Open-domain Question Answering on 2WikiMultihopQA Search-o1 framework (50% held-out test)
39.2AccuracyPROBE&PREFILL
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PROBE&PREFILLN=2502026.05 | 39.2 | 297 | -56 | |
| Best BaselineN=2502026.05 | 38.4 | 561 | -16 | |
| DefaultN=2502026.05 | 36.4 | 670 | — |