Image-to-Text Retrieval on Flickr30k (Recall@1)
97.4Recall@1Gemini Embedding 2
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini Embedding 2Modality=V/A/I/T2026.05 | 97.4 | |
| Voyage-3.5-multimodalModality=V/I/T2026.05 | 94.5 | |
| multimodalembedding@001Modality=I/T2026.05 | 94 | |
| Amazon Nova MMEModality=V/A/I/T2026.05 | 87.5 | |
| FineLIPBackbone=ViT-B/16, Evaluation Protocol=zero-shot, Implementation Source=our implementation2026.07 | 52.23 | |
| HyFL-CLIPBackbone=ViT-B/16, Evaluation Protocol=zero-shot2026.07 | 50.6 | |
| HiMo-CLIPBackbone=ViT-B/16, Evaluation Protocol=zero-shot, Implementation Source=our implementation2026.07 | 50.27 | |
| Long-CLIPBackbone=ViT-B/16, Evaluation Protocol=zero-shot, Implementation Source=author-provided checkpoints2026.07 | 47.19 | |
| TULIPBackbone=ViT-B/16, Evaluation Protocol=zero-shot2026.07 | 46.1 |