Detail Binding on T2I-CompBench++
87.03BLIP-VQA Color ScoreDetail++(SD3)
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Detail++(SD3)Train=No, Base Model=SD32025.07 | 87.03 | 86.99 | 61 | 0.925 | 1.965 | 1.851 | |
| Detail++(Flux)Train=No, Base Model=Flux2025.07 | 84.61 | 83.48 | 60.51 | 1.084 | 1.933 | 1.876 | |
| T2I-R1Train=Yes2025.07 | 84.32 | 76.53 | 61.23 | 0.874 | 1.922 | 1.538 | |
| R-Bind(SD3)Train=No, Base Model=SD32025.07 | 81.43 | 77.36 | 60.21 | 0.667 | 1.899 | 1.274 | |
| TACA(SD3)Train=Yes, Base Model=SD32025.07 | 80.27 | 75.04 | 60.24 | 0.781 | 1.863 | 1.26 | |
| SD3Train=Yes2025.07 | 80.17 | 76.5 | 60.03 | 0.675 | 1.843 | 0.037 | |
| TACA(Flux)Train=Yes, Base Model=Flux2025.07 | 77.52 | 65.84 | 51.71 | 0.821 | 1.908 | 1.65 | |
| Flux.1.devTrain=Yes2025.07 | 74.24 | 64.46 | 57.67 | 0.791 | 1.875 | 1.189 | |
| RPGTrain=No2025.07 | 74 | 65.47 | 58.73 | 0.684 | 1.861 | 1.62 | |
| Detail++(SDXL)Train=No, Base Model=SDXL2025.07 | 73.89 | 72.41 | 55.82 | 0.934 | 1.922 | 1.842 | |
| Iter-CompTrain=Yes2025.07 | 71.14 | 65.17 | 54.47 | 0.746 | 1.829 | 0.714 | |
| ELLA(SD1.5)Train=Yes, Base Model=SD1.52025.07 | 69.11 | 63.08 | 49.38 | 0.4 | 1.73 | 0.016 | |
| Ranni(SDXL)Train=Yes, Base Model=SDXL2025.07 | 68.93 | 63.25 | 49.34 | — | — | — | |
| PixArt-αTrain=Yes2025.07 | 68.86 | 70.44 | 55.82 | -0.584 | 1.709 | 1.02 | |
| GORSTrain=Yes2025.07 | 66.03 | 62.87 | 47.85 | 0.239 | 1.752 | -0.105 | |
| ToMeTrain=No2025.07 | 65.83 | 63.71 | 55.17 | 0.424 | 1.564 | 0.563 | |
| SDXLTrain=Yes2025.07 | 63.69 | 56.37 | 54.08 | 0.503 | 1.663 | -0.075 | |
| Attention RegulationTrain=No2025.07 | 58.6 | 51.73 | 46.72 | 0.201 | 1.454 | -0.794 | |
| Structured DiffusionTrain=No2025.07 | 49.9 | 49 | 42.18 | 0.074 | 0.912 | -0.124 | |
| SD1.5Train=Yes2025.07 | 47.19 | 43.34 | 38.98 | -2.278 | -1.115 | -1.741 | |
| Composable DiffusionTrain=No2025.07 | 40.63 | 36.45 | 32.99 | -2.009 | 0.747 | -0.819 |