T-hat represents the evidence strength of the top-ranked class, F-hat represents the evidence strength of the strongest competitor, I1-hat represents prediction entropy, and I2-hat represents the degree of internal decision disagreement within the ensemble.