Safety Alignment in NLP Tasks: Weakly Aligned Summarization as an In-Context Attack.

Yu Fu, Yufei Li, Wen Xiao, Cong Liu, Yue Dong

Browse the full ACL paper archive.