Text-to-Image Generation on Test (test)
1.17ImageRewardSD3.5-Medium + DRM & Step-GRPO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SD3.5-Medium + DRM & Step-GRPOReward Model=DRM, Optimization Algorithm=Step-GRPO2026.05 | 1.17 | 17.04 | 10.28 | |
| SD3.5-Medium + HPSv3 & GRPOReward Model=HPSv3, Optimization Algorithm=GRPO2026.05 | 1.15 | 16.9 | 9.71 | |
| SD3.5-Medium + PickScore & GRPOReward Model=PickScore, Optimization Algorithm=GRPO2026.05 | 1.14 | 16.94 | 9.64 | |
| SD3.5-Medium + DRM & GRPOReward Model=DRM, Optimization Algorithm=GRPO2026.05 | 1.14 | 16.95 | 10.07 | |
| SD3.5-MediumModel Size=Medium2026.05 | 1.01 | 16.76 | 8.95 |