Image Classification on ImageNet (Base/Novel Accuracy Metrics)
82.63Base AccuracyMMA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MMAVisual Backbone=ViT-L/14, Shots (k)=82025.03 | 82.63 | 76.8 | 79.61 | |
| 2SFSVisual Backbone=ViT-L/14, Shots (k)=82025.03 | 82.43 | 76.46 | 79.34 | |
| CLIPVisual Backbone=ViT-L/14, Shots (k)=82025.03 | 79.18 | 74.04 | 76.53 | |
| PromptSRC2026.03 | 77.6 | 70.73 | 74.01 | |
| MetaPrompt2026.03 | 77.52 | 70.83 | 74.02 | |
| MMA2026.03 | 77.31 | 71 | 74.02 | |
| TCP2026.03 | 77.27 | 69.87 | 73.38 | |
| EvoPrompt2026.03 | 76.98 | 71.8 | 74.29 | |
| 2SFSBackbone=ViT-B/16, Shots=8, Evaluation Protocol=base-to-novel generalization2025.03 | 76.97 | 70.67 | 73.69 | |
| LLaMA-Adapter (Zero-init.)Backbone=ViT-B/162023.03 | 76.7 | 71 | 73.74 | |
| MaPLeBackbone=ViT-B/162023.03 | 76.66 | 70.54 | 73.47 | |
| MaPLe2026.03 | 76.66 | 70.54 | 73.47 | |
| CoOpBackbone=ViT-B/162023.03 | 76.47 | 67.88 | 71.92 | |
| CoOp2026.03 | 76.47 | 67.88 | 71.92 | |
| MMABackbone=ViT-B/16, Shots=8, Evaluation Protocol=base-to-novel generalization2025.03 | 76.43 | 70.07 | 73.11 | |
| CoCoOpBackbone=ViT-B/162023.03 | 75.98 | 70.43 | 73.1 | |
| CoCoOp2026.03 | 75.98 | 70.43 | 73.1 | |
| KgCoOp2026.03 | 75.83 | 69.96 | 72.78 | |
| ProVP2026.03 | 75.82 | 69.21 | 72.36 | |
| ProDA2026.03 | 75.4 | 70.23 | 72.72 | |
| CLIPBackbone=ViT-B/162023.03 | 72.43 | 68.14 | 70.22 | |
| CLIPBackbone=ViT-B/16, Shots=8, Evaluation Protocol=base-to-novel generalization2025.03 | 72.43 | 68.14 | 70.22 | |
| CLIP2026.03 | 72.43 | 68.14 | 70.22 |