Text-to-image reward alignment on GenEval (test)
0.26Reward 1 Score (r1)DB-MPA
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DB-MPAPreference weight (w)=12025.05 | 0.26 | 0.01 | |
| DB-MPA-LSPreference weight (w)=12025.05 | 0.26 | 0.01 | |
| RSPreference weight (w)=12025.05 | 0.26 | 0.01 | |
| DB-MPAPreference weight (w)=0.82025.05 | 0.25 | 0.18 | |
| DB-MPA-LSPreference weight (w)=0.82025.05 | 0.25 | 0.13 | |
| DB-MPA-LSPreference weight (w)=0.52025.05 | 0.19 | 0.37 | |
| DB-MPAPreference weight (w)=0.52025.05 | 0.13 | 0.47 | |
| RSPreference weight (w)=0.82025.05 | 0.12 | 0.07 | |
| CoDePreference weight (w)=0.82025.05 | 0.12 | -0.05 | |
| CoDePreference weight (w)=12025.05 | 0.12 | -0.09 | |
| CoDePreference weight (w)=0.52025.05 | 0.05 | 0.01 | |
| RGGPreference weight (w)=02025.05 | 0.03 | 0.52 | |
| RGGPreference weight (w)=0.22025.05 | 0.02 | 0.44 | |
| CoDePreference weight (w)=02025.05 | -0.03 | 0.07 | |
| DB-MPAPreference weight (w)=0.22025.05 | -0.05 | 0.73 | |
| DB-MPA-LSPreference weight (w)=0.22025.05 | -0.07 | 0.69 | |
| CoDePreference weight (w)=0.22025.05 | -0.07 | 0.05 | |
| RSPreference weight (w)=0.52025.05 | -0.09 | 0.21 | |
| DB-MPAPreference weight (w)=02025.05 | -0.19 | 0.83 | |
| DB-MPA-LSPreference weight (w)=02025.05 | -0.19 | 0.83 | |
| RSPreference weight (w)=02025.05 | -0.19 | 0.83 | |
| SDPreference weight (w)=02025.05 | -0.2 | -0.04 | |
| SDPreference weight (w)=0.22025.05 | -0.2 | -0.04 | |
| SDPreference weight (w)=0.52025.05 | -0.2 | -0.04 | |
| SDPreference weight (w)=0.82025.05 | -0.2 | -0.04 | |
| SDPreference weight (w)=12025.05 | -0.2 | -0.04 | |
| RGGPreference weight (w)=12025.05 | -0.2 | -0.06 | |
| RSPreference weight (w)=0.22025.05 | -0.21 | 0.54 | |
| RGGPreference weight (w)=0.52025.05 | -0.35 | 0.27 | |
| RGGPreference weight (w)=0.82025.05 | -0.41 | 0.1 |