用熵值自动拆分推理步骤，过滤错误中间过程，提升大模型数学推理能力。
EntroCoT: Enhancing Chain-of-Thought via Adaptive Entropy-Guided Segmentation
- 基于熵值动态分割推理链，在不确定节点处切分步骤
- 通过蒙特卡洛回溯评估每步贡献，筛选出真正有效的中间推理
- 适合提升数学推理类模型训练质量，尤其关注逻辑严谨性
Chain-of-Thought（CoT）提示显著提升了大语言模型的数学推理能力。我们发现现有微调数据集普遍存在‘答案正确但推理错误’的问题，即最终答案虽对，但中间步骤存在幻觉、冗余或逻辑无效。本文提出EntroCoT，一个统一框架，用于自动识别并优化低质量的CoT监督轨迹。EntroCoT首先设计一种基于熵的机制，在推理不确定性高的节点自动分割推理链为多个步骤；随后引入蒙特卡洛回溯机制，评估每一步的边际贡献。通过精准过滤误导性推理样本，EntroCoT构建了一个高质量数据集，其中每个推理步骤均真实推动最终答案得出。在多个数学基准测试上的实验表明，使用EntroCoT生成的子集进行微调，性能持续优于全数据集监督基线。
原文摘要 · Abstract (English)
Chain-of-Thought (CoT) prompting has significantly enhanced the mathematical reasoning capabilities of Large Language Models. We find existing fine-tuning datasets frequently suffer from the "answer right but reasoning wrong" probelm, where correct final answers are derived from hallucinated, redundant, or logically invalid intermediate steps. This paper proposes EntroCoT, a unified framework for automatically identifying and refining low-quality CoT supervision traces. EntroCoT first proposes an entropy-based mechanism to segment the reasoning trace into multiple steps at uncertain junctures, and then introduces a Monte Carlo rollout-based mechanism to evaluate the marginal contribution of each step. By accurately filtering deceptive reasoning samples, EntroCoT constructs a high-quality dataset where every intermediate step in each reasoning trace facilitates the final answer. Extensive experiments on mathematical benchmarks demonstrate that fine-tuning on the subset constructed by EntroCoT consistently outperforms the baseslines of full-dataset supervision.
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。