Text-to-Image Generation on TextGround-Bench Mid
29.31CS ScoreGPT-4o
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-4oFine-tuning=None2026.04 | 29.31 | 86.63 | 79.09 | 68.51 | 15.22 | 82.61 | |
| SD-3.5 LargeFine-tuning=None2026.04 | 29.02 | 76.81 | 58.44 | 63.95 | 6.3 | 65.45 | |
| FLUX.1-devFine-tuning=None2026.04 | 28.97 | 70.18 | 64.23 | 73.06 | 9.24 | 60.29 | |
| DALL·E 3Fine-tuning=None2026.04 | 28.6 | 54.45 | 40.48 | 88.43 | 4.58 | 46.38 | |
| Janus Pro 7BFine-tuning=None2026.04 | 28.26 | 14.89 | 17.57 | 82.45 | 1.6 | 12.58 | |
| Layout-posting (Ours†)Fine-tuning=TextGround4M, Resolution=512x512, Supervision Signal=Text & Bbox2026.04 | 27.54 | 21.9 | 27.75 | 81.23 | 4.32 | 19.09 | |
| TextDiffuser-2Fine-tuning=None2026.04 | 27.36 | 15.71 | 22.38 | 90.57 | 0.71 | 8.19 | |
| Pre-Image†Fine-tuning=TextGround4M, Resolution=512x5122026.04 | 27.3 | 20.33 | 26.14 | 82.67 | 4.99 | 19.55 | |
| w/ Bbox Only†Fine-tuning=TextGround4M, Resolution=512x512, Supervision Signal=Bbox2026.04 | 27.03 | 17.95 | 23.03 | 82.81 | 3.67 | 16.97 | |
| Janus Pro 1BFine-tuning=None2026.04 | 26.75 | 4.48 | 5.52 | 88.72 | 0.27 | 3.33 | |
| Janus Pro 1B†Fine-tuning=TextGround4M, Resolution=512x5122026.04 | 26.53 | 10.25 | 13.36 | 86.53 | 1.37 | 7.58 | |
| AnyTextFine-tuning=None2026.04 | 26.47 | 3.88 | 6.23 | 88.94 | — | — | |
| w/ Text Only†Fine-tuning=TextGround4M, Resolution=512x512, Supervision Signal=Text2026.04 | 26.4 | 11.78 | 15.22 | 85.04 | 1.76 | 7.88 | |
| PixArt-ΣFine-tuning=None2026.04 | 25.72 | 0.93 | 0.73 | 82.53 | 0 | 0.15 |