Affective image generation on Pick-a-Pic
31.08HPSV2SPO
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SPOBackbone=SDXL, Training algorithm=SPO, Prompt source (MC-Agent generated)=false2025.12 | 31.08 | 0.817 | 0.512 | 0.036 | 38.6 | 55.8 | |
| SPOBackbone=SDXL, Training algorithm=SPO, Prompt source (MC-Agent generated)=true2025.12 | 30.29 | 0.801 | 0.514 | 0.035 | 40.2 | 62.4 | |
| Emotion-DirectorBackbone=SDXL, Prompt source (MC-Agent generated)=true2025.12 | 28.72 | 0.914 | 0.577 | 0.053 | 64.6 | 78.9 | |
| EmotiCrafterBackbone=SDXL, Prompt source (MC-Agent generated)=true2025.12 | 28.39 | 0.625 | 0.535 | 0.029 | 55.7 | 66.8 | |
| Diffusion-DPOBackbone=SDXL, Training algorithm=Diffusion-DPO, Prompt source (MC-Agent generated)=false2025.12 | 28.13 | 0.882 | 0.543 | 0.04 | 52.3 | 64.5 | |
| Diffusion-DPOBackbone=SDXL, Training algorithm=Diffusion-DPO, Prompt source (MC-Agent generated)=true2025.12 | 28.07 | 0.859 | 0.54 | 0.047 | 59.5 | 70.4 | |
| SFTBackbone=SDXL, Training algorithm=SFT, Prompt source (MC-Agent generated)=false2025.12 | 28.06 | 0.769 | 0.559 | 0.044 | 49.8 | 61.8 | |
| SFTBackbone=SDXL, Training algorithm=SFT, Prompt source (MC-Agent generated)=true2025.12 | 27.91 | 0.714 | 0.537 | 0.051 | 53.3 | 67.7 | |
| EmoGenBackbone=SD v1.5, Prompt source (MC-Agent generated)=false2025.12 | 22.79 | 0.228 | 0.434 | 0.036 | 47.9 | 63.5 |