Long-Text-to-Image Generation on DetailMaster
46.84Character PresencePRISM-Qwen
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| PRISM-QwenBase Model Architecture=Modern Architectures2026.04 | 46.84 | 91.55 | 90.36 | 93.53 | 41.49 | 94.62 | 97.32 | 95.62 | 49.23 | |
| FLUX-Dev.Base Model Architecture=Modern Architectures2026.04 | 42.02 | 91.14 | 89.61 | 90.23 | 38.18 | 95.73 | 96.91 | 95.28 | 44.94 | |
| Qwen-ImageBase Model Architecture=Modern Architectures2026.04 | 40.46 | 90.21 | 89.13 | 91.29 | 40.14 | 92 | 96.93 | 91.53 | 47.02 | |
| StableDiffusion-3.5Base Model Architecture=Modern Architectures2026.04 | 39.01 | 87.6 | 87.57 | 89.55 | 31.91 | 93.82 | 92.53 | 95.31 | 39.36 | |
| ParaDiffusion(SDXL)Base Model Architecture=Modern Architectures2026.04 | 28.63 | 87.4 | 85.34 | 84.66 | 20.62 | 84.83 | 93.59 | 72.16 | 25.95 | |
| PRISMBase Model Architecture=StableDiffusion-1.52026.04 | 28.21 | 84.78 | 83.24 | 84.54 | 16.57 | 82.45 | 92.48 | 64.1 | 20.88 | |
| PRISM w/ tuningBase Model Architecture=StableDiffusion-1.5, Tuning Status=Tuned2026.04 | 25.99 | 86.05 | 86.21 | 86.16 | 16.21 | 90.96 | 91.16 | 84.93 | 24.47 | |
| LongAlignBase Model Architecture=StableDiffusion-1.52026.04 | 25.88 | 85.54 | 83.28 | 83.85 | 14.12 | 78.6 | 87.33 | 70.49 | 21.24 | |
| ELLABase Model Architecture=StableDiffusion-1.52026.04 | 25.57 | 82.38 | 78.75 | 80.33 | 15.04 | 69.15 | 83.12 | 44.17 | 15.17 | |
| LLM4GENBase Model Architecture=StableDiffusion-1.52026.04 | 19.43 | 82.99 | 78 | 81.67 | 9.48 | 28.32 | 68.08 | 50.28 | 8.04 | |
| StableDiffusion-1.5Base Model Architecture=StableDiffusion-1.52026.04 | 19.12 | 84.4 | 76.62 | 80.73 | 8.66 | 24.53 | 69.27 | 84.47 | 7.18 | |
| LLM BlueprintBase Model Architecture=StableDiffusion-1.52026.04 | 18.69 | 81.4 | 76.25 | 76.53 | 18.4 | 56.69 | 83.28 | 67.07 | 14.16 |