data/sft_selector_classifier_v2_rows/dataset_dict.json

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