Learning Inter-Atomic Potentials without Explicit Equivariance

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Elhag, Ahmed A., Raja, Arun, Morehead, Alex, Blau, Samuel M., Zhao, Hongtao, Tyrchan, Christian, Nittinger, Eva, Morris, Garrett M., Bronstein, Michael M.
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911556931944448
author Elhag, Ahmed A.
Raja, Arun
Morehead, Alex
Blau, Samuel M.
Zhao, Hongtao
Tyrchan, Christian
Nittinger, Eva
Morris, Garrett M.
Bronstein, Michael M.
author_facet Elhag, Ahmed A.
Raja, Arun
Morehead, Alex
Blau, Samuel M.
Zhao, Hongtao
Tyrchan, Christian
Nittinger, Eva
Morris, Garrett M.
Bronstein, Michael M.
contents Accurate and scalable machine-learned inter-atomic potentials (MLIPs) are essential for molecular simulations ranging from drug discovery to new material design. Current state-of-the-art models enforce roto-translational symmetries through equivariant neural network architectures, a hard-wired inductive bias that can often lead to reduced flexibility, computational efficiency, and scalability. In this work, we introduce TransIP: Transformer-based Inter-Atomic Potentials, a novel training paradigm for interatomic potentials achieving symmetry compliance without explicit architectural constraints. Our approach guides a generic non-equivariant Transformer-based model to learn SO(3)-equivariance by optimizing its representations in the embedding space. Trained on the recent Open Molecules (OMol25) collection, a large and diverse molecular dataset built specifically for MLIPs and covering different types of molecules (including small organics, biomolecular fragments, and electrolyte-like species), TransIP attains comparable performance in machine-learning force fields versus state-of-the-art equivariant baselines. Further, compared to a data augmentation baseline, TransIP achieves 40% to 60% improvement in performance across varying OMol25 dataset sizes. More broadly, our work shows that learned equivariance can be a powerful and efficient alternative to equivariant or augmentation-based MLIP models. Our code is available at: https://github.com/Ahmed-A-A-Elhag/TransIP.
format Preprint
id arxiv_https___arxiv_org_abs_2510_00027
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learning Inter-Atomic Potentials without Explicit Equivariance
Elhag, Ahmed A.
Raja, Arun
Morehead, Alex
Blau, Samuel M.
Zhao, Hongtao
Tyrchan, Christian
Nittinger, Eva
Morris, Garrett M.
Bronstein, Michael M.
Machine Learning
Artificial Intelligence
Biomolecules
Quantitative Methods
I.2.1; J.3
Accurate and scalable machine-learned inter-atomic potentials (MLIPs) are essential for molecular simulations ranging from drug discovery to new material design. Current state-of-the-art models enforce roto-translational symmetries through equivariant neural network architectures, a hard-wired inductive bias that can often lead to reduced flexibility, computational efficiency, and scalability. In this work, we introduce TransIP: Transformer-based Inter-Atomic Potentials, a novel training paradigm for interatomic potentials achieving symmetry compliance without explicit architectural constraints. Our approach guides a generic non-equivariant Transformer-based model to learn SO(3)-equivariance by optimizing its representations in the embedding space. Trained on the recent Open Molecules (OMol25) collection, a large and diverse molecular dataset built specifically for MLIPs and covering different types of molecules (including small organics, biomolecular fragments, and electrolyte-like species), TransIP attains comparable performance in machine-learning force fields versus state-of-the-art equivariant baselines. Further, compared to a data augmentation baseline, TransIP achieves 40% to 60% improvement in performance across varying OMol25 dataset sizes. More broadly, our work shows that learned equivariance can be a powerful and efficient alternative to equivariant or augmentation-based MLIP models. Our code is available at: https://github.com/Ahmed-A-A-Elhag/TransIP.
title Learning Inter-Atomic Potentials without Explicit Equivariance
topic Machine Learning
Artificial Intelligence
Biomolecules
Quantitative Methods
I.2.1; J.3
url https://arxiv.org/abs/2510.00027