Image Classification on DTD (Accuracy, CovGap, MCCC)
0.11CovGapL-TOPK
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| L-TOPKBackbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.11 | — | 0.14 | |
| L-LAC (naive)Backbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.11 | — | 0.13 | |
| L-LACBackbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.11 | — | 0.19 | |
| L-APS (naive)Backbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.11 | — | 0.13 | |
| L-APSBackbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.11 | — | 0.18 | |
| TOPKBackbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.12 | — | 0.03 | |
| L-TOPK (naive)Backbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.12 | — | 0.04 | |
| LACBackbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.12 | — | 0.09 | |
| APSBackbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.12 | — | 0.1 | |
| RAPSBackbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.12 | — | 0.05 | |
| L-RAPS (naive)Backbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.12 | — | 0.06 | |
| L-RAPSBackbone=RN50, Alpha=0.1, Number of folds=102026.06 | 0.12 | — | 0.06 | |
| Concrete-TABackbone=ViT-B/322026.07 | — | 58.56 | — | |
| DOGE-TABackbone=ViT-B/322026.07 | — | 64.31 | — | |
| FinetuningBackbone=ViT-B/322026.07 | — | 97.29 | — | |
| Fisher MergingBackbone=ViT-B/322026.07 | — | 54.52 | — | |
| PoE-EBMBackbone=ViT-B/322026.07 | — | 76.06 | — | |
| TA-DARE-TIESBackbone=ViT-B/322026.07 | — | 60.9 | — | |
| Task Arithmetic (TA)Backbone=ViT-B/322026.07 | — | 55 | — | |
| Weight averagingBackbone=ViT-B/322026.07 | — | 51.76 | — | |
| Zero-Shot AccuracyBackbone=RN50, Alpha=0.1, Number of folds=102026.06 | — | 42.8 | — |