The evaluation will use the accuracy metric, i.e. the proportion of correctly predicted items, on the test set. This is the case for both the public and private test sets.