cs.LGOct 30, 2026

Towards Out-of-Distribution Uncertainty Quantification for Machine-Learned Interatomic Potentials via Pre-Training

Authors: Stefan P. Schmid, Tobias Setzer, Maximilian Scheurer

Organizations: Merck KGaA, Frankfurter Str. 250, 64293 Darmstadt, Germany · Merck Electronics KGaA, Frankfurter Str. 250, 64293 Darmstadt, Germany

Abstract

Informative uncertainty quantification (UQ) is essential for the practical deployment of machine-learned interatomic potentials (MLIPs), as it provides insight into their reliability across chemical space. Although numerous post-hoc uncertainty estimation methods for MLIPs have recently been proposed, most have been trained in low-data regimes, restricting their applicability to narrow domains of interest. Inspired by the emergence of foundational MLIPs, we propose pre-training uncertainty estimators on large, diverse datasets to enable broader applicability. We pre-train several UQ models on large materials science datasets to quantify the uncertainty of different MLIPs, and evaluate their performance on both in-distribution and out-of-domain (OOD) adsorption-energy predictions. We find that pre-trained uncertainty estimators provide viable zero-shot uncertainty estimates across multiple datasets, while fine-tuning consistently yields positive correlations between predicted uncertainty and prediction error, even on the most challenging datasets. These results lay the groundwork for foundational UQ models for MLIPs, ultimately facilitating their deployment in practical settings by providing users with a priori information about model reliability and applicability.

Figures & tables

Appendix figures & tables20 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

May 1, 2026

Knowing when to trust machine-learned interatomic potentials

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.
Oct 29, 2026

Does System Size Matter in Uncertainty Quantification for Machine Learning Interatomic Potentials?

Machine-learning interatomic potentials (MLIPs) commonly use uncertainty estimates to flag out-of-distribution (OOD) configurations and select structures for active learning. A key question is whether the resulting structure-level uncertainty scores remain comparable across different system sizes. A common practice is to aggregate per-atom uncertainties by their mean or maximum and compare the resulting score against a fixed threshold, implicitly assuming such comparability. We test this assumption across several in-distribution (ID)/OOD pairs by characterizing both aggregations as a function of atom count. The two aggregations exhibit distinct size dependencies. Mean aggregation remains size-independent for homogeneous systems but loses sensitivity to defects in large cells. In contrast, maximum aggregation grows with system size for ID systems while remaining approximately flat for defect-containing systems, narrowing the ID-OOD gap in a manner consistent with extreme-value statistics. Although highly uncertain atoms in large structures may indicate novel local environments, retraining on these structures yields little improvement, in contrast to retraining on OOD structures constructed to lie outside the training distribution. Our results reveal a practical limitation of size-unaware uncertainty criteria and motivate size-aware calibration for more reliable OOD detection in large-scale atomistic simulations.
Jun 6, 2026

Inverse design of bespoke interatomic potentials via active learning by information-matching

Interatomic potentials (IPs) enable large-scale atomistic simulations beyond the reach of first-principles methods, but their predictive reliability depends critically on the selection of training data, quantified uncertainty, and model expressiveness. Active learning (AL) provides a principled framework for constructing efficient and accurate IPs, yet most strategies reduce parameter uncertainty without explicitly accounting for the specific material properties being predicted. The information-matching (IM) approach addresses this limitation by requiring that the selected training data provide at least as much parameter space information as needed to achieve prescribed uncertainty targets for selected quantities of interest (QoIs). Here, we apply IM to develop bespoke IPs specifically tailored for predicting plastic strength in metals. Due to the high computational cost of simulating plastic strength, we employ an indirect IM strategy that targets inexpensive intermediate QoIs that correlate with strength. The IM method enables precise parameter constraints with minimal training data, yielding precise predictions for both the intermediate QoIs and plastic strength. Yet, model error remains a key limitation, and a post hoc uncertainty inflation correction provides a viable means to mitigate this limitation. These findings illustrate both the promise and limits of uncertainty-aware AL for predicting complex material properties.