rankings / model / UMA-M

UMA-M

UMA-M is a machine-learning interatomic potential — version 1.4.2 · Commercial license · 412M parameters. ScoreMol scores it on 9 benchmark suites covering Formation energy, Force MAE, Lattice parameters, Polymorph ranking (ρ), Stable-set discovery (F1), Band gap (GW ref), Band gap (experimental ref), Γ-point frequency error, False imaginary-mode rate, Missed-instability rate, Elastic tensor Cij, Adsorption energy, Ionic conductivity, RDF / structure factor, NVE energy drift, MD stability 1000 K, Liquid density, Self-diffusion, OOD degradation, each run under the same protocol on held-out reference data with cluster-bootstrap confidence intervals.

Suites: FormE-40k, Hull-5k, Gap-2k, Phon-1k, Elas-600, Adsorb-3k, MD-Electro-800, Liq-300, OOD-X.