benchmark/dataset_dict.json

{"splits": ["train", "test"]}