Human Preference on DrawBench
23.53PickScoreFlow-GRPO (PickScore-Only)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Flow-GRPO (PickScore-Only)Model Category=Single-Objective Specialist, Optimization Target=PickScore2026.01 | 23.53 | 1.11 | |
| APEXModel Category=Multi-Objective Generalist2026.01 | 23.03 | 4.49 | |
| Flow-GRPO (Static-Weight)Model Category=Multi-Objective Generalist, Weighting Strategy=Static-Weight2026.01 | 22.62 | 0.41 | |
| Flow-GRPO (OCR-Only)Model Category=Single-Objective Specialist, Optimization Target=OCR2026.01 | 22.44 | 0 | |
| SD3.5-MModel Category=Base Model2026.01 | 21.72 | — |