Text-to-Image Generation Alignment on GenEval 80-Obj
100S-ObjSD3.5-M + UnifiedReward
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SD3.5-M + UnifiedRewardBackbone=SD3.5-M, Reward Model=UnifiedReward2026.03 | 100 | 98 | 90 | 86 | 81 | 78 | 89 | |
| SD3.5-M + SpatialRewardBackbone=SD3.5-M, Reward Model=SpatialReward2026.03 | 100 | 99 | 86 | 96 | 98 | 91 | 95 | |
| FLUX1-dev + SpatialRewardBackbone=FLUX1-dev, Reward Model=SpatialReward2026.03 | 100 | 99 | 89 | 98 | 99 | 94 | 97 | |
| GPT Image 1Type=Proprietary T2I model2026.03 | 99 | 92 | 85 | 92 | 75 | 61 | 84 | |
| Seedream 3.0Type=Proprietary T2I model2026.03 | 99 | 96 | 91 | 93 | 47 | 80 | 84 | |
| Qwen-ImageType=Proprietary T2I model2026.03 | 99 | 92 | 90 | 88 | 76 | 77 | 87 | |
| SD3.5-M + TextOCRBackbone=SD3.5-M, Reward Model=TextOCR2026.03 | 99 | 87 | 59 | 81 | 33 | 58 | 70 | |
| SD3.5-M + QwenVLBackbone=SD3.5-M, Reward Model=QwenVL2026.03 | 99 | 58 | 48 | 81 | 24 | 55 | 61 | |
| SD3.5-MBackbone=SD3.5-M2026.03 | 98 | 79 | 59 | 80 | 28 | 58 | 67 | |
| SD3.5-M + PickScoreBackbone=SD3.5-M, Reward Model=PickScore2026.03 | 98 | 94 | 77 | 83 | 31 | 62 | 74 | |
| FLUX1-devBackbone=FLUX1-dev2026.03 | 97 | 88 | 48 | 82 | 67 | 76 | 76 | |
| SD3.5-M + ImageRewardBackbone=SD3.5-M, Reward Model=ImageReward2026.03 | 95 | 92 | 52 | 92 | 70 | 77 | 80 |