MatRIS: Toward Reliable and Efficient Pretrained Machine Learning Interaction Potentials

2026-03-02

AI Summary

This work proposes MatRIS, an invariant-based machine learning interatomic potential (MLIP) model that overcomes the high computational cost of conventional equivariant models, which rely on expensive higher-order tensor operations and struggle to efficiently capture high-dimensional atomic interactions in large-scale systems. By introducing a separable attention mechanism with linear complexity (O(N)), MatRIS enables, for the first time within an invariant framework, accurate and efficient modeling of three-body interactions. This approach breaks the longstanding trade-off between efficiency and accuracy inherent in equivariant models. Experimental results demonstrate that MatRIS achieves accuracy comparable to state-of-the-art equivariant models on benchmarks such as Matbench-Discovery (F1 = 0.847), while substantially reducing training overhead, thereby establishing the feasibility and competitiveness of high-performance invariant-based MLIPs.

Technology Category

• Machine Learning: Matrix & Tensor Methods
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Computer Vision: Representation Learning for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Security and Privacy: Security and privacy of machine learning and AI applications

Abstract

Foundation MLIPs demonstrate broad applicability across diverse material systems and have emerged as a powerful and transformative paradigm in chemical and computational materials science. Equivariant MLIPs achieve state-of-the-art accuracy in a wide range of benchmarks by incorporating equivariant inductive bias. However, the reliance on tensor products and high-degree representations makes them computationally costly. This raises a fundamental question: as quantum mechanical-based datasets continue to expand, can we develop a more compact model to thoroughly exploit high-dimensional atomic interactions? In this work, we present MatRIS (\textbf{Mat}erials \textbf{R}epresentation and \textbf{I}nteraction \textbf{S}imulation), an invariant MLIP that introduces attention-based modeling of three-body interactions. MatRIS leverages a novel separable attention mechanism with linear complexity $O(N)$, enabling both scalability and expressiveness. MatRIS delivers accuracy comparable to that of leading equivariant models on a wide range of popular benchmarks (Matbench-Discovery, MatPES, MDR phonon, Molecular dataset, etc). Taking Matbench-Discovery as an example, MatRIS achieves an F1 score of up to 0.847 and attains comparable accuracy at a lower training cost. The work indicates that our carefully designed invariant models can match or exceed the accuracy of equivariant models at a fraction of the cost, shedding light on the development of accurate and efficient MLIPs.

Problem

• Machine Learning Interaction Potentials
• Computational Efficiency
• High-dimensional Atomic Interactions
• Invariant Models
• Scalability

Innovation

• invariant MLIP
• separable attention
• three-body interactions
• linear complexity
• efficient materials modeling