[2603.07267] How to Steal Reasoning Without Reasoning Traces

Abstract:Many large language models (LLMs) use reasoning to generate responses but do not reveal their full reasoning traces (a.k.a. chains of thought), instead outputting only final answers and brief reasoning summaries. To demonstrate that hiding reasoning traces does not prevent users from "stealing" a model's reasoning capabilities, we introduce trace inversion models that, given only the inputs, answers, and (optionally) reasoning summaries exposed by a target model, generate detailed, synthetic reasoning traces. We show that (1) traces synthesized by trace inversion have high overlap with the ground-truth reasoning traces (when available), and (2) fine-tuning student models on inverted traces substantially improves their reasoning and enables distillation from proprietary, black-box LLMs.

How to Steal Reasoning Without Reasoning Traces Tingwei Zhang John X. Morris Vitaly Shmatikov Department of Computer Science Cornell Tech arXiv:2603.07267v2 [cs.CR] 12 May 2026 Abstract Many large language models (LLMs) use…

saved by

• Joshua Baek
• Julian H
• Will Anderson

related reading

• [2606.04071] Covert Influence Between Language Modelsarxiv.org
• [2608.09867] Stealing Reasoning Traces from Proprietary LLM APIsarxiv.org
• [2506.19733] Breaking Barriers: Do Reinforcement Post Training Gains Transfer To Unseen Domains?arxiv.org
• Stealing Reasoning Traces from Proprietary LLM APIsarxiv.org
• Stolen Thoughtsstolen-thoughts.com
• [2201.11903] Chain of Thought Prompting Elicits Reasoning in Large Language Modelsarxiv.org
• [2412.06769] Training Large Language Models to Reason in a Continuous Latent Spacearxiv.org
• Stealing Reasoning Traces from Proprietary LLM APIsresearch.snyk.io
• [2203.14465] STaR: Bootstrapping Reasoning With Reasoningarxiv.org
• [2507.11473] Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safetyarxiv.org
• [2604.16812] Introspection Adapters: Training LLMs to Report Their Learned Behaviorsarxiv.org
• Base Models Know How to Reason, Thinking Models Learn Whenarxiv.org