Text-to-Image Generation on MS-COCO (Fidelity Metrics)
35.16ClipScoreTeacher (SANA)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Teacher (SANA)Architecture=Flow-DiT, Steps=202026.06 | 35.16 | 22.4056 | — | — | — | — | — | — | |
| SD3-M-BIMBackbone=SD3-M, Guidance Strategy=BIM2026.03 | 34.56 | 23.1 | 5.76 | 52 | 99 | 1.53 | 30.32 | 30.6 | |
| Teacher (SD v2.1)Architecture=UNet, Steps=502026.06 | 34.56 | 21.7022 | — | — | — | — | — | — | |
| SD3-L-BIMBackbone=SD3-L, Guidance Strategy=BIM2026.03 | 34.43 | 23.28 | 5.91 | 52 | 99 | 1.56 | 30.76 | 31.42 | |
| SiD + GADArchitecture=Flow-DiT, Steps=42026.06 | 34.4 | 22.0735 | — | — | — | — | — | — | |
| SD3-M-IRBackbone=SD3-M, Guidance Strategy=IR2026.03 | 34.15 | 23.14 | 5.84 | 52 | 98 | 1.83 | 30.5 | 31.17 | |
| SD3-L-IRBackbone=SD3-L, Guidance Strategy=IR2026.03 | 34.12 | 23.26 | 5.98 | 51 | 98 | 1.83 | 30.85 | 31.81 | |
| SD2-BIMBackbone=SD2, Guidance Strategy=BIM2026.03 | 33.96 | 22.31 | 5.3 | 51 | 99 | 1.09 | 29.34 | 28.17 | |
| TDM + GADArchitecture=DiT, Steps=42026.06 | 33.52 | 22.2704 | — | — | — | — | — | — | |
| SD2-IRBackbone=SD2, Guidance Strategy=IR2026.03 | 33.5 | 22.43 | 5.37 | 50 | 96 | 1.59 | 29.56 | 29.03 | |
| TDMArchitecture=DiT, Steps=42026.06 | 33.43 | 22.1007 | — | — | — | — | — | — | |
| Teacher (PixArt-α)Architecture=DiT, Steps=202026.06 | 33.41 | 22.2922 | — | — | — | — | — | — | |
| FLUX.1-dev-IRBackbone=FLUX.1-dev, Guidance Strategy=IR2026.03 | 33.04 | 23.6 | 6.14 | 50 | 97 | 1.79 | 30.48 | 31.47 | |
| FLUX.1-dev-BIMBackbone=FLUX.1-dev, Guidance Strategy=BIM2026.03 | 33.03 | 23.49 | 6.07 | 51 | 99 | 1.47 | 30.22 | 30.79 | |
| SiDArchitecture=Flow-DiT, Steps=42026.06 | 32.75 | 21.7629 | — | — | — | — | — | — | |
| LADD + GADArchitecture=UNet, Steps=12026.06 | 32.68 | 21.379 | — | — | — | — | — | — | |
| LADDArchitecture=UNet, Steps=12026.06 | 32.58 | 20.8515 | — | — | — | — | — | — | |
| SD3-MBackbone=SD3-M, Guidance Strategy=None2026.03 | 32.43 | 22.65 | 5.8 | 49 | 89 | 1.12 | 29.68 | 29.64 | |
| SD3-LBackbone=SD3-L, Guidance Strategy=None2026.03 | 32.4 | 22.75 | 5.91 | 49 | 89 | 1.14 | 30 | 30.29 | |
| SD2Backbone=SD2, Guidance Strategy=None2026.03 | 31.95 | 21.82 | 5.26 | 47 | 77 | 0.49 | 28.64 | 26.95 | |
| FLUX.1-devBackbone=FLUX.1-dev, Guidance Strategy=None2026.03 | 30.92 | 22.88 | 6.01 | 46 | 80 | 0.92 | 29.22 | 29.14 |