Text-to-Image Generation on Text-to-Image Evaluation Benchmarks
10.86HPSv3SD 3.5 Medium
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SD 3.5 MediumTeacher Architect.=MM-DiT, Paras (B)=2.50, NFE=502026.06 | 10.86 | 30.02 | 0.96 | 14.13 | 84.5 | |
| KolorsTeacher Architect.=Large U-Net, Paras (B)=2.57, NFE=502026.06 | 10.59 | 30.94 | 0.88 | 14.12 | 76.52 | |
| Bridge (Merged)Student Backbone=SD 1.5, Teacher Model=All 5 Teachers, Paras (B)=0.86, NFE=12026.06 | 10.53 | 29.07 | 0.65 | 12.62 | 66.67 | |
| FLUX.2-klein-4BTeacher Architect.=MM-DiT, Paras (B)=4.00, NFE=502026.06 | 10.05 | 28.9 | 0.8 | 14 | 83.2 | |
| Bridge (Merged)Student Backbone=SD 2.1, Teacher Model=All 5 Teachers, Paras (B)=0.86, NFE=12026.06 | 9.75 | 30 | 0.74 | 12.6 | 68.5 | |
| PixArt-Σ-1024Teacher Architect.=DiT, Paras (B)=0.61, NFE=202026.06 | 9.62 | 30.39 | 0.92 | 14.14 | 80 | |
| BridgeStudent Backbone=SD 1.5, Teacher Model=FLUX.2-klein-4B [17], Teacher Architect.=MM-DiT, Paras (B)=0.86, NFE=12026.06 | 9.49 | 26.64 | 0.4 | 12.35 | 64.06 | |
| BridgeStudent Backbone=SD 1.5, Teacher Model=SD 3.5 Medium [8], Teacher Architect.=MM-DiT, Paras (B)=0.86, NFE=12026.06 | 9.42 | 28.3 | 0.62 | 12.96 | 65.75 | |
| BridgeStudent Backbone=SD 1.5, Teacher Model=Kolors [46], Teacher Architect.=Large U-Net, Paras (B)=0.86, NFE=12026.06 | 9.33 | 27.38 | 0.34 | 11.66 | 64.46 | |
| SDXLTeacher Architect.=Large U-Net, Paras (B)=2.57, NFE=502026.06 | 9.25 | 28.36 | 0.66 | 13.72 | 74 | |
| BridgeStudent Backbone=SD 1.5, Teacher Model=SDXL [36], Teacher Architect.=Large U-Net, Paras (B)=0.86, NFE=12026.06 | 9.04 | 27.37 | 0.3 | 11.93 | 63 | |
| BridgeStudent Backbone=SD 2.1, Teacher Model=FLUX.2-klein-4B [17], Teacher Architect.=MM-DiT, Paras (B)=0.86, NFE=12026.06 | 8.74 | 28.31 | 0.33 | 12.03 | 66 | |
| BridgeStudent Backbone=SD 1.5, Teacher Model=PixArt-Σ-1024 [3], Teacher Architect.=DiT, Paras (B)=0.86, NFE=12026.06 | 8.65 | 26.43 | 0.3 | 11.94 | 63.43 | |
| BridgeStudent Backbone=SD 2.1, Teacher Model=SD 3.5 Medium [8], Teacher Architect.=MM-DiT, Paras (B)=0.86, NFE=12026.06 | 8.64 | 29.11 | 0.45 | 12.92 | 68.52 | |
| BridgeStudent Backbone=SD 2.1, Teacher Model=Kolors [46], Teacher Architect.=Large U-Net, Paras (B)=0.86, NFE=12026.06 | 8.52 | 26.57 | 0.2 | 11.91 | 68.34 | |
| BridgeStudent Backbone=SD 2.1, Teacher Model=PixArt-Σ-1024 [3], Teacher Architect.=DiT, Paras (B)=0.86, NFE=12026.06 | 8.42 | 28.29 | 0.44 | 12.2 | 68.51 | |
| BridgeStudent Backbone=SD 2.1, Teacher Model=SDXL [36], Teacher Architect.=Large U-Net, Paras (B)=0.86, NFE=12026.06 | 8.4 | 26.73 | 0.23 | 11.88 | 67.03 | |
| SiD (init.)Student Backbone=SD 2.1 Student, Paras (B)=0.86, NFE=12026.06 | 6.42 | 23.74 | 0.1 | 11.29 | 61.33 | |
| DMD2 (init.)Student Backbone=SD 1.5 Student, Paras (B)=0.86, NFE=12026.06 | 5.37 | 21.9 | -0.29 | 10.42 | 59.85 |