This repository serves as a comprehensive, curated collection of academic resources on machine unlearning, targeting researchers and practitioners in AI safety, privacy, and model management. It aims to consolidate the rapidly evolving field by cataloging papers, methodologies, datasets, and evaluation metrics, providing a centralized hub for understanding and advancing unlearning techniques.

How It Works

The project categorizes machine unlearning approaches into three main types: Model-agnostic (applicable across various model architectures), Model-intrinsic (designed for specific model types), and Data-driven (leveraging data manipulation for unlearning). It also lists relevant datasets and evaluation metrics used in the field. The collection is actively maintained and updated with the latest research.

Quick Start & Requirements

This is a curated list of resources, not a software package. No installation or execution is required. The primary resource is the survey paper itself, available via arXiv.

Highlighted Details

• Comprehensive catalog of over 100 research papers on machine unlearning, spanning from 2015 to 2025.
• Categorization of unlearning methods by approach (Model-agnostic, Model-intrinsic, Data-driven) and application domain (LLMs, GANs, GNNs, Federated Learning).
• Includes lists of datasets suitable for unlearning research (Image, Tabular, Text, Sequence, Graph) and common evaluation metrics.
• Actively updated with recent publications and welcomes community contributions.

Maintenance & Community

The repository is maintained by tamlhp and actively seeks community contributions via GitHub issues and pull requests for new research papers and updates.

Licensing & Compatibility

The repository itself is licensed under the MIT License, allowing for broad use and distribution of the curated list. The underlying research papers retain their original licenses.

Limitations & Caveats

This repository is a survey and collection of links to external research; it does not provide executable code or a unified framework for performing machine unlearning. The "code" links provided point to external repositories, which may have their own dependencies and setup requirements.