pallest·3 年前·議論One solution is placing the ChatGPT response in another model that determines the accuracy of the information and revises where necessary