Image Classification on BFFHQ (test)
71.33Accuracy (BC)BiasEdit
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BiasEditCheckpoint selection=best bias-conflict test-set performance, Training bias-conflict ratio=0%, Number of runs=32026.05 | 71.33 | 84.33 | 97.33 | |
| DeNetDMCheckpoint selection=best bias-conflict test-set performance, Training bias-conflict ratio=0%, Number of runs=32026.05 | 62.8 | 80.53 | 98.26 | |
| BiasEditCheckpoint selection strategy=best in-distribution validation performance, Training bias-conflict ratio=0%2026.05 | 62.46 | 80.7 | 98.93 | |
| SelecMixCheckpoint selection=best bias-conflict test-set performance, Training bias-conflict ratio=0%, Number of runs=32026.05 | 57.48 | 60.08 | 62.68 | |
| LfF+BECheckpoint selection=best bias-conflict test-set performance, Training bias-conflict ratio=0%, Number of runs=32026.05 | 49.53 | 58.43 | 67.33 | |
| DeNetDMCheckpoint selection strategy=best in-distribution validation performance, Training bias-conflict ratio=0%2026.05 | 49.13 | 74.1 | 99.07 | |
| BiasAdvCheckpoint selection=best bias-conflict test-set performance, Training bias-conflict ratio=0%, Number of runs=32026.05 | 48.46 | 68.13 | 87.8 | |
| LfFCheckpoint selection=best bias-conflict test-set performance, Training bias-conflict ratio=0%, Number of runs=32026.05 | 46.8 | 57.34 | 67.88 | |
| DFACheckpoint selection=best bias-conflict test-set performance, Training bias-conflict ratio=0%, Number of runs=32026.05 | 44.52 | 68.1 | 91.68 | |
| VanillaCheckpoint selection=best bias-conflict test-set performance, Training bias-conflict ratio=0%, Number of runs=32026.05 | 44.04 | 70.54 | 97.04 | |
| SoftConCheckpoint selection=best bias-conflict test-set performance, Training bias-conflict ratio=0%, Number of runs=32026.05 | 43.5 | 68.2 | 92.9 | |
| BiasAdvCheckpoint selection strategy=best in-distribution validation performance, Training bias-conflict ratio=0%2026.05 | 40.66 | 69.86 | 99.06 | |
| LfFCheckpoint selection strategy=best in-distribution validation performance, Training bias-conflict ratio=0%2026.05 | 39.52 | 68.7 | 97.88 | |
| SoftConCheckpoint selection strategy=best in-distribution validation performance, Training bias-conflict ratio=0%2026.05 | 39.4 | 69.35 | 99.3 | |
| SelecMixCheckpoint selection strategy=best in-distribution validation performance, Training bias-conflict ratio=0%2026.05 | 39.2 | 65.7 | 92.2 | |
| VanillaCheckpoint selection strategy=best in-distribution validation performance, Training bias-conflict ratio=0%2026.05 | 37.96 | 68.64 | 99.32 | |
| DFACheckpoint selection strategy=best in-distribution validation performance, Training bias-conflict ratio=0%2026.05 | 35.12 | 66.54 | 97.96 | |
| LfF+BECheckpoint selection strategy=best in-distribution validation performance, Training bias-conflict ratio=0%2026.05 | 34.2 | 62.37 | 90.53 |