safety

A Unifying Lens on Reward Uncertainty in RLHF

arXiv:2606.09073v1 Announce Type: cross Abstract: Reinforcement learning from human feedback (RLHF) is bottlenecked by reward hacking, where the policy exploits errors in a proxy reward model (RM) and

A VideoMAE-v2 Approach to Zero-Shot Traffic Accident Anticipation

arXiv:2606.09542v1 Announce Type: new Abstract: Traffic accident anticipation -- predicting the likelihood of an imminent collision at every frame of a dashcam video -- is safety-critical yet difficul

safetyarxiv-cs-cv

9 Jun 2026

Ablation-Reversible Heads Don't Transfer: A Stress Test for Mechanistic Role Claims in Transformers

arXiv:2606.08292v1 Announce Type: new Abstract: In mechanistic interpretability, attention heads are commonly elevated to role claims (e.g., 'this head represents addition') when they are necessary fo

absolutely correct

absolutely correct @GaryMarcus I've recently learned not to believe anything that anyone says anymore when it comes to AI companies. They're all competing to win against each other while exploiting, s

ActProbe: Action-Space Probe for Early Failure Detection of Generative Robot Policies

arXiv:2606.08508v1 Announce Type: cross Abstract: Generative robot policies fail unpredictably at deployment: they hesitate at critical moments, drift off-task, or commit to unrecoverable actions. Exi

Adaptive Loss Balancing for Noise-Robust GRPO in Generative Recommendation

arXiv:2606.08480v1 Announce Type: cross Abstract: Reinforcement learning (RL) presents a promising avenue for enhancing generative recommendation beyond supervised imitation, leveraging reward signals

AeroSpectra Sentinel: An Auditable LLM Prompt-Chaining Decision-Support Workflow for Acute Asthma Risk Assessment from Respiratory Sounds and Clinical Signals

arXiv:2606.08247v1 Announce Type: cross Abstract: Acute asthma risk assessment requires rapid interpretation of respiratory sounds, oxygenation, airflow limitation, speech ability, work of breathing,

Agent Economics: An Entropy-Controlled Pluralistic Alignment Framework for Preventing Artificial Hivemind in Autonomous Agents

arXiv:2606.09039v1 Announce Type: new Abstract: This study proposes the Behavioral Protocol Framework (BPF), an entropy-controlled pluralistic alignment framework designed to address two critical chal

AgriGov: A Structured Multilingual Dataset Curation for Indian Government Schemes for Farmers

arXiv:2606.08272v1 Announce Type: cross Abstract: AgriGov is a curated, trilingual (English-Hindi-Marathi) dataset designed to address the scarcity of domain-grounded multilingual resources for agricu

AHA-WAM:Asynchronous Horizon-Adaptive World-Action Modeling with Observation-Guided Context Routing

arXiv:2606.09811v1 Announce Type: cross Abstract: World-action models have emerged as a promising paradigm for robot manipulation, jointly modeling visual scene dynamics and actions to inject physical

AI Assurance in UK Defence: Challenges in Operationalising JSP 936

arXiv:2606.09414v1 Announce Type: cross Abstract: This report examines practical challenges in operationalising JSP 936 Part 1 for AI assurance in UK Defence. Using a structured interpretive review of

AI Code Sandboxes: A Comparative Security Study. Part 1 of 2 -- Engine-Level Properties (Attack Surface, Leakage, Stackability, CVE History, Patch Cadence, Fuzzing)

arXiv:2606.08433v1 Announce Type: cross Abstract: This paper reads six engine-level measurements together -- 1.1 host attack surface, 1.2 information leakage, 1.3 defense-in-depth stackability, 1.4 pu

AI-Integrated Learning Management System for Middle School: A Longitudinal Study of Learning Outcomes Through High School and Beyond

arXiv:2606.07544v1 Announce Type: cross Abstract: Middle school is a key window for building core academic skills and the learning routines students carry into later grades, yet many students still fa

alienating your customers before you IPO is maybe not the best idea, @AnthropicAI

alienating your customers before you IPO is maybe not the best idea, @AnthropicAI Brilliant idea! Next up: Apple randomly reboots your Mac if you're building competing tech, Gmail silently edits your

Aligned but Not Partner-Specific: Distinguishing How Multimodal LLM Agents Succeed in Reference Games Without Human-Like Conventions

arXiv:2606.08081v1 Announce Type: cross Abstract: Repeated reference games test whether interlocutors replace their initially long descriptions with shorter, partner-specific conventions grounded in s

AMix-1: A Pathway to Test-Time Scalable Protein Foundation Model

arXiv:2507.08920v4 Announce Type: replace-cross Abstract: We introduce AMix-1, a powerful protein foundation model built on Bayesian Flow Networks and empowered by a systematic training methodology, e

An Agency-Transferring Model-Free Policy Enhancement Technique

arXiv:2606.09825v1 Announce Type: cross Abstract: Training reinforcement learning (RL) policies from scratch is costly: it requires careful reward and environment design, extensive tuning, and substan

Anchor-Conditioned Compositional Control for Landscape Image Generation

arXiv:2606.07638v1 Announce Type: cross Abstract: Image generative models, though widely used as creative tools, offer limited support for the kind of compositional control that photographers and visu

Anthropic didn’t just add guardrails to make Mythos safer; they added guardrails to protect their own IP. Their own IP. They are still as …

Anthropic didn’t just add guardrails to make Mythos safer; they added guardrails to protect their own IP. *Their own IP*. They are still as happy as fuck to build their AI on other people’s IP. Intere

Anthropic played the media like a fiddle. From “untold catastrophe” to “check our latest model”, in two months and a day 🙄 (cc @tomfriedman…

Anthropic played the media like a fiddle. From “untold catastrophe” to “check our latest model”, in two months and a day 🙄 (cc @tomfriedman) This is the scary phase of AI — a model deemed so powerful

Automated Framework to Evaluate and Harden LLM System Instructions against Encoding Attacks

arXiv:2604.01039v2 Announce Type: replace-cross Abstract: System Instructions in Large Language Models (LLMs) are commonly used to enforce safety policies, define agent behavior, and protect sensitive

Autonomous Aerial Manipulation via Contextual Contrastive Meta Reinforcement Learning

arXiv:2606.08533v1 Announce Type: new Abstract: Unmanned aerial vehicles (UAVs) are increasingly being deployed in logistics, service robotics, and other real-world applications, creating a growing de

Autonomous FPV Flight with Translational Optical Flow and Uncertainty Mask

arXiv:2606.09088v1 Announce Type: new Abstract: Autonomous FPV quadrotor flight in complex environments using a monocular RGB camera as the sole exteroceptive sensor remains a fundamental challenge. R

safetyarxiv-cs-ro

Autonomous Incident Resolution at Hyperscale: An Agentic AI Architecture for Network Operations

arXiv:2606.09122v1 Announce Type: cross Abstract: Cloud network infrastructure at hyperscale presents unique operational challenges where traditional human-driven incident response cannot keep pace wi

Autonomous Obstacle Removal for Excavators through Policy Learning with Particle Simulation

arXiv:2606.09183v1 Announce Type: new Abstract: Autonomous obstacle removal from the ground is an important earthwork task, but this is difficult to automate because an excavator must adapt its excava

safetyarxiv-cs-ro

Back to the Familiar Future: Failure Recovery for VLA Policies via Pre-Imagined Milestone Selection

arXiv:2606.09258v1 Announce Type: new Abstract: Vision-language-action (VLA) policies can deviate from nominal trajectories during manipulation, even when tasks remain physically feasible. Recovering

safetyarxiv-cs-ro

Baichuan-M4: A Clinical-Grade Medical Agent System for Continuous Care

arXiv:2606.08982v1 Announce Type: new Abstract: Baichuan-M4 is Baichuan Intelligence's clinical-grade medical large model, designed for continuous care rather than single-turn medical question answeri

Bandits for Efficient Experimentation: Adapting to Control Group, Preferences, and Context Drifts

arXiv:2606.09802v1 Announce Type: cross Abstract: We consider a variant of the linear contextual stochastic multi-armed bandits, where the learner must provide recommendations to a group of users, eac

BareWave: Waveform-Native Flow-Matching Text-to-Speech

arXiv:2606.09048v1 Announce Type: cross Abstract: Removing intermediate representations and separately trained decoding stages has become an important direction in generative modeling. In text-to-spee

Beware of GeeksBearing Gifts: Building True EU Frontier AI Sovereignty

arXiv:2606.07536v1 Announce Type: cross Abstract: Frontier artificial intelligence is reshaping all aspects of society, from economic output or military capability to democratic institutions. The EU i

Beyond Accuracy: Interpreting Topic Representation in Suicide Ideation Detection Models

arXiv:2606.07714v1 Announce Type: cross Abstract: Suicide ideation detection models are typically evaluated using aggregate performance metrics, yet little is known about how they internally represent

Beyond Homophily: Towards Generalized Graph Reconstruction Attack and Defense

arXiv:2606.08067v1 Announce Type: new Abstract: Graph neural networks (GNNs) are widely deployed on relational data, yet they can leak sensitive or proprietary information about the training graph adj

Beyond Linear Activation Steering: Invertible Latent Transformations for Controlling LLM Behavior

arXiv:2606.08454v1 Announce Type: new Abstract: Activation steering provides a lightweight inference-time mechanism for controlling large language models (LLMs) by modifying their internal activation

Beyond Neural Collapse: Task-Intrinsic Geometry Governs Neural Representations in Modular Arithmetic

arXiv:2606.08985v1 Announce Type: new Abstract: While neural collapse (NC) predicts that a K-class-balanced classifier should organize terminal representations as a (K-1)-dimensional simplex equiangul

Beyond Scalar Rewards by Internalizing Reasoning into Score Distributions

arXiv:2606.09076v1 Announce Type: new Abstract: Reward models are central to text-to-image post-training, but visual preference is subjective and better represented as a distribution over rubric score

safetyarxiv-cs-cv

Boundary Variance Inflation Causes Acquisition Bias in Gaussian Processes

arXiv:2606.07561v1 Announce Type: new Abstract: Gaussian processes with stationary kernels on bounded domains exhibit inflated posterior variance near the boundary. Despite being a long-recognized art

Brain-Prompt Injection: A Route-Safety Audit for BCI-LLM Agents

arXiv:2606.09315v1 Announce Type: cross Abstract: BCI-to-agent pipelines turn decoded neural activity into an authorization channel for tool-use agents, exposing a new attack surface we call brain-pro

Breaking the Tokenizer Barrier: On-Policy Distillation across Model Families

arXiv:2606.09456v1 Announce Type: new Abstract: On-Policy Distillation (OPD) has become a core technique in the post-training of Large Language Models (LLMs) for transferring knowledge from domain exp

Bridging Expert Knowledge and Automated Feature Engineering via Self-Evolution

arXiv:2606.08800v1 Announce Type: new Abstract: In high-stakes settings such as brand compliance, clinical care, and content moderation, machine learning cannot be deployed as opaque oracles: practiti

Bridging Traditional Explainability Methods and Multimodal Multilingual Models: An XAI-Based Analysis

arXiv:2606.07533v1 Announce Type: cross Abstract: Multimodal Large Language Models (MLLMs) effectively integrate text and audio to interpret context in complex interactive dialogues. However, the inte

Brilliant idea! Next up: Apple randomly reboots your Mac if you're building competing tech, Gmail silently edits your email if you mention r…

Brilliant idea! Next up: Apple randomly reboots your Mac if you're building competing tech, Gmail silently edits your email if you mention rival platforms, and Tesla Autopilot swerves if it detects yo

safetyjeremy-howard--x

Can Data Work be Reparative?

arXiv:2606.09408v1 Announce Type: cross Abstract: We present an ethnographic study of an alternative approach to data work, developed by a civic-tech initiative that builds datasets for training and b

Can the Environment Speak for Itself? T^{2}-GRPO: A Turn-Trajectory Group Relative Policy Optimization for Caregiver Agents

arXiv:2606.08875v1 Announce Type: new Abstract: Optimizing large language models (LLMs) for long-horizon caregiver agents requires balancing delayed task objectives with immediate environment dynamics

Capability-Aligned Hierarchical Learning for Tool-Augmented LLMs

arXiv:2606.09371v1 Announce Type: new Abstract: Tool learning enables LLMs to invoke external tools to accomplish tasks. Prior studies have demonstrated the effectiveness of a hierarchical structure:

CARE: A Conformal Safety Layer for Medical Summarization

arXiv:2606.08969v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used for medical summarization, but their outputs can omit medically important information and introduce

Causal Semantic Alignment for LLM-based Time Series Forecasting

arXiv:2606.08262v1 Announce Type: new Abstract: Recent advances in Large Language Models (LLMs) have opened new possibilities for time series forecasting by enabling alignment between temporal pattern

Causal Transfer in Medical Image Analysis

arXiv:2603.24388v2 Announce Type: replace Abstract: Medical imaging models frequently fail when deployed across hospitals, scanners, populations, or imaging protocols due to domain shift, limiting the

safetyarxiv-cs-cv

CheXanatomy: Anatomy-Aware Vision-Language Modeling for Chest Radiographs

arXiv:2606.08420v1 Announce Type: new Abstract: Vision-language models (VLMs) pretrained on large-scale image-text pairs demonstrate strong image-level understanding, but are primarily optimized for g

safetyarxiv-cs-cv