Question Answering on HotpotQA (EM, F1, ROUGE-L, Latency)
47.8EMGraphRAG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GraphRAG2026.04 | 47.8 | 67.9 | 67.8 | 105 | |
| HybridRouter2026.04 | 33 | 52.4 | 52.1 | 73 | |
| ReasonRAGuncertainty_type=grounded self-rating, SSC_sampling=K=32026.04 | 21.2 | 41.7 | 41.5 | 1,865 | |
| VanillaRAG2026.04 | 19.6 | 39.2 | 38.9 | 23 | |
| ReasonRAGuncertainty_type=lexical2026.04 | 19.6 | 42.1 | 38.9 | 73 |