Text-to-Image Retrieval on MS-COCO Combined
75.2R@1BEIT-3
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| BEIT-3Config=COCO-Finetuned, Zero-Shot=false2026.02 | 75.2 | 94.4 | 98.4 | |
| Visual Model Checking + BEITZero-Shot=false2026.02 | 68.9 | 91.3 | 96.9 | |
| CLIPBackbone=ViT-H, QuickGELU=true, Zero-Shot=true2026.02 | 60.5 | 89.9 | 96.1 | |
| SigLIPParameters=400M, Zero-Shot=true2026.02 | 58.1 | 85.3 | 94.6 | |
| Visual Model Checking + ALIGNZero-Shot=true2026.02 | 55.6 | 87.2 | 93.4 | |
| ALIGNZero-Shot=true2026.02 | 48.1 | 82.7 | 93.8 | |
| Visual Model Checking (base)Zero-Shot=true2026.02 | 45 | 70 | 86.8 | |
| Visual Model Checking + CLIPZero-Shot=true2026.02 | 42.9 | 63.8 | 67.9 | |
| CLIPBackbone=ViT-B, Zero-Shot=true2026.02 | 39.5 | 71 | 88.4 |