Reproduce “AssemblyBench: Physics-Aware Assembly of Complex Industrial Objects” 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
Danrui Li, Jiahao Zhang, Bernhard Egger, Moitreya Chatterjee, Suhas Lohit, Tim K. Marks, Anoop Cherian
AssemblyDyno predicts assembly order and part trajectories from manuals, diagrams, and 3D geometry. It matches parts to instructions, then predicts 12-step sequences of six-degree-of-freedom poses. On the synthetic AssemblyBench test split, simulator success was 13.57%, versus 1.79% for ManualPA, while a heuristic-path ablation reached 1.43%; this is controlled replay, not physical-robot evidence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.