Image Classification on DomainNet (Quickdraw Source Transfer Accuracy)
69.68Accuracy (Q->C)LoA
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| LoAInitial Server Model Pre-training Domain=Quickdraw2025.12 | 69.68 | 14.91 | 39.15 | 56.38 | 75.25 | 51.07 | |
| FedSCAlInitial Server Model Pre-training Domain=Quickdraw2025.12 | 65.37 | 40.15 | 65.74 | 66.03 | 65.75 | 60.61 | |
| MVT + FTMLLM=MMICL, Backbone=ViT-L2025.12 | 62.5 | 63.8 | 62.4 | 80.1 | 54 | — | |
| MVT + FTMLLM=Otter, Backbone=ViT-L2025.12 | 62 | 63 | 61.3 | 79.7 | 53.4 | — | |
| MVTMLLM=MMICL, Backbone=ViT-L2025.12 | 61.3 | 62.1 | 60.4 | 78.7 | 53.4 | — | |
| CLIPMLLM=None, Backbone=ViT-L2025.12 | 61.1 | 62.1 | 60.2 | 78.4 | 51.1 | — | |
| FedMOONInitial Server Model Pre-training Domain=Quickdraw2025.12 | 60.34 | 30.44 | 52.9 | 66.35 | 56.41 | 53.29 | |
| DualAdaptInitial Server Model Pre-training Domain=Quickdraw2025.12 | 59.78 | 22.02 | 40.68 | 51.85 | 49.58 | 44.78 | |
| FedProxInitial Server Model Pre-training Domain=Quickdraw2025.12 | 59.74 | 35.76 | 51.33 | 57.34 | 60.86 | 53.01 | |
| MVTMLLM=Otter, Backbone=ViT-L2025.12 | 56.5 | 58.6 | 56.3 | 74.1 | 46.5 | — | |
| LADDInitial Server Model Pre-training Domain=Quickdraw2025.12 | 52.72 | 25.94 | 37.19 | 50.8 | 52.95 | 43.92 | |
| FedLoAInitial Server Model Pre-training Domain=Quickdraw2025.12 | 45.21 | 29.56 | 50.02 | 64.08 | 52.35 | 48.24 | |
| FedWCAInitial Server Model Pre-training Domain=Quickdraw2025.12 | 41.49 | 32.7 | 46.22 | 54.74 | 43.29 | 43.68 |