Information Retrieval on HotpotQA
95.4RecallMoG
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MoG2026.05 | 95.4 | — | — | 81.3 | |
| Youtu-GraphRAG2026.05 | 92.3 | — | — | 83.5 | |
| HippoRAG22026.05 | 89.7 | — | — | 85.7 | |
| IMRNNsBackbone=MiniLM2026.01 | 66 | 63 | 56 | — | |
| IMRNNsBackbone=BGE2026.01 | 66 | 63 | 57 | — | |
| MiniLM2026.01 | 64 | 61 | 53 | — | |
| BGE2026.01 | 64 | 61 | 54 | — | |
| IMRNNsBackbone=e5-large-v22026.01 | 63 | 59 | 66 | — | |
| e5-large-v22026.01 | 61 | 54 | 65 | — | |
| DIMEPruning Ratio=20%2026.01 | 60 | 53 | 64 | — | |
| DIMEPruning Ratio=40%2026.01 | 58 | 51 | 62 | — | |
| DIMEPruning Ratio=60%2026.01 | 54 | 47 | 57 | — | |
| UnveilInput Modality=Visual-Textual2026.05 | 52.4 | — | 36.62 | — | |
| MiniCPMModel Type=Text-Based Models2026.05 | 50.93 | — | 36.11 | — | |
| BM25Model Type=Text-Based Models2026.05 | 48.93 | — | 33.64 | — | |
| VisRAGModel Type=Hybrid Models2026.05 | 48.67 | — | 34.81 | — | |
| DSEModel Type=Hybrid Models2026.05 | 48.2 | — | 34.25 | — | |
| UnveilInput Modality=Visual-Only2026.05 | 48.13 | — | 32.95 | — | |
| NV-EmbedModel Type=Text-Based Models2026.05 | 45.63 | — | 31.78 | — | |
| VisRAGModel Type=Visual-Based Models2026.05 | 45.6 | — | 31.06 | — | |
| DSEModel Type=Visual-Based Models2026.05 | 45.53 | — | 30.43 | — | |
| SearchAd2026.01 | 43 | 37 | 46 | — | |
| BGE-LargeModel Type=Text-Based Models2026.05 | 42.55 | — | 29.05 | — | |
| DIMEPruning Ratio=80%2026.01 | 42 | 37 | 45 | — | |
| ColPaliModel Type=Visual-Based Models2026.05 | 40.77 | — | 27.43 | — | |
| GTRModel Type=Text-Based Models2026.05 | 38.23 | — | 25.04 | — | |
| SigLIPModel Type=Visual-Based Models2026.05 | 31.63 | — | 18.89 | — | |
| HYPENCODERLayers=42026.01 | 26 | 29 | 27 | — | |
| HYPENCODERLayers=62026.01 | 26 | 30 | 28 | — | |
| HYPENCODERLayers=82026.01 | 26 | 30 | 29 | — | |
| HYPENCODERLayers=22026.01 | 25 | 29 | 27 | — | |
| BGEModel Type=Base model2023.11 | — | 71.81 | — | — | |
| Fine-tuned modelFine-tune on=HotpotQA2023.11 | — | 75.96 | — | — | |
| LM-Cocktail10Fine-tune on=HotpotQA, Merge components=Base model + 9 Specialists2023.11 | — | 74.67 | — | — | |
| LM-Cocktail2Fine-tune on=HotpotQA, Merge components=Base model + 1 Specialist2023.11 | — | 74.78 | — | — |