Stress Testing Chain-of-Thought Monitoring Against Covert Misalignment
cs.CL, cs.AI, cs.LG
Submitted: 2026-11-05
Updated: 2026-11-05
Comments: Advances in Neural Information Processing Systems (NeurIPS), 2026. 51 pages, 8 figures
Code: https://github.com/T-STAR-Lab/TraceMRT
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks
- CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring
- Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation
- Sabotage Evaluations for Frontier Models
- Can We Predict Alignment Before Models Finish Thinking? Towards Monitoring Misaligned Reasoning Models
- AI Deception: Risks, Dynamics, and Controls
- Reasoning Models Don't Always Say What They Think
- When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors
- Detecting Strategic Deception Using Linear Probes
- Alignment faking in large language models
- Monitoring Monitorability
- Reliable Weak-to-Strong Monitoring of LLM Agents
- BashArena: A Control Setting for Highly Privileged AI Agents
- Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety
- SHADE-Arena: Evaluating Sabotage and Monitoring in LLM Agents
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- AgentDoG: A Diagnostic Guardrail Framework for AI Agent Safety and Security
- AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models
- Agentic Misalignment: How LLMs Could Be Insider Threats
- Natural Emergent Misalignment from Reward Hacking in Production RL
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data