Generating Physically Consistent Molecules with Energy-Based Models
Researchers have developed EBMol, a novel energy-based model for generating physically consistent 3D molecules. This model learns an atom-additive potential without requiring explicit simulations during training, utilizing a Restoring Field Matching objective. EBMol achieves state-of-the-art performance on QM9 and GEOM-Drugs benchmarks and offers a principled quality metric for molecular configurations. AI
IMPACT Introduces a new method for generating physically consistent molecules, potentially advancing drug discovery and materials science.