This repository provides a curated collection of over 150 survey papers on Large Language Models (LLMs), aiming to offer a comprehensive overview for researchers and practitioners. It categorizes surveys across key LLM research areas, facilitating quick understanding and navigation of the rapidly evolving LLM landscape.

How It Works

The project acts as a bibliographical database, compiling and categorizing existing survey papers. It organizes these surveys by topic, including general overviews, transformers, alignment, prompt learning, reasoning, data, evaluation, societal issues, safety, security, misinformation, factuality, model attributes, efficiency, learning methods, multimodality, knowledge integration, tool use, interactions, long sequences, and specific application domains like education, law, healthcare, games, NLP tasks, software engineering, recommender systems, and graphs.

Quick Start & Requirements

• Access the collection via the provided README.
• No installation or specific software requirements are needed to view the survey list.
• Links to papers (arXiv, ACM, etc.) are provided for each entry.

Highlighted Details

• Covers over 150 survey papers, offering broad coverage of LLM research.
• Organizes surveys into 20+ distinct categories, from foundational concepts to specific applications.
• Includes papers published from 2020 to early 2024, ensuring relative recency.
• Provides direct links to papers for easy access.

Maintenance & Community

• Core contributors include Junhao Ruan, Long Meng, Weiqiao Shan, Tong Xiao, and Jingbo Zhu.
• No community links (Discord, Slack) or roadmap are explicitly mentioned in the README.

Licensing & Compatibility

• The repository itself does not appear to have a specific license mentioned.
• The linked survey papers are subject to their respective publication licenses and terms of use.

Limitations & Caveats

This is a curated list of existing surveys, not a new survey or a tool for LLM analysis. The recency of the listed papers varies, and the collection's comprehensiveness is limited by what has been published and included by the contributors.