Uncertainty-Aware Calibrated Clinical Text Classification with Large Language Models
The inferred posterior separates correct predictions from errors more sharply than black-box confidence, with misclassification-detection AUROC on MultiCare of 0.923 for M-SMC against 0.760 for softmax and 0.584 for a consistency baseline.
Mridul Sharma, Adeetya Patel, Z. Dsouza et al.
· 0 citations