data/ModeBench/python_factor_modebench_v1/eval/dataset_dict.json

{"splits": ["multi_answer"]}