Multi-task Scene Classification on SUN-397 8 fine-grained tasks
95Average AccuracyFine-tuned
Evaluation results
| Method | Links | |
|---|---|---|
| Fine-tunedBackbone=ViT-B/32, Note=Task-specific baseline2026.06 | 95 | |
| TM-TIES + SiMBackbone=ViT-B/32, Merging Strategy=Dynamic model merging2026.06 | 93.1 | |
| TSV-MBackbone=ViT-B/32, Merging Strategy=Static model merging2026.06 | 92.7 | |
| Iso-CBackbone=ViT-B/32, Merging Strategy=Static model merging2026.06 | 92.5 | |
| MoW-MergingBackbone=ViT-B/32, Merging Strategy=Dynamic model merging2026.06 | 92.2 | |
| TWIN-MergingBackbone=ViT-B/32, Merging Strategy=Dynamic model merging2026.06 | 92 | |
| TIES-MergingBackbone=ViT-B/32, Merging Strategy=Static model merging2026.06 | 90.8 | |
| DaWinBackbone=ViT-B/32, Merging Strategy=Dynamic model merging2026.06 | 90.7 | |
| Task ArithmeticBackbone=ViT-B/32, Merging Strategy=Static model merging2026.06 | 90.6 | |
| WEMoEBackbone=ViT-B/32, Merging Strategy=Dynamic model merging2026.06 | 90.1 |