Image-to-text retrieval on IMDB
83.4R@1Cosine
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CosineFeature Extractor=CLIP2025.06 | 83.4 | 91.4 | 93.8 | |
| ManhattanFeature Extractor=CLIP2025.06 | 79.1 | 89.7 | 92.5 | |
| CosineFeature Extractor=BLIP2025.06 | 73.4 | 88.1 | 91.1 | |
| EuclideanFeature Extractor=BLIP2025.06 | 63.9 | 82.2 | 86.6 | |
| ManhattanFeature Extractor=BLIP2025.06 | 49.1 | 71 | 78.1 | |
| MLP (Contr.)Feature Extractor=CLIP2025.06 | 35.9 | 58 | 63.9 | |
| Transformer (Contr.)Feature Extractor=BLIP2025.06 | 35.4 | 61.2 | 69 | |
| EuclideanFeature Extractor=CLIP2025.06 | 31 | 46.8 | 52.4 | |
| MLP (Contr.)Feature Extractor=BLIP2025.06 | 19.4 | 39.3 | 45.7 | |
| Transformer (Contr.)Feature Extractor=CLIP2025.06 | 18.9 | 43 | 53.8 | |
| MLP (Contr.)Feature Extractor=ResNet/BERT2025.06 | 2 | 5.9 | 10.8 | |
| Transformer (Contr.)Feature Extractor=ConViT/RoBERTa2025.06 | 1.7 | 4.9 | 11.1 | |
| MLP (MSE)Feature Extractor=CLIP2025.06 | 1 | 3.9 | 7.4 | |
| Transformer (MSE)Feature Extractor=BLIP2025.06 | 0.5 | 2 | 3.2 | |
| Transformer (MSE)Feature Extractor=CLIP2025.06 | 0.2 | 4.2 | 7.6 | |
| MLP (MSE)Feature Extractor=BLIP2025.06 | 0.2 | 2 | 3.2 | |
| Chi-SquareFeature Extractor=CLIP2025.06 | 0.1 | 0.1 | 0.3 | |
| CosineFeature Extractor=Meta-Transf.2025.06 | 0.1 | 0.6 | 1.2 | |
| Chi-SquareFeature Extractor=BLIP2025.06 | 0 | 0.4 | 1.3 | |
| EuclideanFeature Extractor=Meta-Transf.2025.06 | 0 | 0.5 | 1.2 |