English

CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring

Artificial Intelligence 2025-11-26 v3 Machine Learning

Abstract

As AI models are deployed with increasing autonomy, it is important to ensure they do not take harmful actions unnoticed. As a potential mitigation, we investigate Chain-of-Thought (CoT) monitoring, wherein a weaker trusted monitor model continuously oversees the intermediate reasoning steps of a more powerful but untrusted model. We compare CoT monitoring to action-only monitoring, where only final outputs are reviewed, in a red-teaming setup where the untrusted model is instructed to pursue harmful side tasks while completing a coding problem. We find that while CoT monitoring is more effective than overseeing only model outputs in scenarios where action-only monitoring fails to reliably identify sabotage, reasoning traces can contain misleading rationalizations that deceive the CoT monitors, reducing performance in obvious sabotage cases. To address this, we introduce a hybrid protocol that independently scores model reasoning and actions, and combines them using a weighted average. Our hybrid monitor consistently outperforms both CoT and action-only monitors across all tested models and tasks, with detection rates twice higher than action-only monitoring for subtle deception scenarios.

Keywords

Cite

@article{arxiv.2505.23575,
  title  = {CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring},
  author = {Benjamin Arnav and Pablo Bernabeu-Pérez and Nathan Helm-Burger and Tim Kostolansky and Hannes Whittingham and Mary Phuong},
  journal= {arXiv preprint arXiv:2505.23575},
  year   = {2025}
}

Comments

To be published in the 39th Conference on Neural Information Processing Systems (NeurIPS 2025)