Researchflow matchingcrystal structure predictionmolecular crystalsreinforcement learning
PackFlow Generates Lattice-Aware Molecular Crystal Proposals
7.6
Relevance Score
Researchers introduce PackFlow, a flow-matching framework for molecular crystal structure prediction, submitted Feb 23, 2026. PackFlow jointly samples heavy-atom Cartesian coordinates and unit-cell lattice parameters from molecular graphs and integrates with downstream relaxation and lattice-energy ranking. The paper also proposes physics alignment, a post-training RL stage using learned interatomic potentials to steer sampling toward low-energy basins, producing candidates that relax to lower-energy minima in blind-test evaluations.
