Multiple Choice QA on PubMedQA (test)
81.8AUROCTotal variance, experts
Evaluation results
| Method | Links | |
|---|---|---|
| Total variance, expertsUses experts=✓, inference2026.04 | 81.8 | |
| Total variance, CAEUses experts=✓, training2026.04 | 73.2 | |
| Total variance, base2026.04 | 70.9 | |
| Entropy and MI2026.04 | 70.7 | |
| Prediction variance2026.04 | 70.2 | |
| Single model2026.04 | 69.3 | |
| MC Dropout2026.04 | 67.1 | |
| Test-time augmentations2026.04 | 67 | |
| HUQ2026.04 | 66.2 |