Multi-hop Question Answering on 2WikiMultiHopQA l2
0.09Cost per PassStandard RAG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Standard RAG2026.06 | 0.09 | 31.89 | 0.03 | 726 | |
| IRCoT2026.06 | 0.11 | 52.99 | 0.06 | 1,204 | |
| GDP-RAG2026.06 | 0.3 | 72.62 | 0.22 | 4,460 | |
| Search-o12026.06 | 0.34 | 72.66 | 0.25 | 5,172 | |
| PAR-RAG2026.06 | 0.43 | 69.1 | 0.3 | 5,956 | |
| KnowTracemode=inference-time KG-tracing, self-bootstrapping fine-tuning=False2026.06 | 0.51 | 68.31 | 0.35 | 7,692 | |
| Godbole et al.adaptation=QA first planning stage only2026.06 | 0.64 | 44.83 | 0.29 | 5,964 |