Image-to-image retrieval on 13-Dataset Retrieval Suite (test)
93.1Caltech Retrieval ScoreIsoCLIP
Evaluation results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| IsoCLIPIntra-modal=true, Backbone=SigLIP2 B/16, Pre-training dataset=WebLI2026.03 | 93.1 | 41.4 | 54.2 | 77.9 | 74.5 | 64 | 87.2 | 40.9 | 43.8 | 53.2 | 63.6 | 46.9 | 61.8 | 61.7 | |
| IsoCLIPIntra-modal=true, Backbone=EVA-02 B/16, Pre-training dataset=Merged-2B2026.03 | 89.7 | 58.3 | 53 | 80.4 | 55.2 | 62.8 | 92.7 | 25.7 | 40 | 62.3 | 57.5 | 42.2 | 57.9 | 59.8 | |
| Image-ImageIntra-modal=true, Backbone=SigLIP2 B/16, Pre-training dataset=WebLI2026.03 | 89.2 | 38.1 | 53.2 | 76.5 | 70.8 | 56.6 | 89.3 | 41.8 | 39 | 50.8 | 59.2 | 43.6 | 59.2 | 59 | |
| Image-ImageIntra-modal=true, Backbone=EVA-02 B/16, Pre-training dataset=Merged-2B2026.03 | 86.9 | 55.9 | 52.1 | 78.3 | 49.4 | 55.4 | 91.5 | 24.9 | 35.8 | 61.2 | 55.5 | 41.1 | 57 | 57.3 | |
| IsoCLIPIntra-modal=true, Backbone=ViT-B/32-open, Pre-training dataset=DataComp2026.03 | 83.4 | 34.2 | 56.8 | 75.8 | 49.9 | 47.8 | 78.2 | 19.8 | 37.6 | 57.7 | 41.5 | 36.6 | 46.3 | 51.2 | |
| OTI (I→T)Intra-modal=false, Backbone=ViT-B/32-open, Pre-training dataset=DataComp2026.03 | 83.3 | 34.3 | 54.4 | 75.8 | 50.5 | 50.5 | 78 | 20.1 | 40.9 | 54.5 | 42.9 | 37.8 | 48.2 | 51.6 | |
| Image-ImageIntra-modal=true, Backbone=ViT-B/32-open, Pre-training dataset=DataComp2026.03 | 82.3 | 32.1 | 50.8 | 74.7 | 46.7 | 44.1 | 77 | 19.6 | 36.9 | 56.4 | 39.6 | 36.2 | 45.7 | 49.4 |