Reproduce “Matrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game Worlds” 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
Jihai Zhang, Runjia Qian, Kaiqin Kong, Biao Jiang, Puyi Wang, Siru Chen, Tianle Li, Guanjie Chen and 10 more
Matrix-Game 4.0 uses an engine to maintain game state and enforce rules, while an agent adapts game code and captions events. A causal video model renders frames from a simplified scene proxy, captions, and visual history. On one RTX 5090, it reports 112 ms at 480p from key press to first reacting frame, demonstrating mechanics in existing games, not arbitrary creation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.