Multi-hop QA Retrieval on MuSiQue (test)
81.46R@5BRIDGERAG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BRIDGERAGCondition=C (+bridge cond.)2026.04 | 81.46 | — | 3.1 | |
| BRIDGERAGCondition=B (+2-way judge)2026.04 | 80.9 | — | 2.6 | |
| SAGMethod Type=Graph-based Method, Embedding Model=BGE-Large-EN-v1.5, LLM=Qwen3.6-Flash2026.06 | 80 | 64.1 | — | |
| BRIDGERAGCondition=A (SVO only)2026.04 | 79.4 | — | 1.1 | |
| PROPRAG2026.04 | 78.3 | — | — | |
| CoopRAGLLM=Gemma2-27B, Single-step retrieval=true2025.12 | 75.7 | 59.4 | — | |
| CoopRAGLLM=GPT-4o-mini, Single-step retrieval=true2025.12 | 75.7 | 59.6 | — | |
| CoopRAGBackbone LLM=Gemma2-27B2025.12 | 75.7 | 59.4 | — | |
| CoopRAGBackbone LLM=GPT-4o-mini2025.12 | 75.7 | 59.6 | — | |
| CoopRAGLLM=Gemma2-9B, Single-step retrieval=true2025.12 | 75.5 | 59.4 | — | |
| CoopRAGBackbone LLM=Gemma2-9B2025.12 | 75.5 | 59.4 | — | |
| CoopRAGLLM=Llama3.3-70B, Single-step retrieval=true2025.12 | 75.3 | 58.2 | — | |
| CoopRAGBackbone LLM=Llama3.3-70B2025.12 | 75.3 | 58.2 | — | |
| HippoRAG2LLM=Llama3.3-70B, Single-step retrieval=true2025.12 | 74.7 | 56.1 | — | |
| HippoRAG2Backbone LLM=Llama3.3-70B2025.12 | 74.7 | 56.1 | — | |
| HIPPORAG22026.04 | 74.7 | — | 3.6 | |
| HippoRAG2LLM=GPT-4o-mini, Single-step retrieval=true2025.12 | 74.2 | 53.5 | — | |
| HippoRAG2Backbone LLM=GPT-4o-mini2025.12 | 74.2 | 53.5 | — | |
| NV-Embed-v2Parameters=7B2025.12 | 69.7 | 52.7 | — | |
| NV-Embed-v2 (7B)Method Type=Large Embedding Model, Source=Literature2026.06 | 69.7 | 52.7 | — | |
| HopRAGLLM=GPT-4o-mini, Single-step retrieval=true2025.12 | 66.8 | 53.7 | — | |
| HopRAGBackbone LLM=GPT-4o-mini2025.12 | 66.8 | 53.7 | — | |
| GritLM-7B2025.12 | 65.9 | 49.7 | — | |
| GritLM-7BMethod Type=Large Embedding Model, Source=Literature2026.06 | 65.9 | 49.7 | — | |
| HippoRAG 2Method Type=Graph-based Method, Embedding Model=BGE-Large-EN-v1.5, LLM=Qwen3.6-Flash2026.06 | 65.1 | 49.5 | — | |
| SiReRAGLLM=GPT-4o-mini, Single-step retrieval=true2025.12 | 64.9 | 52.5 | — | |
| SiReRAGBackbone LLM=GPT-4o-mini2025.12 | 64.9 | 52.5 | — | |
| GTE-Qwen2-7B-Instruct2025.12 | 63.6 | 48.1 | — | |
| GTE-Qwen2-7B-InstructMethod Type=Large Embedding Model, Source=Literature2026.06 | 63.6 | 48.1 | — | |
| RAPTORBackbone LLM=GPT-4o-mini2025.12 | 61 | 49.1 | — | |
| RAPTORBackbone LLM=Llama3.3-70B2025.12 | 57.8 | 47 | — | |
| BGE-Large-EN-v1.5Method Type=Large Embedding Model2026.06 | 56.2 | 41.6 | — | |
| HippoRAGBackbone LLM=Llama3.3-70B2025.12 | 53.2 | 41.2 | — | |
| HippoRAGBackbone LLM=GPT-4o-mini2025.12 | 52.4 | 41.8 | — | |
| PropositionBase Retriever=ColBERTv22025.12 | 50.1 | 37.8 | — | |
| ColBERTv22025.12 | 49.2 | 37.9 | — | |
| GTRBackbone=T5-base2025.12 | 49.1 | 37.4 | — | |
| GTR (T5-based)Method Type=Simple Baseline, Source=Literature2026.06 | 49.1 | 37.4 | — | |
| Contriever2025.12 | 46.6 | 34.8 | — | |
| ContrieverMethod Type=Simple Baseline, Source=Literature2026.06 | 46.6 | 34.8 | — | |
| BM252025.12 | 43.5 | 32.4 | — | |
| BM25Method Type=Simple Baseline, Source=Literature2026.06 | 43.5 | 32.4 | — |