RECAST: REconstruction of Chain-of-Thought via Auditable Structured Traces

📄 arXiv: 2610.41813v1 📥 PDF

作者: Qi Wang, Kang Xu, Chengcheng Wan, Jiangtao Wang

分类: cs.AI

发布日期: 2026-10-30


💡 一句话要点

提出RECAST框架以重建闭源模型的推理过程

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 推理透明性 闭源模型 链式推理 条件解码器 数据重建 人工智能可解释性 监督微调

📋 核心要点

  1. 现有的推理模型提取方法无法有效获取闭源模型的中间推理过程,导致推理透明性不足。
  2. RECAST框架通过条件解码器重建推理过程,避免依赖隐藏的推理通道,创新性地实现了结构化推理的生成。
  3. 在Claude-Sonnet-5.5上成功重建推理过程,并在GPT-6.1和Gemini 3上获得了良好的定性结果,展示了该方法的有效性。

📝 摘要(中文)

大型闭源推理模型仅暴露最终答案,而其产生答案的中间推理过程则隐藏在API接口、工具调用策略和安全过滤器之后。现有的链式推理(CoT)提取方法假设存在可见的推理通道,但在Claude-Sonnet-5.5的实验中发现,三种主流提取方法均未能返回任何CoT令牌。为此,本文提出了RECAST框架,通过条件解码器重建结构化推理过程,成功在Claude-Sonnet-5.5、GPT-6.1和Gemini 3上实现了高质量的推理重建。该系统为开源,能够生成适用于监督微调的数据。

🔬 方法详解

问题定义:本文旨在解决现有推理模型中间过程不可见的问题,现有方法在Claude-Sonnet-5.5等模型上无法提取有效的链式推理(CoT)信息。

核心思路:RECAST框架通过条件解码器,基于问题和目标模型的最终答案重建推理过程,清洗并生成适用于监督微调的数据,避免了对隐藏推理通道的依赖。

技术框架:该框架主要包括三个模块:条件解码器、推理过程重建和数据清洗。首先,条件解码器接收问题和答案,生成初步的推理结构;接着,重建模块对推理进行结构化处理;最后,清洗模块确保生成数据的质量和适用性。

关键创新:RECAST是首个从前沿闭源模型中重建链式推理的系统,显著不同于现有方法的依赖于可见推理通道的设计,提供了一种全新的思路。

关键设计:在实现过程中,采用了特定的损失函数来优化重建的推理结构,并设计了适应不同模型的参数设置,以确保生成的推理过程具有高质量和一致性。

📊 实验亮点

在Claude-Sonnet-5.5上,RECAST成功重建了推理过程,并在GPT-6.1和Gemini 3上获得了定性重建的成功,展示了该方法在闭源模型推理透明性方面的显著提升,且未使用任何隐藏推理通道。

🎯 应用场景

RECAST框架的潜在应用领域包括教育、智能问答系统和推理透明性研究等。通过重建推理过程,用户能够更好地理解模型的决策依据,从而提升信任度和可解释性,推动人工智能在各个领域的应用。

📄 摘要(原文)

Large closed-source reasoning models expose only their final answers. The intermediate reasoning that produces those answers is hidden behind API surfaces, tool-call policies, and safety filters. Existing chain-of-thought (CoT) extraction methods assume a side channel: a visible reasoning field, a scratchpad carried through tool calls, or a signed/encrypted reasoning block. We reproduce three representative state-of-the-art extraction methods and find that, on Claude-Sonnet-5.5 accessed through the standard chat-completions interface, all three return zero CoT tokens. The failure is structural rather than a bug: the reasoning field is absent, named tool selection is rejected, free-text scratchpad arguments are filtered, and no replayable reasoning block is exposed. We therefore take a different route. We present RECAST, a reconstruction-based framework that conditions a decoder on the question and the target model's final answer, reconstructs a first-person structured rationale, cleans it, and builds SFT-ready data. To the best of our knowledge, RECAST is the first system to reconstruct CoT from frontier closed models such as Claude-Sonnet-5.5, GPT-6.1, and Gemini 3 without using hidden reasoning channels. We instantiate RECAST on Claude-Sonnet-5.5 and additionally obtain successful qualitative reconstructions on GPT-6.1 and Gemini 3 with the same answer-only protocol. The system is open-source and packaged as an agent skill: one prompt deploys it and runs the complete pipeline to produce SFT-ready CoT data. This preprint describes the system, its threat model, and a controlled artifact validation; it does not claim large-scale benchmark improvements.