Question Answering on 2WIKI
86EMOracle
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Oracle2026.04 | 86 | 91.79 | — | — | — | — | — | — | — | — | — | — | |
| EvolveRouterPrompt Optimization=Enabled, Adaptive K=Enabled2026.04 | 70.33 | 77.7 | — | — | — | — | — | — | — | — | — | — | |
| T^2RAGInference LLM=Gemini-2.5-flash2025.08 | 69.3 | 77.5 | — | — | — | — | — | — | — | — | — | — | |
| Ψ-RAG ⍋Type=Tree-RAG, Reranker=Qwen3-Reranker-8B (Q3R), Abstract Type=summative2026.05 | 69.1 | 76.73 | — | — | — | — | — | — | — | — | — | — | |
| Ψ-RAG øType=Tree-RAG, Reranker=Qwen3-Reranker-8B (Q3R), Abstract Type=keyword2026.05 | 69 | 76.94 | — | — | — | — | — | — | — | — | — | — | |
| EvolveRouter (w/o PO)Prompt Optimization=Disabled, Adaptive K=Enabled2026.04 | 68 | 76.69 | — | — | — | — | — | — | — | — | — | — | |
| AgentRouter2026.04 | 67.67 | 75.54 | — | — | — | — | — | — | — | — | — | — | |
| EvolveRouter (w/o Adaptive K)Prompt Optimization=Enabled, Adaptive K=Disabled2026.04 | 67.67 | 76.53 | — | — | — | — | — | — | — | — | — | — | |
| T^2RAGInference LLM=GPT-4o-mini2025.08 | 66.7 | 74.4 | — | — | — | — | — | — | — | — | — | — | |
| CascadeRouting2026.04 | 65.33 | 73.36 | — | — | — | — | — | — | — | — | — | — | |
| RouteLLM2026.04 | 65 | 71.49 | — | — | — | — | — | — | — | — | — | — | |
| HippoRAG 2Type=Graph-RAG2026.05 | 64.8 | 71.35 | — | — | — | — | — | — | — | — | — | — | |
| Best Agent2026.04 | 64.56 | 75.21 | — | — | — | — | — | — | — | — | — | — | |
| MLP Router2026.04 | 64 | 72.59 | — | — | — | — | — | — | — | — | — | — | |
| KNN Router2026.04 | 63 | 71.54 | — | — | — | — | — | — | — | — | — | — | |
| Best LLM2026.04 | 62.72 | 71.77 | — | — | — | — | — | — | — | — | — | — | |
| IRCoTInference LLM=Gemini-2.5-flash2025.08 | 61.6 | 71.7 | — | — | — | — | — | — | — | — | — | — | |
| IRCoTInference LLM=GPT-4o-mini2025.08 | 60.7 | 74.3 | — | — | — | — | — | — | — | — | — | — | |
| HippoRAG 2Backbone=GPT-4o-mini2025.12 | 60.5 | 69.7 | — | — | — | — | — | — | — | — | — | — | |
| HippoRAG 2 + Q3RType=Graph-RAG, Reranker=Qwen3-Reranker-8B (Q3R)2026.05 | 60.1 | 66 | — | — | — | — | — | — | — | — | — | — | |
| HippoRAG2Inference LLM=GPT-4o-mini2025.08 | 59.6 | 69.3 | — | — | — | — | — | — | — | — | — | — | |
| HippoRAGBackbone=GPT-4o-mini2025.12 | 59.4 | 67.3 | — | — | — | — | — | — | — | — | — | — | |
| Majority Vote2026.04 | 58.55 | 67.67 | — | — | — | — | — | — | — | — | — | — | |
| RouteRAGBackbone=Qwen2.5-3B2025.12 | 57.5 | 64.1 | — | — | — | — | — | — | — | — | — | — | |
| RouteRAGBackbone=Qwen2.5-7B2025.12 | 57.1 | 64.6 | — | — | — | — | — | — | — | — | — | — | |
| Quest-GNNLLM Generator=Llama-3.3-70B-Instruct2025.10 | 57.02 | 66.83 | — | — | — | — | — | — | — | — | — | — | |
| HippoRAG2LLM Generator=Llama-3.3-70B-Instruct2025.10 | 56.6 | 66.18 | — | — | — | — | — | — | — | — | — | — | |
| HybridLLM2026.04 | 54.67 | 66.91 | — | — | — | — | — | — | — | — | — | — | |
| GTE-Qwen2-7BLLM Generator=Llama-3.3-70B-Instruct2025.10 | 54.11 | 59.61 | — | — | — | — | — | — | — | — | — | — | |
| NV-Embed-v2LLM Generator=Llama-3.3-70B-Instruct2025.10 | 53.78 | 59.99 | — | — | — | — | — | — | — | — | — | — | |
| ICRLModel=Qwen2.5-7B2026.03 | 53.6 | — | — | — | — | — | — | — | — | — | — | — | |
| StandardInference LLM=GPT-4o-mini2025.08 | 53.1 | 60.2 | — | — | — | — | — | — | — | — | — | — | |
| GraphRouter2026.04 | 53 | 61.98 | — | — | — | — | — | — | — | — | — | — | |
| GraphRAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 52.46 | 63.63 | — | — | — | — | — | — | — | — | — | — | |
| GraphRAGType=Graph-RAG2026.05 | 52.2 | 56.62 | — | — | — | — | — | — | — | — | — | — | |
| Search-R1Backbone=Qwen2.5-7B2025.12 | 51.8 | 58.9 | — | — | — | — | — | — | — | — | — | — | |
| PRAISEApproach=Ours2026.04 | 50.48 | 58.14 | — | — | — | — | — | — | — | — | — | — | |
| M2RModel Family=Mistral-7B-Instruct2026.04 | 50 | — | — | — | — | — | — | — | — | — | — | — | |
| LLM-Blender2026.04 | 49.67 | 64.03 | — | — | — | — | — | — | — | — | — | — | |
| ReSearchModel Family=Mistral-7B-Instruct2026.04 | 49.1 | — | — | — | — | — | — | — | — | — | — | — | |
| GFM-RAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 48.99 | 59.58 | — | — | — | — | — | — | — | — | — | — | |
| KiRAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 48.91 | 59.53 | — | — | — | — | — | — | — | — | — | — | |
| DPR + IRCoT + Q3RType=Traditional RAG, Reranker=Qwen3-Reranker-8B (Q3R), Iterative Retrieval Protocol=IRCoT2026.05 | 48.9 | 53.44 | — | — | — | — | — | — | — | — | — | — | |
| RAPTORLLM Generator=Llama-3.3-70B-Instruct2025.10 | 48.57 | 62.15 | — | — | — | — | — | — | — | — | — | — | |
| NORInference LLM=Gemini-2.5-flash2025.08 | 48.1 | 55.6 | — | — | — | — | — | — | — | — | — | — | |
| ToG2LLM Generator=Llama-3.3-70B-Instruct2025.10 | 47.74 | 62.14 | — | — | — | — | — | — | — | — | — | — | |
| SUREModel Family=Mistral-7B-Instruct2026.04 | 47.7 | — | — | — | — | — | — | — | — | — | — | — | |
| M2RModel Family=Llama-3.1-8B-Instruct2026.04 | 47.2 | — | — | — | — | — | — | — | — | — | — | — | |
| R1-SearcherRetrieval Model=Qwen3-Embedding-0.6B2026.05 | 46.99 | — | — | — | — | — | — | — | — | — | 8.861 | — | |
| G-retrieverLLM Generator=Llama-3.3-70B-Instruct2025.10 | 46.84 | 58.1 | — | — | — | — | — | — | — | — | — | — | |
| LatentRAG△Retrieval Model=Qwen3-Embedding-0.6B2026.05 | 46.14 | — | — | — | — | — | — | — | — | — | 0.607 | — | |
| ReSearchModel Family=Llama-3.1-8B-Instruct2026.04 | 45.8 | — | — | — | — | — | — | — | — | — | — | — | |
| GraphRAGBackbone=GPT-4o-mini2025.12 | 45.7 | 61 | — | — | — | — | — | — | — | — | — | — | |
| COFTModel Family=Mistral-7B-Instruct2026.04 | 45.5 | — | — | — | — | — | — | — | — | — | — | — | |
| Auto-RAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 45.28 | 55.41 | — | — | — | — | — | — | — | — | — | — | |
| LatentRAG△Retrieval Model=jina-embeddings-v5-text-nano2026.05 | 45.24 | — | — | — | — | — | — | — | — | — | 0.539 | — | |
| LatentRAG♢Retrieval Model=jina-embeddings-v5-text-nano2026.05 | 45.06 | — | — | — | — | — | — | — | — | — | 0.641 | — | |
| Random Router2026.04 | 45 | 54.76 | — | — | — | — | — | — | — | — | — | — | |
| LatentRAG♢Retrieval Model=harrier-oss-v1-270m2026.05 | 44.99 | — | — | — | — | — | — | — | — | — | 0.698 | — | |
| LatentRAG♢Retrieval Model=Qwen3-Embedding-0.6B2026.05 | 44.89 | — | — | — | — | — | — | — | — | — | 0.704 | — | |
| Self-RAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 44.76 | 53.27 | — | — | — | — | — | — | — | — | — | — | |
| LatentRAG△Retrieval Model=harrier-oss-v1-270m2026.05 | 44.58 | — | — | — | — | — | — | — | — | — | 0.602 | — | |
| HippoRAG2Inference LLM=Gemini-2.5-flash2025.08 | 44.3 | 51.2 | — | — | — | — | — | — | — | — | — | — | |
| Search-R1♢Retrieval Model=harrier-oss-v1-270m2026.05 | 44.16 | — | — | — | — | — | — | — | — | — | 6.864 | — | |
| LatentRAG△Retrieval Model=F2LLM-v2-330M2026.05 | 44.08 | — | — | — | — | — | — | — | — | — | 0.583 | — | |
| R1-SearcherApproach=Agentic Search2026.04 | 43.97 | 50.64 | — | — | — | — | — | — | — | — | — | — | |
| DeepRAGRetrieval Model=Qwen3-Embedding-0.6B2026.05 | 43.92 | — | — | — | — | — | — | — | — | — | 3.758 | — | |
| AutoRefine△Retrieval Model=harrier-oss-v1-270m2026.05 | 43.31 | — | — | — | — | — | — | — | — | — | 4.91 | — | |
| Search-R1♢Retrieval Model=Qwen3-Embedding-0.6B2026.05 | 43.27 | — | — | — | — | — | — | — | — | — | 6.925 | — | |
| BM25LLM Generator=Llama-3.3-70B-Instruct2025.10 | 43.15 | 49.49 | — | — | — | — | — | — | — | — | — | — | |
| LongLLMLinReader=GPT-5-mini (low)2026.03 | 43.1 | 53.6 | — | — | — | — | — | — | — | — | 1.357 | 39.3 | |
| LatentRAG♢Retrieval Model=F2LLM-v2-330M2026.05 | 43.02 | — | — | — | — | — | — | — | — | — | 0.679 | — | |
| TIPSBackbone=Qwen-2.5 Instruct, Model Size=7B2026.03 | 42.96 | — | — | — | — | — | — | — | — | — | — | — | |
| Search-R1♢Retrieval Model=jina-embeddings-v5-text-nano2026.05 | 42.95 | — | — | — | — | — | — | — | — | — | 6.758 | — | |
| AutoRefine△Retrieval Model=jina-embeddings-v5-text-nano2026.05 | 42.95 | — | — | — | — | — | — | — | — | — | 4.879 | — | |
| BM25Inference LLM=GPT-4o-mini2025.08 | 42.9 | 48.2 | — | — | — | — | — | — | — | — | — | — | |
| LooCompReader=GPT-5-mini (low)2026.03 | 42.8 | 52.6 | — | — | — | — | — | — | — | — | 0.159 | 6.4 | |
| Average2026.04 | 42.73 | 52.15 | — | — | — | — | — | — | — | — | — | — | |
| EXITReader=GPT-5-mini (low)2026.03 | 42.7 | 52.6 | — | — | — | — | — | — | — | — | 1.619 | 37.6 | |
| AutoRefine△Retrieval Model=Qwen3-Embedding-0.6B2026.05 | 42.53 | — | — | — | — | — | — | — | — | — | 5.264 | — | |
| Search-R1Backbone=Qwen2.5-3B2025.12 | 42.4 | 50.8 | — | — | — | — | — | — | — | — | — | — | |
| ParallelSearchModel=Qwen2.5-7B2026.03 | 42.4 | — | — | — | — | — | — | — | — | — | — | — | |
| Search-R1♢Retrieval Model=e5-base-v22026.05 | 42.13 | — | — | — | — | — | — | — | — | — | 6.662 | — | |
| SUREModel Family=Llama-3.1-8B-Instruct2026.04 | 42 | — | — | — | — | — | — | — | — | — | — | — | |
| LatentRAG△Retrieval Model=e5-base-v22026.05 | 41.98 | — | — | — | — | — | — | — | — | — | 0.552 | — | |
| ProvenceReader=GPT-5-mini (low)2026.03 | 41.9 | 50.9 | — | — | — | — | — | — | — | — | 0.197 | 11.9 | |
| AutoRefine△Retrieval Model=e5-base-v22026.05 | 41.9 | — | — | — | — | — | — | — | — | — | 4.695 | — | |
| R1-SearcherBackbone=Qwen2.5-7B2025.12 | 41.7 | 49 | — | — | — | — | — | — | — | — | — | — | |
| BM25 + DPR + Q3RType=Traditional RAG, Reranker=Qwen3-Reranker-8B (Q3R)2026.05 | 41.6 | 45.43 | — | — | — | — | — | — | — | — | — | — | |
| AutoRefine△Retrieval Model=F2LLM-v2-330M2026.05 | 41.58 | — | — | — | — | — | — | — | — | — | 5.101 | — | |
| Search-R1♢Retrieval Model=F2LLM-v2-330M2026.05 | 41.53 | — | — | — | — | — | — | — | — | — | 6.765 | — | |
| Search-R1Model=Qwen2.5-7B2026.03 | 41.4 | — | — | — | — | — | — | — | — | — | — | — | |
| Search-R1-instructBackbone Model=Qwen2.5-7B2025.10 | 41.4 | — | — | — | — | — | — | — | — | — | — | — | |
| DPRType=Traditional RAG2026.05 | 39.9 | 43.13 | — | — | — | — | — | — | — | — | — | — | |
| Search-R1Approach=Agentic Search2026.04 | 39.71 | 47.1 | — | — | — | — | — | — | — | — | — | — | |
| RAPTORBackbone=GPT-4o-mini2025.12 | 39.7 | 48.4 | — | — | — | — | — | — | — | — | — | — | |
| EvolveRBackbone Model=Qwen2.5-7B2025.10 | 39.5 | — | — | — | — | — | — | — | — | — | — | — | |
| COFTModel Family=Llama-3.1-8B-Instruct2026.04 | 39.5 | — | — | — | — | — | — | — | — | — | — | — | |
| ICRLModel=Qwen2.5-3B2026.03 | 39.2 | — | — | — | — | — | — | — | — | — | — | — | |
| LatentRAG♢Retrieval Model=e5-base-v22026.05 | 38.61 | — | — | — | — | — | — | — | — | — | 0.687 | — |