Note Training data for the models in this collection
Note Evaluation benchmark of 22 constituent datasets used to evaluate the models in this collection