Text-to-Image Generation on Pick-a-Pic 1K prompts v1
1.28ImageRewardHyperAlign-P
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HyperAlign-PBackbone=FLUX, Time=17s, Denoising steps=50, CFG scale=3.52026.01 | 1.28 | 6.769 | 22.13 | 0.2506 | 0.353 | — | — | — | — | — | — | — | — | — | |
| HyperAlign-SBackbone=FLUX, Time=20s, Denoising steps=50, CFG scale=3.52026.01 | 1.251 | 6.853 | 22.37 | 0.2602 | 0.3611 | — | — | — | — | — | — | — | — | — | |
| HyperAlign-IBackbone=FLUX, Time=16s, Denoising steps=50, CFG scale=3.52026.01 | 1.242 | 6.733 | 22.18 | 0.2511 | 0.3529 | — | — | — | — | — | — | — | — | — | |
| MixGRPOBackbone=FLUX, Time=15s, Denoising steps=50, CFG scale=3.52026.01 | 1.24 | 6.76 | 22.25 | 0.2499 | 0.346 | — | — | — | — | — | — | — | — | — | |
| DanceGRPOBackbone=FLUX, Time=15s, Denoising steps=50, CFG scale=3.52026.01 | 1.135 | 6.706 | 22.08 | 0.2341 | 0.3527 | — | — | — | — | — | — | — | — | — | |
| BoNBackbone=FLUX, Time=300s, Denoising steps=50, CFG scale=3.5, number of noise candidates=202026.01 | 1.124 | 6.016 | 22.26 | 0.2668 | 0.3114 | — | — | — | — | — | — | — | — | — | |
| ε-greedyBackbone=FLUX, Time=1100s, Denoising steps=50, CFG scale=3.5, local search iterations=20, noise candidates=42026.01 | 1.077 | 6.166 | 22.3 | 0.2662 | 0.3115 | — | — | — | — | — | — | — | — | — | |
| DyMOBackbone=FLUX, Time=300s, Denoising steps=50, CFG scale=3.52026.01 | 1.062 | 6.165 | 21.74 | 0.2714 | 0.3004 | — | — | — | — | — | — | — | — | — | |
| FLUXBackbone=FLUX, Time=15s, Denoising steps=50, CFG scale=3.52026.01 | 1.023 | 6.182 | 22.26 | 0.2651 | 0.3072 | — | — | — | — | — | — | — | — | — | |
| GPSBackbone=Hunyuan-DiT2025.12 | 0.97 | — | — | 0.73 | 0.3056 | — | — | — | — | — | — | — | — | — | |
| Z-SamplingBackbone=Hunyuan-DiT2025.12 | 0.94 | — | — | 0.724 | 0.3012 | — | — | — | — | — | — | — | — | — | |
| StandardBackbone=Hunyuan-DiT2025.12 | 0.92 | — | — | 0.712 | 0.2915 | — | — | — | — | — | — | — | — | — | |
| FreeDoMBackbone=FLUX, Time=240s, Denoising steps=50, CFG scale=3.52026.01 | 0.842 | 5.933 | 21.67 | 0.2672 | 0.297 | — | — | — | — | — | — | — | — | — | |
| SRPOBackbone=FLUX, Time=15s, Denoising steps=50, CFG scale=3.52026.01 | 0.833 | 6.061 | 22.16 | 0.2624 | 0.2917 | — | — | — | — | — | — | — | — | — | |
| GPSBackbone=SDXL2025.12 | 0.79 | — | — | 0.723 | 0.2995 | — | — | — | — | — | — | — | — | — | |
| Z-SamplingBackbone=SDXL2025.12 | 0.75 | — | — | 0.719 | 0.298 | — | — | — | — | — | — | — | — | — | |
| StandardBackbone=SDXL2025.12 | 0.64 | — | — | 0.71 | 0.2899 | — | — | — | — | — | — | — | — | — | |
| Diff-KTOBackbone=SDv15, Prompt set=Pick-a-pic, Training reward=PickScore2026.05 | 0.599 | 5.585 | 20.83 | 0.24 | 0.272 | 101.3 | 0.801 | 0.634 | 22.7 | 39.12 | 26.25 | 4.361 | 11.22 | 1 | |
| DPOKBackbone=SDv15, Prompt set=Pick-a-pic, Training reward=PickScore2026.05 | 0.582 | 5.661 | 20.97 | 0.242 | 0.272 | 99.3 | 0.82 | 0.641 | 22.52 | 39.35 | 12.28 | 4.608 | 11.23 | 3 | |
| AdaScopeBackbone=SDv15, Prompt set=Pick-a-pic, Training reward=PickScore2026.05 | 0.542 | 6.071 | 23.01 | 0.243 | 0.278 | 85.37 | 0.846 | 0.652 | 23.73 | 39.37 | 12.2 | 4.019 | 11.91 | 12 | |
| TDPOBackbone=SDv15, Prompt set=Pick-a-pic, Training reward=PickScore2026.05 | 0.504 | 5.991 | 22.94 | 0.265 | 0.273 | 128.66 | 0.82 | 0.635 | 21.99 | 39.09 | 12.96 | 4.077 | 11.59 | 1 | |
| Diff-DPOBackbone=SDv15, Prompt set=Pick-a-pic, Training reward=PickScore2026.05 | 0.443 | 5.551 | 20.87 | 0.244 | 0.271 | 109.1 | 0.795 | 0.639 | 22.77 | 39.2 | 15.12 | 4.463 | 11.02 | 1 | |
| SPOBackbone=SDv15, Prompt set=Pick-a-pic, Training reward=PickScore2026.05 | 0.282 | 5.613 | 20.76 | 0.241 | 0.218 | 78.76 | 0.855 | 0.649 | 23.18 | 39.32 | 25.67 | 4.103 | 11.09 | 3 | |
| DDPOBackbone=SDv15, Prompt set=Pick-a-pic, Training reward=PickScore2026.05 | 0.196 | 5.704 | 21.79 | 0.242 | 0.212 | 147.5 | 0.539 | 0.629 | 20.01 | 39.18 | 12.46 | 4.328 | 11.74 | 1 | |
| SDv15Backbone=SDv15, Prompt set=Pick-a-pic, Training reward=PickScore2026.05 | 0.181 | 5.412 | 20.48 | 0.243 | 0.262 | — | — | 0.654 | 23.79 | 38.05 | 18.66 | 5.401 | 11.26 | 3 | |
| GPSBackbone=SD-2.12025.12 | -0.18 | — | — | 0.702 | 0.2709 | — | — | — | — | — | — | — | — | — | |
| Z-SamplingBackbone=SD-2.12025.12 | -0.25 | — | — | 0.696 | 0.2686 | — | — | — | — | — | — | — | — | — | |
| StandardBackbone=SD-2.12025.12 | -0.54 | — | — | 0.681 | 0.2541 | — | — | — | — | — | — | — | — | — |