Knowledge catalogue
arXiv:2609.38178v1 Announce Type: cross Abstract: Robots deployed in the physical world must be able to improve beyond their initial training as they encounter new situations and failures. For this im
arXiv:2609.37326v1 Announce Type: new Abstract: On-policy distillation, where a student learns from a stronger teacher's feedback on its own outputs, is a common way to pass reasoning to smaller model
arXiv:2609.36217v1 Announce Type: new Abstract: A deeper understanding of brain function requires a precise, structured characterization of behavior.Yet, extracting behavioral representations from vid
arXiv:2609.36471v1 Announce Type: cross Abstract: World-Action Models (WAMs) improve robotic manipulation by conditioning action generation on predicted future observations, but future prediction adds
arXiv:2609.36900v1 Announce Type: new Abstract: Reward hacking occurs when policy optimization exploits a brittle reward interface or an overly permissive proxy objective, improving the training score
arXiv:2609.36867v1 Announce Type: new Abstract: We propose STRAT, an auxiliary task that trains deep reinforcement learning (RL) agents to predict a short textual trace of their own state. Inspired by
arXiv:2609.34799v2 Announce Type: replace Abstract: Language-conditioned trajectory generation is here, but its evaluation has not kept pace. Existing pedestrian trajectory metrics compare trajectorie
arXiv:2609.38096v1 Announce Type: new Abstract: Policies with similar mean returns can differ sharply in rare failures, yet estimating lower-tail conditional value-at-risk (CVaR) accurately can requir
arXiv:2609.37302v1 Announce Type: new Abstract: Deep vision models often degrade under distribution shift. Test-time adaptation can improve robustness but typically requires iterative optimization, hy
arXiv:2609.37824v1 Announce Type: new Abstract: Transformer language models build predictions through successive residual updates, but how their representations become specific to an eventual outcome
arXiv:2609.36118v1 Announce Type: new Abstract: Vision-language-action (VLA) policies connect a pretrained vision-language backbone to an action head through a latent interface, but which backbone lay
arXiv:2609.37498v1 Announce Type: new Abstract: Wikipedia aims to provide a unified, neutral record of history, yet its independent language editions often function as distinct epistemic communities,
arXiv:2609.36484v1 Announce Type: new Abstract: On-policy distillation (OPD) trains a student to match the teacher's next-token distributions on the student's own trajectories and has yielded substant
arXiv:2609.37372v1 Announce Type: new Abstract: Visual reward models are essential for evaluating and improving visual generation models, yet existing approaches typically map task conditions and cand
Today we're announcing Praxis-1, an open-weight world action model that adapts to new robot embodiments with light finetuning. Video is eating the world Introducing Praxis-1, an open-weight World Acti
arXiv:2609.36816v1 Announce Type: new Abstract: Reinforcement learning (RL) has become a cornerstone for improving the reasoning capabilities of large language models (LLMs), but the need for on-polic
arXiv:2609.35822v1 Announce Type: new Abstract: Sycophantic agreement in language models refers to the tendency to overly affirm a user's stated beliefs or preferences, often at the expense of factual
arXiv:2609.36924v1 Announce Type: cross Abstract: Learning humanoid skills from videos typically requires a successful human demonstration, which often demands custom data collection. Although failure
arXiv:2609.37132v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) improves large language models by letting a self-teacher with privileged information provide dense token-level superv
arXiv:2609.36063v1 Announce Type: cross Abstract: Neural Combinatorial Optimization (NCO) has achieved strong empirical success, yet the internal mechanisms driving model decisions remain largely unex
arXiv:2609.37119v1 Announce Type: cross Abstract: Recent approaches to reinforcement learning (RL) post-training for large language models increasingly remove the critic to reduce training instability
arXiv:2609.37250v1 Announce Type: cross Abstract: World-action models (WAMs) couple future visual-state prediction with action generation. By adapting video generators or image-editing models pretrain
arXiv:2609.36804v1 Announce Type: cross Abstract: Chinese Semantic Error Correction (CSEC) targets semantic errors in Chinese text, which are typically more subtle and complex than spelling and gramma
arXiv:2609.36870v1 Announce Type: new Abstract: Video demonstrations offer a scalable alternative to costly robot data for learning manipulation, yet existing reconstruction-based approaches often rel
arXiv:2609.38086v1 Announce Type: new Abstract: Active multimodal agents use visual tools to acquire task-relevant evidence while reasoning. Although reinforcement learning samples multiple interactio
arXiv:2609.36958v1 Announce Type: cross Abstract: Repeated verifier calls are useful only when they contribute conditional information. We introduce VStress, an auditable replay contract, and VStress-
arXiv:2609.37922v1 Announce Type: new Abstract: Visual Language Action (VLA) models offer unprecedented generalization for autonomous robots; however, their real-world deployment is frequently bottlen
arXiv:2609.36054v1 Announce Type: cross Abstract: In contrast to even a year ago, AI systems now write most of the code inside the companies that build them. As more of the AI research and development
arXiv:2609.37125v1 Announce Type: new Abstract: Prospective memory allows an agent to retain an intention tied to a future condition, but the stored intention does not reveal whether that condition cu
arXiv:2609.36645v1 Announce Type: cross Abstract: Future prediction is increasingly used to improve vision-language-action (VLA) policies, based on the premise that anticipating scene evolution encour
arXiv:2609.36864v1 Announce Type: new Abstract: Group-relative methods for reinforcement learning with verifiable rewards (RLVR) learn from differences in rollout outcomes. Independently sampling comp
arXiv:2609.37991v1 Announce Type: new Abstract: Brain-AI alignment is often interpreted as a sign that model and brain perform similar computations. Whether the aligned units are causally involved in
arXiv:2609.37535v1 Announce Type: new Abstract: In the softmax output layer, a rare token receives a small positive logit gradient on most steps and a much larger negative gradient on the few steps wh
arXiv:2609.38059v1 Announce Type: cross Abstract: Real-world robot learning is constrained by the cost of collecting experience and evaluating candidate behaviors. Video generation models offer a scal
arXiv:2609.37552v1 Announce Type: new Abstract: While contact-rich manipulation requires deliberate regulation of interaction forces, recent approaches to robot manipulation learning predominantly rep
arXiv:2609.37902v1 Announce Type: new Abstract: Existing LLM routers choose among models using static per-model costs. We show that open-weight inference markets introduce a second, largely ignored de
arXiv:2609.31686v1 Announce Type: new Abstract: Existing aviation emissions inventories lack accurate trajectory data for high-resolution social cost and health impact assessment. This paper develops
arXiv:2607.14522v2 Announce Type: replace Abstract: We formulate reinforcement learning (RL) in continuous time with discrete state spaces and possibly arbitrary action spaces via a stochastic control
arXiv:2609.32290v1 Announce Type: new Abstract: It has recently been found that deep learning often occurs at the 'edge of stability (EoS),' where the maximum Hessian eigenvalue of the model is stabil
arXiv:2609.34374v1 Announce Type: new Abstract: Language model post-training is often bottlenecked by the need for human-collected preference data, which is expensive and difficult to scale. Reinforce
arXiv:2601.00416v2 Announce Type: replace Abstract: Functional connectivity (FC) analysis, a valuable tool for computer-aided brain disorder diagnosis, traditionally relies on atlas-based parcellation
arXiv:2609.33832v1 Announce Type: cross Abstract: World-action models can jointly predict future visual observations and robot actions. However, discrepancies may exist between their visual prediction
arXiv:2510.26672v2 Announce Type: replace-cross Abstract: At the heart of reinforcement learning are actions -- decisions made in response to observations of the environment. Actions are equally funda
arXiv:2609.32752v1 Announce Type: cross Abstract: Reward shaping has a theorem: a potential-based term can be removed without changing the optimal policy. The same practice on the action channel, an o
arXiv:2609.33256v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models map visual observations and language instructions directly to robot actions, but they do not explicitly represent th
arXiv:2609.32252v1 Announce Type: cross Abstract: Acquiring data is costly: higher measurement fidelity costs power and storage and risks collecting irrelevant content, while aggressive cost reduction
arXiv:2510.13680v2 Announce Type: replace Abstract: Preconditioned methods are central to deep learning optimization. Predominant approaches include computationally light diagonal preconditioners such
arXiv:2609.32464v1 Announce Type: new Abstract: Multi-objective Bayesian optimization (MOBO) commonly relies on independent Gaussian processes (GPs) with stationary kernels, limiting its ability to re
arXiv:2609.34944v1 Announce Type: cross Abstract: Flow-based Vision-Language-Action (VLA) policies are typically trained by behavior cloning and thus do not explicitly optimize long-term task return.
arXiv:2602.01750v2 Announce Type: replace-cross Abstract: Reinforcement Learning from Human Feedback (RLHF) remains vulnerable to reward hacking, where models exploit spurious correlations in learned
arXiv:2609.35690v1 Announce Type: new Abstract: Robots that learn from a few demonstrations often require two forms of generalization. Compositional generalization recombines skills to solve new tasks
arXiv:2609.33323v1 Announce Type: cross Abstract: Recent advances in Large Language Models (LLMs) have enabled agentic systems capable of solving complex tasks through multi-turn planning, tool use, v
arXiv:2609.34823v1 Announce Type: new Abstract: Augmented dexterity has the potential to reduce the fatigue experienced by surgeons during repetitive surgical tasks. In this paper, we propose the firs
arXiv:2510.14391v3 Announce Type: replace-cross Abstract: Recent neural beat trackers predict beats and downbeats with two independent frame-wise binary classification heads, a convention adopted beca
arXiv:2609.35466v1 Announce Type: new Abstract: Soft Actor-Critic (SAC) is widely used for entropy-regularised reinforcement learning with continuous action spaces, and practical implementations perfo
arXiv:2609.32762v1 Announce Type: cross Abstract: Visual imitation learning is a promising approach to training robot manipulation policies capable of completing a wide variety of tasks. However, poli
arXiv:2604.07666v2 Announce Type: replace Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) is widely used for post-training Large Language Models, but practical verifiers can make error
arXiv:2609.35505v1 Announce Type: new Abstract: We study on-policy distillation (OPD) through the lens of reinforcement learning, establishing a connection between the reverse-KL objective in OPD and
arXiv:2609.32380v1 Announce Type: new Abstract: Data attribution methods used in practice (TRAK, LoGRA, EK-FAC) are post-hoc: after training they make a second pass over the training set to recompute
arXiv:2609.33662v1 Announce Type: cross Abstract: Imperfect verifiers can assign a harmful update direction even when clipping and regularization bound its magnitude. We introduce Audit-First VAPO, wh