Visual Document Retrieval on VisDoc
89.4VDRv1GME
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GMEBackbone=Qwen2VL, Model Size=8.3B2026.02 | 89.4 | 55.6 | 85 | 44.4 | 75.2 | |
| Magic-MM-EmbeddingBackbone=InternVL3-VTC, Model Size=8.1B, Retrieval Strategy=Embedder + Reranker (E+R)2026.02 | 86.9 | 60.4 | 89.2 | 43.1 | 75.8 | |
| GMEBackbone=Qwen2VL, Model Size=2.2B2026.02 | 86.1 | 54 | 82.5 | 43.1 | 72.7 | |
| Magic-MM-EmbeddingBackbone=InternVL3-VTC, Model Size=8.1B, Retrieval Strategy=Embedder only (E)2026.02 | 86.1 | 59.9 | 87.6 | 43.4 | 75 | |
| Magic-MM-EmbeddingBackbone=InternVL3-VTC, Model Size=1.9B, Retrieval Strategy=Embedder + Reranker (E+R)2026.02 | 84.4 | 56.1 | 87.4 | 41.8 | 73.3 | |
| ColPaliBackbone=Paligemma, Model Size=2.9B2026.02 | 83.6 | 52 | 81.1 | 43.1 | 71 | |
| Magic-MM-EmbeddingBackbone=InternVL3-VTC, Model Size=1.9B, Retrieval Strategy=Embedder only (E)2026.02 | 83.4 | 53.3 | 85.6 | 42.2 | 72.1 | |
| Ops-MM-embedding-v1Backbone=Qwen2VL, Model Size=8.3B2026.02 | 80.1 | 59.6 | 79.3 | 43.3 | 70.3 | |
| VLM2Vec-V2Backbone=Qwen2VL, Model Size=8.3B2026.02 | 78.8 | 52.6 | 82.7 | 42.1 | 69.3 | |
| VLM2Vec-V2Backbone=Qwen2VL, Model Size=2.2B2026.02 | 75.5 | 44.9 | 79.4 | 39.4 | 65.4 | |
| LamRA-Qwen2.5Backbone=Qwen2.5VL, Model Size=8.3B2026.02 | 56.3 | 33.3 | 58.2 | 40.1 | 50.2 | |
| LamRA-Qwen2Backbone=Qwen2VL, Model Size=8.3B2026.02 | 22 | 11.5 | 37.4 | 21 | 23.9 |