Reproduce “Efficient Machine Learning Interatomic Potentials” 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

Yury Lysogorskiy, Anton Bochkarev, Ralf Drautz

The authors review how machine-learning interatomic potentials can be described with ACE and GRACE basis expansions, linking local and message-passing architectures. Their recursive evaluation and tensor decomposition provide a framework for trading inference cost against accuracy. On five selected datasets, GRACE-WARP reaches 0.5% of GRACE-2L’s GPU time with about three times its error; this comparison is benchmark- and implementation-specific.

Summary by alphaXiv.

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