Full-data accuracy with fewer labels for training and fine-tuning machine-learning force fields

2026-07-15

AI Summary

This work addresses the limitation of machine learning force fields (MLFFs), whose performance is constrained by the distribution of training data, while the construction of diverse training sets remains costly. The authors propose an active learning framework based on last-layer projection regression (LLPR), introducing LLPR-based uncertainty estimation to MLFFs for the first time. This approach efficiently identifies high-value configurations through a single forward pass, substantially reducing the number of required density functional theory (DFT) labels. By circumventing the need for ensemble-based methods and repeated fine-tuning, the framework enables automatic termination of the learning loop and supports setting absolute force error thresholds. Demonstrated across molecular, condensed-phase, and electrolyte systems, the method achieves accuracy comparable to models trained on full datasets using only a small fraction of DFT labels, outperforming random sampling in fine-tuning and faithfully reproducing reference densities and ionic coordination structures.

Abstract

Machine-learning force fields (MLFFs) are reliable only near their training distribution, making efficient construction of diverse training sets a major bottleneck for both train-from-scratch and foundation fine-tuning workflows. Active learning can reduce this cost, but standard model-committee uncertainty is impractical for foundation MLFFs because each committee member requires a separate fine-tuning run. We present an active-learning workflow based on last-layer-projection regression (LLPR), a forward-pass-cheap per-configuration uncertainty estimator. Across molecular, condensed-phase, and electrolyte systems, LLPR identifies compact, high-value training sets that recover full-data accuracy using only a small fraction of electronic-structure labels. In foundation-model fine-tuning, LLPR-selected configurations reach the full-pool fine-tuning ceiling with substantially fewer labels than random selection. In iterative electrolyte fine-tuning, LLPR detects unphysical local coordination before DFT labelling, provides an absolute force-error threshold, and enables automatic termination of the learning loop. The resulting models reproduce reference density and ion-coordination structure, providing a scalable uncertainty-quantification strategy across MLFF training regimes.

Problem

• machine-learning force fields
• training data efficiency
• label scarcity
• active learning
• uncertainty quantification

Innovation

• last-layer-projection regression
• active learning
• machine-learning force fields
• uncertainty quantification
• foundation model fine-tuning