Overcoming the limitations of body-ordered potentials for atomistic machine learning
Abstract
Machine-learning interatomic potentials have become indispensable tools in atomistic simulations. The most popular models rely heavily on physical priors, such as locality, smoothness and symmetry, that are assumed to improve the accuracy and transferability of the trained models. However, recent evidence suggests that, in the data-rich regime, unconstrained models that learn symmetry from the data can achieve very competitive accuracy and computational efficiency, without sacrificing stability and generalization power. Many equivariant symmetric architectures historically rely on the assumption that the interatomic potential can be approximated well by a convergentcluster expansion, i.e., a decomposition in terms associated to atomic pairs, triplets, quadruplets, and higher-order tuples. We describe an architecture that incorporates locality, smoothness, and symmetry priors, but that -- thanks to a computationally efficient tensor product formulation -- does not assume low-order truncation of the body-ordered series.
Related
- From sparse quantum-computing data to atomistic simulation with universal machine-learning interatomic potentials
- From Nature to Medicine: Overcoming the Limitations of Herbal Therapeutics for Pancreatic Cancer Treatment Based on Nanotechnology
- A Lightweight Universal Machine-Learning Interatomic Potential via Knowledge Distillation for Scalable Atomistic Simulations
- Learning Electron Density with Virtual Atoms for Transferable Atomistic Representations
- Enabling Domain-Specific Atomistic Models: A Machine Learning Potential for the Solid Acid Family
BibTeX
@misc{bigi2026overcoming,
title = {{Overcoming the limitations of body-ordered potentials for atomistic machine learning}},
author = {Filippo Bigi and Michelangelo Domina and Maximilian Lloyd Ach and Sergey N. Pozdnyakov and Paolo Pegolo and Joseph W. Abbott and Kevin K. Huguenin-Dumittan and Michele Ceriotti},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.10830},
eprint = {2610.10830},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.10830},
}
RIS
TY - UNPB
TI - Overcoming the limitations of body-ordered potentials for atomistic machine learning
AU - Filippo Bigi
AU - Michelangelo Domina
AU - Maximilian Lloyd Ach
AU - Sergey N. Pozdnyakov
AU - Paolo Pegolo
AU - Joseph W. Abbott
AU - Kevin K. Huguenin-Dumittan
AU - Michele Ceriotti
PY - 2026
DA - 2026/10/07
PB - arXiv
DO - 10.48550/arxiv.2610.10830
UR - https://arxiv.org/abs/2610.10830
AB - Machine-learning interatomic potentials have become indispensable tools in atomistic simulations. The most popular models rely heavily on physical priors, such as locality, smoothness and symmetry, that are assumed to improve the accuracy and transferability of the trained models. However, recent evidence suggests that, in the data-rich regime, unconstrained models that learn symmetry from the data can achieve very competitive accuracy and computational efficiency, without sacrificing stability and generalization power. Many equivariant symmetric architectures historically rely on the assumption that the interatomic potential can be approximated well by a convergentcluster expansion, i.e., a decomposition in terms associated to atomic pairs, triplets, quadruplets, and higher-order tuples. We describe an architecture that incorporates locality, smoothness, and symmetry priors, but that -- thanks to a computationally efficient tensor product formulation -- does not assume low-order truncation of the body-ordered series.
ER -