Active rejection enables reliable generalization of universal machine-learning interatomic potentials
2026-07-10 • Machine Learning
Machine Learning
AI summaryⓘ
The authors introduce Adaptive Multi-Teacher Routing (ATR), a method to improve how machine learning models predict atomic interactions with high accuracy but less expensive data. ATR uses several pretrained models ('teachers') and decides which one to trust most for each atomic structure, rejecting uncertain predictions. This approach allows them to generate millions of reliable pseudo-labels using very few expensive high-accuracy calculations. Their technique improves prediction accuracy and stability in molecular simulations compared to traditional methods. Overall, the authors demonstrate a way to efficiently create better training data for complex materials modeling.
machine learning interatomic potentialsr2SCANpseudo-labelingmolecular dynamicsmulti-teacher learningactive rejectionstructural descriptorsmaterial simulationCHGNetbenchmarking
Authors
Mingxiang Luo, Xinnan Mao, Lu Wang, Lei Bai, Feng Ding, Yuqiang Li
Abstract
Universal machine learning interatomic potentials (uMLIPs) bridge quantum-mechanical accuracy and large-scale molecular dynamics, but the cost of high-accuracy calculations such as r$^2$SCAN limits training to datasets that remain small relative to the open materials space. Strong average benchmark performance also does not guarantee reliable energy--force predictions for every structure. We propose Adaptive Multi-Teacher Routing (ATR), which reformulates high-fidelity data construction as a structure-wise decision problem under uncertainty. Using a small set of real r$^2$SCAN labels, ATR calibrates multiple pretrained uMLIP teachers and combines structural descriptors, teacher identity, and inter-teacher disagreement to estimate the reliability of each structure--teacher pair. It selects high-confidence predictions for pseudo-label generation and rejects structures for which no teacher is sufficiently reliable. With real r$^2$SCAN labels for only 0.2\% of candidate structures, ATR distils 2.89 million traceable r$^2$SCAN-level pseudo-labels for pretraining. On held-out r$^2$SCAN structures and the MP-r$^2$SCAN benchmark, a lightweight CHGNet trained on the ATR-generated dataset consistently outperforms the baseline and non-routed controls. Finite-temperature molecular dynamics further shows that ATR improves dynamical robustness across multiple material systems, maintaining stable trajectories where baseline simulations undergo catastrophic structural collapse. These results establish active rejection as an effective mechanism for converting multiple pretrained uMLIPs into a scalable and reliable data-construction system for high-fidelity uMLIPs.