Human Preference Alignment on DrawBench Task-specific (test)
24.64PickScore (Task Metric)DenseGRPO
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| DenseGRPO2026.01 | 24.64 | 6.35 | 4.06 | 1.41 | 24.55 | — | 3.39 | |
| Flow-GRPO+CoCARefinement=CoCA2026.01 | 23.63 | 6.22 | 4.16 | 1.32 | 23.8 | — | 3.38 | |
| Flow-GRPO2026.01 | 23.31 | 5.92 | 4.22 | 1.28 | 23.53 | 3.66 | 3.38 | |
| SD3.5-MModel Size=Medium2026.01 | 21.72 | 5.39 | 4.07 | 0.87 | 22.34 | 3.33 | 3.06 |