Multi-task Image Classification on Vision Benchmark 20-task
94.7Average AccuracyFine-tuned
Evaluation results
| Method | Links | |
|---|---|---|
| Fine-tunedBackbone=ViT-L/142026.03 | 94.7 | |
| Fine-tunedBackbone=ViT-B/162026.03 | 93.2 | |
| Fine-tunedBackbone=ViT-B/322026.03 | 91.3 | |
| ACE-MergingBackbone=ViT-L/142026.03 | 89.5 | |
| CARTBackbone=ViT-L/142026.03 | 87.9 | |
| TSV-MBackbone=ViT-L/142026.03 | 87.7 | |
| ACE-MergingBackbone=ViT-B/162026.03 | 82.1 | |
| TSV-MBackbone=ViT-B/162026.03 | 80.6 | |
| CARTBackbone=ViT-B/162026.03 | 80.5 | |
| Consensus TABackbone=ViT-L/142026.03 | 79 | |
| ACE-MergingBackbone=ViT-B/322026.03 | 78.8 | |
| TSV-MBackbone=ViT-B/322026.03 | 77.1 | |
| CARTBackbone=ViT-B/322026.03 | 76.3 | |
| Task ArithmeticBackbone=ViT-L/142026.03 | 74 | |
| Weight AveragingBackbone=ViT-L/142026.03 | 71.6 | |
| Consensus TABackbone=ViT-B/162026.03 | 69.8 | |
| Task ArithmeticBackbone=ViT-B/162026.03 | 65.8 | |
| Consensus TABackbone=ViT-B/322026.03 | 65.4 | |
| Weight AveragingBackbone=ViT-B/162026.03 | 65.3 | |
| Zero-shotBackbone=ViT-L/142026.03 | 65.2 | |
| Weight AveragingBackbone=ViT-B/322026.03 | 61 | |
| Task ArithmeticBackbone=ViT-B/322026.03 | 60.5 | |
| Zero-shotBackbone=ViT-B/162026.03 | 59.7 | |
| Zero-shotBackbone=ViT-B/322026.03 | 56.1 |