Priority Learning for Validation