Text-to-Image Generation on TextGround-Bench (Eazy)
31.39CS ScoreSD-3.5 Large
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SD-3.5 LargeFine-tuning=None2026.04 | 31.39 | 80.6 | 49.72 | 74.55 | 8.61 | 72.24 | |
| Janus Pro 7BFine-tuning=None2026.04 | 30.33 | 34.87 | 33.8 | 83.35 | 4.79 | 30.42 | |
| TextDiffuser-2Fine-tuning=None2026.04 | 29.47 | 30.69 | 34.19 | 88.94 | 4.12 | 26.77 | |
| Pre-Image†Fine-tuning=TextGround4M, Resolution=512x5122026.04 | 29.38 | 38.62 | 37.26 | 82.58 | 10.43 | 34.41 | |
| Layout-posting (Ours†)Fine-tuning=TextGround4M, Resolution=512x512, Supervision Signal=Text & Bbox2026.04 | 29.04 | 34.1 | 33.3 | 83.02 | 7.93 | 30.8 | |
| w/ Bbox Only†Fine-tuning=TextGround4M, Resolution=512x512, Supervision Signal=Bbox2026.04 | 28.9 | 34.29 | 33.77 | 83.83 | 7.95 | 30.61 | |
| Janus Pro 1BFine-tuning=None2026.04 | 28.71 | 10.85 | 10 | 89.86 | 0.8 | 7.22 | |
| Janus Pro 1B†Fine-tuning=TextGround4M, Resolution=512x5122026.04 | 27.92 | 20.27 | 20.09 | 87.54 | 3.8 | 15.02 | |
| PixArt-ΣFine-tuning=None2026.04 | 27.7 | 1.06 | 0.63 | 82.71 | 0 | 0.38 | |
| w/ Text Only†Fine-tuning=TextGround4M, Resolution=512x512, Supervision Signal=Text2026.04 | 27.54 | 20.46 | 20.26 | 86.04 | 3.5 | 16.35 | |
| AnyTextFine-tuning=None2026.04 | 27.37 | 19.7 | 22.61 | 82.99 | — | — | |
| FLUX.1-devFine-tuning=None2026.04 | 26.93 | 79.35 | 43.98 | 75.71 | 9.17 | 82 | |
| GPT-4oFine-tuning=None2026.04 | 26.65 | 84.45 | 50.3 | 73.52 | 18.23 | 84.91 | |
| DALL·E 3Fine-tuning=None2026.04 | 26.48 | 58.82 | 38.1 | 88.38 | 6.47 | 49.09 |