Image Classification on 11 Downstream Classification Datasets (test)
73ImageNet-1K AccSelf-TPT-v
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Self-TPT-vBackbone=ViT-B/16, Supervision=one-shot, Category=Test-Time Prompt Tuning, w/ adapters=false2025.02 | 73 | 94.7 | 68.8 | 49.4 | 51.9 | 27.6 | 71.8 | 85.4 | 91.3 | 68.2 | 69.5 | 68.3 | |
| AWT-AdapterBackbone=ViT-B/16, Supervision=one-shot, Category=Description-Based, w/ adapters=true2025.02 | 72.1 | 95.1 | 73.4 | 59.4 | 76.3 | 33.9 | 85.6 | 85.9 | 92.9 | 72.7 | 78.4 | 75.1 | |
| PromptAlignBackbone=ViT-B/16, Supervision=one-shot, Category=Test-Time Prompt Tuning, w/ adapters=false2025.02 | 71.4 | 94 | 68.5 | 47.2 | 47.9 | 24.8 | 72.4 | 86.7 | 90.8 | 67.5 | 69.5 | 67.3 | |
| ProAPO w/ APEBackbone=ViT-B/16, Supervision=one-shot, Category=Description-Based, w/ adapters=true, Adapter=APE2025.02 | 71.3 | 95.8 | 70.9 | 60.6 | 72.4 | 33.2 | 91.4 | 89.9 | 93.4 | 71 | 77.6 | 75.2 | |
| DiffTPTBackbone=ViT-B/16, Supervision=one-shot, Category=Test-Time Prompt Tuning, w/ adapters=false2025.02 | 70.3 | 92.5 | 67 | 47 | 43.1 | 25.6 | 70.1 | 87.2 | 88.2 | 65.7 | 68.2 | 65.9 | |
| ProAPOBackbone=ViT-B/16, Supervision=one-shot, Category=Description-Based, w/ adapters=false2025.02 | 69.9 | 95.2 | 67.7 | 55.8 | 65.3 | 28.3 | 82.7 | 89.5 | 92.7 | 68.9 | 73.1 | 71.7 | |
| ALIGNBackbone=ViT-B/16, Supervision=one-shot, Category=Vector-based Prompt Tuning, w/ adapters=false2025.02 | 69.8 | 94 | 68.3 | 54.1 | 53.2 | 29.6 | 81.3 | 85.3 | 91.4 | 69.1 | 74.4 | 70.1 | |
| UPTBackbone=ViT-B/16, Supervision=one-shot, Category=Vector-based Prompt Tuning, w/ adapters=false2025.02 | 69.6 | 93.7 | 67.6 | 45 | 66.5 | 28.4 | 75 | 84.2 | 82.9 | 68.8 | 72 | 68.5 | |
| MaPLeBackbone=ViT-B/16, Supervision=one-shot, Category=Vector-based Prompt Tuning, w/ adapters=false2025.02 | 69.6 | 92.6 | 66.6 | 52.1 | 71.8 | 26.7 | 83.3 | 80.5 | 89.1 | 64.8 | 71.8 | 69.9 | |
| CuPLBackbone=ViT-B/16, Supervision=one-shot, Category=Description-Based, w/ adapters=false2025.02 | 69.6 | 94.3 | 66.1 | 53.8 | 55.7 | 26.6 | 73.9 | 88.9 | 91.2 | 69 | 70.3 | 69 | |
| CoCoOpBackbone=ViT-B/16, Supervision=one-shot, Category=Vector-based Prompt Tuning, w/ adapters=false2025.02 | 69.4 | 93.8 | 67.2 | 48.5 | 55.3 | 12.7 | 72.1 | 85.7 | 91.3 | 68.3 | 70.3 | 66.8 | |
| ProAPO w/ AWT-textBackbone=ViT-B/16, Supervision=one-shot, Category=Description-Based, w/ adapters=false2025.02 | 69.4 | 95.3 | 67.8 | 54.3 | 67.1 | 27.4 | 82.1 | 89.6 | 93.2 | 68.5 | 73.1 | 71.6 | |
| TPTBackbone=ViT-B/16, Supervision=one-shot, Category=Test-Time Prompt Tuning, w/ adapters=false2025.02 | 69 | 94.2 | 66.9 | 47.8 | 42.4 | 24.8 | 69 | 84.7 | 87.8 | 65.5 | 68 | 65.5 | |
| AWT-textBackbone=ViT-B/16, Supervision=one-shot, Category=Description-Based, w/ adapters=false2025.02 | 68.9 | 95.2 | 66 | 52 | 52.6 | 26.1 | 74.5 | 89.4 | 91.2 | 68.4 | 69.8 | 68.6 | |
| PromptSRCBackbone=ViT-B/16, Supervision=one-shot, Category=Vector-based Prompt Tuning, w/ adapters=false2025.02 | 68.1 | 93.7 | 69.4 | 56.2 | 73.1 | 27.7 | 85.9 | 84.9 | 92 | 69.7 | 74.8 | 72.3 | |
| Vanilla CLIPBackbone=ViT-B/16, Supervision=one-shot, Category=Base Module, w/ adapters=false2025.02 | 66.9 | 93.2 | 65.5 | 44.3 | 51 | 24.4 | 70.6 | 88.4 | 89 | 62.5 | 67.7 | 65.8 |