Retrieval on HotpotQA
96.9R@5QRRanker-4B
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| QRRanker-4Bmethod_category=Reranking Methods, rerank_top_k=50, initial_retriever=Qwen3-Embedding-8B2026.02 | 96.9 | — | 97.7 | — | 95.05 | |
| HippoRAG-v2method_category=Reranking Methods, rerank_top_k=50, initial_retriever=Qwen3-Embedding-8B2026.02 | 96.3 | — | — | — | — | |
| Ψ-RAG AType=Tree-RAG, Abstract Type=Summative2026.05 | 96 | 88.15 | — | — | — | |
| Ψ-RAG øType=Tree-RAG, Abstract Type=Keyword2026.05 | 95.8 | 88.1 | — | — | — | |
| QRHeads-4B (out-of-box)method_category=Reranking Methods, rerank_top_k=50, initial_retriever=Qwen3-Embedding-8B, status=out-of-box2026.02 | 94.8 | — | 96.9 | — | 90.2 | |
| HippoRAG 2Type=Graph-RAG2026.05 | 94.4 | 84.05 | — | — | — | |
| Qwen-Reranker-4B (out-of-box)method_category=Reranking Methods, rerank_top_k=50, initial_retriever=Qwen3-Embedding-8B, status=out-of-box2026.02 | 94.15 | — | 96.75 | — | 89.8 | |
| Qwen-Reranker-4B (trained)method_category=Reranking Methods, rerank_top_k=50, initial_retriever=Qwen3-Embedding-8B, status=trained2026.02 | 93.95 | — | 96.9 | — | 89.35 | |
| BM25 + DPR + Q3RType=Traditional RAG2026.05 | 93.8 | 85.7 | — | — | — | |
| Quest-GNNLLM Generator=Llama-3.3-70B-Instruct2025.10 | 93.67 | 78.42 | — | — | — | |
| Self-Correcting RAGCategory=Proposed Framework2026.04 | 93.6 | — | — | — | — | |
| RAPTORLLM Generator=Llama-3.3-70B-Instruct2025.10 | 93.19 | 77.83 | — | — | — | |
| NV-Embed-v2LLM Generator=Llama-3.3-70B-Instruct2025.10 | 92.82 | 79.27 | — | — | — | |
| HippoRAG2LLM Generator=Llama-3.3-70B-Instruct2025.10 | 92.56 | 78.25 | — | — | — | |
| DPR + IRCoT + Q3RType=Traditional RAG2026.05 | 92.4 | 84.5 | — | — | — | |
| CRAGCategory=Iterative & Agentic RAG2026.04 | 91.5 | — | — | — | — | |
| HippoRAG 2 + Q3RType=Graph-RAG2026.05 | 90.65 | 74.7 | — | — | — | |
| GroupRank-32B*method_category=Reranking Methods, rerank_top_k=50, initial_retriever=Qwen3-Embedding-8B, run_type=single run evaluation2026.02 | 90.6 | — | 94.5 | — | 82.45 | |
| HippoRAG-v1method_category=Reranking Methods, rerank_top_k=50, initial_retriever=Qwen3-Embedding-8B2026.02 | 90.4 | — | — | — | — | |
| RECOMPCategory=Advanced Selection & Reranking2026.04 | 90.2 | — | — | — | — | |
| ToG2LLM Generator=Llama-3.3-70B-Instruct2025.10 | 90.18 | 75.14 | — | — | — | |
| GTE-Qwen2-7BLLM Generator=Llama-3.3-70B-Instruct2025.10 | 90.15 | 77.24 | — | — | — | |
| HippoRAG2type=GraphRAG Baseline2026.02 | 89.6 | — | — | — | 82.7 | |
| CatRAGRetrieval Strategy=Structure-Aware RAG2026.02 | 89.5 | — | — | — | — | |
| G-retrieverLLM Generator=Llama-3.3-70B-Instruct2025.10 | 89.49 | 74.16 | — | — | — | |
| Qwen3-Embedding-8Bmethod_category=Embedding Methods2026.02 | 89.05 | — | 95.15 | — | 82.85 | |
| GFM-RAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 88.86 | 74.05 | — | — | — | |
| SFT-Embedding-8Bmethod_category=Embedding Methods2026.02 | 88.63 | — | 94.19 | — | 82.36 | |
| FilcoCategory=Advanced Selection & Reranking2026.04 | 88.4 | — | — | — | — | |
| DRAGCategory=Iterative & Agentic RAG2026.04 | 88.3 | — | — | — | — | |
| Self-RAGCategory=Iterative & Agentic RAG2026.04 | 87.7 | — | — | — | — | |
| EA-GraphRAG2026.02 | 87.2 | — | — | — | 80.5 | |
| HippoRAG 2Retrieval Strategy=Structure-Aware RAG2026.02 | 87.1 | — | — | — | — | |
| IRCoTCategory=Iterative & Agentic RAG2026.04 | 86.9 | — | — | — | — | |
| DPRType=Traditional RAG2026.05 | 86.75 | 35.25 | — | — | — | |
| Qwen3-Embedding-4Bmethod_category=Embedding Methods2026.02 | 86.16 | — | 92.33 | — | 78.84 | |
| RAG + MMRCategory=Advanced Selection & Reranking2026.04 | 85.1 | — | — | — | — | |
| LongLLMLinguaCategory=Advanced Selection & Reranking2026.04 | 85 | — | — | — | — | |
| KiRAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 81.69 | 67.18 | — | — | — | |
| GNN-RAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 81.69 | 67.18 | — | — | — | |
| text-embedding-3-smallRetrieval Strategy=Standard Retrieval2026.02 | 81.3 | — | — | — | — | |
| RAPTORRetrieval Strategy=Structure-Aware RAG2026.02 | 79.5 | — | — | — | — | |
| Auto-RAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 79.5 | 65.81 | — | — | — | |
| Self-RAGLLM Generator=Llama-3.3-70B-Instruct2025.10 | 79.27 | 66.36 | — | — | — | |
| RRRCategory=Standard Baselines2026.04 | 78.9 | — | — | — | — | |
| ColBERTv2type=RAG Baseline2026.02 | 77.8 | — | — | — | 71.5 | |
| HippoRAGtype=GraphRAG Baseline2026.02 | 76.2 | — | — | — | 68.9 | |
| BM25LLM Generator=Llama-3.3-70B-Instruct2025.10 | 75.82 | 57.36 | — | — | — | |
| ContrieverRetrieval Strategy=Standard Retrieval2026.02 | 75.3 | — | — | — | — | |
| HyDECategory=Standard Baselines2026.04 | 75.3 | — | — | — | — | |
| NaiveCategory=Standard Baselines2026.04 | 74.8 | — | — | — | — | |
| BM25Type=Traditional RAG2026.05 | 74.8 | 57.25 | — | — | — | |
| Contriever+IRCoTLLM Generator=Llama-3.3-70B-Instruct2025.10 | 74.37 | 56.21 | — | — | — | |
| GTR (T5-base)Retrieval Strategy=Standard Retrieval2026.02 | 73.9 | — | — | — | — | |
| RAPTOR + Q3RType=Tree-RAG2026.05 | 72.25 | 67.1 | — | — | — | |
| BM25type=RAG Baseline2026.02 | 71.8 | — | — | — | 62.6 | |
| Contrievertype=RAG Baseline2026.02 | 68.5 | — | — | — | 60.6 | |
| BM25Retrieval Strategy=Standard Retrieval2026.02 | 64.8 | — | — | — | — | |
| INTRA2026.05 | 59.9 | — | 70.9 | 76.1 | — | |
| BGEvariant=BGE-large2026.05 | 54.8 | — | 63.5 | 69.6 | — | |
| RAPTORtype=GraphRAG Baseline2026.02 | 50.1 | — | — | — | 45.9 | |
| Qwen3-Emb-4B + Jina rerankerparameters=4B, reranker=Jina2026.05 | 48.8 | — | 59.6 | 65.4 | — | |
| Hybrid RAGfusion=reciprocal rank fusion (RRF)2026.05 | 48 | — | 61.8 | 71.2 | — | |
| Qwen3-Emb-4Bparameters=4B2026.05 | 44.6 | — | 53.5 | 61.1 | — | |
| MaxSimtype=late-interaction, style=ColBERT2026.05 | 36.1 | — | 46.3 | 54.2 | — | |
| Qwen3-Emb-0.6Bparameters=0.6B2026.05 | 35.9 | — | 44.2 | 50.4 | — | |
| BM25type=sparse lexical2026.05 | 32.2 | — | 41 | 48.9 | — | |
| TF-IDFtype=sparse lexical2026.05 | 18.2 | — | 25.9 | 35.5 | — | |
| BM25Q=10002026.05 | — | — | 82.8 | — | — | |
| ConvMemoryQ=10002026.05 | — | — | 79.83 | — | — | |
| Dense + lexicalQ=10002026.05 | — | — | 86.21 | — | — | |
| DrKIThops=2, type=dense retrieval2020.09 | — | 38.3 | 67.2 | 71 | — | |
| MDRhops=2, mode=direct2020.09 | — | 65.9 | 77.5 | 80.2 | — | |
| Raw denseQ=10002026.05 | — | — | 76.82 | — | — | |
| TF-IDFhops=22020.09 | — | 10.3 | 29.1 | 36.8 | — | |
| TF-IDF + Linkedhops=2, reranking=BM252020.09 | — | 17.3 | 50 | 62.7 | — |