Multi-hop Document Retrieval on HotpotQA (test)
91.03Recall@KGritHopper
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GritHopperModel Parameters=7B2026.05 | 91.03 | — | — | — | |
| MDRModel Parameters=110M2026.05 | 87.57 | — | — | — | |
| IRCoTK=7.92025.03 | 77 | — | — | — | |
| ThinkGRModel Parameters=8B2026.05 | 76.09 | — | — | — | |
| IterRetGenK=6.02025.03 | 73.2 | — | — | — | |
| SelfAskK=28.62025.03 | 71.1 | — | — | — | |
| w/o ThoughtModel Parameters=8B2026.05 | 69.86 | — | — | — | |
| w/o RLModel Parameters=8B2026.05 | 67.66 | — | — | — | |
| Vanilla RAGK=8.02025.03 | 64.1 | — | — | — | |
| RT-RAGModel Parameters=8B2026.05 | 62.48 | — | — | — | |
| ITER-RETGENModel Parameters=70B2026.05 | 61.94 | — | — | — | |
| L-RAGK=8.02025.03 | 61.3 | — | — | — | |
| BGE-largeModel Parameters=326M2026.05 | 60.48 | — | — | — | |
| R3-RAGModel Parameters=8B2026.05 | 58.56 | — | — | — | |
| SEALModel Parameters=406M2026.05 | 56.15 | — | — | — | |
| IRCoTModel Parameters=70B2026.05 | 55.79 | — | — | — | |
| Auto-RAGModel Parameters=7B2026.05 | 54.52 | — | — | — | |
| Vanilla RAGK=4.02025.03 | 54.3 | — | — | — | |
| Contriever2026.05 | 50.35 | — | — | — | |
| BM252026.05 | 46.39 | — | — | — | |
| L-RAGK=4.02025.03 | 44.9 | — | — | — | |
| SelfaskModel Parameters=70B2026.05 | 44.4 | — | — | — | |
| w/o SFTModel Parameters=8B2026.05 | 35.76 | — | — | — | |
| HyDEK=6.02025.03 | 25 | — | — | — | |
| BGE-Large-EN-v1.5Method Type=Large Embedding Model2026.06 | — | 76 | 88.8 | — | |
| BM25Method Type=Simple Baseline, Source=Literature2026.06 | — | 57.3 | 74.8 | — | |
| BRIDGERAGCondition=A (SVO only)2026.04 | — | — | 97.2 | 0.2 | |
| BRIDGERAGCondition=C (+bridge cond.)2026.04 | — | — | 98.75 | 1.35 | |
| ContrieverMethod Type=Simple Baseline, Source=Literature2026.06 | — | 58.4 | 75.3 | — | |
| CoopRAGLLM=Gemma2-9B, Single-step retrieval=true2025.12 | — | 87.9 | 95.6 | — | |
| CoopRAGLLM=Gemma2-27B, Single-step retrieval=true2025.12 | — | 88.3 | 96.6 | — | |
| CoopRAGLLM=Llama3.3-70B, Single-step retrieval=true2025.12 | — | 86.9 | 96.6 | — | |
| CoopRAGLLM=GPT-4o-mini, Single-step retrieval=true2025.12 | — | 88.8 | 96.8 | — | |
| GritLM-7BMethod Type=Large Embedding Model, Source=Literature2026.06 | — | 79.2 | 92.4 | — | |
| GTE-Qwen2-7B-InstructMethod Type=Large Embedding Model, Source=Literature2026.06 | — | 75.8 | 89.1 | — | |
| GTR (T5-based)Method Type=Simple Baseline, Source=Literature2026.06 | — | 59.3 | 73.9 | — | |
| HippoRAG 2Method Type=Graph-based Method, Embedding Model=BGE-Large-EN-v1.5, LLM=Qwen3.6-Flash2026.06 | — | 78.4 | 94.4 | — | |
| HippoRAG2LLM=Llama3.3-70B, Single-step retrieval=true2025.12 | — | 83.5 | 96.3 | — | |
| HippoRAG2LLM=GPT-4o-mini, Single-step retrieval=true2025.12 | — | 80.5 | 95.7 | — | |
| HIPPORAG22026.04 | — | — | 96.3 | 1.1 | |
| HopRAGLLM=GPT-4o-mini, Single-step retrieval=true2025.12 | — | 81.1 | 96 | — | |
| NV-Embed-v2 (7B)Method Type=Large Embedding Model, Source=Literature2026.06 | — | 84.1 | 94.5 | — | |
| PROPRAG2026.04 | — | — | 97.4 | — | |
| SAGMethod Type=Graph-based Method, Embedding Model=BGE-Large-EN-v1.5, LLM=Qwen3.6-Flash2026.06 | — | 91.6 | 96.5 | — | |
| SiReRAGLLM=GPT-4o-mini, Single-step retrieval=true2025.12 | — | 80 | 94.8 | — |