Reproduce “Fight World Model: Customizable Combat through Visual Interaction Proxies” 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

Jiahua Li, Zixu John, Tom Zhong, Fuping Wu, Tianhao Xu, Ruicheng Zhang, Jun Zhou, Zunnan Xu

FWM generates combat video by executing rules in an engine and encoding actions, reactions, identities, timing, and persistent health in a visual proxy for the renderer. In fixed-view, two-fighter Street Fighter II, it achieves 98.48% joint interaction fidelity versus 76.35% for Spatial Proxy + Text. Under tested limited-hit immunity, rule-specific fidelity is 99.92%; results are limited to these settings.

Summary by alphaXiv.

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