1 paper · 1 filter
Zonghuan Xu, Xiang Zheng, Yutao Wu +1
Evaluating misinformation requires distinguishing whether readers believe content from whether they would share it. Asking large language models (LLMs) both questions yields two sc…