Preprint—not peer reviewed
Contrastive Regularization of Machine Learning Potentials
Contrastive Regularized MSE adds a distribution-aware term to ordinary energy-and-force fitting and uses persistent Langevin samples from the potential itself to expose configurations that should be raised in energy. On ethanol and aspirin, the authors report that the correction restores energy, distance, and free-energy distributions to near-quantitative agreement with density functional theory while preserving force accuracy. The result makes a useful point for molecular simulation: a potential meant to generate trajectories must be trained against the distribution it produces.
AI–chemistryAI–materials Atomistic modelingNeural potentials