Zero-shot Image-Text Retrieval on Flickr
99.5R@5 TROriginal CLIP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Original CLIPBackbone=ViT-H/14, Biases=Gender2024.11 | 99.5 | 94.1 | — | |
| OursBackbone=ViT-H/14, Biases=Gender2024.11 | 99.5 | 93.8 | — | |
| Original CLIPBackbone=ViT-H/14, Biases=Age2024.11 | 99.5 | 94.1 | — | |
| OursBackbone=ViT-H/14, Biases=Age2024.11 | 99.5 | 94 | — | |
| Biased-promptsBackbone=ViT-H/14, Biases=Gender2024.11 | 99.2 | 93.3 | — | |
| CLIP-clipBackbone=ViT-H/14, Biases=Age2024.11 | 99.2 | 93.8 | — | |
| Biased-promptsBackbone=ViT-H/14, Biases=Age2024.11 | 98.8 | 93 | — | |
| CLIP-clipBackbone=ViT-H/14, Biases=Gender2024.11 | 98.4 | 93 | — | |
| OursBackbone=ViT-L/14, Biases=Age2024.11 | 97.3 | 86.8 | — | |
| Original CLIPBackbone=ViT-L/14, Biases=Gender2024.11 | 97.2 | 87.2 | — | |
| Original CLIPBackbone=ViT-L/14, Biases=Age2024.11 | 97.2 | 87.2 | — | |
| OursBackbone=ViT-L/14, Biases=Gender2024.11 | 97 | 87 | — | |
| CLIP-clipBackbone=ViT-L/14, Biases=Gender2024.11 | 96.6 | 85.2 | — | |
| Original CLIPBackbone=ViT-B/16, Biases=Gender2024.11 | 96.4 | 85.5 | — | |
| CLIP-clipBackbone=ViT-B/16, Biases=Gender2024.11 | 96.4 | 83 | — | |
| Original CLIPBackbone=ViT-B/16, Biases=Age2024.11 | 96.4 | 85.5 | — | |
| CLIP-clipBackbone=ViT-L/14, Biases=Age2024.11 | 96.2 | 86.4 | — | |
| CLIP-clipBackbone=ViT-B/16, Biases=Age2024.11 | 96.1 | 84.5 | — | |
| OursBackbone=ViT-B/16, Biases=Gender2024.11 | 95.8 | 84.6 | — | |
| OursBackbone=ViT-B/16, Biases=Age2024.11 | 95.6 | 84.6 | — | |
| Biased-promptsBackbone=ViT-L/14, Biases=Gender2024.11 | 95.5 | 87.2 | — | |
| Original CLIPBackbone=ViT-B/32, Biases=Gender2024.11 | 94.7 | 83.5 | — | |
| OursBackbone=ViT-B/32, Biases=Gender2024.11 | 94.7 | 82.9 | — | |
| Original CLIPBackbone=ViT-B/32, Biases=Age2024.11 | 94.7 | 83.5 | — | |
| Biased-promptsBackbone=ViT-L/14, Biases=Age2024.11 | 94.5 | 86.6 | — | |
| Biased-promptsBackbone=ViT-B/16, Biases=Gender2024.11 | 94.3 | 86.1 | — | |
| CLIP-clipBackbone=ViT-B/32, Biases=Age2024.11 | 94.2 | 83.2 | — | |
| Biased-promptsBackbone=ViT-B/16, Biases=Age2024.11 | 94.1 | 85.2 | — | |
| OursBackbone=ViT-B/32, Biases=Age2024.11 | 94.1 | 82.5 | — | |
| Biased-promptsBackbone=ViT-B/32, Biases=Gender2024.11 | 93.6 | 82.4 | — | |
| CLIP-clipBackbone=ViT-B/32, Biases=Gender2024.11 | 93 | 81 | — | |
| Biased-promptsBackbone=ViT-B/32, Biases=Age2024.11 | 92 | 82.2 | — | |
| Biased-promptsBackbone=ViT-B/162026.07 | — | — | 94.1 | |
| Biased-promptsBackbone=ViT-B/322026.07 | — | — | 91.9 | |
| Biased-promptsBackbone=ViT-B/162026.07 | — | — | 85.8 | |
| Biased-promptsBackbone=ViT-B/322026.07 | — | — | 83.34 | |
| CLIP-clipBackbone=ViT-B/162026.07 | — | — | 95.8 | |
| CLIP-clipBackbone=ViT-B/322026.07 | — | — | 93.6 | |
| CLIP-clipBackbone=ViT-B/162026.07 | — | — | 85.1 | |
| CLIP-clipBackbone=ViT-B/322026.07 | — | — | 81.74 | |
| Joint V-LBackbone=ViT-B/162026.07 | — | — | 96.5 | |
| Joint V-LBackbone=ViT-B/322026.07 | — | — | 94.6 | |
| Joint V-LBackbone=ViT-B/162026.07 | — | — | 83.8 | |
| Joint V-LBackbone=ViT-B/322026.07 | — | — | 82.4 | |
| Original CLIPBackbone=ViT-B/162026.07 | — | — | 96.4 | |
| Original CLIPBackbone=ViT-B/322026.07 | — | — | 94.7 | |
| Original CLIPBackbone=ViT-B/162026.07 | — | — | 85.5 | |
| Original CLIPBackbone=ViT-B/322026.07 | — | — | 83.5 | |
| RG-TTABackbone=ViT-B/162026.07 | — | — | 97.2 | |
| RG-TTABackbone=ViT-B/322026.07 | — | — | 97 | |
| RG-TTABackbone=ViT-B/162026.07 | — | — | 88.3 | |
| RG-TTABackbone=ViT-B/322026.07 | — | — | 86.5 |