This repository serves as a curated collection of research papers on diffusion models, organized according to a comprehensive survey paper. It targets researchers and practitioners in machine learning, computer vision, and natural language processing, providing a structured taxonomy of methods, applications, and connections to other generative models. The primary benefit is a centralized, categorized resource for understanding the rapidly evolving field of diffusion models.

How It Works

The repository categorizes diffusion model research into several key areas: Algorithm Taxonomy (sampling acceleration, likelihood enhancement, data structures, LLM integration, DPO/RLHF), Application Taxonomy (computer vision, NLP, temporal data, multimodal learning, robust learning, molecular modeling, material design, medical imaging), and connections to other generative models (VAEs, GANs, Flows, Autoregressive, EBMs). This structured approach allows users to navigate and discover relevant research efficiently.

Quick Start & Requirements

This repository is a curated list of papers and does not require installation or execution. Links to papers and their associated resources are provided.

Highlighted Details

• Comprehensive taxonomy covering algorithmic advancements and diverse applications.
• Links to papers accepted in top-tier journals and conferences (e.g., ACM Computing Surveys, NeurIPS, CVPR, ICML).
• Includes connections to other generative models, offering a broader perspective.
• Actively updated to reflect the fast-paced development of diffusion models.

Maintenance & Community

The repository is maintained by YangLing0818 and is associated with a survey paper published in ACM Computing Surveys. Updates are ongoing to reflect the latest research.

Licensing & Compatibility

The repository itself is likely under a permissive license (e.g., MIT, Apache 2.0) as it is a collection of links and information. Individual papers retain their original licenses. Compatibility for commercial use depends on the licenses of the cited papers.

Limitations & Caveats

This repository is a curated list of papers and does not provide code implementations or direct access to models. Users must follow links to access the original research and associated code.