Reproduce “Inhabitant: An Agent Framework for Long-Horizon Human-Scene Interaction with Spatial Control and Quantitative Feedback” 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
Yao Zhang, Zhuchenyang Liu, Yu Xiao
Inhabitant enables an agent to generate long-horizon human-scene interactions from one instruction by specifying spatial constraints for a pretrained motion generator and revising executed motion using simulator measurements. On the 75-task Inhabit benchmark, Claude Opus 5.5 succeeds on 74/75 tasks (98.7%), compared with 26.7% for the strongest adapted baseline. These are simulator-based results with kinematic motion and rule-based grasping.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.