<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> This repository serves as a curated knowledge base for "Agentic Reasoning for Large Language Models," addressing the need for structured information in this rapidly advancing field. It targets researchers, engineers, and power users by organizing a comprehensive collection of papers and resources, facilitating a deeper understanding of LLM agents' capabilities in planning, tool use, search, memory, and multi-agent coordination. The primary benefit is a centralized, categorized overview that aids in evaluating current research trends and potential adoption of agentic AI technologies.

How It Works

The project functions as an extensive, categorized bibliography derived from the survey paper "Agentic Reasoning for Large Language Models: A Survey." It structures research across three core layers: Foundational Reasoning (single-agent capabilities like planning, tool use, search), Self-Evolving Reasoning (adaptation through feedback, memory, and learning), and Collective Reasoning (multi-agent coordination and specialization). These layers are further refined by optimization paradigms such as In-Context and Post-Training Reasoning. The repository meticulously lists relevant papers with their publication years under specific thematic sub-sections, providing a detailed map of the research landscape.

Quick Start & Requirements

This repository is a curated list of research papers and resources, not a runnable software project. Therefore, it does not have installation instructions or specific technical requirements beyond standard web browsing capabilities to access the linked papers and survey.

Highlighted Details

• Comprehensive Survey Paper: Features a recently released, in-depth survey titled "Agentic Reasoning for Large Language Models: A Survey," available on arXiv and HuggingFace, serving as a foundational document.
• Granular Research Categorization: Organizes a vast collection of papers into highly specific thematic areas, including Planning Reasoning, Tool-Use Optimization, Agentic Search, Self-evolving Agentic Reasoning (Feedback, Memory), Collective Multi-agent Reasoning, and diverse Applications (Math, Science, Embodied, Healthcare, Web).
• Extensive Bibliographic Data: Provides detailed tables listing hundreds of research papers, categorized by their contribution to specific agentic reasoning mechanisms and applications, often including publication years and venues.
• Active Community Engagement: Actively solicits community contributions via pull requests and issues to continuously expand and refine the repository's coverage of emerging research.

Maintenance & Community

The repository is an ongoing effort, actively maintained and updated with new research. Contributions from the community are welcomed via pull requests or by opening issues. Direct contact is available via email for suggestions and contributions.

Licensing & Compatibility

The repository is licensed under the permissive MIT License, allowing for broad compatibility and use in both academic and commercial contexts without significant restrictions.

Limitations & Caveats

As a curated list of research papers, this repository does not provide executable code or a framework for building agents. Its value is as a comprehensive academic resource, requiring users to access and review the cited papers independently. The rapidly evolving nature of agentic reasoning means the content will need continuous updates to maintain its currency.