Question Answering on MedQA (Accuracy, ECE, AUROC)
0.1145ECERewarding doubt
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Rewarding doubtTraining dataset=TriviaQA2025.03 | 0.1145 | 0.6649 | 51.61 | |
| Trained ProbeTraining dataset=TriviaQA2025.03 | 0.2099 | 0.5513 | 50.51 | |
| VerbalizeTraining dataset=TriviaQA2025.03 | 0.448 | 0.5075 | 50.67 |