Reproduce “EvoWorld: Towards Out-of-Sight World Evolution in Generative Video Models” 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

Yequan Bie, Dengyang Jiang, Zhixuan Chen, Xiaozhuan Liang, Haoxuan Che, Quande Liu, Xi Chen, Hao Chen

EvoWorld tests whether a video world model can show a plausible subject response to an event introduced while the subject is out of view. A reasoning system observes the generated event and guides a renderer using entity and recent-context references. On 120 EvoWorld-Bench cases, it achieves 90.8% Core Task Success versus 21.7% for Evoke-Turbo, for a controlled single-event task.

Summary by alphaXiv.

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