Multimodal Retrieval on Multimodal E-Commerce dataset 134K (test)
8.48H@1Qwen2.5-VL-3B-Instruct
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Qwen2.5-VL-3B-InstructFramework=MLLM, Resolution=224x2242026.05 | 8.48 | 22.07 | 30.16 | 39.89 | 54.03 | 64.68 | 450.16 | |
| TGQ-FormerFramework=MLRM, Base Model=Chinese-CLIP-ViT-B/16 + Qwen3-Embedding-0.6B, Resolution=224x2242026.05 | 8.44 | 22.18 | 31.62 | 42.93 | 58.58 | 69.13 | 35.43 | |
| Qwen3-VL-4B-InstructFramework=MLLM, Resolution=224x2242026.05 | 8.1 | 21.44 | 29.9 | 39.98 | 54.63 | 65.68 | 592.23 | |
| Qwen2-VL-2B-InstructFramework=MLLM, Resolution=224x2242026.05 | 7.18 | 18.86 | 26.33 | 35.93 | 50.54 | 61.59 | 298.91 | |
| InternVL3-2B-InstructFramework=MLLM, Resolution=448x4482026.05 | 7.1 | 18.67 | 26.4 | 35.86 | 50.56 | 61.83 | 724.71 |