Text-to-Image Generation on T2I-CompBench 1.0 (test)
0.344CLIP ScoreOurs (DINO)
Evaluation Results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Ours (DINO)Model=SDXL-Turbo, Noise initialization=pink noise2025.12 | 0.344 | — | — | — | — | — | — | — | — | — | 0.804 | 0.44 | 0.808 | |
| Parmar et al.Model=SANA-Sprint-1.6B, Noise initialization=pink noise2025.12 | 0.335 | — | — | — | — | — | — | — | — | — | 0.705 | 0.346 | 0.736 | |
| i.i.d.Model=SANA-Sprint-1.6B, Noise initialization=pink noise2025.12 | 0.334 | — | — | — | — | — | — | — | — | — | 0.584 | 0.259 | 0.67 | |
| Ours (DINO)Model=PixArt-α, Noise initialization=pink noise2025.12 | 0.333 | — | — | — | — | — | — | — | — | — | 0.77 | 0.377 | 0.748 | |
| Ours (DINO)Model=SANA-Sprint-1.6B, Noise initialization=pink noise2025.12 | 0.33 | — | — | — | — | — | — | — | — | — | 0.75 | 0.443 | 0.773 | |
| Parmar et al.Model=SDXL-Turbo, Noise initialization=pink noise2025.12 | 0.328 | — | — | — | — | — | — | — | — | — | 0.742 | 0.391 | 0.751 | |
| i.i.d.Model=SDXL-Turbo, Noise initialization=pink noise2025.12 | 0.326 | — | — | — | — | — | — | — | — | — | 0.643 | 0.303 | 0.719 | |
| i.i.d.Model=PixArt-α, Noise initialization=pink noise2025.12 | 0.325 | — | — | — | — | — | — | — | — | — | 0.558 | 0.247 | 0.626 | |
| Parmar et al.Model=PixArt-α, Noise initialization=pink noise2025.12 | 0.324 | — | — | — | — | — | — | — | — | — | 0.675 | 0.326 | 0.692 | |
| Teacher modelArchitecture=SDXL-Base, Reference=Table D header (Teacher)2025.04 | 0.2958 | 13.05 | 36.76 | 0.3599 | 0.3542 | 0.4004 | 0.1055 | 0.3095 | 0.3095 | 0.3065 | — | — | — | |
| Random ConditioningRand Cond=true, Real image=true, Architecture=B-Base2025.04 | 0.2933 | 15 | 36.14 | 0.3789 | 0.3576 | 0.4207 | 0.1112 | 0.3075 | 0.3196 | 0.3159 | — | — | — | |
| Random ConditioningRand Cond=true, Real image=false, Architecture=B-Base2025.04 | 0.2896 | 15.76 | 36.03 | 0.3593 | 0.3551 | 0.4053 | 0.0889 | 0.3086 | 0.3148 | 0.3053 | — | — | — | |
| Student BaselineRand Cond=false, Real image=true, Architecture=B-Base2025.04 | 0.2878 | 15.76 | 33.79 | 0.3585 | 0.3397 | 0.3838 | 0.0981 | 0.308 | 0.3124 | 0.3001 | — | — | — | |
| Student BaselineRand Cond=false, Real image=false, Architecture=B-Base2025.04 | 0.2864 | 18.15 | 33.81 | 0.3474 | 0.3448 | 0.3786 | 0.0966 | 0.3092 | 0.3118 | 0.2985 | — | — | — |