TY  - RPRT
TI  - LC-GRPO: Bridging Train-Inference Gap for Flow-Based GRPO with Langevin Correction
AU  - Yingqing Guo
AU  - Hui Yuan
AU  - Zijian He
AU  - Mengdi Wang
AU  - Zheng Ding
PY  - 2026
UR  - https://arxiv.org/abs/2608.05600
ID  - 2608.05600
ER  -