Image-to-text retrieval on MS-COCO (R@1, R@5, R@10)
64.7R@1Cosine
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CosineFeature Extractor=BLIP2025.06 | 64.7 | 90.1 | 96 | |
| CosineFeature Extractor=CLIP2025.06 | 51 | 82.3 | 92.2 | |
| MLP (Contr.)Feature Extractor=CLIP2025.06 | 27.2 | 57.7 | 71.5 | |
| MLP (Contr.)Feature Extractor=ConViT/RoBERTa2025.06 | 4.9 | 19 | 30.1 | |
| Transformer (Contr.)Feature Extractor=ConvNeXt/XLNet2025.06 | 3.3 | 12.7 | 22.9 |