Multi-hop Retrieval on Average (MuSiQue, 2Wiki, HotpotQA)
79.3R@2SAG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SAGMethod Type=Graph-based Method, Embedding Model=BGE-Large-EN-v1.5, LLM=Qwen3.6-Flash2026.06 | 79.3 | 88.2 | |
| GFM-RAGLLM=GPT-4o-mini2026.03 | 72.5 | 80.5 | |
| C2RAGLLM=GPT-4o-mini2026.03 | 69.3 | 82.1 | |
| HippoRAG 2Method Type=Graph-based Method, Embedding Model=BGE-Large-EN-v1.5, LLM=Qwen3.6-Flash2026.06 | 68.2 | 83.3 | |
| NV-Embed-v2 (7B)Method Type=Large Embedding Model, Source=Literature2026.06 | 68 | 80.2 | |
| NeuroPathLLM=GPT-4o-mini2026.03 | 67.1 | 81.3 | |
| GritLM-7BMethod Type=Large Embedding Model, Source=Literature2026.06 | 65.4 | 78.1 | |
| GTE-Qwen2-7B-InstructMethod Type=Large Embedding Model, Source=Literature2026.06 | 63.5 | 75.8 | |
| Iter-RetGenLLM=GPT-4o-mini2026.03 | 62.3 | 75.5 | |
| BGE-Large-EN-v1.5Method Type=Large Embedding Model2026.06 | 59.7 | 71.3 | |
| HippoRAGLLM=GPT-4o-mini2026.03 | 58.4 | 73.9 | |
| HippoRAG2LLM=GPT-4o-mini2026.03 | 56.4 | 71.2 | |
| IRCoTLLM=GPT-4o-mini2026.03 | 55.3 | 69.5 | |
| GTR (T5-based)Method Type=Simple Baseline, Source=Literature2026.06 | 52.3 | 63.6 | |
| G-retrieverLLM=GPT-4o-mini2026.03 | 49.7 | 57.6 | |
| BM25Method Type=Simple Baseline, Source=Literature2026.06 | 48.3 | 61.2 | |
| BM25LLM=GPT-4o-mini2026.03 | 46.6 | 58.2 | |
| ContrieverMethod Type=Simple Baseline, Source=Literature2026.06 | 46.6 | 59.8 | |
| LightRAGLLM=GPT-4o-mini2026.03 | 36.5 | 49.3 |