Reproduce “Lazy-MOSAIC: Accelerating Skill-Centric Manipulation Planning with Learned Skill-Effect World Models” 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

Itamar Mishani, Mukund Krishnakumar, Oren Salzman, Maxim Likhachev

LAZY-MOSAIC reduces manipulation planning cost by searching with learned skill predictions before simulating promising steps. Its model predicts success and resulting states; physics simulation checks selected steps, and mismatches trigger repair of dependent proposals. Every returned plan is certified relative to the simulator. Across eight settings, it solved 104 of 160 trials versus MOSAIC’s 81; evaluation uses a fixed workspace and object types.

Summary by alphaXiv.

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