General Text-to-Image Evaluation
23.15FIDSD3
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SD3Training Iterations=≥ 500k2026.06 | 23.15 | — | — | — | 0.321 | 0.92 | — | — | |
| FocusDiTTraining Time=156 A100 days, Training Iterations=72k2026.06 | 27.81 | — | — | — | 0.307 | 0.29 | — | — | |
| DiverseGRPOBackbone=SD3.5-M, Reward Model=HPSv32025.12 | 29.82 | 0.1851 | 0.1646 | 0.2103 | 0.3343 | 1.3239 | 0.8462 | 3.5894 | |
| PixArt-αTraining Time=753 A100 days2026.06 | 30.5 | — | — | — | 0.315 | 0.74 | — | — | |
| Flow-GRPOBackbone=SD3.5-M, Reward Model=HPSv32025.12 | 34.04 | 0.1625 | 0.0971 | 0.1967 | 0.3309 | 1.3037 | 0.8445 | 3.5825 | |
| DiverseGRPOBackbone=SD3.5-M, Reward Model=Pickscore2025.12 | 43.115 | 0.1517 | 0.1895 | 0.2137 | 0.3339 | 1.4009 | 0.8837 | 3.5852 | |
| Flow-GRPOBackbone=SD3.5-M, Reward Model=Pickscore2025.12 | 56.206 | 0.1278 | 0.0667 | 0.1701 | 0.3278 | 1.365 | 0.8809 | 3.5817 | |
| Opensora-PlanTraining Iterations=346k2026.06 | 62.41 | — | — | — | 0.217 | -1.85 | — | — | |
| DiverseGRPOBackbone=Flux.1-dev, Reward Model=Pickscore2025.12 | 62.505 | 0.1575 | 0.1059 | 0.1818 | 0.3266 | 1.3959 | 0.8779 | 3.6039 | |
| Flow-GRPOBackbone=Flux.1-dev, Reward Model=Pickscore2025.12 | 68.746 | 0.1382 | 0.0766 | 0.1578 | 0.3198 | 1.3497 | 0.875 | 3.5882 |