MatRIS: Toward Reliable and Efficient Pretrained Machine Learning Interaction Potentials

Researchers introduce MatRIS, a new machine learning interaction potential model for materials science that achieves comparable accuracy to leading equivariant models while being significantly more computationally efficient. The model uses attention-based three-body interactions with linear O(N) complexity, demonstrating strong performance on benchmarks like Matbench-Discovery with an F1 score of 0.847.

• →MatRIS introduces attention-based modeling of three-body interactions for materials simulation with linear O(N) complexity.
• →The model achieves comparable accuracy to state-of-the-art equivariant models at a fraction of the computational cost.
• →MatRIS demonstrates strong performance across multiple benchmarks including Matbench-Discovery, MatPES, and MDR phonon datasets.
• →The research suggests invariant models can match equivariant model accuracy while being more efficient for materials science applications.
• →This advancement could accelerate the development of machine learning potentials for computational materials science.