Reproduce “PhysAg3D: Bridging 3D Generative and Agentic Models for High-Fidelity and Physically-Interactive Assets” 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

Bi'an Du, Fanpeng Meng, Jiemin Fang, Chen Yang, Pufan Li, Daizong Liu, Qi Tian, Wei Hu

PhysAg3D turns a photograph into an interactive 3D asset by linking intended actions to physical mechanisms and the visible surfaces they move. It combines generated textured geometry with agent-built interactions and evidence-guided revision. On 50 objects and 110 actions, it reports 91.3% ACR and 88.0% FVSP, versus aDSL’s 66.8% and 42.0%. These video-based results do not establish universal physical correctness.

Summary by alphaXiv.

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