AI RESEARCH

OP-GRPO: Efficient Off-Policy GRPO for Flow-Matching Models

arXiv CS.CV

ArXi:2604.04142v1 Announce Type: new