Text-to-Image Retrieval on VisualNews (VN) (test)
43.5Recall@5ELVA-7B (Ours)
Evaluation results
| Method | Links | |
|---|---|---|
| ELVA-7B (Ours)Supervision Type=Supervised, Model Architecture=MLLM2026.06 | 43.5 | |
| CLIP-LSupervision Type=Zero-shot2026.06 | 43.3 | |
| UniIR-CLIP_SFSupervision Type=Supervised, Model Architecture=Dual Encoder2026.06 | 42.6 | |
| Vision-R1-7BSupervision Type=Supervised, Model Architecture=MLLM2026.06 | 41.9 | |
| LamRA-Ret-7BSupervision Type=Supervised, Model Architecture=MLLM2026.06 | 41.6 | |
| MM-Embed-7BSupervision Type=Supervised, Model Architecture=MLLM2026.06 | 41 | |
| VLM-R1-7BSupervision Type=Supervised, Model Architecture=MLLM2026.06 | 40.5 | |
| Qwen2.5-VL-7BSupervision Type=Zero-shot2026.06 | 40.2 | |
| PUMA-3BSupervision Type=Supervised, Model Architecture=MLLM2026.06 | 35.7 | |
| ELVA-2B (Ours)Supervision Type=Supervised, Model Architecture=MLLM2026.06 | 35.6 | |
| LamRA-Ret-2BSupervision Type=Supervised, Model Architecture=MLLM2026.06 | 30.8 | |
| SigLIPSupervision Type=Zero-shot2026.06 | 30.1 | |
| UniIR-BLIP_FFSupervision Type=Supervised, Model Architecture=Dual Encoder2026.06 | 23.4 | |
| BLIP2Supervision Type=Zero-shot2026.06 | 16.7 | |
| BLIPSupervision Type=Zero-shot2026.06 | 16.4 | |
| Qwen2-VL-7BSupervision Type=Zero-shot2026.06 | 9.3 |