None defined yet.
$\text{G}^2$RPO: Granular GRPO for Precise Reward in Flow Models
A long-context, multimodal document understanding benchmark