Our work “On the Origin of Hallucination in Conversational Models: Is it the Datasets or the Models?” got accepted at NAACL 2022. [Paper] [Data]