Field guide to the world-model paper explosion

What it does This repository is a curated awesome-list that catalogs papers, code repositories, and project websites for World Models research. It organizes the literature into categories like General World Models, Embodied AI, Visual Understanding, Autonomous Driving, and Vision-Language-Action models, serving as a categorized index.

The interesting bit Instead of treating world models as a single discipline, the list acknowledges that the field now spans everything from Runway’s video generators to NVIDIA’s Cosmos driving simulators and 1X’s robot manipulation models. The maintainers run an arXiv candidate pipeline to discover recent submissions for review.

Key highlights

• Covers the full stack: foundational theory, surveys, benchmarks, and applied models for robotics and autonomous driving.
• Tracks both academic papers and industry technical reports from players like NVIDIA, Tencent, Xiaomi, and Wayve.
• Includes direct links to code and project websites where available, not just PDFs.
• Maintainers use an automated pipeline to discover recent arXiv submissions for review.
• Accepts community contributions via pull requests and email.

Caveats

• The repository is essentially a curated index; visitors will find a categorized reading list rather than a reusable framework or library.
• The README is a dense wall of links, so finding a specific paper requires patience or browser search.

Verdict Researchers and engineers drowning in world-model preprints should bookmark this; practitioners hunting for a drop-in training framework will leave empty-handed.