Question Answering on HotpotQA (EM, F1, Trigger rate (T))
42.6EMUncertain-Calibrate
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Uncertain-CalibrateScope=local2026.04 | 42.6 | 52.7 | 67.4 | |
| Verbal-CalibrateScope=global2026.04 | 42 | 52.8 | 61.6 | |
| DRAGIN2026.04 | 34.6 | 48.4 | 87.6 | |
| Ret-All2026.04 | 34 | 44.9 | 100 | |
| ADARAGUE2026.04 | 27.8 | 37.2 | 57 | |
| No-Ret2026.04 | 23.4 | 31.9 | — | |
| Base-UncTok2026.04 | 22.5 | 33.1 | 3.5 | |
| FLARE2026.04 | 21.8 | 35.2 | 99.2 | |
| Base-Verbal2026.04 | 21.2 | 28.8 | 15.2 | |
| SR-7BBackbone=7B2026.04 | 4.2 | 16.6 | 57.8 | |
| SR-13BBackbone=13B2026.04 | 2.6 | 16 | 42.4 |