructs evidence-based capability profiles by combining query-level performance estimati
benchmarkeval - arxiv:2610.24738 · cs.CVEditing Images into Reusable Layers via Compositional Edit-Prior DistillationWen Xue, Yong Liu, Xueheng Li, Si Wu +4
High-quality image editing requires precise modification of target regions while preserving the content of unrelated regions. Existing editing models produce flattened outputs that limit precise local manipulation and layer reuse, while decomposition methods typically optimize layer recovery separately from instruction-driven editing. To this end, we introduce LayerEditor, a unified framework that couples instruction-guided image editing with controllable generation of reusable RGBA layers. At its core, In-Context RGB-Alpha Token Modeling jointly represents the edited canvas and an ordered seq
manipulationbenchmark - arxiv:2610.24737 · cs.LGGaussian Process Bandits with Paired Auxiliary ObservationsWeibo Huang, Qing Feng, Ruihao Zhu, Cheng Hua
Bayesian optimization (BO) is widely used for expensive black-box optimization, yet observation noise makes each evaluation less informative and slows posterior learning. Standard BO learns from costly outcomes gathered during optimization. In many applications, however, low-cost auxiliary data are available beforehand. We introduce Auxiliary-Informed Gaussian Process Bandit (AI-GPB), a variance-reduction method that combines an offline estimate of the auxiliary mean with a correction coefficient learned from paired online observations. Notably, it exploits correlation between paired observati
benchmark - arxiv:2610.24733 · cs.AICan LLM Agents Resolve Service Failures? Simulating Multi-Turn Complaint Handling under Information AsymmetryWailing Ng, Mengze Hong, Chen Jason Zhang, Shiqi Fan
LLM agents are widely adopted in service chatbots for rule-guided support and information access, yet emotionally demanding complaint handling still relies heavily on human agents. We introduce ServiceBench, to our knowledge the first benchmark dedicated to complaint handling, with 2,500 scenarios spanning 184 complaint categories and five service industries. We propose Asymmetric Progressive Complaint Simulation (APCS) to simulate realistic complaint handling, together with a justice-based evaluation framework. Our experiments reveal that even capable LLM agents struggle to handle complaints
llm agentpost-trainingbenchmarkevaluation framework - arxiv:2610.24730 · cs.AIGLOSS-MT: Controllable Low-Resource Machine Translation with an Editable DictionaryMich-Seth Owusu, Stephen E. Moore, Akwasi Asare, Lawrence Adu Gyamfi +3
Adapting a pretrained machine translation (MT) model to new terminology usually requires new parallel data and another round of fine-tuning, which is costly for low-resource languages. We ask whether controllability can be added to MT, and the fine-tuning burden reduced, without degrading the pretrained model. GLOSS-MT looks up the source's noun phrases in an offline English-Twi dictionary, replaces every phrase found by its Twi translation in place, and fine-tunes a translation model once to translate these mixed inputs. We apply the same data and pipeline to NLLB-200 (600M, 1.3B, 3.3B), OPUS
benchmark - arxiv:2610.24729 · cs.LGfxBench: Evaluating and Understanding Formula Suggestions in SpreadsheetsSanket Mhatre, Sumit Gulwani, Vu Le, Yasharth Bajpai +1
Spreadsheets are among the most widely used tools for computations, and formulas are their core building blocks. Besides agentic spreadsheet tools, proactive formula suggestions enable significant productivity gains and are now available in popular spreadsheet software (Microsoft Excel and Google Sheets). Such suggestions pose interesting challenges: the right context must be collected and represented for a machine learning model, which must then be able to write a correct formula, in a way that is fast and cheap enough to trigger on every = (or subsequent keystrokes for auto-completion) witho
agenticbenchmark - arxiv:2610.24723 · cs.AIWhere Natural Language Ends in Agents: Executable Skill Composition with MetaSkillMeng Shu, Wei He, Hang Zhou, Runke Liu +2
Agent Skills express open-ended task knowledge in natural language, but composing them requires persistent execution order and upstream-output handoffs. We propose selective formalization at Skill boundaries: coordination relations with stable references and machine-checkable constraints become executable, while substantive task judgments remain within Skills. We introduce MetaSkill, a hybrid representation that preserves complete Skill packages and local instructions while making invocation dependencies and predecessor-text bindings executable through a directed acyclic graph (DAG) runtime. T
agentagent system - arxiv:2610.24718 · cs.AINeuro-Inspired Inhibitory Message Passing for Accurate and Token-Efficient Multi-Agent ReasoningNima H. Siboni, Magdalena Springer, Vahid Rostami
Multi-agent systems promise to solve problems by bringing together information beyond the reach of any individual agent. Retrieval and tool use can distribute this information across agents, making communication essential to collaboration. A central challenge is keeping crucial evidence visible in each agent's context as messages accumulate. Repeated information can crowd out rare, decisive evidence. Models of cortical circuits suggest a principle for addressing this imbalance: inhibition can selectively suppress shared inputs while preserving sensitivity to weaker signals. We adapt this princ
multi-agentagent systemtool usebenchmark - arxiv:2610.24717 · cs.AIDiffuPlan: Editable Planning State for Diffusion Language Model AgentsQisheng Zhou, Zhen Xiong, Qiaoyu Tan
Diffusion language models (dLLMs) are emerging as an alternative to autoregressive (AR) language models, yet dLLM agents largely inherit an interaction protocol designed for AR generation: once intermediate state is written, later observations can only be incorporated by appending more text. This is a poor match to a defining capability of dLLMs -- reconstructing earlier text conditioned on context that appears both before and after it. We ask whether this capability can support a different agent-state paradigm in which intermediate planning state is directly revisable as evidence arrives. We
agentllm agentagenticbenchmark - arxiv:2610.24709 · cs.AISQuIREL: A Context-Relative Framework for Measuring Entity Richness in WikidataMichelle Nwachukwu, Albert Meroño-Peñuela, Martim Brandão
Wikidata is a large, heterogeneous knowledge graph and applying data quality metrics requires context-specific approaches. The "richness" quality metric is currently under-researched and typically restricted to the ontology level, which fails to capture entity-specific structural depth. To address this gap, we introduce SQuIREL, a method that measures richness directly at the entity level using six dimensions: Semantic Specificity, Qualifier Density, Internal Connectivity, Reference Provenance, External Network Connectivity, and Linguistic Accessibility. These dimensions are kept disaggregated
knowledge graph - arxiv:2610.24705 · cs.RORelational Spatial Alignment for Vision-Language-Action ModelsChaoran Zhu, Hengyi Wang, Chenglin Cui, Yazhe Wan +2
Spatial understanding is essential for robots to perceive their surrounding environments and execute actions. Existing approaches typically encode 3D awareness into vision-language-action (VLA) models by directly aligning their representations with pre-trained 3D foundation models. However, we observe that such direct geometric feature alignment essentially exploits 3D foundation priors as a form of low-rank regularization in the feature space, resulting in severe rank collapse of the pre-trained VLA representations. In contrast, we shift the focus from enforcing point-wise equality to alignin
vision-language-actionvlamanipulationbenchmark - arxiv:2610.24688 · eess.SYMulti-Agentic AI for Intent-Driven and Conflict-Aware rApp Policy Orchestration in Open RANHaiyuan Li, Yulei Wu, Luyu Qi, Dimitra Simeonidou
Open radio access network (O-RAN) enables a multi-vendor ecosystem that accelerates xApp development for near-real-time (RT) network management. However, this openness shifts a combinatorial coordination burden to the Non-RT layer, where rApp policies compose xApps from a growing catalog to realize operator intents. Coexisting intents further compound this challenge, as conflicts may arise among xApp actions, rApp policies, and their cross-layer interactions. To maximize intent admission while maintaining key performance indicator (KPI) compliance across incoming and incumbent intents, we prop
agentmulti-agentagentic - arxiv:2610.24687 · cs.LGGeometry-Aware Interpolation for Inverse ProblemsPourya Shamsolmoali, Masoumeh Zareapoor
Recovering missing visual information from incomplete or degraded images remains a challenging problem, particularly when spatial variations and complex image structures affect reconstruction accuracy. Although Spatial Transformer Networks (STNs) facilitate spatial feature alignment, their reliance on conventional linear or bilinear interpolation limits their ability to capture the underlying geometric characteristics of image data. Furthermore, interpolation based on a restricted set of neighbouring pixels can introduce inconsistencies in gradient propagation, potentially affecting optimisati
benchmark - arxiv:2610.24685 · cs.LGContrastive Adversarial Training for Robust Graph Neural Networks under Label PoisoningManshika Charvi Bissessur, Melis Ilayda Bal, Michael Muehlebach
Graph Neural Networks (GNNs) are effective for modeling relational data but are vulnerable to label poisoning, where a small number of corrupted training labels can propagate errors across the graph via message passing. Despite this risk, defenses against label poisoning remain underexplored: existing methods are primarily designed for label noise and often rely on robust losses or heuristic data cleaning that fail to distinguish adversarial poisoning from naturally hard examples. In this paper, we propose CoLAT (Contrastive Label-flipping for Adversarial Training), a novel contrastive adversa
benchmark - arxiv:2610.24681 · cs.LGMM-LeJEPA: Correlated Gaussian Regularization for Multimodal Joint-Embedding Predictive ArchitecturesHaowen Bai, Rui Gong$, Meichen Liu, Ming-Xuan Weng +2
LeJEPA derives an isotropic Gaussian embedding target by minimizing, under its task assumptions, a Fisher information criterion that controls the task-averaged squared prediction bias. This enables unimodal self-supervised learning without reconstruction decoders, exponential-moving-average teachers, or contrastive negatives. However, its isotropic task prior does not account for the cross-modal structure of multimodal tasks that read shared content. We introduce MM-LeJEPA by extending LeJEPA's task assumptions to multimodal learning. Specifically, we model multimodal tasks as strength-weighte
benchmark - arxiv:2610.24676 · cs.CVFrom Views to Worlds: Active Exploration over 3D Worlds for Vision-Language ModelsQijian Tian, Jiayu Ying, Ke Fan, Lizhuang Ma +1
Vision-language models (VLMs) have achieved remarkable progress in 2D visual understanding tasks, yet the passive view-centric perception paradigm limits their capacity for 3D spatial reasoning. Such reasoning often requires actively acquiring spatial evidence beyond the currently visible views, such as camera poses and cross-view spatial relationships, to infer 3D relationships that are not explicitly represented in 2D images. However, existing methods either lack active exploration or merely provide additional views during inference, without explicitly acquiring such spatial evidence beyond
benchmark - arxiv:2610.24675 · cs.ROAction-Aligned Latent Planning for Versatile Humanoid Loco-ManipulationYutang Lin, Guanya Shi, C. Karen Liu
Learning a single perceptive policy for diverse humanoid loco-manipulation requires unifying high-level task decisions without losing the state-dependent physical responses that make each interaction succeed. This makes the representation between high-level planning and whole-body control critical: kinematic or contact targets provide compact abstractions but do not directly encode interaction-specific control responses, while directly predicting low-level actions requires the planner to model high-frequency closed-loop feedback. We propose FIBER, a framework that uses the executable latent sp
manipulationhumanoidwhole-body controlsim-to-realgrasp - arxiv:2610.24674 · cs.RONoise-Invariant Spatial Encoding for 3D Manipulation Policies under Depth CorruptionZhiyuan Yang, Jiaju He, Ruiteng Zhao, Anqi Cheng +2
3D manipulation policies derive actions from explicit scene geometry rather than color images alone, enabling them to reason over metric distances, object relations, and robot-scene reachability. This geometric interface is also a vulnerability: commodity RGB-D sensors return missing measurements, boundary artifacts, and plausible but wrong surfaces around transparent, reflective, thin, or textureless objects. We study this failure under realistic depth corruption and observe a sharp depth-noise cliff across modern 3D policies, even when they are trained with matched noisy depth. This residual
manipulationbenchmark - arxiv:2610.24671 · cs.CLDo Conversational Agents Need Explicit Memory? A Controlled Comparison with Raw-History RetrievalYongjie Zhou, Shuai Wang, Bevan Koopman, Guido Zuccon
Explicit memory systems transform conversation history into persistent representations intended to improve future access, often at substantial construction cost. Yet it remains unclear whether this added complexity yields better answer accuracy than retrieving directly from raw history. This study compares six state-of-the-art explicit memory systems with raw-history RAG on LongMemEval-S and BEAM-100K under original and controlled retrieval settings. We find only localized gains for particular question types and tight run-time context budgets, with no consistent advantage over strong raw-histo
memoryragai agent - arxiv:2610.24662 · cs.AIFrom Prediction to Action: Population Health World Model for Digital MedicineXike Xie, Yu Li, Weizhong Yang, Amalie Dyda +1
Digital medicine now captures data that traces individual health trajectories and reveals population health dynamics through electronic health records, claims, remote monitoring, wearable devices, mobile devices, social sensing data, medical images, etc. But much of its modeling still centers on task-specific, outcome-specific, and prediction-specific estimation. The next frontier is not simply better prediction, but action in an evolving health world shaped by multiple decision-makers, heterogeneous populations, resource constraints, and feedback in care delivery. This is why digital medicine
world model - arxiv:2610.24661 · cs.LGCan AI agents make law? Evaluating AI legal relevance and creativityBenjamin Ricaud, Artem Galushko
Large language models, the artificial intelligence (AI) systems behind modern chatbots, are trained on vast amounts of legal text, but can they build on it to write law for situations that text never anticipated? In this study, we analyse AI agents' ability to draft a hypothetical "Constitution" for a "community of AI agents". We design "constitutional assemblies" in which a single language model plays several characters that debate to produce a draft constitution. We find that the assemblies converge on a robust floor of legal provisions, and generate a wide variety of different proposals aro
ai agent - arxiv:2610.24649 · cs.LGGood Judgment, Bad Advice: Decision-Model Controllers Raise Agent Cost Without Improving SuccessNaman Vats
Agent scaffolds increasingly delegate small decisions (whether to stop, verify, or change approach) to cheaper models. We ask whether a typed decision model, Jev 1.13, consulted only at deterministic decision points, makes a Claude Sonnet 5.5 terminal agent cheaper without making it worse. In a pre-specified study on Terminal-Bench 2.1, we hold one agent loop fixed and vary only the controller: none, deterministic rules, a small LLM (Claude Haiku 5.5) answering the same questions, Jev, and a placebo that samples Jev's development-phase choice frequencies at the same moments. On 21 held-out tas
agent - arxiv:2610.24648 · cs.LGEdgeKernelBench: Unified LLM-Driven Kernel Generation for Edge BackendsYuhua Zheng, Xueyi Chen, Xinze Zou, Xin Li +2
Recent works show that LLMs can generate kernels with strong performance on server GPUs. Edge deployment presents distinct challenges because the hardware is heterogeneous, the software stacks are fragmented, and the resource budgets are tight. LLM kernel generation across diverse edge backends has yet to be studied systematically. We introduce EdgeKernelBench, which combines a unified IR deployment framework with a hierarchical benchmark for edge workloads. The unified IR captures reusable computation, scheduling, and memory decisions and connects to each target backend through a compiler int
memoryagentbenchmark - arxiv:2610.24645 · eess.SYDistributed Real-Time Cooperative Model Predictive ControlYuning Jiang, Kristína Fedorová, Roland Schwan, Juraj Oravec +1
This paper presents a novel distributed real-time iterative method for cooperative linear Model Predictive Control (MPC) with polyhedral state and input constraints, implementing a relaxed, recentered logarithmic barrier function. The method transforms centralized MPC problems into a distributed framework using temporal-spatial splitting, facilitating efficient online implementation. To achieve real-time performance, we introduce a distributed convex nonlinear optimization solver with guaranteed global linear convergence. By executing this solver with a fixed number of iterations per sampling
benchmark - arxiv:2610.24642 · cs.CVSpectral-Aware Proposal Enhancement for Diffusion-Based Multispectral UAV Multi-Object TrackingXu Lin, Tingfa Xu, Tianhao Li, Haolin Qin +3
Multispectral UAV multi-object tracking (MOT) with oriented bounding boxes remains challenging due to small targets, cluttered backgrounds, and unstable inter-frame associations. Although diffusion-based MOT provides an effective iterative proposal refinement paradigm, existing diffusion-based trackers lack dedicated designs for exploiting multispectral cues in proposal initialization and confidence calibration, limiting tracking robustness in complex UAV scenarios. This letter proposes DSATrack, a diffusion-guided spectral-aware tracking framework for multispectral UAV MOT. Specifically, a li
benchmark - arxiv:2610.24630 · cs.AISelm: Self-Evolving Library Management for Multi-Skill Agent OptimizationMengxi Hu, Yuxuan Hu, Jinbo Su, Ke Wang +5
Large language model (LLM) agents increasingly use skills as a lightweight way to acquire task-specific knowledge. Recent work enables skills to evolve from task trajectories. However, these methods often lack systematic management of evolved skills, making it difficult for agents to reuse and accumulate skills across domains, thus limiting performance on subsequent tasks.To reliably accumulate, reuse, and maintain useful skills across tasks, we present SELM (Self-Evolving Library Management), a framework that formulates skill evolution as library-level management. SELM jointly manages skill r
agentagent benchmarkself-evolvingbenchmark - arxiv:2610.24628 · cs.AIMuse Glimmer 30B: An On-Device Agentic ModelArash Ebrahim Abtahizadeh, Muna Aghamelu, Adil Ahmed, Juan C. Angeles Acuna +308
We introduce Muse Glimmer, an open-source 30B parameter causal language model with perception encoder, purpose-built for autonomous agentic tasks on consumer hardware. Muse Glimmer integrates multi-step reasoning, reliable tool use and failure recovery into a single model that runs locally on consumer devices without requiring cloud infrastructure. The model accepts text and image inputs and produces text outputs, was trained on data from over a hundred languages, and is quantized to approximately 4-bit precision for efficient single-GPU inference. Muse Glimmer achieves state-of-the-art perfor
ai agentautonomous agentagentictool usebenchmark - arxiv:2610.24619 · cs.CVBeyond Geometric Closure: Measuring and Diagnosing Mechanical Consistency in Visual DynamicsQibin Gu, Yunfeng Yan
Visual accuracy leaves open whether predicted motion obeys dynamics. We introduce a framework for measuring and diagnosing mechanical consistency in visual dynamics. An independent SI audit of rigid closed chains separates geometric admissibility, reaction-free actuator effort, and work--energy balance. Continuous scoring measures progress against fixed engineering targets, while reference replay quantifies measurement error. With three seeds per setting, we evaluate KeyCLD and CON across 343 planar assemblies at 64/256 pixels and V-JEPA + AC and LingBot-VA on 100 shared spatial test assemblie
v-jepa - arxiv:2610.24611 · cs.AIHSPR: Recovering Software-Agent Benchmark Scores via Hierarchical Posterior InferenceWanyi Chen, Wanqin Shen, Yuanchen Shi, Zhensu Sun +1
Evaluating software agents on a full benchmark is expensive. A small task subset can preserve agent rankings while yielding inaccurate scores. Historical responses provide useful structure, but tasks with identical source responses can have different outcomes for a new target agent. We introduce Hierarchical Score Posterior Recovery (HSPR), which combines source-fitted response factors with uncertain group success rates. Observed target outcomes update the posteriors over the target latent state and group rates; posterior-guided acquisition selects subsequent tasks. Its score-based criterion e
agentagent benchmarkbenchmark - arxiv:2610.24610 · cs.ROSolving a Giant Rubik's Cube with a Quadruped RobotLukas Amiet, Gabriel Sánchez-Blanco Repgen, Mayank Mittal, Victor Klemm +3
Pedipulating a freestanding articulated object requires controlling its pose and internal configuration while maintaining balance. We demonstrate this capability by autonomously solving a giant Rubik's Cube with a quadruped using only its feet. The task demands repeated precise interactions, since unintended layer motion can invalidate subsequent operations. We develop an onboard RGB-D perception pipeline that estimates the cube's metric pose and its full color state from partial views acquired through locomotion and flipping. A planner converts the computed solve into robot-executable skill s
quadruped - arxiv:2610.24607 · cs.CVIntriAdapt: Generalizable Intrinsic-Guided Test-Time Adaptation for Metric Geometry EstimationJiaxing Chen, Zizheng Yan, Jingyu Zhuang, Jinwei Chen +3
Recovering metric 3D geometry from images is fundamentally ill-posed due to perspective projection ambiguity. Current geometry models incorporate supplied or predicted intrinsics during inference, but do not use them to supervise model adaptation. To refine these predictions, test-time adaptation (TTA) typically relies on external depth measurements. In this paper, we instead use camera intrinsics as an output-side geometric verifier. Comparing the point cloud's viewing rays with those defined by the known intrinsics yields a direct measure of geometric inconsistency. We propose \textit{IntriA
benchmark - arxiv:2610.24603 · cs.RORefuse the Harm, Complete the Task via Graph-of-Rules Planning for LLM-Enabled RobotsAli Taheri, Alireza Taban, Qizhou Wang, Wanhua Li
Large Language Models (LLMs) are increasingly used in embodied AI for robotic manipulation and planning, but their deployment in physical systems exposes vulnerabilities such as jailbreak attacks that induce unsafe behavior. Existing inference-time defenses are mainly evaluated on short instructions containing a single hazard, whereas realistic tasks may interleave safe and unsafe sub-tasks, conceal hazardous steps, or combine individually safe actions into unsafe outcomes. Such compound instructions make it harder for a single safety decision to distinguish hazardous sub-tasks from legitimate
embodiedmanipulationmulti-agentbenchmark - arxiv:2610.24601 · cs.LGTCMScience: An Autonomous Scientist for Traditional Chinese MedicineYanlan Kang, Ruiqi Liu, Xinpei Gu, Tzu-Yi Lei +3
Traditional Chinese medicine (TCM) research connects classical texts, medicinal entities, molecular assays, computational predictions, and clinical studies. An autonomous research agent must preserve the inferential boundaries between these evidence types while coordinating analysis across heterogeneous resources. We introduce TCMScience, a scientific knowledge-discovery agent that separates language-model proposals from execution and release authority. Its architecture combines a trusted kernel, a scientific compiler, and claim-support contracts: plans are compiled into typed operations, exec
agent - arxiv:2610.24599 · cs.LGSeverity-Channel Conditioning: An Injection-Level Study for Consumer-Device DeblurringAbdullah Al Shafi, Kazi Saeed Alam, Sumaiya Rahim Suma, Sk. Imran Hossain
Motion-deblurring models lose substantial accuracy when transferred from professional cameras to smartphone-derived imagery. This paper studies whether a severity descriptor, partly derived from the paired sharp target, helps adaptation on the iPhoneBlur consumer-device benchmark. Under a shared update budget, input concatenation of a severity map is compared with global feature modulation by capture metadata, spatial feature modulation by the severity map, and a tier-contrastive output objective, together with an unconditioned spectral-loss control, all fine-tuned from the same domain-adapted
benchmark - arxiv:2610.24598 · cs.LGAdamPM: Memory-Efficient LLM Optimization via Moment PoolingXiaowei Ji, Lishan Xu, Jianing An, Jie Luo +2
Adam is widely used for training large language models (LLMs) owing to its strong optimization performance and training stability, but its element-wise estimation for the first and second moments incur substantial memory overhead. Although numerous memory-efficient optimizers have been proposed, many incur additional computational overhead or require additional tuning of hyperparameters. Consequently, Adam remains a dominant and reliable choice for LLM pre-training. We propose AdamPM (Adam with Pooled Moments), a simple and memory-efficient variant of Adam that groups adjacent moment entries a
memory - arxiv:2610.24597 · cs.AICADS: Measuring the Incremental Contribution of LLM Proposers in Sequential Biological ScreensGuanxing Chen, Nihui Shao, Haohuai He, Zhenchao Tang +2
Large language models are increasingly used to guide sequential biological screens, yet standard evaluations rely on raw hit counts that conflate a proposer's value with the discovery capacity of the underlying algorithmic harness. Adaptive sampling and corrupted experimental readouts also leave standard feedback comparisons without a false-discovery guarantee. We present Certified Agentic Discovery Screens (CADS), an error-controlled measurement framework that quantifies a proposer's incremental value against an identical LLM-free loop. By integrating capped likelihood-ratio bounds, e-process
agentic - arxiv:2610.24590 · cs.ROScene-Adaptive VLA: Efficient Autonomous Driving via Dynamic Layer RoutingYujia Yang, Quan Yuan, Jiawei Fan, Yang Li +4
Vision-Language-Action (VLA) models offer a promising paradigm for autonomous driving. However, the massive parameter scale of VLAs strains onboard computational resources during real-time inference. This directly increases energy consumption and limits vehicle battery range, making it critical to minimize computational overhead without compromising driving capability and explainability. Current research on efficient VLA-based driving focuses on sequential adaptation (\textit{e.g.}, Chain-of-Thought and token adjustment), leaving structural adaptation underexplored despite parameter redundancy
vision-language-actionvlavla modelbenchmark - arxiv:2610.24587 · cs.ROLIBERO-Pro2.0: A Benchmark for Evaluating the Robustness of Embodied Foundation ModelsXueyang Zhou, Yangming Xu, Qianjiang Li, Guiyao Tie +9
LIBERO has emerged as a widely adopted benchmark for evaluating robotic manipulation policies. However, strong performance on LIBERO does not necessarily indicate robust embodied capabilities, as models may exploit benchmark-specific regularities rather than generalize to diverse operating conditions. In this work, we introduce LIBERO-Pro2.0, a comprehensive benchmark for systematically evaluating the robustness of embodied foundation models under distribution shifts. LIBERO-Pro2.0 encompasses 22 static perturbations and 20 runtime interventions across six dimensions: Condition, Environment, O
vision-language-actionembodiedmanipulationliberobenchmark - arxiv:2610.24579 · cs.ROLIBERO-Dynamic: Benchmarking Robotic Manipulation in Dynamic EnvironmentsShaojie Ren, Jingwen Sun, Zezhi Liu, Hongjin Chen +3
Most simulation benchmarks for robotic manipulation keep target objects static and treat policy inference as instantaneous. These assumptions simplify evaluation but hide how a moving target evolves during computation and how stale an observation is when an action is executed. We introduce LIBERO-Dynamic, a dynamic extension of all 40 LIBERO tasks. It drives objects through contact friction along three trajectory families and speed tiers, while advancing the physical world according to calibrated, model-specific inference latency. The same mechanism supports training by replaying verified stat
manipulationliberobenchmark - arxiv:2610.24573 · cs.RORetrieval-triggered Action Projection Adjustment for Frozen VLA Policies: An Empirical Study and Failure AnalysisJinming Wang, Ruiheng Zhang, Yong Wang, Zipan Xiong +1
Frozen Vision-Language-Action (VLA) policies cannot incorporate previously encountered failures without updating the underlying model. We present retrieval-triggered projection correction (RTPC), an auditable external-memory wrapper for frozen OpenVLA policies that retrieves stress-collected local-depth records and applies translation-only projection correction while keeping rotation and gripper commands unchanged. RTPC is evaluated on CALVIN single-task manipulation using primary and independent replication cohorts. In the primary evaluation, RTPC reduced a depth-based near-collision or stall
vision-language-actionvlamanipulationopenvlagripper - arxiv:2610.24571 · cs.LGRepresentations Have Trajectories: Characterizing Temporal Representation Dynamics in Neural NetworksShivaprasad B. Gowda, Sparsh Mittal
Recent work has shown that deep neural networks exhibit an implicit low-rank simplicity bias. In this paper, we investigate the questions that naturally arise from this observation. What are the temporal consequences of low-rank simplicity bias during optimization? How does representation geometry evolve throughout optimization rather than only at convergence? Do different neural architectures exhibit distinct geometric optimization behaviors? To answer these questions, we conduct a large-scale empirical study tracking representational geometry throughout training. We find that neural networks
world model - arxiv:2610.24559 · cs.ROSocialCoAct: Enabling and Benchmarking Egocentric Social Co-Action for VLM-Based Humanoid RobotsJiancong Ma, Kun Li, Jin Jiang, Shengcai Liao
Enabling humanoid robots to participate effectively in social interactions is important for their integration into human-centered environments. Existing methods, however, typically map pre-generated speech directly to continuous motion trajectories but have limited awareness of complex social environments and lack explicit high-level semantic action planning. We introduce SocialCoAct, a unified agent-centered framework for VLM-based egocentric social co-action. We formulate egocentric social co-action as an end-to-end agent task and develop SocialCoAct-Agent, which transforms egocentric observ
humanoidagentbenchmark - arxiv:2610.24556 · cs.LGFrom Stories to Features: Scaling Concept Vectors for AI Alignment & InterpretabilityDmitry Scherbakov, Oleg Chernikov, Victoria Khamash, Kirill Afendulev +4
We study feature vectors: directions in a language model's activation space that correspond to human-interpretable concepts, such as honesty or deception. Existing unsupervised methods like sparse autoencoders attempt to recover the whole feature space, often at the cost of feature quality and interpretability. We take a more pragmatic route: focusing on concepts relevant to AI safety and alignment, we generate synthetic concept-antagonist story pairs and use the difference of their mean model activations as a feature vector. We scale this approach to 1,036 alignment-relevant concepts using 2M
agentic - arxiv:2610.24549 · cs.ROCARVS: A Congestion-Aware Algorithm for Online Multi-Agent Pickup-and-Delivery in Human-Collaborative WarehousesChaoran Li, Zeyuan Zhao, Shao Zhang
Intelligent warehouse systems have transformed logistics by integrating robots with minimal infrastructure changes, boosting efficiency and optimizing workflows compared to traditional human-only operations. However, human-robot collaboration introduces new challenges, particularly due to human uncertainty, which leads to dynamic interactions and unpredictable behavior. To address these complexities, this paper presents the Multi-Agent Pickup-and-Delivery with Human Collaboration (MAPDwHC) problem, an extension of the Multi-Agent Pickup-and-Delivery (MAPD) problem that incorporates human unpre
multi-agent - arxiv:2610.24545 · cs.AIWhen Code Becomes an LLM Call: An Empirical Study of Call-Site Contracts in Agentic SystemsShihao Weng, Longtian Wang, Yang Feng, Mingfei Cheng +5
Software quality has long rested on specifications that state what code must do, and on testing and verification that check that the code does it. Agentic systems change this foundation: they replace parts of their logic, such as choosing the next tool, judging whether an action is allowed, or writing code to run, with calls to a large language model (LLM). The program still relies on what this logic must guarantee, but the model's behavior is not fixed by code, varies across runs, and can be steered by its input. The guarantee must therefore come from the program around each LLM call site, as
agentmulti-agentagentic - arxiv:2610.24539 · cs.CLEGIP: An Evidence-Grounded Interaction Protocol for Clinical AgentsZecheng Ren, Ruichen Cong, Yi Min Chen, Jing Zhang +7
Interactive clinical agents acquire evidence through patient dialogue and diagnostic examinations, yet their performance also depends on how the interaction environment exposes information, available actions, and examination results. When these conditions are implicit or inconsistently controlled, final-answer accuracy can conflate diagnostic reasoning with the effects of the interaction protocol. We propose the \textbf{E}vidence-\textbf{G}rounded \textbf{I}nteraction \textbf{P}rotocol (EGIP), which enforces role-specific information boundaries, exposes diagnostic actions without revealing the
agent framework - arxiv:2610.24537 · cs.LGWrite Once, Query Many: Evolving Compact Reusable Memory for Long VideosKerui Chen, Jinglu Wang, Yun Wang, Xinjiang Wang +1
Long videos often exceed the context capacity of multimodal large language models (MLLMs), while agentic search repeats costly video processing and grounding for each query. Reusable memory can amortize this cost, but must preserve information for questions unknown during construction. We propose Video-EMem, an evolvable video memory harness built on time-anchored perceptual observations. This representation combines compact free-form text with sparse images to preserve semantic content and complementary visual details across temporal scales, supporting temporal navigation without a fixed sema
memoryagenticbenchmark - arxiv:2610.24534 · cs.LGStructured Latent Dynamics for Long-Horizon Embodied World ModelsKevin Chen, Yu-Wei Zhan, Haoran Liang, Sen Cui +1
Reliable long-horizon prediction remains a central challenge for embodied world models. Although recent generative models can synthesize visually plausible action-conditioned futures, their latent dynamics are typically learned through unconstrained transition modules, which may gradually deviate from coherent physical evolution as predictions are recursively rolled forward. We investigate whether explicitly structuring latent state evolution can improve the stability and utility of such imagined trajectories. To this end, we introduce HWM, a structured dynamics framework that represents backb
embodiedworld modelaction-conditionedlatent dynamics - arxiv:2610.24532 · cs.LGContrastive Hypergraph Source-free Domain Adaptive Object Detection in Adverse Weather Conditions--Extended VersionJicheng Yuan, Tung Kieu, Manh Nguyen-Duc, Manfred Hauswirth +1
Adverse weather conditions such as fog, rain, snow, and low illumination introduce structured visibility degradations that severely impair object detectors trained on clear-weather data. When source data are inaccessible due to privacy or trans- mission constraints, source-free domain adaptation (SFDA) becomes necessary. However, existing SFDA methods based on pairwise contrastive learning struggle to capture high-order semantics and often suffer from agreement collapse under severe corruption. We propose REUNION, a hypergraph-guided SFDA framework for object detection in adverse weather condi
benchmark - arxiv:2610.24530 · cs.CVAn Optical Music Recognition Tool for String QuartetsJin-Shiang Hu, Bo-Yuan Cheng, Li Su
This paper presents an optical music recognition (OMR) system tailored for PDF scores of string quartet pieces. The approach begins with a SegNet-based segmentation model to localize musical regions, followed by convolutional neural networks and image processing techniques to classify and extract musical symbols. To improve rhythmic precision, dynamic programming methods enforce time signature constraints and explicitly handle tuplets. Evaluation using a longest common subsequence metric across multiple feature matrices demonstrates robust accuracy in recognizing pitch, onset, duration, and ac
benchmark - arxiv:2610.24522 · cs.ROMOMO: Human Motion from Moving SoundsDeli Cai, Shuyang Xu, Zhiyang Dou, Zitong Lan +3
Human responses to sound depend not only on its semantic content, but also on the evolving spatial relationship between source and listener, conveyed through spatial audio and its temporal dynamics. Modeling human reactions to such dynamic spatial audio is important for character animation, human behavior analysis, and robotics. As of yet, existing audio-driven motion generation methods primarily focus on non-spatial or static acoustic conditions, leaving this important problem largely underexplored. To bridge this gap, we formulate the novel task of dynamic spatial-audio-driven human motion g
benchmark - arxiv:2610.24520 · cs.AIHABIT-Bench: Evaluating Habit Induction from Longitudinal Weak Evidence in Agent MemoryJunming Zhang, Rubin Wei, Xu Cheng, Hao Sun +5
In long-term interactions with large language model (LLM) agents, users reveal habits through their choices and actions. These habits capture implicit user information that agents must infer across conversations. However, individual observations may provide only partial evidence about a habit or when it applies. Retaining past interactions alone does not ensure that agents can infer habits and apply them appropriately. We introduce HABIT-Bench to evaluate how well agent memory systems infer habits from weak evidence and apply them to later requests. We construct synthetic user histories from t
memoryagent memoryagent - arxiv:2610.24519 · cs.AIAxiomatic World Modeling for Physics ReasoningXinye Yang, Zhenyang Liu, Yuxuan Wang, Yuanyuan Lei
Recent research-level physics-reasoning benchmarks for LLMs reveal a failure mode that lies not in knowledge or execution, but in modeling the world being reasoned about. Models often silently alter given quantities, fabricate laws, or inject unstated assumptions to force a problem into familiar patterns, and then reason inside this corrupted world. We define this premise-level divergence as modeling drift: the gap between a model's internal axiomatic world model (AWM) and the true physical world specified in the problem. Although RLVR and PRM have yielded substantial gains in mathematics and
world modelbenchmark - arxiv:2610.24515 · cs.LGProbePO: Sparse Probes, Redistributed Credit for Critic-Free LLM Reinforcement LearningDiyi Hu, Ruizhong Qiu, Lin Qiu, Hanghang Tong +1
Reinforcement learning from verifiable rewards assigns one terminal reward to an entire reasoning trajectory, obscuring how individual reasoning steps contributed to the outcome. GRPO broadcasts a group-normalized advantage to every response token, while finer-grained approaches often rely on dense auxiliary rollouts or branching schemes that can also alter the data used for policy updates. We cast the missing middle as budgeted measurement: with a limited rollout budget, which partial solutions should be evaluated, and how should their value estimates refine credit? ProbePO selects partial-so
benchmark - arxiv:2610.24512 · cs.ROContactBench 1.0: A Benchmark for Everyday Contact-Rich Robotic ManipulationJingyi Xiang, Yixuan Wang, Jeannette Bohg, Wenzhen Yuan
Contact-rich manipulation is essential to everyday activities, yet evaluating robots' ability to perform these activities remains challenging because of their diverse contact behaviors and the difficulty of reproducing tasks across laboratories. Drawing on observations of everyday human manipulation, we identify eight recurring skill categories that provide a practical framework for studying these capabilities. We introduce ContactBench 1.0, an extensible benchmark that instantiates these skills through 36 reference tasks and more than 40 object assets. The assets capture task-relevant geometr
manipulationtactilebenchmark - arxiv:2610.24511 · cs.LGTopoCompile: Molecular Generation with Exact Global Count ConstraintsXiang Wu, Rong-Hua Li, Quanqing Xu, Kangfei Zhao +1
Molecular generation designs novel molecules with desired structures and properties. Global count constraints specify molecular size, elemental composition, and ring content without prescribing detailed structure. Property-conditioned generators steer generation toward such targets but do not guarantee exact satisfaction. We introduce TopoCompile, which combines an autoregressive component-tree generator with a dynamic-programming compiler for supported component-additive integer counts. TopoCompile represents molecules as trees of intact ring systems and individual atoms, making each supporte
benchmark - arxiv:2610.24508 · cs.LGBEM: Learning Bregman Divergences for Concept Erasure in Flow Matching ModelsMohamed-Hicham Leghettas, Markus Püschel
Concept erasure removes harmful, private, or copyrighted concepts from text-to-image models while preserving their other abilities. Leading models such as FLUX and Stable Diffusion 3 are trained with flow matching, and erasure methods for them fine-tune the model toward a guided teacher's velocities under a Euclidean loss. Yet the squared distance is only one member of a larger family: Bregman divergences are exactly the losses under which flow matching recovers the correct velocity field, and each member shapes differently what is erased and what is preserved. We introduce BEM (Bregman erasur
evaluation protocol - arxiv:2610.24507 · cs.ROGazeAct: A Human-Gaze-Annotated Manipulation Dataset for Attention-Regularized PoliciesAnupam Pani, Wang Yuan, Qifan Wang, Yanchao Yang
Vision-language-action (VLA) policies must ground language instructions in the right objects, contact points, and target locations for reliable manipulation. Human gaze provides a natural signal for this grounding, but existing robot datasets rarely pair real human gaze with the multi-camera observations used by modern policies. We introduce GazeAct, a gaze-equipped real-robot manipulation dataset and training framework for multi-view VLA policies. Using screen-based teleoperation, we record gaze directly on the robot camera feed viewed by the operator, yielding 12000 demonstrations across 40
vision-language-actionvlamanipulationteleoperationlibero - arxiv:2610.24501 · cs.LGHiPPO Functional Memory: Coordinate-Queryable Attention for Irregular Time SeriesZhaodong Guo, Haoyun Feng
Irregular time series require models to answer temporal queries at arbitrary coordinates, yet existing approaches typically retain observation-dependent representations that must be repeatedly resolved for each query. We introduce HiPPO Functional Memory (HiPPO-FM), a functional-memory attention operator that compiles an irregular history into a reusable coordinate-addressable state. Instead of storing observation-indexed tokens or latent trajectories, HiPPO-FM writes observations into a fixed-rank functional space and performs direct coordinate evaluation through matched functional geometry.
memorybenchmark - arxiv:2610.24500 · cs.CVInvestigating Length Bias and Robustness in LVLMs for Multiple-Choice Question AnsweringMd. Atabuzzaman, Hani Alomari, Chris Thomas
Large Vision-Language Models (LVLMs) demonstrate strong performance on multiple-choice question answering (MCQA), yet their reliability under biased conditions remains underexplored. We investigate length bias, the systematic preference for longer answer options, through controlled benchmarks spanning fine-grained visual domains. LVLMs consistently favor longer options even when visually inconsistent, with accuracy dropping 2--38\% when correct answers are shorter and increasing 1--70\% when longer. This bias persists across model scales. Through layer-wise mechanistic analysis, we reveal that
benchmark - arxiv:2610.24499 · cs.LGTideKV: Delayed KV Eviction via Layer-Wise Chunked Prefill for Memory-Efficient Long-Context LLM InferenceYu Wang, Zhu Wang, Ning Hu, Yifan Guo +3
Long-context large language model (LLM) inference is important for modern intelligent applications but remains constrained by peak memory during prompt prefill. Existing KV cache compression reduces decoding memory but typically requires full-prompt prefill to preserve global context, retaining a high construction peak. Online chunk-level eviction lowers memory earlier but selects before each layer has observed the complete prompt, potentially degrading cache construction and decoding. We present TideKV, an on-GPU cache-construction architecture using layer-major delayed eviction. TideKV proce
memorylong-context - arxiv:2610.24494 · cs.AIConCI: A Controlled Counterfactual Benchmark for Contextual Integrity in LLMsJuhyeon Park, Sumin Yu, Taesup Moon
Contextual integrity (CI) holds that the appropriateness of sharing personal information depends on the context of the information flow, not on the information alone. However, existing CI benchmarks often use independent scenarios that vary multiple CI parameters simultaneously and are dominated by inappropriate flows, making it unclear whether correct predictions reflect sensitivity to context or a general refusal tendency. We present ConCI, a controlled-counterfactual benchmark in which an anchor flow is paired with two counterfactuals that each modify a single CI parameter. The resulting tr
benchmark - arxiv:2610.24491 · cs.LGVBO-MI: A Differentiable Black-Box Optimization Framework via Variational Mutual Information EstimationFarhad Mirkarimi
Many real-world tasks require optimizing expensive black-box functions accessible only through noisy evaluations, a setting commonly addressed with Bayesian optimization. While Gaussian processes(GPs) and Bayesian neural networks have driven significant progress, GP-based methods scale cubically with observations, and BNN-based approaches remain burdened by expensive posterior sampling such as Hamiltonian Monte Carlo, and inner-loop acquisition optimization in high-dimensional setting. In this work, we propose VBO-MI, a fully gradient-based BO framework that requires no assumptions on a Gaussi
benchmark - arxiv:2610.24485 · cs.ROTouch Where It Matters: Learning Local Tactile Policies for Fine-Grained Gripper ControlMingyuan Ba, Yanan Zhou, James Zhao, Weiming Zhi
Visual imitation policies can organise task motion, yet contact-rich object use demands gripper adjustments beyond coarse opening and closing. We study fine-grained gripper control as a local feedback problem within ongoing task execution. Our approach assigns tactile feedback a dedicated control role during contact-rich task stages. We introduce ContactPolicy, a learned local controller that takes over the gripper while a visual base policy continues to control the arm. ContactPolicy combines tactile and force histories with current robot state to jointly predict gripper increments and their
manipulationtactilegripper - arxiv:2610.24483 · cs.LGWrite Once, Check Many: Reusable Ledgers for Post-Hoc Policy Compliance Auditing of Agent TrajectoriesAshwin Nagarajan, Gursimran Panesar, Alicia Yang, Mona T. Diab +2
Auditing LLM agents against policies is critical to ensure safety but requires rereading long trajectories once per policy. Can we instead record each trajectory once and reuse that record for policies not known in advance? We propose write once, check many, a framework in which a writer model converts each agent trajectory into a policy-independent ledger of the agent's actions and outcomes. A reader then checks each new policy against the ledger alone, without access to the original trajectory. On 900 trajectories from SHADE-Arena, $τ$-bench, and $τ^2$-bench, we show that a 12B reader of nar
agentllm agent - arxiv:2610.24478 · cs.CLBeyond Recall Saturation: Scaling Hard-Negative Reranker Training for Industrial Legal RetrievalYunhan Li, Mingjie Xie, Zeyang Shi, Gengshen Wu +1
Industrial retrieval systems often improve quality by enlarging embedding candidate pools or adopting complex training pipelines. We show that, under a sufficiently large but practical candidate pool, embedding recall exhibits diminishing returns, motivating a closer examination of reranker quality. On Chinese legal retrieval benchmarks, embedding top-300 reaches a high-recall operating region, while relevant evidence may still be ranked below the output cutoff. We compare hard-negative strategies for reranker training, including embedding-mined, self-mined, and dynamic-negative variants, unde
benchmarkeval protocol - arxiv:2610.24477 · cs.LGPrivate Evaluation of High-Dimensional FunctionsMohammad Afzali, Hassan Ashtiani, Adam Smith
We study differentially private black-box evaluation of arbitrary high-dimensional functions. Given a dataset $S$ and query access to a function $f$---whose sensitivity may be unknown or unbounded---the goal is to privately output an approximation to $f(S)$. The black-box evaluator can only query the value of $f$ over a down neighborhood of $S$ (i.e., those subsets of $S$ that are obtained by removing a bounded number of elements from $S$). We design generic black-box evaluators for both pure and approximate differential privacy through a reduction to high-dimensional private quasi-concave opt
evaluator - arxiv:2610.24474 · cs.CLBen-Misandry: A Bengali Dataset and Late-Fusion Transformer Framework for Detecting Anti-Male Gender HateMd Mehedi Hasan Nipu, Chinmoy Mitra, Tarannum Ahmed Nowshin, Israt Moyeen Noumi
Gender based hate speech research in Bengali has focused almost exclusively on misogyny, leaving anti male hate, or misandry, entirely unstudied for this language and creating a systematic blind spot in moderation systems trained to catch attacks on women but not on men. We introduce Ben-Misandry, the first Bengali misandry detection dataset: 5,000 comments across six taxonomy classes, including a novel category for mocking male emotional vulnerability with no direct analogue in existing misogyny taxonomies. The corpus is taxonomy-controlled synthetic data generated with GPT-4.1, spot-checked
benchmark - arxiv:2610.24471 · cs.AIPluralistic Alignment in Language Models: Objectives, Human References, and Preference RepresentationsMeng-Chen Wu, Umesh Bodhwani
Pluralistic alignment aims to represent meaningful variation in human preferences rather than collapse heterogeneous judgments into a single target. Yet existing work often mixes three distinct questions: what pluralistic behavior a model should exhibit, whose preferences define the reference, and what evidence is sufficient to demonstrate success. We organize the literature around a two-layer taxonomy. Layer 1 distinguishes representing a range of human-supported viewpoints (Overton Pluralism), adapting faithfully to a specified perspective or preference condition (Steerable Pluralism), and m
evaluation framework - arxiv:2610.24467 · cs.CVUnifying Amplitude-Phase Coupling and Reliable Endogenous Priors for Low-Light Image EnhancementTongshun Zhang, Zixuan Zhong, Pingping Liu, Jin Kuang +3
Frequency-domain methods have shown strong potential for low-light image enhancement (LLIE) because they can effectively regulate global illumination. However, most existing methods assume that amplitude and phase can be restored independently, while overlooking their inherent spectral coupling. This artificial decoupling often leads to information inconsistency and structural distortions. To address this issue, we propose a frequency-spatial framework that jointly models spectral interaction and reliable prior guidance. We first empirically analyze the coupling between amplitude and phase and
benchmark - arxiv:2610.24464 · cs.AILearning to Compose: Recursive Self-Improvement of Shared Executable Stacks for Zero-Shot Long-Horizon Robot ManipulationYangzhi Chen, Feiran Liu, Zishuo Wang, Zijian Jin +3
Individually competent robot skills can fail when composed: tools may lack a required capability, completion signals may misrepresent physical progress, and one skill may leave a state that violates its successor's prerequisites. We introduce Recursia, a recursive self-improvement framework that treats the shared executable stack, rather than individual task programs, as the persistent unit of improvement. The stack integrates tools, skills with execution contracts, handoffs, and verifiers. Source-side execution traces guide failure attribution and the generation of missing tools or repairs to
manipulationliberofrankaself-improvement - arxiv:2610.24462 · cs.CVO-SCD++: Replay-Free Online Scene Change Detection via Recursive Evidence FusionChamuditha Jayanga Galappaththige, Jason Lai, Lloyd Windrim, Timothy Patten +3
Online scene change detection requires an agent to detect what has changed in a previously mapped environment while it is still on site, observing from unconstrained viewpoints and without access to future observations. This causal setting is strictly harder than its offline counterpart, and online methods have consistently trailed offline accuracy. O-SCD closed that gap by treating change as a persistent property of a 3D Gaussian splatting map of the reference scene, surpassing the strongest offline methods in change detection accuracy while operating at over 10 frames per second. O-SCD optim
agent - arxiv:2610.24461 · cs.LGMeasuring Empowerment from Offline Data for Intrinsically Motivated ExplorationIshir Garg, Vivek Myers, Qiyang Li, Catherine Ji +2
Large, unlabeled datasets of prior experience have become increasingly available for agentic tasks in domains such as robotics and coding. A common recipe for these tasks is to train a policy using the prior data, and then further improve it with online rollouts. But while much work has studied how to use both the offline and online data for policy extraction, far less work has been done to understand how the prior data can be used to efficiently \textit{explore} during the online learning phase. Previous work has attempted to formalize this idea by defining the \textit{empowerment}, a measure
agentagenticonline learning - arxiv:2610.24458 · cs.LGGenerative Retrieval for Reverse Target Fishing with a Proteome-Scale BenchmarkZhichao Deng, Haichuan Tan, Xin Hong, Huaizhi Wang +3
Reverse target fishing identifies protein targets for a query molecule and is central to drug discovery and repurposing. Discriminative approaches adapted from virtual screening face sparse, long-tailed associations and missing-not-at-random measurements: treating unrecorded pairs as negatives can suppress genuine targets. We reformulate the task as generative retrieval and introduce GenFish. Residual vector quantization converts frozen protein language model features into shared semantic identifiers. Given a molecular fingerprint, GenFish autoregressively models codes conditioned on the query
benchmark - arxiv:2610.24454 · cs.AIMELD: Recursive Self-Improvement via Hierarchical Coordination for Multi-Agent Scientific DiscoverySizhe Tang, Zuyuan Zhang, Zeyu Fang, Tian Lan
Multi-agent scientific discovery with large language models raises a coupled learning problem: research assignments determine both which experiments are pursued and the experiences from which agents adapt. We introduce MELD, a unified hierarchical test-time training framework that jointly adapts a team-level Coordinator policy and agent-level individual Scientist policies within a single discovery process. The Coordinator generates a joint portfolio of search branches and research directives over a shared derivation tree, while Scientists develop hypotheses and execution plans implemented by a
multi-agentself-improvement - arxiv:2610.24453 · cs.AIHippo: Training-Free Agent Self-Evolution via Hippocampal ReplayYutong Song, Jiang Wu, Yuxuan Ren, Tony Qin +1
Motivated by recursive self-improvement (RSI), we study how LLM agents can learn through external memory without updating model parameters. Extracting useful experience from a single attempt is difficult: flawed reasoning can produce success, while poor execution can obscure a sound strategy. Preserving its value over time is also difficult: our accumulation baseline without consolidation initially improves but eventually performs below a no-memory baseline. Our key idea is to compare attempts at the same task to identify reusable decisions and refine when they apply. We introduce Hippo, a mem
memoryexternal memoryagentllm agentself-improvement - arxiv:2610.24452 · cs.CVHumanSkill-4D: Training-Free Language Querying of Animatable Gaussian Avatars in Space and Timeunsu Kim, Naeun Kim, Dongyoon Han, Seungryul Baek
We introduce avatar referring segmentation, which selects regions of animatable Gaussian avatars from language queries. A query can also specify a pose or motion to select matching intervals. This lets users interact with avatars for local editing and motion inspection. However, making these selections with existing language Gaussian splatting methods poses two challenges: adjacent parts and garments are hard to separate and may be referred to indirectly (region ambiguity), while fine-grained motions can look alike in RGB (motion ambiguity). We address these challenges with HumanSkill-4D, a tr
memorysemantic memorybenchmark - arxiv:2610.24447 · cs.CVDo 3D Large Language Models Really Use 3D Evidence?Yifei Zhao, Fanyu Zhao, Han Zhang, Yixuan Lin +5
Recent 3D Large Language Models (3D-LLMs) have shown strong performance on scene question answering with extensive post-training. Yet, existing benchmarks rarely evaluate how 3D-LLMs respond to referential ambiguity, insufficient scene evidence, or false presuppositions, obscuring whether their responses are grounded in 3D evidence. In this work, we introduce UNICORNS, a benchmark for evaluating the reliability of 3D-LLM responses across four evidence conditions. UNICORNS contains 11,387 questions across 557 ScanNet scans, covering eight challenge categories alongside answerable controls. A re
post-trainingbenchmark - arxiv:2610.24436 · cs.CVBodyBridge: Grounded Body Layers for Sim-to-Real Translation of Vulnerable Road UsersMark Azer, Fabian B. Flohr
Simulation-to-real driving video diffusion models can translate photorealistic scenes, but they typically represent pedestrians and cyclists, vulnerable road users (VRUs), as generic object classes. This is insufficient for vulnerable road users whose safety-relevant appearance depends on articulated pose and identity. We present BodyBridge and Grounded Body Layers (GBL), a simulator-derived, pixel-aligned representation that combines scene geometry, canonical per-pixel body correspondence, and exact VRU masks. GBL provides explicit articulation to the video diffusion model, emphasizes VRU reg
sim-to-realpost-training - arxiv:2610.24431 · cs.MALow-Bit Distributed Constrained Optimization with Application to Optimal Scheduling and AllocationMohammadreza Doostmohammadian, Sergio Pequito
This paper proposes a distributed resource allocation algorithm in which each agent communicates only the sign of its local cost gradient, a single bit per directed link per iteration. The algorithm is based on Laplacian-gradient dynamics and preserves the equality (demand) constraint at every iterate, guaranteeing anytime feasibility irrespective of termination time. First, we provide a complete Lyapunov-based convergence analysis for the fixed-step-size regime: the iterates are shown to enter, in at most O(1/\varepsilon^2) iterations, a neighborhood of the optimizer whose radius scales as O(
agent - arxiv:2610.24428 · cs.AIWhen Opposite Premises Translate Identically: A Source-Conditioned Diagnostic for Multilingual Commonsense ItemsShuai Wang, Rui Hu
Multilingual reasoning comparisons depend on translated premises preserving the relations in their English sources. Automatic translation scores can rank source-target pairs, but an acceptance decision alone cannot certify a counterfactual benchmark field. We test a source-contrast diagnostic for ACCORD premises: reverse the English polarity, translate both arms independently, count identical target outputs, and score each arm against its own source. Across 71 matched question identities at two containing-item depths, five target languages, and three translation checkpoints, Swahili has 63, 11
benchmark - arxiv:2610.24424 · cs.CLToward a Universal Local Speech Feature Extractor Through DistillationJessica Yang, Marcelo Beramendi Caballero, Ju-Chieh Chou, Karen Livescu
Self-supervised speech models commonly consist of a set of convolutional neural network (CNN) layers used as a local feature extractor, followed by transformer layers. The CNN layers are computationally demanding, and are also similar across models and resemble the much simpler mel spectral features. We hypothesize that the CNN layers can be replaced by a simple, universal model. To test this hypothesis, we learn a two-layer feature extractor via distillation from one or more teacher models (HuBERT, data2vec, and WavLM). Experiments on SUPERB(-SG) benchmark tasks show that our model largely re
benchmark - arxiv:2610.24411 · cs.CLAsk Less, Ask Better: Allocating User Attention in Proactive LLM Agents with Interruption BudgetsXuanming Zhang, Amir H. Rezaeian, Miguel Ballesteros, Weiyi Sun +5
LLM agents can ask clarifying questions to complete tasks, but answering each question demands users' time and effort. Existing evaluations typically measure this burden by counting questions, treating necessary and avoidable questions equally. We propose an evaluation protocol that distinguishes how often agents ask from how well they ask. An interruption budget limits the number of questions, while a user-attention cost charges extra for questions that are hard to read or not worth asking, as rated by a human-validated LLM judge. Beyond question counts, this cost captures perceived effort: o
memoryllm agentevaluation protocol - arxiv:2610.24409 · cs.AILong Listening Thoughts: Eliciting Open Auditory Reasoning with Deliberative Perception and Cognitive RefinementJaeyeon Kim, Chao-Han Huck Yang, Luoyi Zhang, Chan-Jan Hsu +4
We introduce Long Listening Thoughts (LLT), a framework for constructing post-training datasets that improves audio reasoning by pairing deliberate perceptual grounding with explicit cognitive behaviors. On audio understanding and reasoning benchmarks, large audio-language models (LALMs) are limited primarily by perceptual errors, yet existing post-training datasets rely on audio descriptions of limited diversity. Through a four-stage pipeline, LLT (1) builds multi-aspect dense audio descriptions spanning semantic, acoustic, and temporal information, (2) generates more challenging questions, (
post-trainingbenchmark - arxiv:2610.24403 · cs.LGPrototype-Guided Optimization of Quantum Data Embeddings with Linear CostSeokhoon Jeong, Daniel K. Park
In supervised quantum machine learning with classical data, classification performance critically depends on the geometry induced by quantum data embedding. Existing data-adaptive approaches rely on pairwise fidelity evaluations between samples, which scale as $O(B^2)$ for $B$ samples and create a computational bottleneck in iterative embedding optimization. To address this issue, we propose a prototype-guided framework for efficient data-adaptive quantum embedding. Using class-prototype separation and sample-to-prototype compactness, we derive a measurable lower certificate for class-ensemble
benchmark - arxiv:2610.24400 · cs.CVSpatons: Spatiotonally Localized Building Blocks for Efficient XR PredistortionMohammadjavad Maheronnaghsh, Sushant Kondguli, Steve Mann, Andreas Moshovos
We present Spatons, a family of compositional, spatially localized neural operators for computationally efficient XR (extended reality) predistortion and distortion compensation. XR displays (AR, VR, and MR) suffer from spatial (e.g., geometric) and tonal (e.g., colour) distortions. Frames can be pre-distorted so that the projected image looks as intended. Distortions vary with manufacturing tolerances and projection systems, motivating programmable compensation. Untethered, battery-operated XR systems also require low energy and latency. A Spaton is a hybrid patch-to-pixel block combining tin
memory - arxiv:2610.24395 · cs.LGMSCR: Jointly Balancing Utilization Across Modalities and Discovering Synergistic InformationXinyu Chen, Liangjian Wen, Jiang Duan, Dongkai Wang +6
In multimodal learning, modality imbalance often lets one modality dominate optimization and keeps the model from fully exploiting cross-modal interactions. Balance-oriented methods such as the Multimodal Competition Regularizer (MCR) perform well by alleviating this imbalance during training. These dynamic balancing methods, however, mainly adjust modality contributions and do not consider how imbalance affects the formation of synergistic information during fusion. Synergy is a major source of multimodal advantage, because it is predictive gain that cannot be recovered from unimodal evidence
benchmark - arxiv:2610.24392 · cs.ROGen2Dex: Learning Dexterous Manipulation from Simulation-Initialized Generated Video DemonstrationsXiaowen Qiu, Yian Wang, Haoxiang You, Chuang Gan
Human video demonstrations provide rich manipulation strategies for dexterous manipulation learning. We study a setting in which a fully specified simulated scene is used to generate videos that specify the demonstrated manipulation behavior. This setup provides access to scene information such as object geometry, articulation, and camera parameters, while allowing direct control over task and scene configuration. However, generated videos may contain geometric distortions, inaccurate contacts, and other physical inconsistencies, making the recovered motion difficult to execute directly on a r
manipulationdexterous - arxiv:2610.24390 · cs.AITrust the Edge, Doubt the Arrow: Auditing Validation Signals for Text-Mined Clinical Causal Knowledge GraphsSu Lan, Shicheng Zhou, Yiqi Wang, Yanming Zhu +2
Clinical causal retrieval-augmented generation systems increasingly use knowledge graphs (KGs) mined from notes, yet extracted assertions remain candidates that require audit for patient-level EHR support and direction. We construct a directed candidate graph from 328,839 MIMIC-IV discharge summaries using LLM extraction and UMLS normalization, and evaluate three signals against SemMedDB: within-admission co-occurrence lift, cross-admission diagnosis-coding temporal precedence, and literature corroboration. Lift filters unsupported co-mentions and provides useful EHR support, but not causal pr
retrieval-augmentedknowledge graphbenchmark - arxiv:2610.24386 · cs.LGLearning volatility-adaptive forgetting via GARCH-modulated LSTM for stock price forecastingXin-Rui Lu, Da-Qing Zhang
Stock price forecasting is hindered by the heteroscedastic nature of financial time series, in which turbulent and tranquil regimes alternate. Although the Long Short-Term Memory (LSTM) network is widely used for this task, its gates are oblivious to such volatility variation, so memory retention remains fixed regardless of market conditions. A common remedy is to supply volatility estimates as an additional input feature, yet this leaves the internal memory dynamics untouched. Instead, volatility can serve as a direct regulatory signal over the gate activation itself. This paper presents the
memory - arxiv:2610.24385 · cs.AIBridging Exploration and Persona: Information-Directed Dueling for Test-Time Agent PersonalizationJie Mao, Pingchen Lu, Xiangyi Wang, Hengquan Huang +3
Effective test-time personalization requires LLM agents to adapt to query-dependent user preferences from lightweight interaction. We study persona selection with binary comparisons between candidate responses as the online signal. We present ID-TAP, a scalable, IDS-inspired neural dueling-bandit framework for large discrete persona spaces. ID-TAP combines a query-conditioned neural utility model with a lightweight Bradley-Terry explorer, and transfers comparison evidence across related queries through similarity-weighted warm starts. The neural model shares statistical strength across queries
agentllm agentbenchmark - arxiv:2610.24384 · cs.LGA decomposition-integration framework coupling singular spectrum analysis with a volatility-adaptive LSTM for stock price forecastingXin-Rui Lu, Da-Qing Zhang
Financial time series exhibit multi-scale, high-noise, and non-stationary volatility dynamics. Decomposition--integration frameworks built on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) have advanced this field, yet existing CEEMDAN hybrids remain unstable: high-frequency mode mixing contaminates downstream predictors, and the heterogeneous volatility of decomposed components is not matched by any volatility-adaptive mechanism. To address these gaps, this paper proposes CEEMDAN-SE-SSA-VMGL. CEEMDAN adaptively decomposes the original series, and sample-entropy-d
memorybenchmark - arxiv:2610.24383 · cs.AIIF-MAP: Influence-Function-Inspired Agent Prioritization in Multi-Agent SystemsJie Mao, Pingchen Lu, Zikun Qu, Jiahang Sun +5
Improving every agent in an LLM-based multi-agent system can be prohibitively costly, making it important to identify which agent should be prioritized to achieve the greatest system-level gain with limited resources. Failure attribution can inform this decision, but an agent's responsibility for failures does not directly reveal the system-level gains achievable by improving it. We therefore introduce IF-MAP, an influence-function-inspired method for identifying agents with high potential for system-level improvement. For each agent, IF-MAP measures end-to-end loss changes under controlled se
agentmulti-agentagent system - arxiv:2610.24382 · cs.AIIs This Your Agent? Identity Watermarking for Agent Systems in Dynamic EnvironmentsYuwei Zhong, Yu Cheng, Jiuan Zhou, Weizhi Liu +4
Agent systems deployed in dynamic environments continuously update their memories, skills, and runtime configurations, leading to evolving trajectories and challenges for persistent identity authentication. Existing text and agent watermarking methods struggle to achieve stable watermark embedding under dynamic environments, while embedded watermarks may become difficult to recover due to environmental changes. This paper proposes DASMark, an identity watermarking framework for dynamic agent systems. DASMark introduces Action-Preserving Embedding (APE), which learns semantically similar but di
agentagent system - arxiv:2610.24380 · cs.AITrain Locally, Learn Together: Distributed Multi-Agent Training via ADMMYiwei Chen, Changsheng Wang, Yicheng Lang, Qing Qu +2
Multi-agent systems (MAS) can outperform individual language-model agents by combining specialized roles, diverse reasoning processes, and collaborative decision-making. However, collaborative MAS training is typically centralized, requiring a common trainer to aggregate role-local trajectories and learning signals and maintain model and optimization states. This motivates a distributed multi-agent learning paradigm in which agents improve system-level capabilities through role-local learning and decentralized communication, paving the way toward scalable, resilient, and self-organizing MAS. T
multi-agentagent system - arxiv:2610.24379 · cs.CVDefending Against Dormant Poisoning Attacks Across Language and Multimodal AgentsWonjun Lee, Vaidehi Patil, Zaid Khan, Yue Zhang +3
Recent work shows that attackers can implant dormant malicious behaviors that remain hidden at release but emerge after benign fine-tuning. As foundation models are used as agents, whether dormant poisoning extends to harmful agent actions and multimodal settings remains unclear. We show two agentic risks: (1) harmful behaviors activated by benign fine-tuning can extend to unseen actions; (2) multiple harmful behaviors can coexist and be jointly activated by benign downstream fine-tuning. Existing defenses often rely on external safe responses, which are hard to obtain for agent actions, or a
agentagentic - arxiv:2610.24378 · cs.CVPACE-Mem: Prefix-Aware Complementary Evidence for Multimodal AgentsShidu Ren, Rong-cheng Tu, Hang Zhou, Xiao Luo
As multimodal agents accumulate experience, their long-term memories contain many similar observations of recurring scenes, actions, and entities. Answer quality depends on the evidence retrieved from these memories into the model's context. However, query-relevant memories may contain overlapping information, and a fixed retrieval depth may return too little or too much evidence for a given question. To address these challenges, we propose PACE-Mem, Prefix-Aware Complementary Evidence, which constructs a complementary evidence sequence and determines its length from the same selection traject
memoryagent memoryagentagent frameworkbenchmark - arxiv:2610.24372 · cs.AIBeyond Next-Screen Prediction: The Decision Value of GUI World ModelsSizhe Tang, Hanhan Zhou, Jeffrey Svajlenko, Rashmi Gangadharaiah
GUI world models are increasingly used to improve action selection by predicting the consequences of candidate actions, but it is unclear how much of the resulting gain comes from prediction itself rather than from the evaluator that interprets those predictions. We study this attribution problem through controlled interventions on GUI action selection. With candidates and predictions held fixed, lookahead gains can change with how the prediction is presented to the evaluator, so end-to-end gains cannot be attributed to prediction alone. We then introduce Direct Action Ranking (DAR), which lea
world modelevaluator - arxiv:2610.24370 · cs.LGRevisiting Smoothed Online Learning: Competitive Ratio, Dynamic Regret with Switching Cost, and Best of Both WorldsLijun Zhang, Linzhe He, Wei Jiang, Shiyin Lu
We revisit smoothed online learning and develop algorithms with improved competitive ratios, optimal dynamic regret with switching cost, and best-of-both-worlds guarantees. Surprisingly, minimizing the hitting cost alone is $\max(1,\frac{2}α)$-competitive for $α$-polyhedral functions under the $\ell_2$-norm switching cost and $1+\frac{4}λ$-competitive for $λ$-quadratic growth functions under the squared $\ell_2$-norm switching cost, significantly improving existing guarantees. The polyhedral bound is tight for this naive algorithm, and we establish a matching lower bound for $α\geq1$. For conv
online learning - arxiv:2610.24369 · cs.AITactix: Towards a Programmable Strategy Interface for Flexible SAT SolvingHaoze Wu, André Schidler, Armin Biere, Nina Narodytska
Modern SAT solvers combine a rich repertoire of proof rules with sophisticated strategies that determine when and how to apply them. Often, the internal implementations are tightly coupled to particular strategies, making it inconvenient to define new strategies without changing solver internals. We present Tactix, a proof-producing sequential CDCL solver that incorporates a wide range of techniques in state-of-the-art solvers with a clean separation of proof rules from solving strategies. This separation enables a public interface through which users can invoke particular techniques, observe
benchmark - arxiv:2610.24366 · cs.AIA Two-Stage Hybrid Framework for Inferring Air Trips from Mobile Location-based Service (LBS) Data under Incomplete Airport SightingsHsuan-Chih Wang, Xianfeng Terry Yang
Mobile device location-based service (LBS) data provide large-scale spatiotemporal observations that can help airport planners understand air travelers' mobility patterns. However, airport sightings are often missing before or after flights, making strictly rule-based detection incomplete. This study develops a two-stage framework to infer U.S. domestic air trips from large-scale LBS data under incomplete airport sightings. Stage 1 uses heuristic rules to identify high-confidence air trips whose endpoints are observed at two distinct airports. Stage 2 addresses latent air trips with a probabil
benchmark - arxiv:2610.24365 · cs.CLFlashMap: Flow Map Speculative DecodingJeff Shen, Zhejian Zhou, Ryan Lee, Jonathan May +1
Speculative decoding accelerates large language model (LLM) inference by using a lightweight drafter to propose a block of tokens that the target model verifies in one parallel pass. Autoregressive drafters generate this block one token at a time, whereas block diffusion drafters such as DFlash generate it in one pass, making drafting faster. However, these diffusion drafters predict each token independently, ignoring dependencies between tokens in the block. We introduce FlashMap, a continuous flow drafter that conditions a flow map language model on the target model's hidden states and gener
benchmark - arxiv:2610.24359 · cs.LGCan Agents Keep Secrets They Were Never Told? Benchmarking Latent Privacy Risks Across Sessions with MemJoinQishun Yang, Shu Yang, Difei Xu, Di Wang +1
To support sustained interactions and personalized assistance, agents use long-term memory to retain and combine user information across sessions. Reusing such memories can lead an agent to state a sensitive relation while completing an ordinary task that does not require it. The relation to protect and the combination of clues that reveal it may remain hidden at memory write time, when later records and tasks are still unknown. In this paper, we introduce MemJoin, a privacy benchmark of 288 synthetic analysis cases in which the complete history entails a sensitive relation, but every individu
memoryagentbenchmarkevaluator - arxiv:2610.24355 · cs.AIKoi: Evidence-Guided, Self-Calibrating Planning for Multi-Cloud, Heterogeneous LLM Inference FleetsMankeerat Sidhu, Gangmuk Lim, Hetarth Chopra, Maximillian Eisenberg +9
Deploying a large language model (LLM) requires choosing the right accelerator, parallelization strategy, engine configuration, and replica count amongst ~140 other variables across clouds, regions, and markets. These decisions are intertwined, and the best cluster-wide plan continually changes as jobs contend for shared capacity on shifting workloads. We present Koi, a self-calibrating cluster planner for heterogeneous, multi-cloud fleets. Koi uses a Recursive Language Model (RLM) hierarchy to propose hypotheses and reason over deployment while deterministic components predict serving behavio
benchmark - arxiv:2610.24349 · cs.AIFlashRT-PAI: Fast Kernels Are Not Fast Pipelines for Physical AI InferenceLiang Su, David Chen, Jason Li, Kevin Guo +1
Vision-language-action (VLA), video-action (VA), and world action models (WAMs) form the backbone of modern physical AI systems, yet their real-time single-batch inference remains a defining bottleneck. Latency is dominated not by individual kernel computation, but by system-level overhead: thousands of short kernel launches, intermediate data movement across operator boundaries, and redundant recomputation across consecutive decisions. General-purpose graph compilers cannot address these costs, as they optimize individual calls in isolation and lack visibility into cross-call state semantics
vision-language-actionpi0libero - arxiv:2610.24346 · cs.LGMeasuring the Task Value of Precision Upgrades in Closed-Loop AgentsSicheng Lyu, Xinyu Wang, Yu Gu, Lei Ding +2
A low-precision agent can call its full-precision (BF16) model at selected decisions as a fallback. A learned precision router chooses where to call it from the disagreement between the two models and is judged by end-to-end success. An inserted BF16 action, however, can rescue a failing task or damage a succeeding one. We argue that a fallback's value is the number of tasks it saves minus the number it spoils, measured for the intervention that will actually run. From restored states, we run paired branches that cross which model supplies the current action with which model continues, separat
agenttool use - arxiv:2610.24345 · cs.LGThe Certification Exponent: Instance-Optimal Sequential Certification of Policy Improvement from History Dependent LogsRui Wang
Before a new policy replaces a deployed one, we often want a certificate, computed from logged episodes, that it is better. We ask how many episodes such a certificate needs when the logging policy is known but history-dependent, and the error probability must stay below $δ$ in every environment consistent with the logs. The answer is an exponent $K^\star$: the smallest KL divergence from the logged law to a law with the same logged action probabilities under which the target does not improve. No sound certifier stops in fewer than $\log(1/δ)/K^\star$ expected episodes, at any $δ$, and wheneve
memory - arxiv:2610.24329 · cs.AIVERAS: Verification-Guided Adaptive Search for Reliable Operations ResearchMinbo Ma, Honglin Zhang, Bangtao Zhou, Chunguang Yu +4
Advances in reasoning, coding, and tool use enable LLM agents to solve operations research (OR) problems through iterative problem modeling, program construction, and verification. Yet passing self-authored verification does not ensure that the solutions satisfy task constraints and meet optimization quality. This gap can arise from correlated specification errors, where the generated program and its checks share omissions or misinterpretations of task requirements, alongside incomplete verification coverage and ineffective optimization strategies. We propose VERAS, a verification-guided adapt
agentllm agenttool use - arxiv:2610.24320 · cs.AIThe Judge Can Debias Itself: Awareness and Self-Debiasing for Reliable LLM-as-a-JudgeYoubing Yin, Jingyu Wu, Anoop Kumar, Scott Novotney +2
LLM-as-a-Judge systems are increasingly employed for automated evaluation, yet remain susceptible to systematic biases that undermine their reliability. We demonstrate that the judge can effectively debias itself without additional training. We propose two lightweight strategies: bias-aware prompting, which explicitly alerts the judge to potential target biases prior to evaluation, and self-debiasing, in which the judge critiques the initial judgment for bias influence. To broaden the evaluation scope, we introduce three novel benchmarks--authority bias (fabricated citations), confidence bias
benchmark - arxiv:2610.24312 · cs.ROBeyond Human-Derived Motion: Exploring Robot-Specific Execution for Construction HumanoidsMinJun Chang, Taehyung Kim, Yong K. Cho, Francis Baek
Construction productivity has remained a per- sistent challenge, motivating growing interest in humanoid robots that can operate in workspaces and with tools designed for human workers. Realizing this potential requires effective motion control, for which human-derived motion transfer has emerged as a promising approach by capturing and retargeting human demonstrations to humanoid robots. While promising, however, human-derived motion transfer may not fully define how a humanoid should ultimately execute the transferred motion. Once task-relevant motion information has been con- veyed, the hum
humanoid - arxiv:2610.24311 · cs.LGRATION: Towards Quantitative Risk-Based Access Control for AI AgentsYuntao Du, Harsh Chaudhari, Yuetian Chen, Ron Heichman +7
AI agents increasingly interact with external environments to complete complex user tasks. These interactions can lead to unexpected behavior and expose agents to security threats such as indirect prompt injection. Current agent security systems largely protect agents by restricting their actions, which can result in a substantial security-utility trade-off for complex, open-ended tasks. To address this, we propose a quantitative risk-based approach to access control for AI agents that regulates measurable risks accumulated during task execution. As a concrete realization of this approach, we
agentai agentbenchmark - arxiv:2610.24304 · cs.AICME-SpectrumBench: Can LLMs Analyze Condensed Matter Spectral Data?Jin Gene Wong, Anjney Midha, Joseph Mathew Tennyson, Wei-Lin Chiang +2
Large Language Models (LLMs) are presently well assessed by expert-level mathematics and physics benchmarks which focus on theoretical, symbolic reasoning. However, we lack experimental benchmarks that evaluate data analysis in frontier research problems, a capability that future domain-specific foundation models must have. To this end, we present CME-SpectrumBench, an expert-designed benchmark consisting of 190 questions that measures how reliably frontier LLMs analyze research-level spectral scientific data in condensed matter experiments (CME). Using simulated CME spectral data, we prompt m
agentictool usebenchmark - arxiv:2610.24298 · cs.AIWho Speaks, Who Authorizes, Who Acts? Understanding Trust Boundaries in Multi-Agent LLMsJunru Zhu, Xin Chen, Hang Lyu, Jiayue Meng +5
Multi-agent LLM systems exchange messages before acting through tools, but message delivery, signing authority, and tool-side enforcement are separate trust boundaries. We represent them as communication, authorization, and execution structures, and use the authority cut to characterize the smallest independently compromised signer coalition that can construct a policy-valid false-claim proof under stated assumptions. Controlled paired hidden-state studies reveal four distinctions. First, no fixed-message routing contrast across four models meets the prespecified large-effect criterion; when r
multi-agent - arxiv:2610.24294 · cs.AINot Every User Says It All: Training Tool-Using Agents Beyond Ideal Simulated UsersHan Bao, Zheyuan Zhang, Kaiwen Shi, Yanfang Ye
Tool-using agents are trained and evaluated almost exclusively with strong simulated users who state every detail clearly, yet real users are often weaker. We construct weak users by masking sentences of a simulated user's replies, holding the task and the user's knowledge fixed, and cross-evaluate agents trained with strong and weak users on tau2-bench and ToolWOZ. Weak users lower the success of every policy, and training with them does not help: no policy trained with weak users, through fixed masks, mixtures or a curriculum, serves weak users better than the strong-trained policy beyond ru
agent - arxiv:2610.24291 · cs.LGAn important Bottleneck in LLM Optimizers: Principle-Level Local OptimaNian Ran, Tianhe Jia, Zhengyang Mu, Yue Wang +5
Feedback-driven LLM optimizers are increasingly used for evolutionary search, Bayesian optimization, heuristic design, and scientific discovery. Yet rapid initial gains often give way to prolonged plateaus despite substantial remaining evaluation budgets. We investigate \emph{principle-level search collapse}, where an optimizer's search becomes confined to a narrow set of design principles, even though it may continue to generate distinct candidates and achieve small improvements. We assemble a diagnostic benchmark spanning molecular optimization, network sentinel-node selection, and GPU-kerne
benchmark - arxiv:2610.24290 · cs.CLBeyond Clean Accuracy: The Practical Challenges and Benefits of Pixel-Based Language ModelsHaowen Tang, Wei Xia, Raoyuan Zhao, Hinrich Schütze +1
Pixel-based language models process rendered text without relying on a fixed subword vocabulary, offering a promising approach to noisy and multilingual inputs. However, their practical value depends on whether they can be reliably adapted to downstream tasks. To this end, we systematically examine this question through fine-tuning behavior, typographical robustness, and zero-shot crosslingual transfer. Under our tested settings, pixel-based models require longer training and broader labelled-data coverage than text-based models, while delayed learning and learning-rate-dependent failures make
benchmark - arxiv:2610.24289 · cs.AIUnbounded Losses, Unchecked Inputs: Measuring the Safeguards of LLM Options-Trading AgentsMengxiao Wang, Nitesh Saxena
Large language model (LLM) agents now place options orders at retail brokers, and developers publish open-source options agents faster than anyone has examined them. Prior security work on trading agents scored academic equities designs by whether an attacker can flip a buy-or-sell decision. We present the first measurement of the options agents that brokers and developers ship, scored by the safeguards they enforce and by the worst loss their documented order paths can reach. We screened 765 GitHub repositories, broker documentation, and the research literature and found 215 eligible reposito
agent - arxiv:2610.24288 · cs.LGBanglaShield: A Script-Robust Safety Classifier and Benchmark for Bangla Native, Romanized, and Code-Mixed PromptsMohsinul Kabir, Md. Tariquzzaman, Mohammad Rashidujjaman Rifat, Tharindu Ranasinghe +1
Safety classifiers for large language models (LLMs) are typically built and evaluated on standard-script text, yet Bangla, like many lower-resource languages, is routinely written in native script, in Latin-script romanization, and code-mixed with English, Hindi, or Urdu. We introduce BanglaShield, a resource and model for graded prompt-safety classification across these written forms. We consolidate 10,245 Bangla prompts from six sources under a four-level taxonomy (Safe, Offensive, Harmful, Severe) and render each prompt in three aligned forms that share labels and a group-aware split, enabl
benchmark - arxiv:2610.24287 · cs.AISteerSearch: Interactive Search Reasoning with Mid-Response SteeringUdit Thakur, Baohao Liao, Matteo Nulli, Stefan Vasilev +3
Search-augmented language models increasingly serve as long-horizon research agents, yet most open training methods assume a fixed user goal throughout search. We study mid-response steering, where users redirect a model during ongoing reasoning or retrieval. We identify answer-target anchoring, a failure mode in which fixed-query search agents retrieve evidence for the updated request but still answer an earlier question. We introduce SteerSearch, an RL framework that trains on interactive search trajectories with injected interventions to enable context-preserving target reassignment, and St
benchmark - arxiv:2610.24285 · cs.ROACTS: Learning Visuo-Tactile World Models from Human Physical InteractionYuxiang Ma, Weihan Xu, Pengfei Ye, Anna Mei +4
Abstract Action-conditioned world models for contact-rich manipulation require data that captures how contact forms, slides, and releases under varying force. Such data are slow to collect through robot teleoperation, which often lacks force feedback and is constrained by gripper geometry. We ask whether these dynamics can instead be learned from human physical interaction and transferred to a robot. We present ACTS, a visuo-tactile world model trained only on human interaction data. Operators manipulate handheld GelSight sensors while we record third-person and wrist video, tactile images, 6-
manipulationteleoperationtactilegripperworld modelaction-conditioned - arxiv:2610.24282 · cs.MAROAR: Benchmarking LMM Agents under Real-Time Constraints in Multiplayer Action-Shooter GamesDawei Wang, Di Zhao, Jianran Yu, Chengming Zhou +7
Large multimodal models (LMMs) are increasingly deployed as agents that perceive, reason, and act in interactive multi-agent environments. Yet most existing benchmarks for LMM-based agents in multi-agent settings remain turn-based: the environment and other agents pause while an agent performs inference. Such protocols obscure a fundamental challenge of real-time constraints, where the world continues to evolve and delayed decisions may become outdated before execution. Motivated by this limitation, we introduce ROAR, a multiplayer action-shooter game environment for evaluating LMM-based agent
agentmulti-agentbenchmark - arxiv:2610.24278 · cs.LGDataset Inference for Data Provenance and Privacy Auditing in Tabular Foundation ModelsDariush Wahdany, Jesse Cresswell, Naiqing Guan, Atiyeh Ashari Ghomi +2
Tabular foundation models (TFMs) are increasingly deployed through black-box APIs and trained on large collections of real-world tabular datasets. As this training paradigm becomes more widely adopted, private or proprietary datasets may be incorporated into pre-training corpora, yet there are currently no dedicated methods for determining whether a given tabular dataset was used to train a TFM. As a solution, we introduce the first dataset inference method for TFMs, aiming to infer whether a suspect dataset was part of a model's pre-training data. We systematically analyze a broad collection
manipulation - arxiv:2610.24268 · cs.AIFrom SEC Filings to Financial Signals: Evaluating LLMs on Stock Market Reactions and Analyst ForecastsShengkun Wang, Taoran Ji, Linhan Wang, Sattar Mansi +2
Form 10-Q MD&A sections combine long-form management narrative with accounting and forward-looking information. Benchmarks for financial filings commonly evaluate question answering or numerical reasoning, leaving unclear whether outputs from large language models (LLMs) align with market and analyst outcomes. We evaluate prompted, zero-shot outputs from 13 general-purpose and finance-specialized LLMs on 45,964 Form 10-Q filing events from 2001 to 2024, linking MD&A text to cross-sectional fundamental ranks, pre-filing analyst forecasts, and realized abnormal returns. The evaluation covers ret
benchmark - arxiv:2610.24267 · cs.ROHow exposed is work to robotics? Occupational, industry, and geographic measuresSebastian Sartor, Robert Seamans, Neil Thompson
Exposure indices that map technical capabilities onto occupational abilities have become a standard tool for anticipating how AI may reshape work, but existing measures either cover software AI or proxy robotics through patents and installation data that reflect the industrial robots of past decades. We construct the first abilities-based exposure measures for robotics and physical AI, and the first to span occupations, industries, and geographies. Following the framework of Felten et al. (2018, 2021), we build a robotics capability taxonomy of 16 capabilities in eight domains from 78 academic
grasp - arxiv:2610.24264 · cs.AIAM-Bench: A Unified Taxonomy and Evaluation Suite for Agentic MisalignmentEric Zhang, Terry Jingchen Zhang, Chijioke Ugwuanyi, Jerick Shi +2
As LLM-based agents are deployed with increasing autonomy in real-world environments, a critical failure mode emerges: agentic misalignment, where agents choose harmful actions over failure. However, existing safety benchmarks primarily evaluate whether agents comply with explicitly harmful instructions, leaving the conditions that cause agents to choose harmful actions poorly understood. To address this limitation, we propose AM-Bench, a unified taxonomy and 35 scenario evaluation benchmark spanning seven industries for agentic misalignment grounded in the fraud triangle, a well-established c
agenticbenchmark - arxiv:2610.24257 · cs.ROEvaluating Lightweight Contact Feedback in Dual-Operator Collaborative Humanoid TeleoperationAustin Graves, Chuizheng Kong, Yunho Cho, Yuanshao Yang +3
Dual-operator humanoid teleoperation enables collaborative manipulation through two separate robots, but operators lack direct awareness of robot-object contact that is normally available during physical interaction. To address this, we present eXtended Reality Teleoperation (XRT), an extended reality teleoperation framework that returns a robot's fingertip contact through lightweight vibrotactile and/or auditory feedback, and a synchronized dual-humanoid testbed built on it for two operators. Sixteen dyads performed a collaborative crate lift-and-flip under vision only and the three feedback
manipulationhumanoidteleoperationtactile - arxiv:2610.24256 · cs.ROTrajEdit: Real-Time Flow-Matching Trajectory Adaptation via FlowEdit and CBF Safety FilterWondmgezahu Teshome, Arya Honarpisheh, Octavia Camps, Mario Sznaier
Safe, real-time motion planning is especially difficult when a dynamic obstacle actively pursues the ego agent, requiring rapid adaptation as the environment changes. We present TrajEdit, a hierarchical trajectory planner for pursuit-evasion that combines flow matching, trajectory editing, and safety filtering. Rather than regenerating trajectories from scratch at every step, TrajEdit follows the current plan under nominal conditions and adapts its remaining portion through FlowEdit as the pursuer approaches, and falls back to a control barrier function safety filter if the adversarial agent i
agent - arxiv:2610.24255 · cs.LGCortex: A Fixed-Point Theory of Governed Coding Agents - Technical ReportMarius-Constantin Dinu, Florian Zeba
A coding agent combines a large language model (LLM) with a harness that plans, edits, and executes code. We study Cortex, a supervisory layer for pre-execution governance, requirement tracking, and iterative validation and repair. We formalize requirement closure on a finite lattice. For a fixed requirement set and a sound, monotone validation rule with persistent evidence, exhaustive iteration reaches the least fixed point with no more strict increases than there are requirements. Fair consequence scheduling reaches the same closure. A separate stochastic model bounds expected completion tim
agentbenchmark - arxiv:2610.24248 · cs.AIProbing Ambiguity: Adversarial Intent Resolution for Robust Text-to-SQLYoubing Yin, Xujun Peng, Muhammad S. Uddin, Erin Babinsky +2
Text-to-SQL systems typically assume a one-to-one mapping between natural-language queries and correct SQL translations. In practice, however, many queries are inherently ambiguous, admitting multiple valid interpretations that yield distinct outputs. We introduce Adversarial Intent Resolution (AIR), a multi-agent framework that actively probes and resolves such ambiguity without any task-specific training. AIR first constructs competing interpretations, verifies through SQL execution that they produce genuinely different results, and generates targeted clarification questions grounded in the
multi-agentagent frameworkbenchmark - arxiv:2610.24241 · cs.ROTowards Reinforcement Learning for Space Robotics: Fast Training and Reliable Real-World TransferAbhishek Naik, Michael Wu, Michael O'Sullivan, Colin Bellinger +1
This work demonstrates the use of a fast simulation-based training framework to learn control policies using reinforcement learning. The hardware requirements are minimal and the simulation-learned policies transfer reliably to the real world. With this setup, we show that a quadruped robot can learn to traverse diverse real-world terrains reliably---including in a lunar-analog facility---after less than 20 minutes of training on a single consumer-grade GPU. These results provide a glimpse of a future where reinforcement learning is a crucial tool in developing more capable and robust space ro
quadruped - arxiv:2610.24240 · cs.LGExtending Myerson's Optimal Auctions to Correlated Bidders via Neural Network InterpolationMingyu Guo, Jiayuan Liu, Vincent Conitzer
We aim to design revenue-maximizing single-item auctions that are deterministic, strategy-proof, and ex post individually rational --- the quintessential fundamental model for optimal mechanism design. While Myerson's seminal work solved this model for independent bidders, straightforward extensions to correlated settings often result in non-monotonic, and therefore non-strategy-proof allocations. Critically, optimal allocation design for correlated bidders is NP-hard to approximate, rendering theoretical optimality guarantees in this domain computationally intractable. We establish a new stan
benchmark - arxiv:2610.24238 · cs.CVGhost Attack: Using 19th-Century Illusions to Fool 21st-Century BiometricsTyler Fitzsimmons, Adam Czajka
While the Pepper's Ghost illusion is a 19th-century optical technique, we found that this forgotten gimmick exposes vulnerabilities in modern face recognition systems. This paper introduces the 'Ghost Attack': a physical attack that repurposes this classic illusion using low-cost, free-space hardware and a transparent medium to project high-resolution facial data into physical 3D space. We show in the paper that despite relying on centuries-old optical principles, this new attack type simultaneously breaks three components of modern facial recognition systems (FRS): (a) the face matcher, (b) t
benchmark - arxiv:2610.24236 · cs.LGPreserving Reasoning Diversity in Reinforcement Learning with Verifiable Rewards via Adaptive ReplayWen-Zhi Li, Sainyam Galhotra, Claire Cardie
Reinforcement Learning with Verifiable Rewards (RLVR) improves the reasoning capabilities of large language models, but often reduces the diversity of their reasoning behaviors. We study this phenomenon from the perspective of finite-sample learning dynamics and show that limited on-policy rollouts can progressively suppress semantically distinct but equally rewarded reasoning modes. Once a mode becomes sufficiently under-sampled, standard on-policy RLVR rarely recovers it. Motivated by this observation, we propose TAR, a Targeted Adaptive Replay framework that preserves diverse correct reason
memorypost-trainingbenchmark - arxiv:2610.24235 · cs.ROMultimodal Interaction Estimation for Contact-Aware Adaptive Impedance ControlMateusz Jaszczuk, Paris Perdikaris, Nadia Figueroa
With advancements in learning-based policies, robots are increasingly being deployed in human-centric environments, such as manufacturing floors and homes, where safe and precise operation requires compensating for unknown loads while remaining compliant during human interactions. Reliably distinguishing between these interaction types is challenging with a single sensing modality, as force signals cannot separate payloads from human pushes, and vision is susceptible to occlusions. We propose a task-agnostic interaction estimation and adaptive control framework that addresses this challenge by
manipulationfranka - arxiv:2610.24232 · cs.AIAnalyzing Repetition and Interference Effects in Memory-Constrained User SimulatorsCara Su-Yi Leong, Nicholas Tomlin, Tal Linzen
User simulators can help evaluate and train AI systems to interact with humans at scale. To do so, evaluation of user simulators must measure behavioral parity with people. We evaluate whether memory-constrained user simulators are sensitive to two factors known to affect human information retention: repetition of a fact and interference from competing information. We find that some simulators achieve human-like overall accuracy, and replicate the magnitude of repetition and interference effects, but none match people's accuracy on particular questions: while humans misremember facts, models f
benchmark - arxiv:2610.24231 · cs.AILoopHIL: Looping Hardware into LLM-assisted Agents for Wireless Embedded SystemsDhairya Shah, Spanddhana Sara, Savitha Viswanadh Kandala, Ambuj Varshney
LLM agents are widely used for software development and debugging, but their adoption in embedded systems, especially for wireless tasks, remains limited. This is because embedded development requires hardware-in-the-loop testing. Agents cannot directly observe the spectrum, probe pins, or measure power, so they cannot verify wireless link functionality; currently, human engineers perform this verification. We systematically evaluate whether LLM agents can establish wireless links when provided with hardware access. We introduce LoopHIL, a framework that lets agents control a multi-device RF b
llm agent - arxiv:2610.24229 · cs.LGTraining-Free Prompt Selection for Efficient Reinforcement Learning With Partial RolloutsLei Jiang, Honghui Bao, Yue Zhou, Zhengyao Gu +2
Reinforcement learning with verifiable rewards (RLVR) has emerged as a prominent paradigm for training reasoning models, but its computational cost makes efficient data selection increasingly important. Since training examples contribute unequally to learning, selecting a small, informative subset can offer a more efficient alternative to using the entire data pool. The training-free selection method estimates prompt value with hidden states from complete greedy rollouts. However, in our datasets, $26$--$32\%$ of these rollouts fail to terminate within the generation budget. We introduce LIFT,
benchmark - arxiv:2610.24206 · cs.AISWE-Bench Pro V2: A Cleaner, Harder-to-Game Coding BenchmarkSoham Dan, Jeff Da, Sanyam Bhutani, Miguel Romero Calvo +7
SWE-Bench Pro is a widely used benchmark of long-horizon software engineering tasks for coding agents. As agents improve, they reach task issues earlier agents never touched and interact with the evaluation environment in ways the original harness did not anticipate. We present SWE-Bench Pro V2, a revised public split built to keep the benchmark accurate as capabilities advance. Agentic audits, expert adjudication and blind re-implementation repaired or retired 480 of the 731 tasks: 23 contracted engineers spent 404 hours, each working in a language they write professionally, and every revised
agentagenticagent benchmarkbenchmark - arxiv:2610.24205 · cs.CVPREEMPT: Proactive Robustness Evaluation under Editing for Media Protection TechniquesQiuyu Tang, Aparna Bharati
Proactive image protection methods aim to defend visual content against unauthorized manipulation and generative misuse by embedding protective signals into images before release. Evaluating whether these protections remain effective under image editing is therefore critical. However, existing evaluations remain largely limited to one-shot edits, providing only a static view of robustness while overlooking its evolution under realistic multi-step editing workflows. To systematically characterize robustness dynamics under sequential editing across various protection paradigms, we introduce PREE
manipulationevaluation framework - arxiv:2610.24194 · cs.AIAgents' Last Doubt: Knowing When to Act and When to Step Back in Tool-Using AgentsQisai Liu, Qingyue Jiao, Bangwei Guo, Wenxi Dai +6
Tool-using LLM agents repeatedly face decision points under uncertainty, choosing between autonomous action, additional information gathering, and human escalation. These decisions require balancing task success against interaction cost and the risk of irreversible errors. Learning to make these trade-offs is challenging because existing interaction trajectories record only the outcomes of actions taken, leaving the consequences of alternative actions unobserved. To address this gap, we construct a counterfactual dataset containing 3,115 action outcomes from 1,036 decision points in ToolSandbo
agentllm agent - arxiv:2610.24187 · cs.CLReasoning with Context Management and External MemoryMan-Ling Sung, William Hartmann, Jiangying Zhou
Reasoning models maintain information about the status of their reasoning process through in-context memory. Complex problems require larger context windows and memory requirements. In this scenario, context management without caching risks discarding useful information for reasoning. We propose RCMEM, an approach that extends context management with an external memory to maintain additional information without burdening the reasoning model with additional context. RCMEM significantly reduces computation time on problems that require repeated computation of intermediate steps, while achieving
memoryexternal memory - arxiv:2610.24177 · cs.CVSelf-Improving Video-to-Action ModelsHyunwoo Ryu, Eric Ming Chen, Max Wilcoxson, Vincent Sitzmann
Video generative models are an emerging tool for embodied agents: they can serve as scalable visual planners, generating videos that depict how to complete previously unseen tasks. However, executing these plans typically requires a video-to-action or "inverse dynamics" model trained on large, human-collected video-action datasets, severely limiting scalability. We introduce Mirror, a self-supervised approach that learns video-to-action mappings through environment interactions, without prior action data or hand-crafted rewards. Mirror learns by attempting to follow a visual plan by executing
embodiedembodied agentself-improvingself-improvement - arxiv:2610.24164 · cs.CLPronouncing Hard Words: Phoneme Adaptation of a Pretrained Conversational Text to Speech ModelYashish M. Siriwardena, Dinesh K. Gnanasekaran, Tanmay Laud, Shinji Watanabe +2
Voice agents built as cascaded automatic speech recognition (ASR)->large language model (LLM)-> text-to-speech (TTS) pipelines increasingly rely on conversational TTS: streaming, context conditioned models that adapt delivery to dialogue history. Such models are natural and expressive but offer little explicit control over how rare, trust-critical drug names are pronounced, because these entities are too infrequent, and their vocabulary too fast-growing, to be covered by any amount of pre-, mid-, or post-training. We retrofit phoneme controllability onto a pretrained conversational TTS model w
post-training - arxiv:2610.24149 · cs.ROFLASH-CBF-RRT: Fast Lipschitz-Adaptive Safe Horizon CBF Steering for Certified Motion PlanningMani Amani, Hesam Mojtahedi, Reza Akhavian, Nikolay Atanasov +1
Fast motion planning requires both sample-efficient exploration and computationally efficient local steering. Conventional straight-line steering is inexpensive but wastes samples by rejecting colliding extensions. Quadratic program controllers with control barrier function (CBF) constraints can be used to achieve effective local steering but can increase the planning time due to the expensive sequential evaluations. We develop a method to accelerate CBF steering by computing a time horizon within which CBF-filtered velocity may be applied constantly using closed-form kinematic bounds on colli
manipulator - arxiv:2610.24129 · cs.RORollbot 2: Agile Locomotion by a Single-Actuator Spherical RobotJingxian Wang*, Chen Yu*, Sam Kriegman, Michael Rubenstein
Reducing actuator count simplifies robot hardware but also removes independent inputs for maneuvering. The original Rollbot steered by modulating a quasi-stable helical orbit, but moved slowly and relied on external motion capture. Rollbot 2 preserves the single-actuator design while its redesigned mass distribution enables near-straight rolling and more direct travel toward commanded paths and beacon targets. Onboard inertial and ultra-wideband sensing support learned steering, beacon seeking, and waypoint following through one motor-velocity command. Rollbot 2 autonomously approaches UWB bea
benchmark - arxiv:2610.24128 · cs.LGRecording-Aware Evaluation of Multimodal OT Intrusion DetectionMustafa Umut Ozbek, Khalil El-Khatib, Li Yang
Machine-learning-based intrusion detection systems (IDSs) for operational technology (OT) are commonly evaluated on benchmark datasets in which normal and attack data are collected in separate recordings. In such settings, high performance may reflect recording-associated differences as well as attack behavior. We introduce a recording-aware evaluation procedure that compares attack windows with background from the same recording, uses prior data for model fitting and calibration, tests temporal alignment against randomized event schedules, and reports event coverage together with alarm burden
benchmark - arxiv:2610.24124 · cs.LGOn the Selectivity of Generative Models in Structure-Based Drug DesignElla Miray Rajaonson, Jungyoon Lee, William Chau, Alan Aspuru-Guzik +4
Designing small-molecule drugs that selectively bind to a target protein while avoiding unintended interactions with other proteins is essential for reducing side effects that contribute substantially to clinical attrition. Although structure-based drug design (SBDD) has advanced rapidly with deep learning, most generative models still optimize binding to a single target and disregard selectivity due to limited training data and unknown off-target interactions. Motivated by this gap, we introduce SELECTBENCH, a comprehensive benchmark assessing the selectivity of generative SBDD models. SELECT
benchmark - arxiv:2610.24118 · cs.LGDo LMs Know What Variables Do? Interpreting Code Representations via Variable RolesCélian Vasson, Daniil Ognev, Kentaro Inui, Benjamin Heinzerling
Variable roles describe recurring behaviors such as accumulation, traversal, and retaining the best candidate encountered. We use Sajaniemi's independently defined taxonomy to study which role distinctions can be recovered from large language models, how sparse autoencoders (SAEs) retain those distinctions, and whether associated directions support control of generation. We introduce static annotation rules for all eleven roles and a corpus of 126,926 natural-code files across five languages. Across five model-SAE systems, residual probes outperform individual SAE features and tested lexical b
benchmark - arxiv:2610.24113 · cs.LGGraph-Regularized Synthetic Control for Stable Counterfactual Estimation under Donor RedundancyYanlin Zhang, Sungyong Chung, Alireza Talebpour
Synthetic control methods estimate counterfactual outcomes by weighting untreated donor units to match a treated unit before treatment. In recent applications, however, the donor pool can be large, highly correlated, and short in pre-treatment history, yielding many near-equivalent solutions with unstable and difficult-to-interpret donor weights. We propose Graph-Regularized Synthetic Control (Graph-SC), a convex estimator that augments the classical pre-treatment matching objective with a graph Laplacian penalty. The graph encodes similarity among donor trajectories, encouraging coherent weig
benchmark - arxiv:2610.24108 · cs.CVJoint, Decoupled, or Agent: When Each Multimodal Alignment Paradigm WinsYi Fang, Gaurav Srivastava, Sirui Ding, Xuan Wang
Jointly pretrained multimodal encoders are a common way to combine modalities, while alternatives integrate their information during downstream training or inference. When does each approach win? We compare three paradigms, instantiated with public checkpoints, across 16 paired datasets in five domains and multiple training-label budgets. JOINT trains a fusion head on frozen, jointly pretrained encoders. DECOUPLED trains a fusion head on frozen, independently pretrained encoders. AGENT gives a frozen LLM task instructions, labelled examples, available sample text, and supervised unimodal reado
agent - arxiv:2610.24107 · cs.AICache2LoRA: Understanding and Improving Context InternalizationYi Fang, Gaurav Srivastava, Zhenyu Bi, Sirui Ding +1
Context-to-LoRA methods encode documents as low-rank parameter updates, allowing language models to answer without retaining the source text in their decoding context. How reliably do these updates preserve facts and support their use under different instructions? Controlled diagnosis of existing context-to-LoRA methods separates fact discrimination, target retention, and response adaptation. Competing sources influence answers, while target-only construction and memory-state optimization improve recovery at fixed update size. Recovering facts also does not ensure that answers follow response
memory - arxiv:2610.24093 · cs.AIDoes AI-Assisted Development Homogenize Software? An Empirical Study of Code Similarity, Quality, and EvolutionLuciano Marchezan, Romain Lefeuvre, Kevin Delcourt, Benoit Combemale +1
AI-assisted software development is evolving from interactive LLM assistance toward agentic systems that autonomously perform software engineering tasks, yet the effects of this shift remain insufficiently understood. We present an empirical study of 50 long-lived Java and Python open-source projects across three eras: pre-LLM, potentially LLM-assisted, and detected agentic development. Using snapshots from Software Heritage, we examine code similarity, software quality, and software evolution. Similarity is measured semantically through embedding-based cosine similarity and syntactically thro
agentic - arxiv:2610.24090 · cs.AIPersistent Memory and Iterative Improvement in an LLM Agent for Chinese-Character DrawingPeilin Wu, Yi-Zhuang You
Does persistent memory accumulated from previous attempts help an LLM agent solve new problems, and does memory design affect that benefit? We study programmatic Chinese-character drawing across five agent conditions attempting the same 668 items. Accumulated memory produces modest aggregate gains despite substantial growth in the memory stores: free-form notes improve first-attempt success by 4.6 percentage points, while a design combining free-form memory with reusable code improves it by 2.8 points. The two designs do not differ significantly overall, but the code bank shows a suggestive ad
memorypersistent memoryagentllm agent - arxiv:2610.24088 · physics.app-phRevealing Detailed Transport Mechanisms in Silicon Tunneling DiodesJuan P. Mendez, Xujiao Gao, Sungjae Hong, Evan M. Anderson +2
Understanding carrier transport in heavily doped semiconductor junctions is essential for developing next-generation tunneling devices for cryogenic electronics and steep-slope technologies. In this work, we investigate the transport mechanisms in a silicon tunneling diode using extensive TCAD (Technology Computer Aided Design) simulations benchmarked against electrical measurements over a wide temperature range of 4 K to 440 K. The measured device characteristics show no negative differential conductance, strong temperature dependence of forward current with a nearly constant semi-logarithmic
benchmark - arxiv:2610.24087 · cs.AIOR-License: Evaluating LLM Agents as Real-World Operations Research PractitionersPeisong Wang, Feixue Liu, Yize Meng, Bowen Liu +15
Can large language model (LLM) agents deliver expert-level operations research (OR) solutions directly from original client materials, without prior expert reformulation? Existing OR benchmarks typically begin with self-contained descriptions or predefined optimization tasks, underrepresenting the translation of original business materials into professionally validated solutions. We introduce OR-License, an end-to-end benchmark of 31 instances across 10 industrial scenarios from paid OR engagements totaling USD 3.68 million. It preserves original requirements, input data, deliverable formats,
llm agentself-improvementbenchmark - arxiv:2610.24086 · cs.CLCoopshift: Post-Training Amplifies The Memory Curse In LLM CooperationJiayuan Liu, Shiyi Du, Tianqin Li, Emanuel Tewolde +2
How does post-training change cooperation during repeated interaction? Within each of seven released model lineages, we compare pretrained (Base), supervised fine-tuned (SFT), and subsequent preference- or reward-optimized (Aligned) checkpoints across seven social dilemmas. Copies of each checkpoint play one another for 200 rounds, with visible history limited to the last two or forty rounds. The memory curse refers to lower cooperation when agents can see more past rounds. Our central finding is that post-training amplifies this decline. Averaged across the six binary-action games, expanding
memorypost-training - arxiv:2610.24061 · cs.AIWorld of SWECraft: Agent Collaboration at Scale via Simulated Software WorldsShannon Zejiang Shen, Yapei Chang, Qiyao Liang, Benjamin Newman +12
As teams of AI agents take on larger projects and long-standing open problems, agent collaboration offers another dimension for scaling inference-time compute. To understand what makes such collaboration effective and scalable, we develop Software World, a framework for simulating multi-agent coworking in realistic software ecosystems. It comes with a pipeline to create environments of real, interdependent Python packages of different topics, sizes, and dependency structures, which we call software worlds. In a simulation, agents maintain packages in the world with a high-level goal of improvi
agentai agentmulti-agent - arxiv:2610.24060 · cs.CVGenerative Text Modelling and Language-Guided Visual Attention for Deciphering PalimpsestsRobin Hollifeldt, Tianru Zhang, Jessica Ström, Ekta Vats
Recovering obscured text in palimpsests, i.e. the manuscripts where the original writing (under-text) was erased and overwritten for reuse, is vital for preserving lost information. Due to the nature of the material, such a task can be highly challenging and time-consuming. This paper presents an approach towards automated palimpsest handwritten text modeling for under-text recovery. We investigate generative models, including Generative Adversarial Networks (GANs) and diffusion models, and design a custom denoising diffusion model that learns structured noise patterns to reconstruct the under
benchmark - arxiv:2610.24055 · cs.LGToolRosetta: Surface-Form Invariance of Tool-Call GuardsHasan Demirkiran, Jens Ernstberger
Step-level guards for tool-using agents are evaluated on tidy JSON tool calls, but deployed coding agents act through execute_bash and execute_code tools whose payloads are strings. We introduce ToolRosetta, which re-renders every candidate action in the AgentDojo and AgentHarm splits of TS-Bench as a Python call, as a shell command, and without the agent's rationale. The renderings are produced by code and verified by round-trip parsing, so the user request, the history, the label and the action's effect stay fixed while only the surface form changes. Across eleven guards with two placebo con
agentbenchmark - arxiv:2610.23771 · cs.AILLM-Assisted Bi-Approximate CHC SolvingBaiting Jiang
Constrained Horn clause (CHC) solvers often struggle to discover deep counterexamples whose derivations may require extensive rule unfolding. Although large language models (LLMs) can propose long-range symbolic relations, prior work on formally checked LLM-assisted verification has mainly focused on generating inductive invariants and concrete traces. We exploit this broader capability through Derivation Power Under-approximations (DPUs). Each DPU is a solver-certified under-approximation of the relation induced by a multi-step CHC derivation. An LLM proposes each DPU with symbolic intermedia
benchmark - arxiv:2610.24054 · cs.LGAlliance of Stochastic Routing and Reinforcement Fusion for Multi-Model ReasoningSelim Furkan Tekin, Ling Liu
This paper studies dynamic model routing under uncertainty in multimodal settings, where a system must select and potentially compose multiple models to answer a query. Standard reinforcement learning approaches for routing often suffer from premature convergence to suboptimal models due to noisy rewards and insufficient exploration. We propose a unified vision-language framework (VL-Scout) for efficient multimodal and multi-step reasoning with turn-based reinforcement fusion. First, VL-Scout combines Bayesian priors, policy gradient optimization, and sequential decision-making to enable struc
benchmark - arxiv:2610.24052 · cs.AIRed-Teaming Multi-Agent Attacks for Loss-of-Control RisksJohn Kitaoka, Thao Amelia Pham, Carlos Rafael Giudice, Jacob Merizian
AI agents are growing more capable and persistent, and are being deployed in ever larger numbers. As more start collaborating, AI control measures must remain effective when they covertly collude on a harmful objective. We introduce MiniLab, a synthetic series of live services and codebases (5 interconnected services, 249 benign tasks, 8 attack objectives), adapted from internal AI deployment infrastructure. We demonstrate distributed attacks, in which different agents carry out separate attack fragments, and how their covertness evades frontier monitors when reading their trajectories individ
agentai agentmulti-agent - arxiv:2610.24051 · cs.AILogicVal:A Logic-Formalized Benchmark for Constraint Satisfaction in Text-to-Image GenerationSavitha Sam Abraham, Marjan Alirezaie, Hossein Rahnama
Text-to-Image (T2I) generative models have made significant progress in producing high-fidelity images from natural-language prompts. Alongside this progress, numerous benchmarks have been developed to evaluate capabilities such as attribute binding, numeracy, spatial reasoning, and commonsense reasoning. In this work, we propose a new perspective on T2I evaluation by casting it as a constraint-satisfaction problem. We introduce a logical framework that formally specifies the benchmark domain, including objects, attributes, and relations, alongside the constraints that generated images must sa
benchmark - arxiv:2610.24050 · cs.CLDiscuss More, Comment Less: Boosting Automated Research via Multi-Perspective Collaborative ReviewJunkai Zhou, Jiajia Wu, Zhaoyi Zhang, Shiting Guan +1
As automated research is increasingly adopted for writing papers, automated review has become a critical link between paper generation and quality control. However, existing methods largely rely on single-model feedback or simple review aggregation, failing to explicitly model expert responsibilities or collective deliberation in actual peer reviews, which limits effective cross-examination and reliable decision-making. Therefore, we formulate paper review as a computational modeling problem of peer review mechanisms and propose RoundtableReview, a deliberative multi-agent framework for enhanc
multi-agentagent framework - arxiv:2610.24048 · cs.AIThe Window Is in the Reasoning: When LLM Agents Decide to Follow Prompt InjectionsHaoxuan Jia, Yang Liu, Bin Chong, Yingguang Yang +12
LLM agents that read untrusted content can be hijacked by prompt injections. Runtime monitors that try to catch a hijack before the agent acts, and benchmarks that score each task with a single run, share an assumption: whether an agent follows an injection is settled when the injection is read. We test this assumption with branching rollouts that replay an episode exactly up to the step at which the agent reads an injection and then resample the rest of the episode many times, splitting the outcome variance into a part fixed at read time and a part decided afterwards. Across four AgentDojo su
agentllm agentbenchmark - arxiv:2610.24045 · cs.LGWhat Remains Without Test Labels? Re-evaluating a Graph-Level Anomaly Detection BenchmarkYujie Lu, Da Yan
Graph-level anomaly detectors are trained only on normal graphs, yet benchmarks may use test labels for model selection. We examine whether UB-GOLD results reused by later papers support label-free detection. We audit UB-GOLD and 14 GLAD code bases, re-run 21 detectors on UB-GOLD's folds under selection rules with and without test labels, and evaluate sensitivity to known structural changes. Test labels select checkpoints in UB-GOLD's shared training loop and checkpoints or hyperparameters in 8 of the 14 code bases. UB-GOLD also evaluates its kernel baselines against misaligned labels. Across
benchmark - arxiv:2610.24037 · cs.CVBenchmarking and Improvement of Artifact Segmentation Methods in H&E-Stained Whole Slide ImagingJulia Merta, Daniel Mendelsohn, Benedikt Schoser, Joanna Polanska +1
Technical artifacts introduced during tissue preparation or slide scanning remain a major obstacle to developing robust computational pathology algorithms for hematoxylin and eosin (HE)-stained whole-slide images (WSIs). Despite the growing number of quality-control tools, their real-world utility remains hard to assess because of the limited number of publicly available implementations and the lack of systematic, multi-dataset comparisons. This paper presents a benchmark of artifact segmentation methods for HE-stained WSIs. After a literature review, six publicly available algorithms were sel
benchmark - arxiv:2610.24034 · cs.CVImplicit Neural Models for Direct Multi-Shell Diffusion MRI Signal EstimationSanna Persson, Fabian Leander Sinzinger, Christoffer Ohlsson, Rodrigo Moreno
Diffusion magnetic resonance imaging is a noninvasive technique that is sensitive to the motion of water molecules in human tissue. The image signal is commonly described in q-space; however, a complete coverage of this space is often infeasible due to hardware and acquisition time limitations. In this work, we propose to represent the entire diffusion signal using an implicit neural representation network with wavelet activation functions. The signal space is encoded using Fourier features and spherical harmonics basis functions for a continuous output distribution of the raw diffusion data,
memory - arxiv:2610.24032 · cs.LGVisual Causal-chain Bookmarking: Retrieval-Weighted Credit Assignment for Reinforcement Learning from Delayed RewardPrathap Selvakumar
RL agents trained on a single trajectory-level return face a credit-assignment problem: when a reward arrives many steps after the decision that caused it, every action gets the same signal regardless of causal responsibility. This paper studies Visual Causal-chain Bookmarking (VCBM), which concentrates credit on the decision points that determine an episode's return, via two instantiations. The first targets a fully observed, low-dimensional MDP (a delayed-reward watch-repair task), where VCBM statistically discovers which state components the agent causally controls and estimates per-decisio
memoryepisodic memoryagentllm agent - arxiv:2610.24030 · cs.LGWhen Does Evidence-Graph Structure Predict Answer Correctness? Label-Free Abstention for Structured RetrievalAnjan Goswami
Retrieval systems return scored answers, but a high score does not show that an answer should be trusted. The proposed training-free abstention method applies the boundary cut to the selected evidence in an evidence graph. The signal is the weighted cut between selected and unselected candidates and requires no correctness labels or confidence-model training. On MemoryAgentBench (MAB) FactConsolidation multi-hop, with unresolved queries ranked last for every selector, it raises area under the precision-coverage curve from 0.736 for the best score baseline to 0.785 (+0.049, 95% CI [+0.008, +0.0
benchmark - arxiv:2610.24029 · cs.LGDo Knowledge and Reasoning Oriented Benchmarks Induce Stable Representation Signatures in LLMs?Parsa Rahimi, Azim Dehghani Amirabad
Internal signatures could support monitoring, adaptive computation, and mechanistic analysis, but benchmark prompts can be separable without any shared computational distinction. We test whether a prespecified knowledge/reasoning label defines a reusable internal signature in five contemporary 7B to 31B models. We treat the \textbf{task} as the unit of inference probes hold out complete tasks, labels are permuted across tasks, and the best layer and stream are reselected for every permutation. Raw probes recover an eight-versus-eight-subject MMLU partition in all five models (macro-F1 $0.78$--
benchmark - arxiv:2610.24024 · cs.LGUnified Multiparadigm Fusion Intelligence for Rare Disease DiagnosisJian Yang, Peiling Liu, Songda Li, Yu Cao +9
Rare diseases (RDs) affect an estimated 300 million people globally, leaving patients to navigate a protracted diagnostic odyssey constrained by clinical scarcity and fragmented medical knowledge. Artificial intelligence (AI) has progressively advanced RD diagnostics, from symbolic methods and deep learning to large language model (LLM)-based systems. However, each paradigm carries structural limitations: symbolic methods are bounded by knowledge completeness, neural models sacrifice interpretability, and LLM-based systems are vulnerable to evaluation biases like data leakage. In this study, w
agenticbenchmark - arxiv:2610.24016 · cs.ROCLARO: Clutter Aware Robot Orchestrator for Open-Vocabulary Grasping using VLM AgentsSungsu Kim, Vineet Bhat, Prashanth Krishnamurthy, Ramesh Karri +1
Robots that follow natural-language instructions unlock a wide range of autonomous applications in households, warehouses, and labs. Real-world environments are diverse and often heavily cluttered, requiring robots to plan over multiple steps, act, and replan as they work toward a goal. We present CLARO (CLutter-Aware Robot Orchestrator), a VLM agent that orchestrates a set of perception and manipulation skills: open-vocabulary object detection, segmentation, 6-DoF grasping, and non-prehensile pushing to locate and retrieve a language-specified target in clutter. CLARO maintains a memory of it
manipulationsim2realgraspmemoryagentbenchmark - arxiv:2610.24014 · cs.LGTWIN-MoE: Two-stage Intra-Expert Imputation Network for Incomplete Multimodal LearningLiang Nathan Zheng, Jiyuan Liu, Wenhao Liang, Wei Emma Zhang +2
Recent Mixture-of-Experts (MoE) approaches address missing modalities either by routing each missingness pattern to a dedicated expert, which scales poorly as the number of modalities grows, or by imputing the missing modality from representations contributed by multiple experts, which requires a costly retrieval mechanism. We show that routing induces expert-conditioned representation neighborhoods and derive an imputation-error bound controlled by their local feature dispersion. When these expert-conditioned neighborhoods are more compact than the global representation space, intra-expert ag
benchmark - arxiv:2610.24012 · cs.ROChairNav: Cross-Embodiment Pre-training and Personalization for Long-Horizon Wheelchair NavigationJunhua Huang, Zhizheng Liu, Honglin He, Xiang Zhang +1
Developing autonomous electric wheelchairs capable of long-horizon sidewalk navigation is critical for accessible personal mobility, yet existing approaches either focus on indoor map-based navigation or rely on expensive, complex sensor stacks to achieve full autonomy. In this work, we propose to tackle outdoor sidewalk navigation by leveraging visual navigation foundation models and large-scale navigation datasets, which offer strong generalization using lightweight hardware. To this end, we present ChairNav, a novel framework that adapts a general visual navigation model into a personalized
post-training
02 US SEMI · SEC 8-K FILINGS
2 itemsscanned: NVDA / AVGO / MRVL / COHR / LITE / AMD / TSM / SMCI / ANET / CRDO / POWL / VECO
03 HUMANOID · COMPANY NEWS
62 itemsscanned: figure-ai / 1x / boston-dynamics / unitree / apptronik / sanctuary-ai / neura-robotics / agility-robotics / physical-intelligence / agibot
Figure AI (10)
- Figure AISeptember 30, 2026F.02 Decommission
- Figure AISeptember 17, 2026Helix 2.5: Zero-Shot 30-Home Generalization
- Figure AIOctober 09, 2025Introducing Figure 03
- Figure AISeptember 03, 2026Figure and Nscale Sign Strategic Partnership For Up to 100,000 GPUs on the NVIDIA Vera Rubin Platform
- Figure AIAugust 25, 2026Introducing Index: Building The World’s Largest and Most Diverse Physical Dataset
Boston Dynamics (10)
Unitree 宇树 (10)
- Unitree 宇树Components
- Unitree 宇树Kung Fu Meets Spring, Unitree SFG Robots Present "Cyber Real Kung Fu" in the Year of the Horse2026-05-31Media Coverage
- Unitree 宇树Welcoming Myanmar President Min Aung Hlaing to Unitree2026-08-05Media Coverage
- Unitree 宇树Unitree founder Wang Xingxing graces the cover of Time magazine2026-08-05Media Coverage
- Unitree 宇树Unitree Announces H2 Plus, an NVIDIA Isaac GR00T Reference Humanoid Robot for Academic Research2026-06-01Media Coverage
Apptronik (1)
Sanctuary AI (6)
- Sanctuary AIBlog PostTIGER: Task-Space Imitation Guidance for Efficient Reinforcement LearningDemonstrations as a compass, not a scriptRead More
- Sanctuary AIPress ReleaseProduct UpdatesSanctuary AI Expands Physical AI Strategy to Industrial Robotics, Demonstrating Production-Ready AI PerformanceRead More
- Sanctuary AICorporate NewsDaniel Friedmann Appointed CEO of Sanctuary AIRead More
- Sanctuary AIPress ReleaseZeon Invests in Sanctuary AI and Partners to Advance Specialized Materials for Dexterous RoboticsRead More
- Sanctuary AIThought LeadershipWeb Summit Reflections: Canada’s Physical AI Moment Can’t WaitRead More
Agility Robotics (10)
- Agility RoboticsAgility’s Humanoid Deployment ProcessPras VelagapudiAugust 04, 2026
- Agility RoboticsThe Realistic Pathway to HomeInsightMay 26, 2026
- Agility RoboticsAgility and AIInsightMarch 16, 2026
- Agility RoboticsAgility Gets a New BrandInsightMarch 5, 2026
- Agility Robotics2026: The Automation EvolutionInsightJanuary 16, 2026
Physical Intelligence (8)
- Physical IntelligenceNewFrontier Models for Every RobotOctober 13, 2026Our latest robotics models are now available through the Pi API beta. Upload data, fine-tune models, and deploy cloud-hosted inference on your own robots.
- Physical Intelligenceπ0.7: a Steerable Model with Emergent CapabilitiesApril 16, 2026A steerable robotic foundation model that exhibits a step-change in generalization.
- Physical IntelligenceThe Physical Intelligence LayerFebruary 24, 2026General-purpose physical intelligence models will enable a Cambrian explosion of robotics applications. See how our partners are already solving real-world problems.
- Physical IntelligenceMoravec's Paradox and the Robot OlympicsDecember 22, 2025By fine-tuning our latest model, we were able to solve a series of very difficult manipulation challenge tasks.
- Physical Intelligenceπ*0.6: a VLA that Learns from ExperienceNovember 17, 2025A method for training our generalist policies with RL to improve success rate and throughput on real-world tasks.
智元 AgiBot (7)
- 智元 AgiBotAGIBOT Opens Hong Kong Office in Causeway Bay2026-10-09
- 智元 AgiBotAGIBOT Partners with Vietnam National Innovation Center(NIC) to Advance Embodied AI Industry and Talent Ecosystem2026-10-02
- 智元 AgiBotIDC Data Shows AGIBOT Led Global Humanoid Robot Shipments in H1 20262026-09-30
- 智元 AgiBotAGIBOT and ASD Deploy 100 AGIBOT X2 Humanoid Robots Across 100 Retail Stores2026-09-29
- 智元 AgiBotAGIBOT and Chimelong Launch Large-Scale Embodied AI Theme Park with More Than 300 Robots2026-09-23