Invariance Makes LLM Unlearning Resilient Even to Unanticipated Downstream Fine-Tuning.

Changsheng Wang, Yihua Zhang, Jinghan Jia, Parikshit Ram, Dennis Wei, Yuguang Yao, Soumyadeep Pal, Nathalie Baracaldo, Sijia Liu

Browse the full ICML paper archive.