Open-domain Question Answering on 2WikiMultiHopQA (500 randomly chosen questions)
3.47BM25 QueriesIRCoT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| IRCoT2025.06 | 3.47 | 3.47 | 59,000 | — | — | |
| Self-Ask2025.06 | 3.44 | 3.44 | 16,000 | — | — | |
| π-CoT2025.06 | 2.14 | 4.14 | 22,000 | — | — | |
| Standard RAG2025.06 | 1 | 1 | 3,700 | — | — | |
| IRCoTPrompting-based=true2025.06 | — | — | — | 32.4 | 42.5 | |
| Self-AskPrompting-based=true2025.06 | — | — | — | 15.8 | 21.8 | |
| Standard RAGPrompting-based=true2025.06 | — | — | — | 37.2 | 40.4 | |
| π-CoTPrompting-based=true2025.06 | — | — | — | 49.4 | 57.5 |