Multi-hop Question Answering on HotpotQA and 2WikiMultiHopQA Distractor Sets Macro Average (val)
65.26Macro F1Oracle (UB)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Oracle (UB)type=Upper Bound2026.06 | 65.26 | 1.61 | 11.28 | |
| Fixed k=5k=52026.06 | 60.52 | 5 | 6.54 | |
| TASR (AS_m25)rule=AS_m252026.06 | 57.4 | 3.13 | 3.42 | |
| Fixed k=3k=32026.06 | 53.98 | 3 | — | |
| Fixed k=1k=12026.06 | 35.29 | 1 | -18.69 |