Classification on Oxford Pets zero-shot
97.6Accuracy (Zero-Shot)DFN Zero-Shot
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DFN Zero-ShotBackbone=DFN, Protocol=10-way 0-shot2026.03 | 97.6 | — | |
| SigLIP Zero-ShotBackbone=SigLIP, Protocol=10-way 0-shot2026.03 | 97 | — | |
| Qwen2-VL HEC-TBackbone=DFN+LLM (Qwen2-VL), Protocol=10-way 0-shot2026.03 | 95.2 | — | |
| WCABackbone=ViT-B/16, Zero-shot=true2026.01 | 92.13 | — | |
| BiFTABackbone=ViT-B/16, Zero-shot=true2026.01 | 91.67 | — | |
| CuPLBackbone=ViT-B/16, Zero-shot=true2026.01 | 91.14 | — | |
| GRACEBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 90.4 | 54.6 | |
| Vanilla FTBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 90.2 | 0 | |
| Qwen2-VL STBackbone=DFN+LLM (Qwen2-VL), Protocol=10-way 0-shot2026.03 | 90.1 | — | |
| TPGMBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 89.8 | 0 | |
| FLYPBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 89.5 | 0 | |
| SPDBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 89.4 | 0 | |
| CLIP-EBackbone=ViT-B/16, Zero-shot=true2026.01 | 89.1 | — | |
| WISE-FTBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 88.6 | 0 | |
| CLIPBackbone=ViT-B/16, Zero-shot=true2026.01 | 88.14 | — | |
| CLIP-DBackbone=ViT-B/16, Zero-shot=true2026.01 | 87.52 | — | |
| WaffleBackbone=ViT-B/16, Zero-shot=true2026.01 | 86.51 | — | |
| CLIPBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 86.13 | 0 | |
| Qwen2-VL BaselineBackbone=DFN+LLM (Qwen2-VL), Protocol=10-way 0-shot2026.03 | 84.1 | — | |
| LLaVA-OV HEC-TBackbone=SigLIP+LLM (LLaVA-OV), Protocol=10-way 0-shot2026.03 | 83.8 | — | |
| FAREBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 82.1 | 28.6 | |
| LLaVA-OV BaselineBackbone=SigLIP+LLM (LLaVA-OV), Protocol=10-way 0-shot2026.03 | 79.6 | — | |
| PMG-AFTBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 75 | 50.2 | |
| LAATBackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 73.2 | 48.3 | |
| TeCoABackbone=ViT-B/16, Training=ImageNet, Evaluation=Zero-shot2026.03 | 70.31 | 45.7 | |
| LLaVA-OV STBackbone=SigLIP+LLM (LLaVA-OV), Protocol=10-way 0-shot2026.03 | 69.8 | — |