arXiv 2410.056952024-10-28cs.CL

释放思维能力：一个用于量化和优化思维链的推理边界框架

Unlocking the Capabilities of Thought: A Reasoning Boundary Framework to Quantify and Optimize Chain-of-Thought

School of Computer Science and Engineering, Central South University（中南大学计算机科学与工程学院）；The Chinese University of Hong Kong（香港中文大学）

Qiguang Chen, Libo Qin, Jiaqi Wang, Jinxuan Zhou, Wanxiang Che

备注：: Accepted at NeurIPS 2024 (Oral)

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出推理边界框架 RBF，通过定义并组合推理边界来量化 CoT 上界、分类并优化推理路径，在27个模型、5项任务上验证其合理性并解释10种CoT策略。

摘要：

思维链（Chain-of-Thought，CoT）推理已成为提升大型语言模型（LLM）在复杂推理任务上性能的一种颇具前景的方法。近来，一系列研究试图解释 CoT 背后的机制，旨在深化对其有效性的理解。然而，现有研究面临两大挑战：（1）缺乏用于评估 CoT 能力的量化指标；（2）缺乏关于优化 CoT 性能的指导。受此启发，本文提出一种新颖的推理边界框架（reasoning boundary framework，RBF）来应对这些挑战。为解决量化不足的问题，我们首先定义推理边界（reasoning boundary，RB）以量化 CoT 的上界，并建立 RB 的组合定律，从而形成一种可应用于各类真实 CoT 任务的实用量化方法。为解决优化不足的问题，我们提出三类 RB。我们进一步利用以 RB 提升和推理路径优化为重点的组合定律来优化这些类别，以改进 CoT。通过在 27 个模型和 5 项任务上的大量实验，本研究验证了所提框架的存在性与合理性。此外，该框架解释了 10 种 CoT 策略的有效性，并从两个角度指导优化。我们希望这项工作能够为理解 LLM 中推理的边界与优化策略提供全面认识。我们的代码和数据可在 https://github.com/LightChen233/reasoning-boundary 获取。

英文摘要：

Chain-of-Thought (CoT) reasoning has emerged as a promising approach for enhancing the performance of large language models (LLMs) on complex reasoning tasks. Recently, a series of studies attempt to explain the mechanisms underlying CoT, aiming to deepen the understanding of its efficacy. Nevertheless, the existing research faces two major challenges: (1) a lack of quantitative metrics to assess CoT capabilities and (2) a dearth of guidance on optimizing CoT performance. Motivated by this, in this work, we introduce a novel reasoning boundary framework (RBF) to address these challenges. To solve the lack of quantification, we first define a reasoning boundary (RB) to quantify the upper-bound of CoT and establish a combination law for RB, enabling a practical quantitative approach applicable to various real-world CoT tasks. To address the lack of optimization, we propose three categories of RBs. We further optimize these categories with combination laws focused on RB promotion and reasoning path optimization for CoT improvement. Through extensive experiments on 27 models and 5 tasks, the study validates the existence and rationality of the proposed framework. Furthermore, it explains the effectiveness of 10 CoT strategies and guides optimization from two perspectives. We hope this work can provide a comprehensive understanding of the boundaries and optimization strategies for reasoning in LLMs. Our code and data are available at https://github.com/LightChen233/reasoning-boundary.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。