通过提示诱导暴露大模型内部推理过程，提升可解释性。
Hidden Thoughts Are Not Secret: Reasoning Trace Exposure in LLMs
- 用辅助代码格式的演示诱导模型显式输出推理路径。
- 使用户可见的推理痕迹与内部真实路径相似度显著提升。
- 适合关注模型可解释性与知识蒸馏的研究者使用。
推理痕迹已成为提升和迁移大语言模型能力的重要学习信号。特别是详细推理痕迹有助于将强教师模型的推理行为蒸馏到弱学生模型中。由于能力迁移的价值，许多部署的推理系统会隐藏原始内部痕迹，仅向用户提供摘要和答案。因此，我们提出：这种接口层面的痕迹隐藏是否能阻止用户通过提示获取有用的推理监督？为此，本文提出轻量级上下文内诱导方法——推理暴露提示（REP），利用影子模型生成的示范，以辅助代码类格式包装，从而从目标模型中唤醒用户可见的推理痕迹。在多个常见推理数据集、不同目标模型及学生模型蒸馏任务中，REP 显著提升了暴露痕迹与 REP 条件下内部痕迹之间的相似度，同时保留了有效的推理信号。
原文摘要 · Abstract (English)
Reasoning traces have become a valuable form of learning signals for improving and transferring the capabilities of large language models. In particular, detailed traces can help distill reasoning behavior from stronger teacher models into weaker student models. The value of capability transfer has motivated many deployed systems with reasoning models to hide raw internal traces and expose at most summaries and answers to users. As a result, we ask whether such interface-level trace hiding prevents users from obtaining useful reasoning supervision through prompting. We study this question with Reasoning Exposure Prompting (REP), a lightweight in-context elicitation method that uses shadow-model-generated demonstrations wrapped in auxiliary code-like formats to raise user-visible reasoning traces from a victim model. Across the common reasoning dataset, different victim models, and different student model distillation, REP substantially increases similarity between exposed and REP-conditioned internal traces while preserving useful reasoning signals.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。