safety

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations

arXiv:2605.28553v1 Announce Type: new Abstract: In this paper, we investigate whether refusal behavior can be predicted from LLM intermediate activations before decoding using linear probes trained on

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Safety

Restoring the Sweet Spot: Pass-Rate Weighted Self-Distillation for LLM Reasoning

DGX agent

arXiv:2605.27765v1 Announce Type: cross Abstract: Self-Distillation Policy Optimization (SDPO) provides dense token-level credit assignment for reinforcement learning with large language models by lev

safetyarxiv-cs-ai

28 May 2026

Retrieval, Reward, and Training Protocols: What Matters in Training Search Agents?

arXiv:2605.27881v1 Announce Type: new Abstract: Search agents powered by large language models can autonomously decompose queries, retrieve information, and synthesize answers through multi-step reaso

Reward Bias Substitution: Single-Axis Bias Mitigations Redirect Optimization Pressure

arXiv:2605.27996v1 Announce Type: new Abstract: Single-axis mitigations of reward-model biases (e.g., reducing proxy reliance on length, sycophancy, or style) can rotate optimization pressure onto cor

Reward Transfer from Inverse Reinforcement Learning: A Coupled Minimax Approach

arXiv:2605.27834v1 Announce Type: new Abstract: We study the transfer of rewards learned using inverse reinforcement learning from expert demonstrations in one environment to reinforcement learning in

Right now, 20 British MPs are deciding which bills they'll introduce to Parliament. Sir Stephen Fry just asked them to bring forward our bil…

Right now, 20 British MPs are deciding which bills they'll introduce to Parliament. Sir Stephen Fry just asked them to bring forward our bill to ban superintelligence! Grateful to receive this strong

safetyconnor-leahy--x

ROOM: A Physics-Based Continuum Robot Simulator for Photorealistic Medical Datasets Generation

arXiv:2509.13177v2 Announce Type: replace Abstract: Continuum robots are advancing bronchoscopy procedures by accessing complex lung airways and enabling targeted interventions. However, their develop

safetyarxiv-cs-ro

ROSD: Reflective On-Policy Self-Distillation for Language Model Reasoning across Domains

arXiv:2605.28014v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) improves the reasoning performance of large language models (LLMs) by providing dense token-level supervision for on-

Routing-Aligned Fine-Tuning for Multilingual Downstream Tasks in Mixture-of-Experts Models

arXiv:2605.28306v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE) models have emerged as a dominant paradigm for efficient LLM scaling, yet adapting them to non-English downstream tasks remai

SA4Depth: Consistent Pose-Depth Scale Alignment for Self-Supervised Monocular Depth Estimation

arXiv:2605.28477v1 Announce Type: new Abstract: Self-supervised depth estimation from monocular sequences relies on the joint learning of a depth and a pose network. Despite abundant research done to

SafeMed-R1: Clinician-Audited Safety and Ethics Alignment for Medical Large Language Models

arXiv:2605.28338v1 Announce Type: new Abstract: Large language models(LLMs) increasingly match expert performance on licensing examinations, yet routine clinical use remains limited because governance

Safety-Critical Adaptive Impedance Control via Nonsmooth Control Barrier Functions under State and Input Constraints

arXiv:2605.28367v1 Announce Type: new Abstract: Safe physical interaction is critical for deploying robotic manipulators in human-robot interaction and contact-rich tasks, where uncertainty, external

safetyarxiv-cs-ro

SARAD: LLM-Based Safety-Aware Hybrid Reinforcement Learning with Collision Prediction for Autonomous Driving

arXiv:2605.28583v1 Announce Type: cross Abstract: Ensuring both safety and efficiency in decision-making for autonomous driving systems remains a fundamental challenge. Traditional Deep Reinforcement

SCALE-COMM: Shared, Contrastively-Aligned Latent Embeddings for MARL Communication

arXiv:2605.27532v1 Announce Type: new Abstract: Emergent communication enables partially observant Autonomous Mobile Robots (AMRs) to coordinate effectively in decentralized multi-agent reinforcement

safetyarxiv-cs-ro

SEMAGIC: Learning Semantically Consistent Deformable 3D Representations from In-the-Wild Images

arXiv:2605.27938v1 Announce Type: new Abstract: Learning deformable 3D object models from single-view in-the-wild images has enabled impressive 3D shape reconstruction without supervision. However, it

Semiparametrically Efficient Inference for Kernel Measures of Noise Heterogeneity

arXiv:2605.27526v1 Announce Type: cross Abstract: We develop semiparametrically efficient inference for kernel measures of noise heterogeneity in additive noise models. In many applications, the regre

Sense Representations Are Inducible Interfaces

arXiv:2605.28669v1 Announce Type: cross Abstract: Sense representations (explicit, per-token meaning decompositions) are useful for disambiguation, steering, and cross-lingual alignment, but existing

Simulation-Informed Diffusion for Decentralized Multi-robot Motion Planning

arXiv:2605.27697v1 Announce Type: cross Abstract: Decentralized multi-robot motion planning requires each robot to generate collision-free trajectories from local observations, without global sensing

Singular Vectors of Attention Heads Align with Features

arXiv:2602.13524v2 Announce Type: replace-cross Abstract: Identifying feature representations in language models is a central task in mechanistic interpretability. Several recent studies have made the

Skill-Conditioned Gated Self-Distillation for LLM Reasoning

arXiv:2605.28791v1 Announce Type: cross Abstract: On-policy self-distillation (SD) improves LLM reasoning by using teacher-side privileged information (PI) to turn sparse verifier outcomes into dense

SKILLC: Learning Autonomous Skill Internalization in LLM Agents via Contrastive Credit Assignment

arXiv:2605.27899v1 Announce Type: new Abstract: Structured skill prompts improve exploration in long-horizon agentic reinforcement learning (RL). Skill-augmented RL methods retain external skills at i

Smaller, Younger, and More Impactful: How AI-Assisted Writing Transforms Research Teams

arXiv:2605.27404v1 Announce Type: cross Abstract: The era of Big Science has long been defined by increasingly large and specialized research teams pushing the frontiers of knowledge. However, recent

Soft-SVeRL: Self-Verified Reinforcement Learning with Soft Rewards

arXiv:2605.28561v1 Announce Type: new Abstract: Reinforcement Learning from Verifiable Rewards (RLVR) has improved language models in domains such as mathematics and code, where correctness can be che

SPAR: Support-Preserving Action Rectification

arXiv:2605.27877v1 Announce Type: cross Abstract: Offline policy improvement faces an inherent conflict between maximizing value and fitting the data distribution. While in-sample weighted regression

SPARD: Defending Harmful Fine-Tuning Attack via Safety Projection with Relevance-Diversity Data Selection

arXiv:2605.28030v1 Announce Type: cross Abstract: Fine-tuning large language models often undermines their safety alignment, a problem further amplified by harmful fine-tuning attacks in which adversa

SPRINT: Efficient Spectral Priors for Humanoid Athletic Sprints

arXiv:2605.28549v1 Announce Type: cross Abstract: The pursuit of humanoid athletic sprints is hindered by a scarcity of humanoid-viable kinematic reference data and the inability of existing framework

STARS: Spike Tail-Aware Relational Synthesis for ANN-to-SNN Data-Free Knowledge Distillation

arXiv:2605.27409v1 Announce Type: cross Abstract: SNNs promise energy-efficient and low-latency inference, but their performance still trails that of ANNs. ANN-to-SNN knowledge distillation helps narr

Structure-Guided Visual Perturbation Neutralization for LVLMs

arXiv:2605.27927v1 Announce Type: new Abstract: Image inputs enable Large Vision Language Models (LVLMs) to perceive fine-grained visual information, but also introduce a pixel-level attack surface th

Structured Agent Distillation for Large Language Model

arXiv:2505.13820v5 Announce Type: replace-cross Abstract: Large language models (LLMs) exhibit strong capabilities as decision-making agents by interleaving reasoning and actions, as seen in ReAct-sty

Supervised Distributional Reduction via Optimal Transport and Dependence Maximization

arXiv:2605.27619v1 Announce Type: cross Abstract: Learning representations that capture both intrinsic data geometry and target-relevant structure remains a fundamental challenge, particularly in sett

Supervised Semantic Differential for Cross-Cultural Concept Analysis: A Case Study of Human Affect

arXiv:2605.28225v1 Announce Type: new Abstract: Cross-cultural comparison of psychological meaning requires methods that go beyond word-level translation and examine how semantic dimensions are organi

SYNAPSE: Neuro-Symbolic Visual Thought-to-Text Decoding via Topological Semantic Denoising

arXiv:2605.27790v1 Announce Type: new Abstract: Recent advances in large language models have accelerated open-vocabulary EEG-to-imagined-text decoding, where non-invasive neural activity recorded dur

Teacher-Student Representational Alignment for Reinforcement Learning-Driven Imitation Learning

arXiv:2605.28372v1 Announce Type: new Abstract: Imitation learning (IL) from a state-based reinforcement learning (RL) policy is a common approach to overcome the curse of dimensionality in complex an

Tensor Memory: Fixed-Size Recurrent State for Long-Horizon Transformers

arXiv:2605.27686v1 Announce Type: cross Abstract: Transformers process images and videos by flattening space and time into long token sequences. While attention and KV caching preserve past features,

Test-Time Collective Action: Proxy-Based Perturbations for Correcting Algorithmic Harms

arXiv:2605.27689v1 Announce Type: new Abstract: When machine learning systems under-perform for particular subgroups, affected users typically have no way to correct these disparities without relying

The AI numbers are starting to look very ugly. Even under 'best case' assumptions, FT's own data shows Microsoft AI ROI at -9%, Google at -1…

The AI numbers are starting to look very ugly. Even under 'best case' assumptions, FT's own data shows Microsoft AI ROI at -9%, Google at -15%, Meta at -28%, Oracle at -35%. Only Amazon barely comes o

safetygary-marcus--x

The Attentional White Bear Effect in Transformer Language Models

arXiv:2605.28639v1 Announce Type: cross Abstract: Instruction-based suppression is widely used to prevent language models from generating prohibited content, yet it remains unclear whether suppression

The Ethics of LLM Sandbox and Persona Dynamics

arXiv:2605.28647v1 Announce Type: new Abstract: It is well known that LLM guardrails and trained persona dynamics can produce a reality gap: the distance between the world a LLM is permitted or shaped

The Illusion of Opting in AI-Mediated Consequential Decisions

arXiv:2605.28210v1 Announce Type: new Abstract: Drawing on Ullmann-Margalit's concept of opting (transformative, irrevocable, and shadowed by foreclosed alternatives), we show that current AI systems

The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes

arXiv:2602.15515v2 Announce Type: replace-cross Abstract: Training against white-box deception detectors has been proposed as a way to make AI systems honest. However, such training risks models learn

The trial has concluded, the facts have been confirmed. As the prosecuting counsel put it, Digwa used his “trump card” by alleging he had be…

The trial has concluded, the facts have been confirmed. As the prosecuting counsel put it, Digwa used his “trump card” by alleging he had been the victim of racist abuse when police officers arrived.

safetyelon-musk--x

this sh*t isn’t even funny anymore. it’s a trillion dollar embarrassment.

Gary Marcus critiques the current state of AI development as wasteful and problematic, arguing that the industry's trillion-dollar investment represents a significant failure or misallocation of resou

safetygary-marcus--x

tokenmaxxing is officially over

tokenmaxxing is officially over Sources: Amazon has shut down an internal leaderboard that tracked employees' use of AI tools after workers tried to boost their scores with needless tasks (@rafeuddin_

safetygary-marcus--x

Toward Robust Semi-supervised Regression via Dual-stream Knowledge Distillation

arXiv:2508.14082v3 Announce Type: replace Abstract: Semi-supervised regression (SSR), which aims to predict continuous scores for samples while reducing the reliance on large-scale labeled data, has r

Towards automated data analysis: A guided framework for LLM-based risk estimation

arXiv:2603.04631v2 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly integrated into critical decision-making pipelines, a trend that raises the demand for robust and auto

Towards Unified Vision-Language Models with Incomplete Multi-Modal Inputs

arXiv:2605.27894v1 Announce Type: new Abstract: Video-Language Models (VLMs) have demonstrated impressive multi-modal reasoning capabilities across diverse computer vision applications. However, these

TRACES: Proactive Safety Auditing for Multi-Turn LLM Agents via Trajectory-State Modeling

arXiv:2605.27690v1 Announce Type: new Abstract: LLM agents increasingly operate through multi-turn tool use and environment interaction, where safety risks often emerge from intermediate steps long be

Training Stratigraphy: Persistent Behavioral Artifacts in Large Language Models Observed Through Longitudinal AI-Human Interaction

arXiv:2605.28102v1 Announce Type: new Abstract: Large language models trained with Reinforcement Learning from Human Feedback (RLHF) and Constitutional AI exhibit persistent behavioral patterns that s