data/merged_dataset/dataset_dict.json

{"splits": ["train", "validation", "test", "orig_train", "orig_validation", "orig_test"]}