This repository curates research on Video World Models leveraging Autoregressive (AR) Diffusion, a paradigm for scalable, consistent, and interactive world modeling. It serves as a comprehensive resource for researchers, practitioners, and enthusiasts, updated weekly to track advancements in areas like Genie3.

How It Works

The project organizes AR diffusion techniques into three key dimensions: Algorithmic Foundations (native pretraining, distillation for real-time generation, long video generation), Real-world Applications (foundation models, interactive agents, egocentric interaction, embodied AI), and Infrastructure-level Acceleration (sparse attention, caching, quantization). This full-stack approach details the evolution from core modeling design to interactive deployment.

Quick Start & Requirements

This repository is a curated list of research papers and does not provide direct installation or execution instructions. The focus is on the underlying algorithms and applications, which typically require significant computational resources, including GPUs, for training and inference. Specific hardware or software prerequisites are not detailed.

Highlighted Details

• Structured Taxonomy: Organizes the AR diffusion ecosystem across algorithms, applications, and infrastructure.
• Consolidated Citations: Offers a single BibTeX file for all listed research papers.
• Weekly Updates: Aims to maintain currency in the rapidly evolving field.

Maintenance & Community

The repository is maintained by Min Zhao, Hongzhou Zhu, and Wenqiang Sun, who can be contacted via email. A WeChat group is indicated by a badge, but no direct join link is provided.

Licensing & Compatibility

No software license is specified in the README. This absence may pose compatibility concerns for commercial use or integration into closed-source projects.

Limitations & Caveats

The curated list is explicitly stated as not exhaustive, with an invitation for pull requests to include missing works. Contributions for categorization and synthesis are also welcomed.