Reproduce “PRISM: Polynomial Representations for Interaction-Structured Motor Control” 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

Seung Hyun Lee, Stella X. Yu

PRISM gives robot policies a means to represent interactions among existing inputs by multiplying features and adding them to a direct path, without physical supervision. In Humanoid-Gym, 24-second survival was 92.50% for PRISM versus 52.25% for a larger parameter-matched MLP across five seeds. This supports an advantage beyond tested capacity increase, not a general claim that model size never matters.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.