Reproduce “EMPIRIC: Experiment-Driven Learning of Residual World Models for Robot Planning” 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
Yichao Liang, Amber Li, Dat Nguyen, Emily Bunnapradist, Michelangelo Naim, Sreela Kodali, Matteo Merler, Bowen Li and 8 more
EMPIRIC extends a physics engine with executable programs for missing mechanisms, hidden state, and uncertain parameters, then uses observations to revise the model and plan or experiment. Across five simulated tabletop domains, it solved all 25 complete runs, compared with 16/25 for each of two strongest baselines. This measures task success and environment steps, not computation or broad physical-robot performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.