Text-to-Image Generation on GenEval 2 (Component Accuracy and Soft-TIFA Scores)
85.41Attribute AccuracyELDiff
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| ELDiffFine-tuned with=Qwen-Image2026.06 | 85.41 | 70.67 | 99.58 | 61.97 | 39.4 | 83.13 | 37.94 | |
| TokenComFine-tuned with=Qwen-Image2026.06 | 84.42 | 68.55 | 99.4 | 61.54 | 37.88 | 81.79 | 35.83 | |
| LLDMFine-tuned with=Qwen-Image2026.06 | 84.33 | 68.47 | 99.46 | 60.93 | 38.07 | 81.9 | 35.11 | |
| Qwen-Image2026.06 | 84.11 | 67.98 | 99.3 | 60.82 | 37.5 | 81.58 | 34.57 |