Multi-hop QA Retrieval on Average (MuSiQue, 2Wiki, MoreHopQA, FRAMES)
0.791NDCGPAR2-RAG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PAR2-RAG2026.03 | 0.791 | 86.3 | 69.3 | |
| Iterative Chain2026.03 | 0.765 | 78.7 | 56.9 | |
| IRCoT2026.03 | 0.747 | 79.4 | 59.5 | |
| Coverage Anchor2026.03 | 0.746 | 82 | 61.2 | |
| ReAct2026.03 | 0.733 | 70.7 | 44.1 |