Awesome-Video-World-Models-with-AR-Diffusion focuses on Video World Models leveraging Autoregressive (AR) Diffusion, a promising approach for generating consistent and interactive video content. This repository aims to provide a comprehensive and regularly updated resource for researchers, practitioners, and enthusiasts in this rapidly evolving field. We categorize models based on algorithmic foundations, applications, and infrastructure considerations, reflecting the full spectrum of AR diffusion technology.
This repository provides a structured taxonomy of AR diffusion models, categorizing them by algorithmic approach, real-world applications, and underlying infrastructure optimizations. It includes a consolidated BibTeX file for easy citation management and welcomes contributions to expand and refine the list. We are committed to staying current with the latest research and appreciate pull requests for additions and improvements.
- AR Diffusion Methods: Focuses on foundational AR diffusion techniques like Diffusion Forcing, Pyramid Flow, and AR-Diffusion.
- Real-world Applications: Showcases practical applications like interactive video action models, avatar control, and embodied AI.
- Infrastructure Optimizations: Covers techniques like sparse attention, caching, and quantized attention for efficient video generation.
- Comprehensive Citation: Includes a consolidated BibTeX file to facilitate research and reproducibility.
- Regular Updates: The repository is updated weekly to incorporate the latest advancements in the field.
- Community Contributions: Actively encourages pull requests to expand the list and improve organization.
- Diverse Model Coverage: Features a variety of models addressing different aspects of AR video world modeling.
The repository is actively maintained, with frequent updates reflecting the ongoing research in AR diffusion models. The core list is continuously updated with new papers and relevant resources. While the list is comprehensive, it acknowledges that the field is rapidly evolving, and welcomes community contributions to ensure completeness and accuracy.
This repository is valuable for researchers seeking to understand the landscape of AR diffusion models, practitioners looking for implementation examples, and anyone interested in the future of video generation. It offers a central resource for navigating the complexities of AR diffusion and understanding its potential in creating interactive and scalable video experiences. It streamlines research by providing organized access to key publications and related resources.