Reproduce “Causal world models under physical parameter shifts” 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
Chu Lin, Yi Sun
The paper tests robot world models forecasting motion beyond trained link-mass ranges. Models assign mass an explicit role in inertia and dynamics bias, routing actions through learned torques or a known PD controller. In simulation, actual masses raise the known-PD model’s 150-step mean joint R² from 0.609 to 0.727, within tested tasks and configurations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.