Text-to-Image Generation on Pick-a-Pic v2
32.96CLIP ScoreDiffusion-DPO (CLIP)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Diffusion-DPO (CLIP)status=upper-bound2025.01 | 32.96 | — | — | — | — | — | — | — | |
| PPDconditioning=reward-conditioned2025.01 | 32.66 | 5.92 | 27.51 | — | — | — | — | — | |
| Diffusion-DPO2025.01 | 32.48 | 5.46 | 25.96 | — | — | — | — | — | |
| SFTconditioning=reward-conditioned2025.01 | 32.26 | 5.56 | 25.78 | — | — | — | — | — | |
| Stable Cascadebase_model=Stable Cascade Stage C2025.01 | 31.97 | 5.33 | 23.87 | — | — | — | — | — | |
| ArenaPOBackbone=Stable Diffusion XL2026.05 | 0.3764 | 6.026 | 0.2883 | 0.9172 | — | 0.2268 | — | — | |
| DSPOBackbone=Stable Diffusion XL2026.05 | 0.3756 | 5.8797 | 0.2834 | 0.8818 | — | 0.2228 | — | — | |
| SDPOBackbone=Stable Diffusion XL2026.05 | 0.3746 | 5.9812 | 0.2876 | 0.884 | — | 0.2257 | — | — | |
| Diff.-DPOBackbone=Stable Diffusion XL2026.05 | 0.3732 | 6.0115 | 0.2868 | 0.8357 | — | 0.2251 | — | — | |
| Pre-TrainedBackbone=Stable Diffusion XL2026.05 | 0.3684 | 5.997 | 0.2846 | 0.7382 | — | 0.2242 | — | — | |
| MaPOBackbone=Stable Diffusion XL2026.05 | 0.3684 | 6.0979 | 0.2871 | 0.8359 | — | 0.2242 | — | — | |
| SFTBackbone=Stable Diffusion XL2026.05 | 0.3658 | 5.7922 | 0.2809 | 0.5974 | — | 0.2183 | — | — | |
| ArenaPOBackbone=Stable Diffusion 1.52026.05 | 0.348 | 5.7317 | 0.2781 | 0.6063 | — | 0.2151 | — | — | |
| SDPOBackbone=Stable Diffusion 1.52026.05 | 0.3458 | 5.7172 | 0.2772 | 0.5546 | — | 0.2143 | — | — | |
| DSPOBackbone=Stable Diffusion 1.52026.05 | 0.3428 | 5.6825 | 0.2769 | 0.5642 | — | 0.2131 | — | — | |
| SFTBackbone=Stable Diffusion 1.52026.05 | 0.3408 | 5.6888 | 0.2765 | 0.5767 | — | 0.2128 | — | — | |
| MaPOBackbone=Stable Diffusion 1.52026.05 | 0.3365 | 5.5572 | 0.2702 | 0.2435 | — | 0.2097 | — | — | |
| Diff.-DPOBackbone=Stable Diffusion 1.52026.05 | 0.3357 | 5.4958 | 0.269 | 0.102 | — | 0.2109 | — | — | |
| Pre-TrainedBackbone=Stable Diffusion 1.52026.05 | 0.3299 | 5.3907 | 0.2651 | -0.1376 | — | 0.2073 | — | — | |
| Diff-DPOModel Architecture=SDXL2026.04 | 0.292 | 6.031 | — | 0.953 | 0.299 | 22.467 | 12.21 | — | |
| Semi-DPOModel Architecture=SDXL2026.04 | 0.287 | 6.21 | — | 1.056 | 0.304 | 22.352 | 12.548 | — | |
| MaPOModel Architecture=SDXL2026.04 | 0.286 | 6.189 | — | 0.852 | 0.29 | 22.091 | 11.887 | — | |
| SDXLModel Architecture=SDXL2026.04 | 0.283 | 6.055 | — | 0.716 | 0.282 | 22.031 | 11.809 | — | |
| Diff-KTOModel Architecture=SD1.52026.04 | 0.281 | 5.659 | — | 0.629 | 0.281 | 21.064 | 10.226 | — | |
| Semi-DPOModel Architecture=SD1.52026.04 | 0.281 | 5.801 | — | 0.801 | 0.288 | 21.524 | 11.03 | — | |
| Diff-DPOModel Architecture=SD1.52026.04 | 0.279 | 5.549 | — | 0.297 | 0.261 | 20.948 | 10.144 | — | |
| SD1.5Model Architecture=SD1.52026.04 | 0.273 | 5.421 | — | 0.085 | 0.25 | 20.566 | 9.635 | — | |
| Diffusion-DPOBase Model=SDXL2026.05 | — | — | — | — | 0.2868 | 0.2271 | — | 7.157 | |
| Diffusion-DPO (Aesthetic)status=upper-bound2025.01 | — | 6.42 | — | — | — | — | — | — | |
| Diffusion-DPO (HPS)status=upper-bound2025.01 | — | — | 28.61 | — | — | — | — | — | |
| Linear-DPOBase Model=SDXL2026.05 | — | — | — | — | 0.2884 | 0.2283 | — | 7.1615 | |
| MaPOBase Model=SDXL2026.05 | — | — | — | — | 0.283 | 0.2232 | — | 6.8585 | |
| SDXLBase Model=SDXL2026.05 | — | — | — | — | 0.2805 | 0.2229 | — | 6.6575 | |
| SFTBase Model=SDXL2026.05 | — | — | — | — | 0.2768 | 0.2171 | — | 6.0146 |