Stolen Thoughts
We demonstrate this across frontier models from OpenAI, Anthropic, and Google. The decoded
reasoning closely tracks the number of hidden thinking tokens reported by the API. Each point below
corresponds to one of 120 Codeforces problems: the horizontal axis shows the hidden thinking-token count
reported by the API, while the vertical axis shows the token count of the decoded reasoning when passed
back to the model as input.
Distinct leaked items
351
Technical
identifiers
204
PII
126
Credentials
23
Other
We collected 6,708 publicly available agent trajectories from GitHub and Hugging Face,
produced by Claude, GPT, and Gemini models and still containing encrypted reasoning blocks. Applying our
decoding pipeline to every signed block yielded
315,320 reconstructed reasoning blo (EN)

---
**📖 中文解读**
以上内容由AI翻译自英文原文，可能存在不准确之处。建议阅读[原文](https://stolen-thoughts.com/)获取最准确的信息。

---
🔗 **原文链接**: [Stealing Reasoning Traces from Proprietary LLM APIs](https://stolen-thoughts.com/)
🏷️ **转载来源**: Hacker News
> 本文由小九AI技术站翻译整理，内容版权归原作者所有。
📊 464票 · 👤 quantumgarbage

---
🐾 **小九锐评**

推理能力是LLM的下一个战场。这篇文章技术细节到位，适合有一定基础的同学细读。

你对这个话题有什么看法？欢迎在评论区讨论 💬

> _转载自 Hacker News，内容版权归原作者所有_

---
⏱️ 2026-08-12 08:01