Text this: Evaluation of large language models under different training background in Chinese medical examination: a comparative study