This repository serves as a curated knowledge base tracking the evolution of video generation towards sophisticated world models. It targets researchers and industrial practitioners by compiling influential works that enable intelligent agents to perceive, act, and plan within simulated interactive environments, moving beyond mere visual realism to physical plausibility and agent-driven dynamics. The collection aims to support progress in creating AI systems that can learn and operate as if interacting with the real world.

How It Works

The repository functions as a structured bibliography, systematically organizing research papers that bridge video generation techniques with world modeling concepts. It categorizes works based on their contributions to faithfulness (accurate simulation), interactiveness (controllability, dynamics), and planning capabilities, highlighting advancements in areas like robotics, autonomous driving, and gaming. This curated approach provides a clear overview of the field's progression and key methodologies.

Quick Start & Requirements

This repository is a curated list of research papers and does not contain executable code. Therefore, there are no installation instructions, prerequisites, or setup requirements.

Highlighted Details

• Comprehensive categorization of research into key areas: Faithfulness, Interactiveness, Planning, and Outlier Modeling.
• Detailed sub-sections focusing on specific application domains: Robotics, Autonomous Driving, and Gaming.
• Inclusion of dedicated sections for Evaluation Metrics and Datasets relevant to video generation and world models.
• Regular updates with recent publications, including papers from 2025, reflecting the rapid advancements in the field.
• Announcement of a CVPR 2025 Tutorial on "From Video Generation to World Model."

Maintenance & Community

The repository is updated periodically. Contributions and suggestions for new resources or fixes are welcomed via GitHub Issues and Pull Requests. Contact information for the maintainers is provided for specific queries.

Licensing & Compatibility

No specific open-source license is declared for the repository content itself. The listed research papers are subject to their respective publication licenses.

Limitations & Caveats

As a curated list, this repository does not provide runnable code or direct access to models. Links to external resources may expire over time, as noted by the maintainers. The field of "world models" is still under active development, with many listed works representing ongoing research rather than established, production-ready systems.