MatRIS: Toward Reliable and Efficient Pretrained Machine Learning Interatomic Potentials

Yuanchang Zhou, Siyu Hu, Xiangyu Zhang, Hongyu Wang, [0.05in] Guangming Tan, Weile Jia, [0.05in] State Key Lab of Processors, Institute of Computing Technology, Chinese Academy of Sciences, [0.05in] University of Chinese Academy of Sciences, [0.05in] School of Advanced Interdisciplinary Sciences, UCAS[0.05in] {zhouyuanchang23s, jiaweile}@ict.ac.cn · 2026

Read MatRIS: Toward Reliable and Efficient Pretrained Machine Learning Interatomic Potentials by Yuanchang Zhou, Siyu Hu, Xiangyu Zhang, Hongyu Wang, [0.05in] Guangming Tan, Weile Jia, [0.05in] State Key Lab of Processors, Institute of Computing Technology, Chinese Academy of Sciences, [0.05in] University of Chinese Academy of Sciences, [0.05in] School of Advanced Interdisciplinary Sciences, UCAS[0.05in] {zhouyuanchang23s, jiaweile}@ict.ac.cn (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-02002.txt.

Description

Yuanchang Zhou, Siyu Hu, Xiangyu Zhang, and colleagues build a fast interatomic potential that skips costly tensor math. Their 2026 model, MatRIS, predicts atomic interactions for materials screening across many chemistries. Top equivariant models are accurate but expensive, since they lean on tensor products and high-degree features.

MatRIS stays invariant instead and models three-body interactions with a separable attention mechanism. That attention scales linearly with atom count, which keeps large systems affordable. Accuracy matches leading equivariant models across a wide range of popular benchmarks in their tests. On a discovery benchmark, it reaches a strong score at lower training cost than those rivals. Careful invariant design, the authors argue, can equal equivariant accuracy at a fraction of the price.

Catalog record

CREATORYuanchang Zhou, Siyu Hu, Xiangyu Zhang, Hongyu Wang, [0.05in] Guangming Tan, Weile Jia, [0.05in] State Key Lab of Processors, Institute of Computing Technology, Chinese Academy of Sciences, [0.05in] University of Chinese Academy of Sciences, [0.05in] School of Advanced Interdisciplinary Sciences, UCAS[0.05in] {zhouyuanchang23s, jiaweile}@ict.ac.cn
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE66421
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.02002

Browse the full library · About the Tower of Babel