Text-to-Image Generation on TextAlign General Generation Benchmark
31.56CLIPScoreZ-Image
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Z-ImageAlignment=DPO2026.05 | 31.56 | 0.898 | 21.54 | 12.12 | |
| Z-ImageAlignment=GRPO2026.05 | 31.47 | 0.928 | 21.55 | 12.13 | |
| Z-ImageAlignment=Base2026.05 | 31.36 | 0.926 | 21.54 | 12.16 | |
| Qwen-ImageAlignment=Base2026.05 | 31.21 | 0.915 | 21.54 | 11.65 | |
| SD3.5Alignment=Base2026.05 | 31.04 | 0.682 | 21.12 | 10.6 | |
| FLUXAlignment=GRPO2026.05 | 29.65 | 0.724 | 21.41 | 11.88 | |
| FLUXAlignment=DPO2026.05 | 29.49 | 0.721 | 12.33 | 11.76 | |
| FLUXAlignment=Base2026.05 | 29.38 | 0.744 | 21.41 | 11.84 | |
| TextDiffuserAlignment=Base2026.05 | 28.53 | -0.328 | 19.93 | 8.097 | |
| AnyTextAlignment=Base2026.05 | 20.77 | -0.902 | 19.11 | 6.01 |