Question Answering on TriviaQA (EM, F1, Trigger rate (T))
66.6EMUncertain-Calibrate
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Uncertain-CalibrateScope=local2026.04 | 66.6 | 73.2 | 34 | |
| Verbal-CalibrateScope=global2026.04 | 63.2 | 72.5 | 28.8 | |
| No-Ret2026.04 | 54.8 | 61.2 | — | |
| DRAGIN2026.04 | 53.6 | 60.6 | 54 | |
| ADARAGUE2026.04 | 52.2 | 58.1 | 29 | |
| Ret-All2026.04 | 43 | 49.1 | 100 | |
| Base-Verbal2026.04 | 42.4 | 43.8 | 4.6 | |
| FLARE2026.04 | 41 | 49.9 | 97.4 | |
| Base-UncTok2026.04 | 34.8 | 36.8 | 4.5 | |
| SR-13BBackbone=13B2026.04 | 6.4 | 36.5 | 29.8 | |
| SR-7BBackbone=7B2026.04 | 5.6 | 24.9 | 53.6 |