Image-to-Text Retrieval on FlowVQA hard negatives (test)
47.527R@1GRAN - SaCLIP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GRAN - SaCLIPData preparation strategy=GRAN, Finetuning strategy=SaCLIP2026.02 | 47.527 | 47.527 | 47.527 | |
| GRAN - CLIPData preparation strategy=GRAN, Finetuning strategy=CLIP2026.02 | 46.667 | 46.667 | 46.667 | |
| PSEUDO - SaCLIPData preparation strategy=PSEUDO, Finetuning strategy=SaCLIP2026.02 | 38.817 | 38.817 | 38.817 | |
| PSEUDO - CLIPData preparation strategy=PSEUDO, Finetuning strategy=CLIP2026.02 | 35.269 | 35.269 | 35.269 | |
| CROP - CLIPData preparation strategy=CROP, Finetuning strategy=CLIP2026.02 | 20.887 | 20.887 | 20.887 | |
| w/o FTData preparation strategy=w/o FT2026.02 | 16.613 | 16.613 | 16.613 |