Separate the Wheat from the Chaff: A Post-Hoc Approach to Safety Re-Alignment for Fine-Tuned Language Models.

Di Wu, Xin Lu, Yanyan Zhao, Bing Qin

Browse the full ACL paper archive.