Text-to-Image on PartiPrompts 42 (test)
91.4VQAScoreSD 3 M + CaPO
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SD 3 M + CaPOBackbone=SD 3 M2025.11 | 91.4 | 13.58 | 5.943 | |
| SD 3 MBackbone=SD 3 M2025.11 | 90.8 | 13.39 | 5.793 | |
| MapReduce LoRABackbone=SD 3.5 M2025.11 | 88.5 | 12.25 | 6.266 | |
| MORL-DBackbone=SD 3.5 M2025.11 | 88.1 | 12 | 6.113 | |
| Individual Expert (Flow-GRPO)Backbone=SD 3.5 M, Reward=PickScore2025.11 | 88 | 12.66 | 6.755 | |
| Individual Expert (Flow-GRPO)Backbone=SD 3.5 M, Reward=GenEval2025.11 | 87.9 | 11.81 | 6.101 | |
| MapReduce LoRA + RaTEBackbone=SD 3.5 M2025.11 | 87.7 | 12.44 | 7.238 | |
| Rewarded SoupBackbone=SD 3.5 M2025.11 | 87.5 | 12.13 | 6.194 | |
| Individual Expert (Flow-GRPO)Backbone=FLUX.1-dev, Reward=PickScore2025.11 | 87.5 | 12.76 | 5.514 | |
| MapReduce LoRABackbone=FLUX.1-dev2025.11 | 87.5 | 12.95 | 6.572 | |
| Individual Expert (Flow-GRPO)Backbone=SD 3.5 M, Reward=OCR2025.11 | 86.3 | 11.75 | 6.014 | |
| Rewarded SoupBackbone=FLUX.1-dev2025.11 | 86.2 | 12.73 | 6.65 | |
| SD 3.5 MBackbone=SD 3.5 M2025.11 | 86.1 | 11.68 | 6.034 | |
| MORL-DRBackbone=SD 3.5 M2025.11 | 86 | 11.71 | 6.014 | |
| Individual Expert (Flow-GRPO)Backbone=FLUX.1-dev, Reward=GenEval2025.11 | 85.4 | 12.45 | 6.661 | |
| Individual Expert (Flow-GRPO)Backbone=FLUX.1-dev, Reward=OCR2025.11 | 84.1 | 12.32 | 6.591 | |
| MORL-DRBackbone=FLUX.1-dev2025.11 | 84 | 12.41 | 6.534 | |
| MORL-DBackbone=FLUX.1-dev2025.11 | 83.4 | 12.36 | 6.513 | |
| FLUX.1-devBackbone=FLUX.1-dev2025.11 | 83 | 12.37 | 6.629 |