If we have enough data to partition the dataset into training, validation, and test samples, which one of the following classification models is most likely to be the best when applied to the test sample?

(A) A model with 18% training error, 22% validation error, and 75% sensitivity.
(B) A model with 17% training error, 20% validation error, and 74% sensitivity.
(C) A model with 21% training error, 21% validation error, and 75% sensitivity.
(D) A model with 19% training error, 20% validation error, and 75% sensitivity.