Multi-modal to Text Retrieval on OVEN (held-out)
46.5R@5ELVA-7B*
Evaluation results
| Method | Links | |
|---|---|---|
| ELVA-7B*Evaluation Protocol=Zero-shot, Training Exposure=Trained on other tasks without exposure to the three held-out tasks2026.06 | 46.5 | |
| UniIR-CLIPSFEvaluation Protocol=Supervised2026.06 | 45.5 | |
| LamRA-Ret*Evaluation Protocol=Zero-shot, Training Exposure=Trained on other tasks without exposure to the three held-out tasks2026.06 | 44.7 | |
| UniIR-BLIPFFEvaluation Protocol=Supervised2026.06 | 41 | |
| Vision-R1Evaluation Protocol=Zero-shot2026.06 | 39.8 | |
| Qwen2.5-VLEvaluation Protocol=Zero-shot2026.06 | 38.5 |