1 paper · 1 filter
Jiahao Xu, Wei Liu, Yang Liu +5
Large language models increasingly generate summaries from collections of documents to support sensemaking and reporting, but verifying whether summary statements are grounded in s…