Reproduce “Continuous-Action World Models: Learning Cross-Embodiment Spatiotemporal Dynamics Priors” 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
Liangyu Li, Qingwen Liu
CoAWM represents robot motions as functions of physical time, allowing commands to be queried at different rates without changing motion duration. It learns a shared function space from typed trajectories, then predicts motion from visual and language features and robot state. Against LaWAM, it lowers first-second translational and rotational jerk RMS while translational speed RMS remains nearly the same; cross-embodiment transfer remains untested.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.