Multi-hop QA on 2WikiMultihopQA
56.4Exact Match (EM)SDGA-Phase
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SDGA-PhaseFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 56.4 | — | — | — | |
| SDGA-AutoFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 55.8 | — | — | — | |
| HippoRAGBackbone=Qwen2.5-72B-Instruct2026.03 | 55.5 | 63.6 | — | — | |
| GRPOmemory=none2026.02 | 53.8 | 62.35 | 0.93 | 0.32 | |
| GraphRAG-RouterSystem=Our Method2026.03 | 52.3 | — | — | — | |
| TaSR-RAGBackbone=Qwen2.5-72B-Instruct2026.03 | 51.8 | 66.2 | — | — | |
| SDGA-AutoFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 51.8 | — | — | — | |
| MemPO2026.02 | 50.2 | 59.17 | 0.37 | 0.15 | |
| SDGA-PhaseFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 49.5 | — | — | — | |
| GRPO-FullFramework=ZeroSearch, Model=Qwen2.5-7B2026.05 | 47.8 | — | — | — | |
| TaSR-RAGBackbone=Qwen2.5-7B-Instruct2026.03 | 45.9 | 65.6 | — | — | |
| Router-R1System=RL-based Agent2026.03 | 45.6 | — | — | — | |
| DeepResearcher2026.02 | 43.9 | 51.44 | 1.24 | 0.34 | |
| LinearRAGSystem=RAG-Based2026.03 | 43.5 | — | — | — | |
| SDGA-PhaseFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 43.2 | — | — | — | |
| GraphSearchSystem=Training-free Agent2026.03 | 42.7 | — | — | — | |
| GRPO-HalfFramework=ZeroSearch, Model=Qwen2.5-7B2026.05 | 42.6 | — | — | — | |
| HyperGraphRAGBackbone=Qwen2.5-72B-Instruct2026.03 | 42.1 | 51.5 | — | — | |
| ReSearch2026.02 | 41.9 | 50.07 | 0.55 | 0.28 | |
| SDGA-AutoFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 41.6 | — | — | — | |
| GRPO-FullFramework=ZeroSearch, Model=Qwen2.5-3B2026.05 | 41.2 | — | — | — | |
| Graph-R1System=RL-based Agent2026.03 | 40.7 | — | — | — | |
| SDGA-AutoFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 38 | — | — | — | |
| SDGA-PhaseFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 37.6 | — | — | — | |
| HippoRAG2System=RAG-Based2026.03 | 37.4 | — | — | — | |
| SDGA-AutoFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 37.3 | — | — | — | |
| SDGA-AutoFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 37.1 | — | — | — | |
| SDGA-PhaseFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 37.1 | — | — | — | |
| SDGA-PhaseFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 36.8 | — | — | — | |
| GRPO-FullFramework=Search-R1, Model=Qwen2.5-7B2026.05 | 36.8 | — | — | — | |
| GRPO-HalfFramework=ZeroSearch, Model=Qwen2.5-3B2026.05 | 36.6 | — | — | — | |
| qwen2.5mode=ReAct2026.02 | 35.9 | 45.21 | 1.12 | 0.29 | |
| Vanilla RAGSystem=RAG-Based2026.03 | 34.8 | — | — | — | |
| GRPO-FullFramework=Search-R1, Model=Qwen2.5-3B2026.05 | 33.7 | — | — | — | |
| HyperGraphRAGSystem=RAG-Based2026.03 | 33.6 | — | — | — | |
| GraphRAGSystem=RAG-Based2026.03 | 32.7 | — | — | — | |
| GRPO-FullFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base2026.05 | 32.6 | — | — | — | |
| GRPO-HalfFramework=Search-R1, Model=Qwen2.5-3B2026.05 | 32.5 | — | — | — | |
| GRPO-HalfFramework=Search-R1, Model=Qwen2.5-7B2026.05 | 32.4 | — | — | — | |
| RAPTORSystem=RAG-Based2026.03 | 31.8 | — | — | — | |
| GraphRAGBackbone=Qwen2.5-72B-Instruct2026.03 | 30.9 | 38.5 | — | — | |
| GRPO-FullFramework=Search-R1, Model=LLaMA-3.2-3B-Base2026.05 | 30.6 | — | — | — | |
| GraphRAGBackbone=Qwen2.5-7B-Instruct2026.03 | 30.5 | 37.8 | — | — | |
| Direct InferSystem=Basic LLM2026.03 | 29.5 | — | — | — | |
| CoTSystem=Basic LLM2026.03 | 29.3 | — | — | — | |
| GRPO-HalfFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base2026.05 | 28.7 | — | — | — | |
| GRPO-HalfFramework=Search-R1, Model=LLaMA-3.2-3B-Base2026.05 | 27.6 | — | — | — | |
| Search-R1System=RL-based Agent2026.03 | 26.9 | — | — | — | |
| SDGA-AntiFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 26.8 | — | — | — | |
| IRCoTBackbone=Qwen2.5-72B-Instruct2026.03 | 25.1 | 32.1 | — | — | |
| SDGA-AntiFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 24.7 | — | — | — | |
| StructRAGBackbone=Qwen2.5-72B-Instruct2026.03 | 24.5 | 37.2 | — | — | |
| SDGA-AntiFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 24.2 | — | — | — | |
| Search-o1System=Training-free Agent2026.03 | 23 | — | — | — | |
| RAGBackbone=Qwen2.5-72B-Instruct2026.03 | 22.7 | 29.1 | — | — | |
| CoTBackbone=Qwen2.5-72B-Instruct2026.03 | 20.5 | 26.2 | — | — | |
| HippoRAGBackbone=Qwen2.5-7B-Instruct2026.03 | 19.7 | 38.2 | — | — | |
| SDGA-AntiFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 18.3 | — | — | — | |
| StructRAGBackbone=Qwen2.5-7B-Instruct2026.03 | 16.7 | 23.9 | — | — | |
| SDGA-AntiFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 16.4 | — | — | — | |
| SDGA-AntiFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 15.7 | — | — | — | |
| Direct InferenceBackbone=Qwen2.5-72B-Instruct2026.03 | 15.3 | 19.6 | — | — | |
| IRCoTBackbone=Qwen2.5-7B-Instruct2026.03 | 14.9 | 21.3 | — | — | |
| RAGBackbone=Qwen2.5-7B-Instruct2026.03 | 13.5 | 47.7 | — | — | |
| Direct InferenceBackbone=Qwen2.5-7B-Instruct2026.03 | 12.6 | 30.1 | — | — | |
| CoTBackbone=Qwen2.5-7B-Instruct2026.03 | 10.8 | 15.1 | — | — | |
| HypergraphRAGBackbone=Qwen2.5-7B-Instruct2026.03 | 1.6 | 1.9 | — | — |