Slice discovery and debiasing on Waterbirds
92.4Worst-group AccuracyLADDER
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LADDERWeights=false, Data=true, GT=false2026.03 | 92.4 | 93.1 | |
| DRO B2TWeights=false, Data=true, GT=false2026.03 | 90.7 | 92.1 | |
| DEXTERWeights=true, Data=false, GT=false2026.03 | 90.5 | 92 | |
| DROWeights=true, Data=true, GT=true2026.03 | 89.9 | 91.5 | |
| UNBOXWeights=false, Data=false, GT=false2026.03 | 88.6 | 91 | |
| CNCWeights=true, Data=true, GT=false2026.03 | 88.5 | 90.9 | |
| JTTWeights=true, Data=true, GT=false2026.03 | 83.8 | 89.3 | |
| LfFWeights=true, Data=true, GT=false2026.03 | 78 | 91.2 | |
| GEORGEWeights=true, Data=true, GT=false2026.03 | 76.2 | 95.7 | |
| ERMWeights=true, Data=true, GT=false2026.03 | 62.6 | 97.3 |