Reproduce “GPWM: Modeling the World as 3D Particle Motion” 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
Robert Ren, Matthew Bronars, He Zhu, Yikai Tang, Jianing Zhang, Sergey Zakharov, Pavel Tokmakov, Katerina Fragkiadaki
GPWM generates 3D predictions of physical scenes and plans contact-rich actions by modeling particle motion with a multi-resolution transformer. For planning, it jointly generates actuator and object trajectories, then searches generator noise rather than raw actuator commands. Across six domains it reports 31% error reduction; in Pushing, noise search reaches 0.57 task progress versus 0.00 for action search in simulation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.