Enhancing Zero-Shot Chain-of-Thought Reasoning in Large Language Models through Logic
Abstract
Recent advancements in large language models have showcased their remarkable generalizability across various domains. However, their reasoning abilities still have significant room for improvement, especially when confronted with scenarios requiring multi-step reasoning. Although large language models possess extensive knowledge, their reasoning often fails to effectively utilize this knowledge to establish a coherent thinking paradigm. These models sometimes show hallucinations as their reasoning procedures are unconstrained by logical principles. Aiming at improving the zero-shot chain-of-thought reasoning ability of large language models, we propose LoT (Logical Thoughts), a self-improvement prompting framework that leverages principles rooted in symbolic logic, particularly Reductio ad Absurdum, to systematically verify and rectify the reasoning processes step by step. Experimental evaluations conducted on language tasks in diverse domains, including arithmetic, commonsense, symbolic, causal inference, and social problems, demonstrate the efficacy of enhanced reasoning by logic. The implementation code for LoT can be accessed at: https://github.com/xf-zhao/LoT. Keywords: Large
中文速览
大语言模型（LLM）在多步推理时容易"一本正经地胡说八道"，因为它们逐步生成文字时缺乏对每一步逻辑正确性的约束，一旦某步出错就会层层传播、越错越远。本文提出了一个名为 LoT（Logical Thoughts，逻辑思维）的零样本提示框架，核心思路是把逻辑学中的"归谬法"（Reductio ad Absurdum）引入推理过程：让模型在生成每一步推断后，先假设该步结论为假，再检验它与已知前提是否产生矛盾，从而自动判断这步推理是否可信，若发现问题则就地修正，再继续往下走。在算术、常识、符号推理、因果推断和社会问题等多个基准测试上，LoT 相比普通零样本思维链（Zero-shot CoT）取得了一致的提升。这项工作的意义在于，它无需人工标注示例，也无需调整模型参数，仅凭借将经典逻辑原则"嵌入"提示词，就能系统性地压制幻觉、提升推理可靠性，为让 LLM 真正"讲逻辑"提供了一条轻量可行的路径。
原文 arXiv:2309.13339;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2309.13339v4