Abstract:The past few years have seen the development of ``universal'' machine-learning interatomic potentials (uMLIPs) capable of approximating the ground-state potential energy surface across a wide range of chemical structures and compositions with reasonable accuracy. While these models differ in the architecture and the dataset used, they share the ability to compress a staggering amount of chemical information into descriptive latent features. Herein, we systematically analyze what the different uMLIPs have learned by quantitatively assessing the relative information content of their latent features with feature reconstruction errors, and observing how the trends are affected by the choice of training set and training protocol. We find that uMLIPs encode the chemical space in significantly distinct ways, with substantial cross-model feature reconstruction errors. When variants of the same model architecture are considered, trends become dependent on the dataset, target, and training protocol of choice. We also observe that fine-tuning of a uMLIP retains a strong pre-training bias in the latent features. Finally, we discuss how atom-level features, which are directly output by MLIPs, can be compressed into global structure-level features via concatenation of progressive cumulants, each adding significantly new information about the variability across the atomic environments within a given system.
| Subjects: | Chemical Physics (physics.chem-ph); Materials Science (cond-mat.mtrl-sci); Machine Learning (cs.LG) |
| Cite as: | arXiv:2512.05717 [physics.chem-ph] |
| (or arXiv:2512.05717v3 [physics.chem-ph] for this version) | |
| https://doi.org/10.48550/arXiv.2512.05717 arXiv-issued DOI via DataCite |
|
| Related DOI: | https://doi.org/10.1002/aisy.202501497
DOI(s) linking to related resources |
From: Sofiia Chorna [view email]
[v1]
Fri, 5 Dec 2025 13:45:01 UTC (5,513 KB)
[v2]
Thu, 22 Jan 2026 15:53:36 UTC (5,570 KB)
[v3]
Fri, 17 Apr 2026 06:47:13 UTC (5,163 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。