Table 2 Uncertainty quantification results

From: Uncertainty-aware and causal test-time adaptive foundation model for robust colorectal cancer pathology diagnosis

Method

ECE ↓

Brier ↓

NLL ↓

MC Dropout

0.089

0.201

0.462

Deep Ensembles

0.052

0.173

0.381

SWAG

0.061

0.184

0.397

UAD-FM (Ours)

0.031

0.157

0.352

  1. UAD-FM achieves the best calibration across ECE, Brier Score, and NLL.
  2. Bold values indicate the best performance in each column.
Back to article page