For Dataset Creation, Humans Should Evaluate Machine Text — New Research
As more language service providers explore data collection and annotation, a new study shows large language models generate more diverse content, while humans excel at rating it.