Reproduce “UniPixie: Unified and Probabilistic 3D Physics Learning via Flow Matching” 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
Qilin Huang, Quynh Anh Huynh, Long Le, Chen Wang, Chuhao Chen, Ryan Lucas, Eric Eaton, Lingjie Liu
UNIPIXIE predicts controllable physical parameters across a soft-to-stiff range. It conditions flow-matching decoders on a scalar, α, and uses shared visual features with specialized outputs for MPM, LBS, and Spring-Mass simulators. On a 41-object held-out MPM test split, it reports log-Young’s-modulus MSE of 0.0091, versus 0.0205 for PIXIE, about 56% lower under this protocol—not evidence of real-world material accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.