Image Classification on TA-8
94.3AccuracyFine-tuned
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Fine-tunedBackbone=CLIP ViT-L/142025.12 | 94.3 | 100 | |
| SAMergingBackbone=CLIP ViT-L/14, k=16002025.12 | 92.6 | 98.2 | |
| MTLBackbone=CLIP ViT-L/142025.12 | 92.3 | 98 | |
| AdaMerging LWBackbone=CLIP ViT-L/14, k=160002025.12 | 91 | 96.5 | |
| Fine-tunedBackbone=CLIP ViT-B/322025.12 | 90.3 | 100 | |
| Isotropic MergingBackbone=CLIP ViT-L/142025.12 | 90.3 | 95.8 | |
| RegMeanBackbone=CLIP ViT-L/14, k=16002025.12 | 89 | 94.4 | |
| MTLBackbone=CLIP ViT-B/322025.12 | 88.5 | 98 | |
| RegMean++Backbone=CLIP ViT-L/14, k=16002025.12 | 88.3 | 93.6 | |
| SAMergingBackbone=CLIP ViT-B/32, k=16002025.12 | 87.1 | 96.5 | |
| AdaMerging LWBackbone=CLIP ViT-L/14, k=16002025.12 | 85.1 | 90.2 | |
| RegMean++Backbone=CLIP ViT-B/32, k=16002025.12 | 84.2 | 93.2 | |
| PCB MergingBackbone=CLIP ViT-L/142025.12 | 84.2 | 89.3 | |
| TIES-MergingBackbone=CLIP ViT-L/142025.12 | 83.8 | 88.9 | |
| AdaMerging LWBackbone=CLIP ViT-B/32, k=160002025.12 | 82.6 | 91.5 | |
| Task ArithmeticBackbone=CLIP ViT-L/142025.12 | 82.1 | 87.1 | |
| RegMeanBackbone=CLIP ViT-B/32, k=16002025.12 | 80.5 | 89.1 | |
| Simple AveragingBackbone=CLIP ViT-L/142025.12 | 79.9 | 86.5 | |
| Isotropic MergingBackbone=CLIP ViT-B/322025.12 | 78.8 | 87.3 | |
| PCB MergingBackbone=CLIP ViT-B/322025.12 | 75.4 | 83.5 | |
| AdaMerging LWBackbone=CLIP ViT-B/32, k=16002025.12 | 73.7 | 81.6 | |
| FisherBackbone=CLIP ViT-L/14, k=16002025.12 | 73.3 | 77.7 | |
| TIES-MergingBackbone=CLIP ViT-B/322025.12 | 71.9 | 79.6 | |
| FisherBackbone=CLIP ViT-B/32, k=16002025.12 | 70.5 | 78.1 | |
| Task ArithmeticBackbone=CLIP ViT-B/322025.12 | 67.5 | 74.8 | |
| Simple AveragingBackbone=CLIP ViT-B/322025.12 | 66.3 | 73.4 | |
| PretrainedBackbone=CLIP ViT-L/142025.12 | 64.6 | 68.5 | |
| PretrainedBackbone=CLIP ViT-B/322025.12 | 48.2 | 53.4 |