Multi-hop Question Answering on MuSiQue (F1 Token Overlap, Gap Closed)
43.6F1 (token-overlap)Oracle
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Oracletype=upper bound2026.04 | 43.6 | — | |
| GraphRAG2026.04 | 40.8 | — | |
| LR+DirectGR2026.04 | 33.2 | 70.4 | |
| HybridRouter2026.04 | 18.2 | — | |
| ReasonRAGmode=lex2026.04 | 11.2 | — | |
| VanillaRAG2026.04 | 8.3 | — |