MetaShift
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Image Classification | MetaShift (test) | Average Accuracy92.1 | 60 | |
| Image Classification | MetaShift | Average Accuracy93 | 46 | |
| Classification | MetaShift | Average Worst Group Accuracy95.17 | 39 | |
| Image Classification | MetaShift Cat vs. Dog 18 (test) | Test Accuracy92.4 | 22 | |
| Domain Generalization | MetaShift Animals 8-class (test) | Test Accuracy91.4 | 22 | |
| Error Slice Discovery | MetaShift | Precision@1091 | 17 | |
| O.O.D. Classification | MetaShift Animals | OOD Accuracy91.4 | 13 | |
| O.O.D. Classification | MetaShift Cat vs. Dog | Accuracy92.4 | 13 | |
| Image Classification | MetaShift Animals | Worst Case Accuracy76.6 | 13 | |
| Image Classification | MetaShift Cat vs. Dog | WGA88.8 | 13 | |
| Group Robustness | MetaShift (test) | WGA79.4 | 13 | |
| Binary Classification | MetaShift (test) | Worst Group Accuracy (WGA)83.2 | 8 | |
| Multi-class debiasing | MetaShift 10-class | Worst-group Acc (p=12%)70.08 | 3 | |
| Multi-class debiasing | MetaShift 2-class | Worst-group Accuracy (p=12%)77.78 | 3 |
Showing 14 of 14 rows