Composed Retrieval on Bookset Multi-Image to Text
79.8R@1HOMIE (Qwen3-VL-8B)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| HOMIE (Qwen3-VL-8B)Model Category=MLLM-Based2025.02 | 79.8 | 96.7 | 98.3 | |
| HOMIE (Qwen3-VL-2B)Model Category=MLLM-Based2025.02 | 75.2 | 94.2 | 96.7 | |
| Patho-CLIP-LFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 44.3 | 77.5 | 88 | |
| MUSKFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 44.3 | 80.3 | 88.3 | |
| Patho-CLIPFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 43.3 | 75.8 | 85.7 | |
| Patho-CLIP-LFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 43.2 | 76.8 | 86.5 | |
| MUSKFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 43 | 72.7 | 83 | |
| Patho-CLIPFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 42.8 | 74.3 | 85.8 | |
| CONCHFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 41.8 | 71.7 | 79.7 | |
| CONCHFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 36.5 | 68.5 | 78.3 | |
| Pathgen-CLIPFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 23 | 52.8 | 64.3 | |
| Pathgen-CLIP-LFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 22.5 | 47.7 | 57.5 | |
| Pathgen-CLIP-LFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 22.2 | 56.7 | 71.8 | |
| Pathgen-CLIPFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 19.7 | 51.3 | 66.5 | |
| BiomedCLIPFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 19 | 40 | 57.2 | |
| BiomedCLIPFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 15.8 | 38.3 | 51.5 | |
| PathoR1-7BModel Category=MLLM-Based2025.02 | 14.7 | 33.2 | 45 | |
| QuiltNetFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 12.3 | 28.2 | 37.7 | |
| PathCLIPFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 11.2 | 36.8 | 48.5 | |
| QuiltNetFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 10.3 | 29.5 | 41 | |
| PathCLIPFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 9.8 | 27 | 42.2 | |
| PLIPFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 6.8 | 20.2 | 31 | |
| Lingshu-7BModel Category=MLLM-Based2025.02 | 6.8 | 22.8 | 33.2 | |
| LamRA-7BModel Category=MLLM-Based2025.02 | 5.8 | 15.7 | 24 | |
| PLIPFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 5 | 17.8 | 23.5 | |
| GME-7BModel Category=MLLM-Based2025.02 | 4.3 | 14.3 | 22.2 | |
| Qwen2.5-VL-7BModel Category=MLLM-Based2025.02 | 3.5 | 8.7 | 16 | |
| HuatuoGPT-V-7BModel Category=MLLM-Based2025.02 | 3.5 | 11.3 | 18.8 | |
| E5-V-7BModel Category=MLLM-Based2025.02 | 2.3 | 8.8 | 12.5 | |
| Qwen3-VL-8BModel Category=MLLM-Based2025.02 | 1.7 | 6 | 10.3 | |
| PubmedCLIPFusion Strategy=Add, Model Category=Dual-Encoder2025.02 | 0.8 | 1.8 | 3 | |
| Qwen3-VL-2BModel Category=MLLM-Based2025.02 | 0.5 | 2.5 | 4 | |
| PubmedCLIPFusion Strategy=RRF, Model Category=Dual-Encoder2025.02 | 0.4 | 1.7 | 3 |