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

Open in new window