I remember a while back some people tried it in English and got lib answers. But what if that's because English datasets are lib? Has anyone tried a "translate this to Chinese, answer, then translate back to English" approach?