Text-to-Image Retrieval on RCLIP 5K V3 (test)
47.2Recall@1ReasonSigLIP (Stage 2)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ReasonSigLIP (Stage 2)Model Scale=ViT Giant Opt (1B), Resolution=384/16, Training Data Scale=+16M, Base Model=SigLIP22026.06 | 47.2 | 64.8 | 70.2 | |
| ReasonSigLIP (Stage 1)Model Scale=ViT Giant Opt (1B), Resolution=384/16, Training Data Scale=+42M, Base Model=SigLIP22026.06 | 43.6 | 61.1 | 66.6 | |
| SigLIP2Model Scale=ViT Giant Opt (1B), Resolution=384/16, Training Data Scale=10.0B2026.06 | 40 | 57 | 62.5 |