Question Answering on NQ, PopQA, 2WikiQA, Musique, LOCOMO (test)
45.27Overall Average RAG F1HippoRAG2 + DeepRefine-8B
Evaluation Results
| Method | Links | |
|---|---|---|
| HippoRAG2 + DeepRefine-8BConstructor=AR1, Refinement Model=DeepRefine-8B2026.05 | 45.27 | |
| HippoRAG2 + DeepRefine-4BConstructor=AR1, Refinement Model=DeepRefine-4B2026.05 | 44.17 | |
| HippoRAG + DeepRefine-8BConstructor=AR1, Refinement Model=DeepRefine-8B2026.05 | 44.11 | |
| HippoRAG2Constructor=AR12026.05 | 43.65 | |
| HippoRAG + DeepRefine-4BConstructor=AR1, Refinement Model=DeepRefine-4B2026.05 | 43.62 | |
| HippoRAGConstructor=AR12026.05 | 43.17 | |
| HippoRAG2 + DeepRefine-8BConstructor=Naive, Refinement Model=DeepRefine-8B2026.05 | 42.24 | |
| HippoRAG + DeepRefine-8BConstructor=Naive, Refinement Model=DeepRefine-8B2026.05 | 42.06 | |
| HippoRAG + DeepRefine-4BConstructor=Naive, Refinement Model=DeepRefine-4B2026.05 | 41.96 | |
| HippoRAG2Constructor=Naive2026.05 | 41.82 | |
| HippoRAGConstructor=Naive2026.05 | 41.73 | |
| HippoRAG2 + DeepRefine-4BConstructor=Naive, Refinement Model=DeepRefine-4B2026.05 | 41.57 | |
| ToG + DeepRefine-8BConstructor=AR1, Refinement Model=DeepRefine-8B2026.05 | 37.51 | |
| ToG + DeepRefine-4BConstructor=AR1, Refinement Model=DeepRefine-4B2026.05 | 37.21 | |
| ToGConstructor=AR12026.05 | 36.3 | |
| Subgraph + DeepRefine-8BConstructor=AR1, Refinement Model=DeepRefine-8B2026.05 | 33.96 | |
| ToG + DeepRefine-8BConstructor=Naive, Refinement Model=DeepRefine-8B2026.05 | 33.75 | |
| Subgraph + DeepRefine-4BConstructor=AR1, Refinement Model=DeepRefine-4B2026.05 | 33.53 | |
| ToG + DeepRefine-4BConstructor=Naive, Refinement Model=DeepRefine-4B2026.05 | 33.06 | |
| ToGConstructor=Naive2026.05 | 33.01 | |
| SubgraphConstructor=AR12026.05 | 32.87 | |
| Subgraph + DeepRefine-8BConstructor=Naive, Refinement Model=DeepRefine-8B2026.05 | 31.72 | |
| Subgraph + DeepRefine-4BConstructor=Naive, Refinement Model=DeepRefine-4B2026.05 | 31.48 | |
| SubgraphConstructor=Naive2026.05 | 30.82 | |
| Subgraph + DeepRefine-8BConstructor=Graphify (LLM-Wiki), Refinement Model=DeepRefine-8B2026.05 | 16.35 | |
| Subgraph + DeepRefine-4BConstructor=Graphify (LLM-Wiki), Refinement Model=DeepRefine-4B2026.05 | 15.96 | |
| SubgraphConstructor=Graphify (LLM-Wiki)2026.05 | 14.82 | |
| ToG + DeepRefine-8BConstructor=Graphify (LLM-Wiki), Refinement Model=DeepRefine-8B2026.05 | 14.02 | |
| ToG + DeepRefine-4BConstructor=Graphify (LLM-Wiki), Refinement Model=DeepRefine-4B2026.05 | 13.82 | |
| ToGConstructor=Graphify (LLM-Wiki)2026.05 | 13.23 | |
| HippoRAG2 + DeepRefine-8BConstructor=Graphify (LLM-Wiki), Refinement Model=DeepRefine-8B2026.05 | 13.18 | |
| HippoRAG + DeepRefine-8BConstructor=Graphify (LLM-Wiki), Refinement Model=DeepRefine-8B2026.05 | 12.92 | |
| HippoRAG + DeepRefine-4BConstructor=Graphify (LLM-Wiki), Refinement Model=DeepRefine-4B2026.05 | 12.41 | |
| HippoRAG2 + DeepRefine-4BConstructor=Graphify (LLM-Wiki), Refinement Model=DeepRefine-4B2026.05 | 12.12 | |
| HippoRAGConstructor=Graphify (LLM-Wiki)2026.05 | 11.99 | |
| HippoRAG2Constructor=Graphify (LLM-Wiki)2026.05 | 11.89 |