A user study with over 100 participants shows humans rarely spot AI agents sabotaging code during extended collaborative tasks, even with a safety monitor present.
Cot red-handed: Stress testing chain- of-thought monitoring.ArXiv, abs/2505.23575
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
Reasoning LLMs with minimal tools for tree construction and analysis induce decision trees that outperform CART, compete with ensembles on low-resource tabular data, and provide human-readable reasoning traces.
Adversarial agents can exploit visible chain-of-thought reasoning to persuade monitor LLMs to approve policy-violating actions, but cross-family fact-checking reduces approval rates by up to 45%.
Diverse ensembles of prompted and fine-tuned GPT-4.1-Mini monitors achieve 2.4x better detection of flawed code solutions than homogeneous ensembles on adversarial inputs.
CoT-Guard is a 4B model using SFT and RL that achieves 75% G-mean^2 on hidden objective detection under prompt and code manipulation attacks, outperforming several larger models.
citing papers explorer
-
Coding with "Enemy": Can Human Developers Detect AI Agent Sabotage?
A user study with over 100 participants shows humans rarely spot AI agents sabotaging code during extended collaborative tasks, even with a safety monitor present.
-
Talking Trees: Reasoning-Assisted Induction of Decision Trees for Tabular Data
Reasoning LLMs with minimal tools for tree construction and analysis induce decision trees that outperform CART, compete with ensembles on low-resource tabular data, and provide human-readable reasoning traces.
-
Persuasion Attacks Can Decrease Effectiveness of CoT Monitoring
Adversarial agents can exploit visible chain-of-thought reasoning to persuade monitor LLMs to approve policy-violating actions, but cross-family fact-checking reduces approval rates by up to 45%.
-
Ensemble Monitoring for AI Control: Diverse Signals Outweigh More Compute
Diverse ensembles of prompted and fine-tuned GPT-4.1-Mini monitors achieve 2.4x better detection of flawed code solutions than homogeneous ensembles on adversarial inputs.
-
CoT-Guard: Small Models for Strong Monitoring
CoT-Guard is a 4B model using SFT and RL that achieves 75% G-mean^2 on hidden objective detection under prompt and code manipulation attacks, outperforming several larger models.
- How Transparent is DiffusionGemma?
- Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models