ParentFull threadsimonw·Which summarization models aren't susceptible to hallucination? Is there any good writing out there about this issue?View on HN