TY  - RPRT
TI  - CrystalGRPO: Target-Aligned and Coverage-Preserving Reinforcement Learning for Flow-Based Crystal Structure Prediction
AU  - Kaixiang Su
AU  - Hongfei Xue
AU  - Qiang Zhu
PY  - 2026
UR  - https://arxiv.org/abs/2608.06582
ID  - 2608.06582
ER  -