Image Classification on 20 Image Classification Tasks
59.9Accuracy (4 tasks merged)DOGE TA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DOGE TABackbone=ViT-B/32, Method Category=Data-Free Methods2025.08 | 59.9 | 62.4 | 63 | 62.7 | 64 | |
| TIES-MergingBackbone=ViT-B/32, Method Category=Data-Free Methods2025.08 | 59.7 | 59.9 | 60.4 | 59.4 | 60.8 | |
| Iso-CBackbone=ViT-B/32, Method Category=Data-Free Methods2025.08 | 59.2 | 60.7 | 58.7 | 53.4 | 44.4 | |
| RegMean++Backbone=ViT-B/32, Method Category=Training-Free Methods2025.08 | 59 | 64.8 | 65.8 | 66.9 | 68.9 | |
| Model SoupsBackbone=ViT-B/32, Method Category=Data-Free Methods2025.08 | 58.9 | 59.3 | 60.1 | 59.1 | 60 | |
| Fisher MergingBackbone=ViT-B/32, Method Category=Training-Free Methods2025.08 | 58.9 | 59.5 | 59.7 | 58.9 | 60.4 | |
| Iso-CTSBackbone=ViT-B/32, Method Category=Data-Free Methods2025.08 | 58.7 | 57.8 | 47.9 | 32.8 | 28.3 | |
| RegMeanBackbone=ViT-B/32, Method Category=Training-Free Methods2025.08 | 58.4 | 64.1 | 65.4 | 66 | 67.5 | |
| Task ArithmeticBackbone=ViT-B/32, Method Category=Data-Free Methods2025.08 | 58.2 | 57.5 | 58.8 | 58.3 | 58.1 | |
| Layer-wise AdaMergingBackbone=ViT-B/32, Method Category=Test-Time Adaption Methods2025.08 | 57.7 | 60.1 | 61.7 | 61.4 | 61.1 | |
| DOGE AMBackbone=ViT-B/32, Method Category=Test-Time Adaption Methods2025.08 | 57 | 61.2 | 62.4 | 63.9 | 63.1 | |
| TSV-MBackbone=ViT-B/32, Method Category=Data-Free Methods2025.08 | 56.9 | 61.4 | 62 | 64 | 63.5 |