Merging Triggers, Breaking Backdoors: Defensive Poisoning for Instruction-Tuned Language Models.

San Kim, Gary Lee

Browse the full ACL paper archive.