Reproduce “Hy-Nomas: Towards Free-form Interactive World Generation” 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

Qiaoling Chen, Jiaxiang Cheng, Xueqing Deng, Tiankai Hang, Ji Hou, Jiaqi Li, Yifan Liu, Tao Lu and 30 more

Hy-Nomas enables interactive video generation by turning user requests and interaction history into segment instructions. It preserves world descriptions across segments while limiting action instructions to their intended temporal windows. In 60 interleaved text-and-camera scenarios, evaluators preferred its prompt following over adapted LingBot-World v2 in 75.0% of pairwise judgments; camera control was less decisive, and spatial memory trailed Genie 3.

Summary by alphaXiv.

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