Distillation Robustifies Unlearning

Neural Information Processing Systems 

Current LLM unlearning methods are not robust.