Multi-task Classification on 30 Image Classification Tasks Average (test)
93.1Average AccuracyFine-tuned
Evaluation results
| Method | Links | |
|---|---|---|
| Fine-tunedBackbone=ViT-B/32, Category=Task-specific baseline2026.06 | 93.1 | |
| IndividualBackbone=ViT-B/162025.12 | 93.05 | |
| DTS-DBackbone=ViT-B/162025.12 | 92.41 | |
| DTS-TBackbone=ViT-B/162025.12 | 92.37 | |
| DTS-D*Backbone=ViT-B/162025.12 | 92.08 | |
| DTS-T*Backbone=ViT-B/162025.12 | 91.99 | |
| T-SwitchBackbone=ViT-B/162025.12 | 91.96 | |
| TSV-C + SiMBackbone=ViT-B/32, Merging Type=Dynamic model merging, Base Framework=TSV-C2026.06 | 90 | |
| EMR-MERGINGBackbone=ViT-B/162025.12 | 89.54 | |
| EMR + SiMBackbone=ViT-B/32, Merging Type=Dynamic model merging, Base Framework=EMR2026.06 | 89 | |
| TM-TA + SiMBackbone=ViT-B/32, Merging Type=Dynamic model merging, Base Framework=TM-TA2026.06 | 87.1 | |
| TM-TIES + SiMBackbone=ViT-B/32, Merging Type=Dynamic model merging, Base Framework=TM-TIES2026.06 | 85.6 | |
| TSV-MBackbone=ViT-B/32, Merging Type=Static model merging2026.06 | 77.3 | |
| Iso-CBackbone=ViT-B/32, Merging Type=Static model merging2026.06 | 72.8 | |
| WEMoEBackbone=ViT-B/32, Merging Type=Dynamic model merging2026.06 | 67.1 | |
| TWIN-MergingBackbone=ViT-B/32, Merging Type=Dynamic model merging2026.06 | 60.1 | |
| TIES-MergingBackbone=ViT-B/32, Merging Type=Static model merging2026.06 | 59.6 | |
| Task ArithmeticBackbone=ViT-B/32, Merging Type=Static model merging2026.06 | 58 | |
| MoW-MergingBackbone=ViT-B/32, Merging Type=Dynamic model merging2026.06 | 56.4 | |
| Task-ArithmeticBackbone=ViT-B/162025.12 | 48.89 | |
| Weight-AveragingBackbone=ViT-B/162025.12 | 42.54 | |
| DaWinBackbone=ViT-B/32, Merging Type=Dynamic model merging2026.06 | 40.3 | |
| Ties-MergingBackbone=ViT-B/162025.12 | 37.53 |