List your company

Confidence Calibration on SimpleQA (test)

6.8ECEHTC
6.5888.0199.4510.881Jan 22, 2026

Evaluation results

MethodLinks
HTCVariant=Reduced2026.016.81475.2
LastStep-TPTemperature Scaling=true2026.017.117.869.8
HTCVariant=Full2026.017.51572.7
GlobalTrace-TPTemperature Scaling=true2026.017.718.169.1
LastStep-TPTemperature Scaling=false2026.0110.118.669.9
GlobalTrace-TPTemperature Scaling=false2026.011119.369.2
Verbalized Conf2026.0112.119.665.5