Text-to-Image Instruction Following on TIFF-Bench short descriptions v1
88.5Overall ScorePromptEcho
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PromptEchobackbone=Z-Image2026.04 | 88.5 | 90.3 | 94.5 | 90.9 | 85.5 | 83.3 | 83.5 | 80.1 | 84.4 | 83.3 | 99.1 | 95.2 | |
| Qwen-Iter+Parstrategy=iterative refinement with parallel comparison2026.01 | 87.4 | 88.1 | 90.5 | 88.1 | 85.4 | 81.5 | 81.4 | 82 | 80.5 | 90 | 97.7 | 92 | |
| Seedream 3.02026.01 | 86 | 87.1 | 90.5 | 89.9 | 80.9 | 79.2 | 79.8 | 77.2 | 75.6 | 100 | 97.2 | 83.2 | |
| PromptEchobackbone=QwenImage-25122026.04 | 85.5 | 87.1 | 91 | 84.7 | 85.5 | 80.6 | 79.1 | 80.2 | 81.3 | 72.4 | 99.1 | 96.2 | |
| Qwen-Iterstrategy=iterative refinement2026.01 | 85.4 | 85 | 92 | 80.5 | 82.3 | 81.3 | 80.8 | 80.1 | 80.2 | 86.2 | 97.6 | 88.4 | |
| Qwen-Parallelstrategy=parallel generation2026.01 | 85.2 | 85.2 | 89.7 | 88.3 | 77.7 | 80.6 | 81.9 | 79.6 | 77.8 | 89.7 | 93.7 | 90.4 | |
| Z-Imagebaseline=true2026.04 | 84.91 | 86.4 | 89.5 | 90.3 | 79.3 | 79.9 | 80.8 | 75.8 | 80.9 | 82.8 | 93.2 | 91.6 | |
| QwenImage-2512baseline=true2026.04 | 84.89 | 85.5 | 92.2 | 84.9 | 79.4 | 80.3 | 84.5 | 73.1 | 81.6 | 80 | 95.5 | 92.9 | |
| FLUX.1 [dev]2026.01 | 71.1 | 83.1 | 87.1 | 87.3 | 75 | 65.8 | 67.1 | 73.8 | 69.1 | 66.7 | 43.8 | 70.7 | |
| MidJourney v72026.01 | 68.7 | 77.4 | 77.6 | 82.1 | 72.6 | 64.7 | 67.2 | 81.2 | 60.7 | 83.3 | 24.8 | 68.8 | |
| SD 32026.01 | 67.5 | 78.3 | 83.3 | 82.1 | 71.1 | 61.5 | 61.1 | 68.8 | 51 | 66.7 | 59.8 | 63.2 | |
| Janus-Pro-7B2026.01 | 66.5 | 79.3 | 79.3 | 78.3 | 80.3 | 59.7 | 66.1 | 70.5 | 67.2 | 60 | 28.8 | 65.8 |