BranchIP: Learning Adaptive Equivariant Computation for Interatomic Potentials
Abstract
Equivariant machine learning interatomic potentials (MLIPs) have revolutionized atomistic modeling, but accurate treatment of complex materials and molecular systems demands expensive models. This limits simulation length- and time-scales, with tensor products a key computational bottleneck. The recent emergence of foundation-scale MLIPs further exacerbates this challenge. We present Branch Interatomic Potential (BranchIP), a single-model framework for learned adaptive tensor product computation, trained with a novel distillation loss. In our experiments on two systems of physical interest, a heterogeneous catalysis system and a proton-conducting solid acid electrolyte, BranchIP accelerates MLIPs across model sizes by up to 2.4× while reducing memory usage by up to 2.6×. This is achieved while maintaining physical fidelity. Furthermore, the learned adaptive computation provides model interpretability by revealing which interactions demand deeper computation and showing how computational depth relates to chemical complexity and dynamics.
Related
- Correcting the Liquid Densities of Universal Machine-Learning Interatomic Potentials Using Local-Number-Density Repulsion
- pdb2reaction: End-to-End Reaction-Path Elucidation from PDB Structures Using Machine-Learning Interatomic Potentials
- Reference Conventions and Structure Types, Rather Than 4f Occupancy, Shape Lanthanide Errors in Universal Machine-Learning Interatomic Potentials
- MLIP Detective: Active Failure Mode Discovery Beyond Benchmark Scores for Machine-Learning Interatomic Potentials
- Hessian-Informed Conformation Augmentation for Potential Energy Surface Curvature in Machine Learning Interatomic Potentials
BibTeX
@misc{zichi2026branchip,
title = {{BranchIP: Learning Adaptive Equivariant Computation for Interatomic Potentials}},
author = {Laura Zichi and Gil Harari and Chuin Wei Tan and Marc L. Descoteaux and Albert Zhu and Menghang Wang and Yoel Zimmermann and H. T. Kung and Boris Kozinsky},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.02013},
eprint = {2610.02013},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.02013},
}
RIS
TY - UNPB
TI - BranchIP: Learning Adaptive Equivariant Computation for Interatomic Potentials
AU - Laura Zichi
AU - Gil Harari
AU - Chuin Wei Tan
AU - Marc L. Descoteaux
AU - Albert Zhu
AU - Menghang Wang
AU - Yoel Zimmermann
AU - H. T. Kung
AU - Boris Kozinsky
PY - 2026
DA - 2026/10/01
PB - arXiv
DO - 10.48550/arxiv.2610.02013
UR - https://arxiv.org/abs/2610.02013
AB - Equivariant machine learning interatomic potentials (MLIPs) have revolutionized atomistic modeling, but accurate treatment of complex materials and molecular systems demands expensive models. This limits simulation length- and time-scales, with tensor products a key computational bottleneck. The recent emergence of foundation-scale MLIPs further exacerbates this challenge. We present Branch Interatomic Potential (BranchIP), a single-model framework for learned adaptive tensor product computation, trained with a novel distillation loss. In our experiments on two systems of physical interest, a heterogeneous catalysis system and a proton-conducting solid acid electrolyte, BranchIP accelerates MLIPs across model sizes by up to 2.4× while reducing memory usage by up to 2.6×. This is achieved while maintaining physical fidelity. Furthermore, the learned adaptive computation provides model interpretability by revealing which interactions demand deeper computation and showing how computational depth relates to chemical complexity and dynamics.
ER -