Reproduce “World2HOSI: A Multi-Expert Harness for Long-Horizon Human-Object-Scene Interaction 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
Jiushu Li, Haoran Wang, Yumeng Liu, Zhiyang Dou, Yuan Liu, Yuexin Ma, Cheng Lin
World2HOSI coordinates navigation, object-manipulation, and sitting specialists to generate multi-stage human–object–scene interactions. It maintains a shared scene state, updates object locations after actions, and conditions each next stage on that changed world. On evaluated tasks, it scores 76.93 for task progress and 75.61 for aggregate geometry, versus 62.54 and 48.16 for single-expert HOI-FHLI; these results do not establish universal superiority.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.