CrystalGRPO post-trains flow-based crystal generators with joint coordinate-lattice stochastic policies and a hybrid MACE-energy plus structure-matching reward, improving Top-1 recovery in one mode and Top-20 coverage in the other.
Tadmor, and Stefano Martiniani
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
CrystalGRPO: Target-Aligned and Coverage-Preserving Reinforcement Learning for Flow-Based Crystal Structure Prediction
CrystalGRPO post-trains flow-based crystal generators with joint coordinate-lattice stochastic policies and a hybrid MACE-energy plus structure-matching reward, improving Top-1 recovery in one mode and Top-20 coverage in the other.