A question about Data-centric AI competition

would it be acceptable to add new labels to the dataset on both train and val sections?