Abstract:Prevailing machine-learned interatomic potential (MLIP) uncertainty-quantification methods rely on ensembles of independently trained backbones. These methods scale unfavorably with foundation-scale MLIPs, and their member-disagreement signals correlate weakly with per-molecule prediction error. Here we probe the frozen per-atom representations of a pretrained MLIP with a compact discriminative classifier, recasting MLIP uncertainty quantification as selective classification rather than error regression. The resulting method, PROBE (Post-hoc Reliability frOm Backbone Embeddings), produces a per-prediction reliability probability that monotonically tracks actual error without modification to the underlying model. Across large held-out evaluation sets and two structurally distinct MLIP architectures, PROBE outperforms ensemble disagreement as a binary reliability signal, which strengthens with the expressiveness of the backbone representation, implying a favorable scaling trajectory toward foundation-scale MLIPs. Multi-head self-attention additionally yields per-atom importance maps, providing chemically interpretable diagnostics at no additional computational cost. PROBE is post-hoc and architecture-agnostic, and is directly deployable on any MLIP that exposes per-atom representations.
| Subjects: | Machine Learning (cs.LG); Chemical Physics (physics.chem-ph) |
| Cite as: | arXiv:2605.00640 [cs.LG] |
| (or arXiv:2605.00640v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.00640 arXiv-issued DOI via DataCite (pending registration) |
From: Shams Mehdi [view email]
[v1]
Fri, 1 May 2026 13:21:56 UTC (10,540 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。