Open Materials Generation with Inference-Time Reinforcement Learning

Philipp Hoellmer, Stefano Martiniani

Continuous-time generative models for crystalline materials enable inverse materials design by learning to predict stable crystal structures, but incorporating explicit target prop…