Image Classification on ImageNet and 10 Target Datasets (1-shot)
73.6ImageNet AccuracyCLIP-LoRA
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CLIP-LoRAShots=16, Backbone=ViT-B/162026.02 | 73.6 | 96.4 | 72 | 92.1 | 54.7 | 84.2 | 98 | 92.4 | 86.3 | 76.1 | 86.7 | 83 | |
| TIP-Adapter-FShots=16, Backbone=ViT-B/162026.02 | 73.4 | 95.7 | 70.8 | 85.9 | 44.6 | 86.8 | 96.2 | 92.6 | 82.3 | 76 | 83.9 | 80.7 | |
| MMAShots=16, Backbone=ViT-B/162026.02 | 73.2 | 95.8 | 72.7 | 85 | 44.7 | 87 | 96.8 | 93.9 | 80.2 | 76.6 | 85 | 81 | |
| TaskResShots=16, Backbone=ViT-B/162026.02 | 73 | 95.8 | 71.5 | 82.7 | 44.9 | 86.9 | 97.5 | 92.4 | 83.5 | 76.1 | 84 | 80.8 | |
| LP++Shots=16, Backbone=ViT-B/162026.02 | 73 | 95.8 | 71.9 | 85.5 | 42.1 | 87.2 | 96.3 | 92.6 | 80.8 | 76 | 83.9 | 80.5 | |
| TaskResShots=8, Backbone=ViT-B/162026.02 | 72.3 | 95.3 | 66.7 | 77.5 | 40.3 | 86.4 | 96 | 92 | 79.6 | 74.6 | 81.6 | 78.4 | |
| CLIP-LoRAShots=8, Backbone=ViT-B/162026.02 | 72.3 | 95.6 | 67.5 | 89.7 | 45.7 | 83.1 | 96.3 | 91.7 | 82.1 | 74.7 | 84.1 | 80.3 | |
| ProGradShots=16, Backbone=ViT-B/162026.02 | 72.1 | 95.9 | 68.8 | 83.6 | 43 | 85.8 | 96.6 | 92.8 | 82.9 | 75.1 | 82.7 | 79.9 | |
| LP++Shots=8, Backbone=ViT-B/162026.02 | 72.1 | 95.5 | 67.7 | 78.2 | 39 | 86.8 | 95.2 | 91.8 | 76.4 | 75.1 | 81.9 | 78.2 | |
| CoOpShots=16, Backbone=ViT-B/16, ctx=162026.02 | 71.9 | 95.8 | 69.7 | 85 | 43.2 | 84.2 | 96.8 | 92 | 82.9 | 74.9 | 83.1 | 80 | |
| MaPLeShots=16, Backbone=ViT-B/162026.02 | 71.9 | 95.4 | 68.4 | 87.5 | 36.8 | 87.4 | 94.2 | 93.2 | 74.3 | 74.5 | 81.4 | 78.6 | |
| MMLoPShots=16, Backbone=ViT-B/162026.02 | 71.9 | 95.7 | 72.6 | 92.2 | 45.5 | 87.4 | 97.2 | 93.9 | 80.4 | 75.8 | 84.3 | 81.5 | |
| MMAShots=8, Backbone=ViT-B/162026.02 | 71.9 | 95.6 | 66.9 | 69.7 | 38.9 | 86.4 | 94.6 | 92.9 | 76.8 | 74.7 | 82.9 | 77.4 | |
| TIP-Adapter-FShots=8, Backbone=ViT-B/162026.02 | 71.7 | 95.2 | 66.7 | 81.3 | 39.5 | 86.9 | 94.3 | 91.8 | 78.3 | 73.5 | 82 | 78.3 | |
| MMLoPShots=8, Backbone=ViT-B/162026.02 | 71.5 | 95.3 | 68.9 | 88.2 | 40.7 | 87 | 95.8 | 93.3 | 77.8 | 74.7 | 82.7 | 79.6 | |
| CLIP-LoRAShots=4, Backbone=ViT-B/162026.02 | 71.4 | 95.2 | 63.8 | 84.9 | 37.9 | 82.7 | 93.7 | 91 | 77.4 | 72.8 | 81.1 | 77.4 | |
| MaPLeShots=8, Backbone=ViT-B/162026.02 | 71.3 | 95.1 | 63 | 82.8 | 33.8 | 87.2 | 90.5 | 93.1 | 71.3 | 73.2 | 79.5 | 76.4 | |
| ProGradShots=8, Backbone=ViT-B/162026.02 | 71.3 | 94.8 | 63.9 | 77.8 | 37.7 | 86.1 | 95 | 92.2 | 78.7 | 73 | 80.5 | 77.4 | |
| CoCoOpShots=16, Backbone=ViT-B/162026.02 | 71.1 | 95.1 | 63.7 | 73.6 | 33.3 | 87.4 | 89.1 | 93.4 | 72.3 | 72.6 | 77.2 | 75.4 | |
| TaskResShots=4, Backbone=ViT-B/162026.02 | 71 | 95 | 60.1 | 74.2 | 33.4 | 86 | 85 | 91.9 | 76 | 72.7 | 76.2 | 74.7 | |
| CoCoOpShots=8, Backbone=ViT-B/162026.02 | 70.8 | 94.9 | 60.1 | 69.1 | 32.4 | 87 | 86.3 | 93.3 | 70.4 | 71.5 | 75.9 | 73.8 | |
| LP++Shots=4, Backbone=ViT-B/162026.02 | 70.8 | 95.1 | 62.4 | 73.6 | 34 | 85.9 | 93 | 90.9 | 74 | 73.2 | 79.2 | 75.6 | |
| MMLoPShots=4, Backbone=ViT-B/162026.02 | 70.8 | 95.2 | 64.7 | 84.5 | 36.1 | 86.1 | 93.3 | 93.2 | 74.9 | 73.1 | 80.2 | 77.5 | |
| TIP-Adapter-FShots=4, Backbone=ViT-B/162026.02 | 70.7 | 94.8 | 59.8 | 76.8 | 35.7 | 86.5 | 92.1 | 91.9 | 74.1 | 70.8 | 78.1 | 75.6 | |
| CoOpShots=8, Backbone=ViT-B/16, ctx=162026.02 | 70.6 | 94.5 | 64.8 | 77.1 | 38.5 | 82.7 | 94.9 | 91.3 | 79 | 71.9 | 80 | 76.8 | |
| CoCoOpShots=4, Backbone=ViT-B/162026.02 | 70.6 | 94.8 | 55.7 | 61.7 | 30.6 | 86.3 | 81.5 | 92.7 | 69.5 | 70.4 | 75.3 | 71.7 | |
| MaPLeShots=4, Backbone=ViT-B/162026.02 | 70.6 | 95 | 59 | 69.9 | 30.1 | 86.7 | 84.9 | 93.3 | 70.1 | 71.4 | 77.1 | 73.5 | |
| MMAShots=4, Backbone=ViT-B/162026.02 | 70.5 | 94.5 | 60.1 | 42.4 | 35 | 86 | 91.3 | 92.9 | 73.3 | 72.9 | 79 | 72.5 | |
| KgCoOpShots=16, Backbone=ViT-B/162026.02 | 70.4 | 95.2 | 68.7 | 76.2 | 36.5 | 87.2 | 93.4 | 93.2 | 74.8 | 73.3 | 81.7 | 77.3 | |
| KgCoOpShots=8, Backbone=ViT-B/162026.02 | 70.2 | 95.1 | 65.6 | 73.9 | 34.8 | 87 | 91.5 | 93 | 72.8 | 72.6 | 80 | 76 | |
| ProGradShots=4, Backbone=ViT-B/162026.02 | 70.2 | 94.4 | 59.7 | 69.6 | 34.1 | 85.4 | 91.1 | 92.1 | 75 | 71.7 | 77.9 | 74.7 | |
| KgCoOpShots=4, Backbone=ViT-B/162026.02 | 69.9 | 95 | 58.7 | 71.8 | 32.2 | 86.9 | 87 | 92.6 | 69.5 | 71.5 | 77.6 | 73.9 | |
| CLIP-AdapterShots=16, Backbone=ViT-B/162026.02 | 69.8 | 94.9 | 59.4 | 71.4 | 34.2 | 87.1 | 92.9 | 92.3 | 74 | 74.2 | 80.2 | 75.5 | |
| CLIP-AdapterShots=8, Backbone=ViT-B/162026.02 | 69.1 | 94.5 | 50.5 | 61.6 | 30.5 | 86.9 | 83.3 | 91.9 | 70.7 | 71.7 | 76.2 | 71.5 | |
| CoOpShots=4, Backbone=ViT-B/16, ctx=162026.02 | 68.8 | 94.5 | 59.5 | 69.7 | 30.9 | 84.5 | 92.2 | 92.5 | 74.4 | 69.7 | 77.6 | 74 | |
| CLIP-AdapterShots=4, Backbone=ViT-B/162026.02 | 68.6 | 94 | 46.1 | 51.2 | 27.9 | 86.5 | 73.1 | 90.8 | 67.5 | 68 | 70.6 | 67.7 | |
| Zero-ShotShots=16, Backbone=ViT-B/162026.02 | 66.7 | 92.9 | 43.6 | 47.5 | 24.7 | 86.1 | 71.4 | 89.1 | 65.3 | 62.6 | 66.7 | 65.1 | |
| Zero-ShotShots=8, Backbone=ViT-B/162026.02 | 66.7 | 92.9 | 43.6 | 47.5 | 24.7 | 86.1 | 71.4 | 89.1 | 65.3 | 62.6 | 66.7 | 65.1 | |
| Zero-ShotShots=4, Backbone=ViT-B/162026.02 | 66.7 | 92.9 | 43.6 | 47.5 | 24.7 | 86.1 | 71.4 | 89.1 | 65.3 | 62.6 | 66.7 | 65.1 | |
| ReHARKShots=12026.03 | 61.88 | 90.13 | 49.23 | 69.19 | 21.45 | 77.55 | 80.82 | 86.34 | 59.18 | 63.53 | 64.83 | 65.83 | |
| GDAShots=12026.03 | 60.68 | 87.29 | 46.26 | 58.3 | 17.78 | 77.42 | 72.08 | 85.49 | 56.78 | 59.93 | 62.65 | 62.24 | |
| ProKeRShots=12026.03 | 60.6 | 88.17 | 47.99 | 59.75 | 20.65 | 77.4 | 78.85 | 86.44 | 56.79 | 59.66 | 65.13 | 63.77 | |
| Tip-AdapterShots=12026.03 | 60.58 | 88.09 | 45.9 | 56.76 | 19.06 | 77.54 | 75.06 | 86.02 | 57.11 | 60.85 | 64.4 | 62.85 | |
| Zero-Shot CLIPShots=02026.03 | 60.35 | 85.68 | 42.91 | 36.27 | 17.01 | 77.37 | 66.02 | 85.72 | 55.75 | 58.82 | 61.78 | 58.88 |