Question Answering on 2WikiMultihopQA (EM/F1/P/R)
47.7EMHippoRAG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| HippoRAGType=Mem., Backbone=GPT-3.52026.01 | 47.7 | — | — | — | |
| ProRAGMethod Category=Reinforcement Learning-based Methods2026.01 | 46 | 51.1 | — | — | |
| SAPO-3B-InstructRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 44.8 | — | — | — | |
| ReasonRAGMethod Category=Reinforcement Learning-based Methods2026.01 | 43.6 | 50.4 | — | — | |
| InForageRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 42.8 | — | — | — | |
| Search-R1Backbone=Qwen2.5-7B-Instruct2026.02 | 41.4 | — | — | — | |
| SLATEBackbone=Qwen2.5-7B-Base2026.02 | 41.3 | — | — | — | |
| SAPO-3B-BaseRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 41.2 | — | — | — | |
| CriticSearchRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 40.9 | — | — | — | |
| Search-R1Type=RL Training2026.01 | 40.1 | — | — | — | |
| Search-R1Method Category=Reinforcement Learning-based Methods2026.01 | 39.9 | 45.2 | — | — | |
| AutoRefine-BaseRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 39.3 | — | — | — | |
| StepSearchBackbone=Qwen2.5-7B-Base2026.02 | 38.5 | — | — | — | |
| SEARCH-R1Backbone=Qwen2.5-7B-Base2026.02 | 38.2 | — | — | — | |
| AutoRefine-InstructRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 38 | — | — | — | |
| O2-SearcherRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 37.4 | — | — | — | |
| EXSEARCH-BaseRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 37.2 | — | — | — | |
| HiPRAGMethod Category=Reinforcement Learning-based Methods2026.01 | 37 | 43.2 | — | — | |
| StepSearchType=RL Training2026.01 | 36.6 | — | — | — | |
| EXITBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 36.5 | 42.2 | — | — | |
| SE-SearchRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 36.1 | — | — | — | |
| ITER-RETGENType=Mem.2026.01 | 35.5 | — | — | — | |
| EXITBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 35.4 | 41 | — | — | |
| Original DocsBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=-2024.12 | 35.4 | 42.4 | — | — | |
| LongLLMLinguaBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 35.4 | 41.1 | — | — | |
| LongLLMLinguaBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 34.7 | 40.2 | — | — | |
| ZeroSearch-BaseRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 34.6 | — | — | — | |
| ACRType=SFT, Reasoner=Qwen-2.5-7B-Instruct2026.01 | 34.32 | — | — | — | |
| StepSearch-BaseRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 33.9 | — | — | — | |
| BAR-RAGBackbone=LLaMA-3.1-8B-Instruct2026.02 | 33 | — | — | — | |
| CompActBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 32.2 | 38.7 | — | — | |
| Search-R1Backbone=Qwen2.5-3B-Instruct2026.02 | 31.9 | — | — | — | |
| Original DocsBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=-2024.12 | 31.4 | 37.5 | — | — | |
| Search-R1-InstructRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 31 | — | — | — | |
| RefinerBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 30.5 | 36.6 | — | — | |
| RECOMP-AbstBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 30.3 | 35.7 | — | — | |
| RefinerBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 30.2 | 35.6 | — | — | |
| SEAKRType=RAG2026.01 | 30.2 | — | — | — | |
| BAR-RAGBackbone=Qwen2.5-7B-Instruct2026.02 | 29.8 | — | — | — | |
| RECOMP-AbstBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 27.8 | 33.4 | — | — | |
| R1-InstructionType=RL Training2026.01 | 27.5 | — | — | — | |
| R1-InstructRetrieval Strategy=None, Backbone Model=Qwen2.5-3B2026.03 | 27.5 | — | — | — | |
| ZeroSearchBackbone=Qwen2.5-7B-Base2026.02 | 27.5 | — | — | — | |
| Search-R1-BaseRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 27.4 | — | — | — | |
| ReSearch-InstructRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 27.2 | — | — | — | |
| ReSearch-BaseRetrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 27.2 | — | — | — | |
| R1 (no search)Backbone=Qwen2.5-7B-Base2026.02 | 27 | — | — | — | |
| SFTBackbone=Qwen2.5-7B-Base2026.02 | 26.9 | — | — | — | |
| RECOMP-AbstBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 26.8 | 32.4 | — | — | |
| R1-BaseRetrieval Strategy=None, Backbone Model=Qwen2.5-3B2026.03 | 26.8 | — | — | — | |
| DioRRetrieval Strategy=SBERT, Backbone=LLaMA2-7B-CHAT2025.04 | 26.6 | 32.8 | 33.4 | 33.2 | |
| DioRBackbone=LLaMA2-7B-CHAT2025.04 | 26.6 | 33.5 | — | — | |
| EXITBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 26.6 | 32.9 | — | — | |
| DioRType=RAG2026.01 | 26.6 | — | — | — | |
| RECOMP-ExtrBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 26.5 | 31.9 | — | — | |
| SEAKRBackbone=LLaMA2-7B-CHAT2025.04 | 26.4 | 33 | — | — | |
| BAR-RAGBackbone=Qwen2.5-3B-Instruct2026.02 | 26.4 | — | — | — | |
| ReSearchBackbone=Qwen2.5-7B-Base2026.02 | 26.4 | — | — | — | |
| RECOMP-AbstBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 26.2 | 32.2 | — | — | |
| CoTBackbone=Qwen2.5-7B-Base2026.02 | 26.1 | — | — | — | |
| EXITBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 25.9 | 32 | — | — | |
| SFTType=SFT2026.01 | 25.9 | — | — | — | |
| CompActBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 25.6 | 30.5 | — | — | |
| DioRRetrieval Strategy=BM25, Backbone=LLaMA2-7B-CHAT2025.04 | 25.4 | 33.5 | 34.2 | 34.5 | |
| RADIOBackbone=LLaMA2-7B-CHAT2025.04 | 25.4 | 31.7 | — | — | |
| LongLLMLinguaBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 25.3 | 31.6 | — | — | |
| IRCoTMethod Category=Advanced Methods2026.01 | 25.1 | 30.4 | — | — | |
| Direct inferenceType=Prompt2026.01 | 25 | — | — | — | |
| Original DocsBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=-2024.12 | 24.8 | 32.4 | — | — | |
| SFTRetrieval Strategy=None, Backbone Model=Qwen2.5-3B2026.03 | 24.8 | — | — | — | |
| RECOMP-ExtrBackbone=Llama-3.1-70B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 24.7 | 29.9 | — | — | |
| DirectBackbone=Qwen2.5-7B-Base2026.02 | 24.6 | — | — | — | |
| Direct GenerationRetrieval Strategy=None, Backbone Model=Qwen2.5-3B2026.03 | 24.4 | — | — | — | |
| FLAREMethod Category=Advanced Methods2026.01 | 24.1 | 29.5 | — | — | |
| CompActBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 23.9 | 31 | — | — | |
| RefinerBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Abs.2024.12 | 23.4 | 30 | — | — | |
| DraginRetrieval Strategy=SBERT, Backbone=LLaMA2-7B-CHAT2025.04 | 23.1 | 29.4 | 29.8 | 29.9 | |
| DraginBackbone=LLaMA2-7B-CHAT2025.04 | 23.1 | 29.4 | — | — | |
| RAGBackbone=Qwen2.5-7B-Base2026.02 | 23.1 | — | — | — | |
| LongLLMLinguaBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 23 | 28.8 | — | — | |
| RefinerBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 22.8 | 29.4 | — | — | |
| DioRRetrieval Strategy=SGPT, Backbone=LLaMA2-7B-CHAT2025.04 | 22.6 | 29.2 | 29.8 | 29.9 | |
| Search-o1Method Category=Advanced Methods2026.01 | 22.6 | 27.9 | — | — | |
| Naive RAGRetrieval Strategy=Naive RAG, Backbone Model=Qwen2.5-3B2026.03 | 22.6 | — | — | — | |
| Original DocsBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=-2024.12 | 22.2 | 29.1 | — | — | |
| DRAGINType=RAG2026.01 | 22 | — | — | — | |
| Search-o1Backbone=Qwen2.5-3B-Instruct2026.02 | 21.8 | — | — | — | |
| Search-o1Retrieval Strategy=Agent, Backbone Model=Qwen2.5-3B2026.03 | 21.8 | — | — | — | |
| DraginRetrieval Strategy=BM25, Backbone=LLaMA2-7B-CHAT2025.04 | 21.4 | 28.2 | 28.8 | 28.5 | |
| RECOMP-ExtrBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-5, Compressor Type=Ext.2024.12 | 21.2 | 27.5 | — | — | |
| CompActBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Abs.2024.12 | 21 | 28.3 | — | — | |
| DraginRetrieval Strategy=SGPT, Backbone=LLaMA2-7B-CHAT2025.04 | 20.9 | 27.8 | 28.4 | 28.1 | |
| RECOMP-ExtrBackbone=Llama-3.1-8B-Instruct, Top-k Documents=Top-20, Compressor Type=Ext.2024.12 | 19.8 | 26 | — | — | |
| FS-RAGBackbone=LLaMA2-7B-CHAT2025.04 | 18.9 | 26.5 | — | — | |
| Search-o1Backbone=Qwen2.5-7B-Base2026.02 | 18.1 | — | — | — | |
| Search-o1Backbone=Qwen2.5-7B-Instruct2026.02 | 17.6 | — | — | — | |
| IRCoTRetrieval Strategy=IRCoT, Backbone Model=Qwen2.5-3B2026.03 | 17.1 | — | — | — | |
| SR-RAGBackbone=LLaMA2-7B-CHAT2025.04 | 16.9 | 25.5 | — | — | |
| Standard RAGMethod Category=Standard Methods2026.01 | 16.1 | 23 | — | — | |
| Naïve GenerationMethod Category=Standard Methods2026.01 | 15.8 | 21 | — | — |