Multi-hop Question Answering on GraphRAG-Bench
87.1LLM AccuracyDTKG
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DTKG2025.10 | 87.1 | — | 14.5 | |
| TOG2025.10 | 84.5 | — | 13.5 | |
| MoGBackbone=DeepSeek-V32026.05 | 84.3 | 80.5 | — | |
| KGR2025.10 | 82 | — | 13.5 | |
| Youtu-GraphRAGBackbone=DeepSeek-V32026.05 | 81.5 | 77.2 | — | |
| HippoRAG2Backbone=DeepSeek-V32026.05 | 79.4 | 75.6 | — | |
| COT2025.10 | 77.2 | — | 14.4 | |
| CRITIC2025.10 | 74.8 | — | 13.4 | |
| RAPTORBackbone=DeepSeek-V32026.05 | 73.1 | 70.8 | — | |
| HippoRAGBackbone=DeepSeek-V32026.05 | 72.9 | 71 | — | |
| Original2025.10 | 71.8 | — | 13.9 | |
| Zero-shot LLMBackbone=DeepSeek-V32026.05 | 70.9 | 63.3 | — | |
| LightRAGBackbone=DeepSeek-V32026.05 | 70.8 | 67.5 | — | |
| G-retrieverBackbone=DeepSeek-V32026.05 | 70.6 | 68.2 | — | |
| E2GraphRAGBackbone=DeepSeek-V32026.05 | 68.7 | 65.3 | — |