This repository serves as a comprehensive survey and resource hub for "World Models in Artificial Intelligence." It addresses the fragmentation of research in this rapidly evolving field by providing a systematic review, categorization, and overview of current paradigms, applications, and challenges. Aimed at researchers, engineers, and practitioners, it offers a unified reference to understand, compare, and advance the state-of-the-art in predictive AI modeling.

How It Works

The project categorizes world models into four primary branches: reinforcement learning-based, observation-level generative, latent space, and object-centric. It systematically reviews their applications across domains like robotics, autonomous driving, scientific discovery, and virtual simulations. This structured approach aims to unify fragmented research by providing formal mathematical formulations and a broad overview of methodologies, benchmarks, and evaluation protocols.

Quick Start & Requirements

This repository is a curated collection of research papers and resources related to a survey on World Models. It does not provide direct installation or execution instructions for a specific software tool. Users can access the survey paper and linked resources for in-depth study.

Highlighted Details

• Features a comprehensive survey paper, "Learning to Model the World: A Survey of World Models in Artificial Intelligence," with recent updates (May 2026) including new papers.
• Organizes world models into four core paradigms: reinforcement learning-based, observation-level generative, latent space, and object-centric.
• Covers a wide array of applications, including robotics, autonomous driving, scientific discovery, virtual game simulation, and GUI-based agents.
• Includes extensive lists of benchmark datasets, evaluation metrics, and physics engines/simulation platforms relevant to world modeling research.

Maintenance & Community

No specific details regarding project maintenance, community channels (e.g., Discord, Slack), or active contributors are provided in the README.

Licensing & Compatibility

The README does not specify a software license. Therefore, compatibility for commercial use or closed-source linking cannot be determined from this information.

Limitations & Caveats

The survey itself highlights key challenges in the field, including achieving long-horizon consistency, ensuring controllability and robustness, addressing evaluation limitations, and improving generalization. As a survey repository, it does not represent a runnable system but rather a curated collection of research.