This repository provides a comprehensive survey of Chain of Thought (CoT) reasoning in Large Language Models (LLMs), targeting researchers and practitioners in Natural Language Processing (NLP) and Artificial Intelligence (AI). It aims to consolidate advances, explore frontiers, and identify future research directions in CoT reasoning, offering a structured overview of the field.

How It Works

The repository serves as a curated collection of academic papers, blogs, and projects related to CoT reasoning. It categorizes research by methodology (e.g., manual construction, automatic construction, structural variants, enhancement methods, question decomposition, external knowledge) and application domains (e.g., mathematical, commonsense, symbolic, multimodal reasoning). This structured approach allows users to navigate the rapidly evolving landscape of CoT research efficiently.

Quick Start & Requirements

This repository is a curated list of resources, not a software package. No installation or execution is required. All resources are linked to their respective papers or platforms.

Highlighted Details

• ACL 2024 Publication: The survey paper is accepted to the Association for Computational Linguistics (ACL) 2024 conference.
• Extensive Categorization: Covers a wide array of CoT advancements, including manual and automatic construction, structural variants (Tree, Graph of Thoughts), enhancement methods (verification, refinement, decomposition), external knowledge integration, and frontier applications like tool use and planning.
• Regular Updates: The reading list is periodically updated with new research, indicated by a 💡 icon for recently added papers.
• Community Contribution: Encourages community suggestions and submissions via GitHub issues or email.

Maintenance & Community

The project is maintained by researchers from Harbin Institute of Technology and Huawei Inc. Suggestions for missed papers can be submitted via GitHub issues or by contacting zchu@ir.hit.edu.cn.

Licensing & Compatibility

The repository itself contains links to external resources, and the licensing of those resources is determined by their original publishers. The code for the repository is likely under a permissive license, but specific details are not provided in the README.

Limitations & Caveats

As a survey and resource collection, this repository does not offer executable code or models. The breadth of the topic means that while comprehensive, it may not cover every niche aspect of CoT reasoning. The rapid pace of research means the list will require continuous updates to remain fully current.