Human Preference Alignment on FLUX.1 (dev)
0.36HPS-v2.1TMPO
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| TMPOIter. Time (s)=72.12026.05 | 0.36 | 1.613 | 23.842 | 0.132 | 0.243 | |
| TreeGRPOIter. Time (s)=87.62026.05 | 0.355 | 1.603 | 23.951 | -0.282 | 0.178 | |
| MixGRPOIter. Time (s)=106.52026.05 | 0.351 | 1.59 | 23.717 | -0.308 | 0.175 | |
| Flow-GRPOIter. Time (s)=119.42026.05 | 0.342 | 1.622 | 22.769 | -0.102 | 0.199 | |
| GARDOIter. Time (s)=125.92026.05 | 0.341 | 1.574 | 22.927 | -0.184 | 0.209 | |
| FLUX.1-dev2026.05 | 0.31 | 1.119 | 22.604 | -0.056 | 0.221 |