Classification on MetaShift
95.17Average Worst Group AccuracyDFR
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DFRGroup Info Train=x, Group Info Val=✓2026.06 | 95.17 | — | — | — | — | 75.23 | |
| CBDebugModel Framework=Post-hoc CBM2025.09 | 89.3 | — | — | — | — | 93.4 | |
| LCGroup Info Train=x, Group Info Val=✓2026.06 | 88.71 | — | — | — | — | 76.16 | |
| GB ERMGroup Info Train=✓, Group Info Val=✓2026.06 | 88.48 | — | — | — | — | 75.69 | |
| GroupDROGroup Info Train=✓, Group Info Val=✓2026.06 | 87.62 | — | — | — | — | 74.07 | |
| Reweight OnlyModel Framework=Post-hoc CBM2025.09 | 87.3 | — | — | — | — | 93.1 | |
| CB ERMGroup Info Train=x, Group Info Val=x2026.06 | 87.27 | — | — | — | — | 73.15 | |
| Augment OnlyModel Framework=Post-hoc CBM2025.09 | 86.3 | — | — | — | — | 92.6 | |
| TCSL-CS (ERM)Group Info Train=x, Group Info Val=x2026.06 | 84.95 | — | — | — | — | 79.4 | |
| OriginalModel Framework=Post-hoc CBM2025.09 | 84.5 | — | — | — | — | 92.9 | |
| RetrainModel Framework=Post-hoc CBM2025.09 | 84.4 | — | — | — | — | 93.1 | |
| TCSL-CSGroup Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 82.64 | — | — | — | — | 87.27 | |
| RGbalGroup Info Train=true, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 78.7 | — | — | — | — | 89.35 | |
| SupERGroup Info?=false, Train Twice?=false2025.08 | 76.8 | 79.8 | 78.4 | 77.6 | 71.4 | — | |
| EL2N (Top)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 75.46 | — | — | — | — | 88.54 | |
| D2 (EL2N+ResNet)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 75.46 | — | — | — | — | 88.25 | |
| RemoveModel Framework=Post-hoc CBM2025.09 | 73.9 | — | — | — | — | 89 | |
| EL2N (Hist)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 71.76 | — | — | — | — | 85.19 | |
| D2 (SelfSup+CLIP)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 71.76 | — | — | — | — | 86.05 | |
| SelfSup (Top)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 71.3 | — | — | — | — | 85.76 | |
| RandomGroup Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 71.3 | — | — | — | — | 85.94 | |
| SelfSup (Hist)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 71.06 | — | — | — | — | 85.94 | |
| D2 (SelfSup+ResNet)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 70.6 | — | — | — | — | 85.36 | |
| D2 (EL2N+CLIP)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 70.6 | — | — | — | — | 85.88 | |
| LfFGroup Info?=false, Train Twice?=false2025.08 | 70 | 77.2 | 73.9 | 69.5 | 59.5 | — | |
| CVaR DROGroup Info?=false, Train Twice?=false2025.08 | 67.5 | 77.8 | 72.5 | 65.1 | 54.7 | — | |
| JTTGroup Info?=false, Train Twice?=true2025.08 | 67.5 | 76.7 | 73.2 | 67.1 | 53 | — | |
| ERMGroup Info?=false, Train Twice?=false2025.08 | 67.3 | 78.8 | 75.8 | 61.9 | 52.6 | — | |
| CnCGroup Info?=false, Train Twice?=true2025.08 | 66.9 | 81.1 | 71.4 | 65.4 | 49.6 | — | |
| GALSGroup Info?=false, Train Twice?=false2025.08 | 66 | 74.8 | 68.8 | 70.6 | 50 | — | |
| SelfSup (Bot)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 61.34 | — | — | — | — | 84.84 | |
| EL2N (Bot)Group Info Train=false, Selection Ratio (r)=0.4, Retraining Method=CB ERM2026.06 | 60.42 | — | — | — | — | 80.21 | |
| CBDebugModel Framework=PIP-Net2025.09 | 57.3 | — | — | — | — | 82.3 | |
| Reweight OnlyModel Framework=PIP-Net2025.09 | 56.1 | — | — | — | — | 81.8 | |
| Augment OnlyModel Framework=PIP-Net2025.09 | 55.6 | — | — | — | — | 82.2 | |
| RemoveModel Framework=PIP-Net2025.09 | 55 | — | — | — | — | 81.4 | |
| RetrainModel Framework=PIP-Net2025.09 | 53.3 | — | — | — | — | 81.2 | |
| OriginalModel Framework=PIP-Net2025.09 | 52.4 | — | — | — | — | 80.9 | |
| ProtoPDebugModel Framework=PIP-Net2025.09 | 52.4 | — | — | — | — | 80.9 | |
| CnCGroup Info?=false, Train Twice?=true2025.08 | — | 2.3 | 71.5 | 262.5 | 769.6 | — | |
| CVaR DROGroup Info?=false, Train Twice?=false2025.08 | — | 21.3 | 97.5 | 237.7 | 599.7 | — | |
| ERMGroup Info?=false, Train Twice?=false2025.08 | — | 10.3 | 14.2 | 411.8 | 722.1 | — | |
| GALSGroup Info?=false, Train Twice?=false2025.08 | — | 82.1 | 197.7 | 157.7 | 955.6 | — | |
| JTTGroup Info?=false, Train Twice?=true2025.08 | — | 31.2 | 24.9 | 128.6 | 699.8 | — | |
| LfFGroup Info?=false, Train Twice?=false2025.08 | — | 15.3 | 30.3 | 73.3 | 258.8 | — | |
| SupERGroup Info?=false, Train Twice?=false2025.08 | — | 9.3 | 13.1 | 7.7 | 49.4 | — |