List your company

Math Reasoning on GSM8K (Brier Score)

4.33Brier ScoreCAGE-CAL
4.09565.67787.268.8422May 28, 2026

Evaluation results

MethodLinks
CAGE-CALCalibration category=Trained calibrators2026.054.33
+ PlattCalibration category=Post-hoc plurality calibrators2026.054.37
DiscoUQ-LLMCalibration category=Trained calibrators2026.054.46
+ IsotonicCalibration category=Post-hoc plurality calibrators2026.054.5
+ Scaling-bin.Calibration category=Post-hoc plurality calibrators2026.054.54
Scalar + GBTCalibration category=Trained calibrators2026.054.87
Collab. Cal.Calibration category=LLM-elicited confidence estimators2026.058.41
LLM-Cal (+topo)Calibration category=LLM-elicited confidence estimators2026.058.96
GraphCalCalibration category=Trained calibrators2026.059.25
Plurality shareCalibration category=Post-hoc plurality calibrators2026.059.61
LLM-Cal (no topo)Calibration category=LLM-elicited confidence estimators2026.0510.19