safety

Monocular Biomechanical Tracking of Fingers with Inverse Kinematics to Foundation Models

arXiv:2605.09258v1 Announce Type: cross Abstract: Accurate hand and finger tracking from video has significant clinical applications for monitoring activities of daily living and measuring range of mo

Safety

Morphology-Aware Graph Reinforcement Learning for Tensegrity Robot Locomotion

DGX agent

arXiv:2510.26067v2 Announce Type: replace Abstract: Tensegrity robots combine rigid rods and elastic cables, offering high resilience and deployability but at the same time posing major challenges for

safetyarxiv-cs-ro

12 May 2026

MTA-RL: Robust Urban Driving via Multi-modal Transformer-based 3D Affordances and Reinforcement Learning

arXiv:2605.10177v1 Announce Type: cross Abstract: Robust urban autonomous driving requires reliable 3D scene understanding and stable decision-making under dense interactions. However, existing end-to

Multi-layer attentive probing improves transfer of audio representations for bioacoustics

arXiv:2605.10494v1 Announce Type: cross Abstract: Probing heads map the representations learned from audio by a machine learning model to downstream task labels and are a key component in evaluating r

Multi-scale Predictive Representations for Goal-conditioned Reinforcement Learning

arXiv:2605.09364v1 Announce Type: new Abstract: This paper investigates robust representation learning in offline goal-conditioned reinforcement learning (GCRL). Particularly in sparse reward scenario

Multimodal Representation Learning Conditioned on Semantic Relations

arXiv:2508.17497v2 Announce Type: replace-cross Abstract: Multimodal representation learning has been largely driven by contrastive models such as CLIP, which learn a shared embedding space by alignin

Muninn: Your Trajectory Diffusion Model But Faster

arXiv:2605.09999v1 Announce Type: new Abstract: Diffusion-based trajectory planners can synthesize rich, multimodal robot motions, but their iterative denoising makes online planning and control prohi

MURPHY: Feedback-Aware GRPO with Retrospective Credit Assignment for Multi-Turn Code Generation

arXiv:2511.07833v3 Announce Type: replace-cross Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has become a standard recipe for post-training LLMs on reasoning tasks, with Group Relat

Mutual Information Optimal Density Control of Linear Systems and Generalized Schrodinger Bridges with Reference Refinement

arXiv:2605.09349v1 Announce Type: cross Abstract: We consider a mutual information (MI) regularized version of optimal density control of a discrete-time linear system. MI optimal control has been pro

MVB-Grasp: Minimum-Volume-Box Filtering of Diffusion-based Grasps for Frontal Manipulation

arXiv:2605.09672v1 Announce Type: new Abstract: State-of-the-art 6-DoF grasp generators excel on tabletop benchmarks with overhead cameras but struggle in frontal grasping scenarios on low-cost manipu

Natural Policy Gradient as Doubly Smoothed Policy Iteration: A Bellman-Operator Framework

arXiv:2605.10671v1 Announce Type: new Abstract: In this work, we show that natural policy gradient, a core algorithm in reinforcement learning, admits an exact formulation as a smoothed and averaged f

Navigating EU AI Act requirements for LLM fine-tuning on Amazon SageMaker AI

In this post, we show you how to set up FLOPs tracking during LLM fine-tuning using the open source Fine-Tuning FLOPs Meter toolkit on Amazon SageMaker AI. You learn how to determine your compliance s

safetyaws-ml-blog

Navigating LLM Valley: From AdamW to Memory-Efficient and Matrix-Based Optimizers

arXiv:2605.09176v1 Announce Type: cross Abstract: Training large language models requires optimization algorithms that are not only statistically effective, but also computationally and memory efficie

NEO: No-Optimization Test-Time Adaptation through Latent Re-Centering

arXiv:2510.05635v2 Announce Type: replace-cross Abstract: Test-Time Adaptation (TTA) methods are often computationally expensive, require a large amount of data for effective adaptation, or are brittl

Neural at ArchEHR-QA 2026: One Method Fits All: Unified Prompt Optimization for Clinical QA over EHRs

arXiv:2605.10877v1 Announce Type: new Abstract: Automated question answering (QA) over electronic health records (EHRs) demands precise evidence retrieval, faithful answer generation, and explicit gro

safetyarxiv-cs-cl

Neural Co-state Policies: Structuring Hidden States in Recurrent Reinforcement Learning

arXiv:2605.05373v2 Announce Type: replace Abstract: A key capability of intelligent agents is operating under partial observability: reasoning and acting effectively despite missing or incomplete stat

Neuromorphic Reinforcement Learning for Quadruped Locomotion Control on Uneven Terrain

arXiv:2605.09595v1 Announce Type: cross Abstract: Reinforcement learning (RL) has enabled robust quadruped locomotion over complex terrain, but most learned controllers are trained offline with backpr

NEXUS: Continual Learning of Symbolic Constraints for Safe and Robust Embodied Planning

arXiv:2605.09387v1 Announce Type: new Abstract: While Large Language Models (LLMs) have catalyzed progress in embodied intelligence, a fundamental gap between their inherent probabilistic uncertainty

Not All Turns Matter: Credit Assignment for Multi-Turn Jailbreaking

arXiv:2605.08778v1 Announce Type: new Abstract: Deploying LLMs in multi-turn dialogues facilitates jailbreak attacks that distribute harmful intent across seemingly benign turns. Recent training-based

not every day @scaling01 and I agree. but he’s right. and most people won’t notice it happening, per the work of @informor.

not every day @scaling01 and I agree. but he’s right. and most people won’t notice it happening, per the work of @informor. hot take: unrestricted LLMs are as dangerous as weapons of mass destruction

note that I said “Even @haider1” because he is often optimistic, but he has informed me (with receipts that he shared) that he expressed som…

Gary Marcus references a conversation with someone named Haider, noting that despite Haider's typically optimistic outlook, he has shared documented evidence of expressing concerns about a particular

Now You See That: Learning End-to-End Humanoid Locomotion from Raw Pixels

arXiv:2602.06382v2 Announce Type: replace Abstract: Achieving robust vision-based humanoid locomotion remains challenging due to two fundamental issues: the sim-to-real gap introduces significant perc

Offline Preference Optimization for Rectified Flow with Noise-Tracked Pairs

arXiv:2605.09433v1 Announce Type: new Abstract: Existing preference datasets for text-to-image models typically store only the final winner/loser images. This representation is insufficient for rectif

On-Policy Distillation with Best-of-N Teacher Rollout Selection

arXiv:2605.09725v1 Announce Type: new Abstract: On-policy distillation (OPD), which supervises a student on its own sampled trajectories, has emerged as a data-efficient post-training method for impro

On the Generation and Mitigation of Harmful Geometry in Image-to-3D Models

arXiv:2605.09606v1 Announce Type: cross Abstract: Recent advances in image-to-3D models have significantly improved the fidelity and accessibility of 3D content creation. Such a powerful reconstructio

On Uniform Error Bounds for Kernel Regression under Non-Gaussian Noise

arXiv:2605.09757v1 Announce Type: new Abstract: Providing non-conservative uncertainty quantification for function estimates derived from noisy observations remains a fundamental challenge in statisti

On Variance Reduction in Learning Mean Flows

arXiv:2605.09235v1 Announce Type: cross Abstract: One-step generative modeling has emerged as a leading approach to amortize the inference cost of diffusion and flow-matching models. Among distillatio

Open Ontologies: Tool-Augmented Ontology Engineering with Stable Matching Alignment

arXiv:2605.09184v1 Announce Type: new Abstract: We present Open Ontologies, an open-source ontology engineering system implemented in Rust that integrates LLM-driven construction with formal OWL reaso

OpenAI’s Sam Altman’s personal investments are coming under intensifying scrutiny from Republicans following an April article in The Wall St…

Sam Altman's personal investments faced increased scrutiny from Republican lawmakers following reporting by The Wall Street Journal in April regarding potential conflicts of interest. The controversy

OpenClaw-RL: Train Any Agent Simply by Talking

arXiv:2603.10165v2 Announce Type: replace-cross Abstract: Every agent interaction generates a next-state signal, namely the user reply, tool output, terminal or GUI state change that follows each acti

Opinion: It is rare for the US public to agree on anything these days. Fear of AI is as close to a national consensus as it gets. A clear ma…

Opinion: It is rare for the US public to agree on anything these days. Fear of AI is as close to a national consensus as it gets. A clear majority says that AI will do more harm than good. https://ft.

Overcoming Catastrophic Forgetting in Visual Continual Learning with Reinforcement Fine-Tuning

arXiv:2605.09640v1 Announce Type: new Abstract: Recent studies suggest that Reinforcement Fine-Tuning (RFT) is inherently more resilient to catastrophic forgetting than Supervised Fine-Tuning (SFT). H

Path-Coupled Bellman Flows for Distributional Reinforcement Learning

arXiv:2605.08253v1 Announce Type: cross Abstract: Distributional reinforcement learning (DRL) models the full return distribution, but existing finite-support or quantile-based methods rely on project

PATRA: Pattern-Aware Alignment and Balanced Reasoning for Time Series Question Answering

arXiv:2602.23161v2 Announce Type: replace Abstract: Time series reasoning demands both the perception of complex dynamics and logical depth. However, existing LLM-based approaches exhibit two limitati

Pay attention to this one if you build research or knowledge-work agents. Most research-agent systems produce uniform outputs regardless of …

Pay attention to this one if you build research or knowledge-work agents. Most research-agent systems produce uniform outputs regardless of who is driving them. This new work, NanoResearch, argues tha

safetydair-ai--x

Perception Without Engagement: Dissecting the Causal Discovery Deficit in LMMs

arXiv:2605.09422v1 Announce Type: new Abstract: Although Large Multimodal Models (LMMs) have achieved strong performance on general video understanding, their susceptibility to textual prior shortcuts

safetyarxiv-cs-cl

Personalizing LLMs with Binary Feedback: A Preference-Corrected Optimization Framework

arXiv:2605.10043v1 Announce Type: cross Abstract: Large Language Model (LLM) personalization aims to align model behaviors with individual user preferences. Existing methods often focus on isolated us

PersonaTeaming: Supporting Persona-Driven Red-Teaming for Generative AI

arXiv:2605.05682v2 Announce Type: replace-cross Abstract: Recent developments in AI safety research have called for red-teaming methods that effectively surface potential risks posed by generative AI

PFN-TS: Thompson Sampling for Contextual Bandits via Prior-Data Fitted Networks

arXiv:2605.10137v1 Announce Type: cross Abstract: Thompson sampling is a widely used strategy for contextual bandits: at each round, it samples a reward function from a Bayesian posterior and acts gre

PHAGE: Patent Heterogeneous Attention-Guided Graph Encoder for Representation Learning

arXiv:2605.10073v1 Announce Type: new Abstract: Patent claims form a directed dependency structure in which dependent claims inherit and refine the scope of earlier claims; however, existing patent en

safetyarxiv-cs-cl

PHMForge: Evaluating LLM Agents on Industrial Prognostics through MCP-Native, Algorithm-Grounded Tools

arXiv:2604.01532v2 Announce Type: replace Abstract: LLM agents are beginning to invoke industrial asset-management tools through the Model Context Protocol (MCP), yet whether they can act reliably on

PhysEDA: Physics-Aware Learning Framework for Efficient EDA With Manhattan Distance Decay

arXiv:2605.10547v1 Announce Type: new Abstract: Electronic design automation (EDA) addresses placement, routing, timing analysis, and power-integrity verification for integrated circuits. Learning met

Plan in Sandbox, Navigate in Open Worlds: Learning Physics-Grounded Abstracted Experience for Embodied Navigation

arXiv:2605.10118v1 Announce Type: new Abstract: Vision-Language Models (VLMs) have demonstrated exceptional general reasoning capabilities. However, their performance in embodied navigation remains hi

Plan2Cleanse: Test-Time Backdoor Defense via Monte-Carlo Planning in Deep Reinforcement Learning

arXiv:2605.09638v1 Announce Type: new Abstract: Ensuring the security of reinforcement learning (RL) models is critical, particularly when they are trained by third parties and deployed in real-world

PMCTS: Particle Monte Carlo Tree Search for Principled Parallelized Inference Time Scaling

arXiv:2605.08982v1 Announce Type: new Abstract: Monte Carlo Tree Search (MCTS) is a widely used approach for policy improvement through search with increasing popularity for real world applications. D

Policy Gradient Methods for Non-Markovian Reinforcement Learning

arXiv:2605.10816v1 Announce Type: cross Abstract: We study policy gradient methods for reinforcement learning in non-Markovian decision processes (NMDPs), where observations and rewards depend on the

Political Plasticity: An Analysis of Ideological Adaptability in Large Language Models

arXiv:2605.08415v1 Announce Type: new Abstract: Since the advent of Large Language Models (LLMs), a significant area of research has focused on their intrinsic biases, particularly in political discou

Position: Academic Conferences are Potentially Facing Denominator Gaming Caused by Fully Automated Scientific Agents

arXiv:2605.09915v1 Announce Type: cross Abstract: The implicit policy of maintaining relatively stable acceptance rates at top AI conferences, despite exponentially growing submissions, introduces a c