Multi-hop QA Retrieval on MuSiQue
54.8R@2G-reasoner
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| G-reasoner2025.09 | 54.8 | 74.9 | |
| HippoRAG 22025.09 | 53.5 | 74.2 | |
| GFM-RAGLLM=GPT-4o-mini2026.03 | 49.8 | 57.6 | |
| C2RAGLLM=GPT-4o-mini2026.03 | 48.5 | 65 | |
| NeuroPathLLM=GPT-4o-mini2026.03 | 47.3 | 62.3 | |
| Qwen3-EmbParameters=8B2025.09 | 46.8 | 62.1 | |
| CoRe-RBackbone Model=Granite-3.2 8B2025.10 | 46.8 | 58.4 | |
| CoRe-RBackbone Model=Llama-3.1 8B2025.10 | 46.7 | 58.9 | |
| CoRe-RBackbone Model=Phi-42025.10 | 46.3 | 58.5 | |
| QR-RBackbone Model=Granite-3.2 8B2025.10 | 46.1 | 57.2 | |
| QR-RBackbone Model=Llama-3.1 8B2025.10 | 45.9 | 58.3 | |
| QR-RBackbone Model=Phi-42025.10 | 45.9 | 57.3 | |
| Iter-RetGenLLM=GPT-4o-mini2026.03 | 45.8 | 60 | |
| ICRBackbone Model=Granite-3.2 8B2025.10 | 45.7 | 56.4 | |
| ICRBackbone Model=Llama-3.1 8B2025.10 | 44.5 | 57 | |
| CoRe-RBackbone Model=Mistral 7B2025.10 | 43.9 | 56.2 | |
| GFM-RAG2025.09 | 43.5 | 57.6 | |
| HippoRAGLLM=GPT-4o-mini2026.03 | 42.3 | 53.5 | |
| KAG2025.09 | 42.2 | 62.4 | |
| QR-RBackbone Model=Mistral 7B2025.10 | 41.9 | 54.4 | |
| IRCoTLLM=GPT-4o-mini2026.03 | 41.7 | 56.4 | |
| HippoRAG2025.09 | 41.2 | 53.2 | |
| HippoRAG2LLM=GPT-4o-mini2026.03 | 41.2 | 56 | |
| ICRBackbone Model=Phi-42025.10 | 40.9 | 54.8 | |
| SubgraphRAG2025.09 | 40.6 | 48.1 | |
| ICRBackbone Model=Mistral 7B2025.10 | 40 | 53.9 | |
| ColBERTv22025.09 | 37.9 | 49.2 | |
| Retriever BaselineBackbone Model=N/A2025.10 | 37.9 | 49.2 | |
| RAPTOR2025.09 | 35.7 | 45.3 | |
| G-retriever2025.09 | 35.6 | 43.5 | |
| GraphRAGVariant=MS2025.09 | 35.4 | 49.3 | |
| G-retrieverLLM=GPT-4o-mini2026.03 | 35 | 44.1 | |
| BM25LLM=GPT-4o-mini2026.03 | 32.6 | 40.7 | |
| BM252025.09 | 32.3 | 41.2 | |
| LightRAGLLM=GPT-4o-mini2026.03 | 25.4 | 34 | |
| LightRAG2025.09 | 24.8 | 34.7 |