Image Classification on 9 Vision Tasks + ImageNet-1K (test)
77SUN397 AccuracyEMR-MERGING
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| EMR-MERGINGBackbone=ViT-B/322024.05 | 77 | 75.2 | 92.9 | 92.7 | 79.7 | 90.2 | 97.6 | 76.2 | 79.8 | 84.6 | |
| IndividualBackbone=ViT-B/322024.05 | 75.3 | 77.7 | 96.1 | 99.7 | 97.5 | 98.7 | 99.7 | 79.4 | 82 | 89.6 | |
| Weight AveragingBackbone=ViT-B/322024.05 | 61.8 | 56.4 | 65.9 | 66.2 | 62.7 | 44.5 | 81.8 | 49 | 61.5 | 61.1 | |
| Ties-MergingBackbone=ViT-B/322024.05 | 53.3 | 34.1 | 57 | 55.8 | 72.3 | 43.2 | 90.5 | 46.5 | 68.9 | 58 | |
| Task ArithmeticBackbone=ViT-B/322024.05 | 51.8 | 30.9 | 55.8 | 64.3 | 69 | 42.2 | 92.7 | 46.8 | 66.6 | 57.8 |