Visual Information Retrieval on MVRB
69.63SRUniSE-MLLM
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| UniSE-MLLMBackbone=Qwen2-VL, #params=2.21B2025.02 | 69.63 | 54.49 | 43.2 | 48.26 | 55.72 | |
| ColPaliBackbone=Paligemma, #params=2.92B2025.02 | 61.73 | 35 | 35.32 | 31.04 | 43.64 | |
| GMEBackbone=Qwen2-VL, #params=2.21B2025.02 | 61.62 | 37.68 | 37.78 | 47.98 | 48.14 | |
| DSEBackbone=Phi-3-V, #params=4.15B2025.02 | 61.54 | 37.78 | 39.24 | 31.51 | 45.21 | |
| BGEBackbone=BERT-Base, #params=109M2025.02 | 45.67 | 36.73 | 36.29 | 47.91 | 41.99 | |
| BM252025.02 | 40.71 | 28.51 | 38.46 | 6.23 | 30.81 | |
| E5-MistralBackbone=Mistral-7B, #params=7.11B2025.02 | 38.68 | 51.44 | 43 | 45.08 | 45.51 | |
| SigLIPBackbone=SOVIT-400m, #params=878M2025.02 | 38.33 | 34.48 | 19.6 | 40.64 | 33.34 | |
| UniSE-CLIPBackbone=CLIP-Large, #params=428M2025.02 | 35.95 | 43.38 | 28.13 | 40.62 | 36.41 | |
| E5-VBackbone=LLaVA-1.6, #params=8.35B2025.02 | 34.11 | 26.59 | 5.23 | 32.85 | 25.13 | |
| MM-EmbedBackbone=LLaVA-1.6, #params=7.57B2025.02 | 25.86 | 40.93 | 42.83 | 32.67 | 34.48 | |
| DPRBackbone=BERT-Base, #params=109M2025.02 | 24.98 | 18.73 | 27.35 | 22.08 | 23.74 | |
| CLIPBackbone=CLIP-Large, #params=428M2025.02 | 18.89 | 25.39 | 23.9 | 30.4 | 23.75 | |
| VLM2VecBackbone=Phi-3.5-V, #params=4.15B2025.02 | 15.93 | 48.05 | 49.42 | 23.24 | 32.19 | |
| Uni-IRBackbone=CLIP-Large, #params=428M2025.02 | 12.35 | 35.92 | 29.68 | 20.06 | 19.63 | |
| VISTABackbone=BERT-Base, #params=196M2025.02 | 5.21 | 11.29 | 25.78 | 16.61 | 13.85 |