The evaluation will use the accuracy metric, i.e. the proportion of correctly predicted items, on the test set. There will be both a public and a private test set.