[2310.18512] Preventing Language Models From Hiding Their Reasoning

Large language models (LLMs) often benefit from intermediate steps of reasoning to generate answers to complex problems. When these intermediate steps of reasoning are used to monitor the activity of the model, it is e…

Preventing Language Models From Hiding Their Reasoning Fabien Roger ∗ Ryan Greenblatt Redwood Research Abstract Large language models (LLMs) often benefit from intermediate steps of reasoning to generate answers to complex problems. When these intermediate steps of reasoning are used to monitor the activity of the model, it is essential that this explicit reasoning is faithful, i.e. that it reflects what the model is actually reasoning about. In this work, we focus on one potential way intermediate steps of reasoning could be unfaithful: encoded reasoning, where an LLM could encode intermediat

related reading

• [2603.07267] How to Steal Reasoning Without Reasoning Tracesarxiv.org
• [2412.06769] Training Large Language Models to Reason in a Continuous Latent Spacearxiv.org
• [2201.11903] Chain of Thought Prompting Elicits Reasoning in Large Language Modelsarxiv.org
• [2602.23163] A Decision-Theoretic Formalisation of Steganography With Applications to LLM Monitoringarxiv.org
• [2312.06942] AI Control: Improving Safety Despite Intentional Subversionarxiv.org
• [2510.09714] All Code, No Thought: Current Language Models Struggle to Reason in Ciphered Languagearxiv.org
• [2608.09867] Stealing Reasoning Traces from Proprietary LLM APIsarxiv.org
• [2604.16812] Introspection Adapters: Training LLMs to Report Their Learned Behaviorsarxiv.org
• [2512.11949] Neural Chameleons: Language Models Can Learn to Hide Their Thoughts from Unseen Activation Monitorsarxiv.org
• Training Large Language Models to Reason in a Continuous Latent Spacearxiv.org
• [2606.04071] Covert Influence Between Language Modelsarxiv.org
• [2511.08579] Training Language Models to Explain Their Own Computationsarxiv.org