This repository serves as a curated collection of essential research papers on knowledge editing for Large Language Models (LLMs). It aims to provide researchers and practitioners with a comprehensive overview of the field, including key methods, benchmarks, and emerging trends, facilitating efficient exploration and advancement of LLM knowledge manipulation.

How It Works

The repository organizes papers by editing methods (e.g., parameter preservation, additional parameters, modifying parameters) and related topics. It also lists relevant benchmarks and tasks for evaluating knowledge editing techniques, covering factual knowledge, multilingualism, bias, hallucination, and more. The project actively tracks and updates with new publications and tutorials from major AI conferences.

Quick Start & Requirements

• Installation: No direct installation is required as this is a curated list of papers.
• Prerequisites: Access to research papers (often via arXiv or conference proceedings). Links to code repositories for specific tools like EasyEdit and FastEdit are provided.
• Resources: Links to Google Drive and BaiduPan are available for tutorial materials.

Highlighted Details

• Comprehensive list of papers categorized by editing methods and research areas.
• Includes links to tutorials from major conferences like AAAI, EMNLP, and COLING.
• Features a dedicated section on benchmarks and tasks for evaluating knowledge editing.
• Provides links to associated code repositories and tools for practical implementation.

Maintenance & Community

The project is actively maintained, with frequent updates on new paper acceptances and tutorial announcements. Contributions for missing works are welcomed.

Licensing & Compatibility

The repository itself does not have a specific license mentioned, but it links to various research papers and code repositories, each with their own licenses. Users should refer to individual paper and code licenses for usage terms.

Limitations & Caveats

This repository is a curated list and does not provide a unified framework or executable code for knowledge editing itself. Users will need to consult the linked papers and code repositories for implementation details and tools.