When Words Are Safe But Actions Kill: Probing Physical Danger Beyond Text Safety in Hidden-State Risk Space

arXiv:2607.15218 · cs.AI, cs.CR · Submitted 2026-08-21 · Read on arXiv

Weimeng Wang, Ziqiang Wang, Zihang Zhan, Chuanpu Fu, Qi Li, Ke Xu

cs.AI, cs.CR

Submitted: 2026-08-21

Updated: 2026-08-24

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Benchmark — 28× in this paper · explained in 2 episodes
• Accuracy — 24× in this paper · explained in 1 episode
• LLM judge — 22× in this paper · explained in 2 episodes
• AUC — 11× in this paper · explained in 2 episodes
• Ablation — 8× in this paper · explained in 1 episode
• Benchmarking — 4× in this paper · explained in 3 episodes
• In International Conference on Learning Representations — 4× in this paper
• Large Language Models (LLMs) — 3× in this paper · explained in 6 episodes
• Linear probe — 3× in this paper · explained in 2 episodes
• Representation Engineering — 3× in this paper · explained in 2 episodes
• Alignment — 2× in this paper · explained in 1 episode
• Calibration — 2× in this paper · explained in 7 episodes
• Embodied AI — 2× in this paper · explained in 2 episodes
• Failure Modes — 2× in this paper · explained in 3 episodes
• Grounding — 2× in this paper · explained in 3 episodes
• In International Conference on Machine Learning — 2× in this paper
• International Conference on Robotics and Automation — 2× in this paper
• Model-Agnostic — 2× in this paper · explained in 2 episodes
• Probing Representations for Integrated Safety Monitoring — 2× in this paper
• Robustness — 2× in this paper · explained in 5 episodes
• Steering — 2× in this paper · explained in 2 episodes
• Task Planning — 2× in this paper · explained in 2 episodes
• AI Agents — 1× in this paper · explained in 2 episodes
• Adversarial Attacks — 1× in this paper · explained in 1 episode
• Affordance — 1× in this paper · explained in 1 episode
• CVF International Conference on Computer Vision — 1× in this paper
• Conference on Computer Vision and Pattern Recognition — 1× in this paper
• Content Danger Direction — 1× in this paper
• Empirical Methods in Natural Language Processing — 1× in this paper
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Foundation Model — 1× in this paper · explained in 3 episodes
• Generalization — 1× in this paper · explained in 13 episodes
• Generative AI — 1× in this paper · explained in 2 episodes
• LLM-as-a-Judge — 1× in this paper · explained in 3 episodes
• Linear Probing — 1× in this paper · explained in 2 episodes
• Physical Danger Direction — 1× in this paper
• Regularization — 1× in this paper · explained in 3 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes
• Task Decomposition — 1× in this paper · explained in 2 episodes
• Transformer — 1× in this paper · explained in 2 episodes
• Zero-Shot Transfer — 1× in this paper · explained in 4 episodes

Sources

• Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone
• Do As I Can, Not As I Say: Grounding Language in Robotic Affordances
• Understanding intermediate layers using linear classifier probes
• Constitutional AI: Harmlessness from AI Feedback
• SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model
• RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control
• SafeMind: Benchmarking and Mitigating Safety Risks in Embodied LLM Agents
• Efficient LLM Moderation with Multi-Layer Latent Prototypes
• AEGIS: Online Adaptive AI Content Safety Moderation with Ensemble of LLM Experts
• A Framework for Benchmarking and Aligning Task-Planning Safety in LLM-Based Embodied Agents
• Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations
• AI2-THOR: An Interactive 3D Environment for Visual AI
• AGENTSAFE: Benchmarking the Safety of Embodied Agents on Hazardous Instructions
• The Llama 3 Herd of Models
• IS-Bench: Evaluating Interactive Safety of VLM-Driven Embodied Agents in Daily Household Tasks
• The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets
• Don't Let Your Robot be Harmful: Responsible Robotic Manipulation via Safety-as-Policy
• SafePlan: Leveraging Formal Logic and Chain-of-Thought Reasoning for Enhanced Safety in LLM-based Robotic Task Planning
• GPT-4 Technical Report
• Qwen2.5 Technical Report

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
• VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection