Zero-shot Image-Text Retrieval on IIW
81.8Zero-shot Accuracy (IIW)CLIPS (ViT-L/14)
Evaluation Results
| Method | Links | |
|---|---|---|
| CLIPS (ViT-L/14)Backbone=ViT-L/142026.03 | 81.8 | |
| SigLIP2-ViT-B/16Backbone=ViT-B/162026.03 | 78.1 | |
| SigLIP ViT-L/16-res256Backbone=ViT-L/16, Input Resolution=256x2562026.03 | 77.6 | |
| C^2LIPBackbone=ViT-B/162026.03 | 76.4 | |
| CLIP-A (ViT-L/14)Backbone=ViT-L/142026.03 | 75.4 | |
| FLAVABackbone=ViT-B2026.03 | 72.9 | |
| BLIP-BBackbone=ViT-B2026.03 | 72.4 |