This repository serves as a comprehensive survey and resource hub for Long Chain-of-Thought (Long CoT) reasoning in Large Language Models (LLMs). It aims to provide a structured overview of the field, distinguishing Long CoT from traditional Short CoT, categorizing reasoning paradigms, and exploring key characteristics like deep reasoning, extensive exploration, and feasible reflection. The project is primarily for researchers and practitioners in AI and LLMs seeking to understand and advance complex reasoning capabilities.

How It Works

The repository organizes research papers and concepts into a taxonomy covering analysis, evaluation, deep reasoning, reflection, and exploration. It delves into the characteristics that enable Long CoT, such as deep reasoning, extensive exploration, and feasible reflection, contrasting them with the shallower nature of Short CoT. The survey also investigates phenomena like "overthinking" and "test-time scaling" and identifies future research directions.

Quick Start & Requirements

This repository is a curated list of research papers and concepts, not a runnable software package. No installation or specific requirements are needed to access the information.

Highlighted Details

• Over 900 reviewed papers on Long CoT reasoning.
• Includes a survey paper titled "Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models."
• Categorizes Long CoT phenomena like overthinking, test-time scaling, and emergence.
• Covers future directions including multimodal, agentic, and efficient Long CoT.

Maintenance & Community

The project is actively maintained, with recent updates in April 2025. The primary contributors are listed as authors of the survey paper.

Licensing & Compatibility

The repository itself does not specify a license, but it references and links to numerous research papers, each with its own licensing and usage terms.

Limitations & Caveats

This repository is a literature survey and does not provide code or tools for direct implementation or experimentation with Long CoT reasoning. Its value lies in its comprehensive cataloging and analysis of existing research.