Reproduce “Code Plans, Diffusion Renders: Open-Ended Generative World Modeling” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Zixun Fang, Yawen Shao, Kai Zhu, Jie Xiao, Shihan Chen, Yu Liu, Xueyang Fu, Yang Cao and 2 more

CoDeR builds executable worlds in which language models define rules, coding agents implement them, and video models render observations. Its registration mechanism carries generated appearance back into the world, supporting state across views. In a cup-and-tap example, water rises after the camera returns. WorldScore averages 87.88, versus 74.79, but rollouts degrade around 5,000 frames, limiting claims of open-ended simulation.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.