Text-to-image alignment on prompt dataset (test)
55.05GenEvalPromptLoop
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PromptLoopBase Model=SDXL [36]2025.10 | 55.05 | 0.742 | 29.06 | |
| PromptLoopTraining Setup=SDXL-turbo & RePrompt2025.10 | 54.83 | 0.8516 | 29.38 | |
| GPT-5Training Setup=SDXL-turbo & RePrompt2025.10 | 54.56 | 0.7593 | 29.15 | |
| SDXL-turboTraining Setup=SDXL-turbo & RePrompt2025.10 | 54.45 | 0.7769 | 29.15 | |
| SDXLConfiguration=Vanilla2025.10 | 54.31 | 0.5518 | 28.86 | |
| Qwen2.5-VL-3BTraining Setup=SDXL-turbo & RePrompt2025.10 | 52.12 | 0.6417 | 28.93 | |
| RePromptTraining Setup=SDXL-turbo & RePrompt2025.10 | 51.01 | 0.7876 | 29.12 | |
| PromptLoopBase Model=SD1.5 [39]2025.10 | 43.99 | -0.0375 | 27.93 | |
| SD1.5Configuration=Vanilla2025.10 | 42.06 | -0.1315 | 27.83 |