There is currently no effective method for unlearning information - specially not when you don't have access to the original training datasets (as is the case w

There is currently no effective method for unlearning information - specially not when you don't have access to the original training datasets (as is the case with open weight models), see: Rethinking Machine Unlearning for Large Language Models https://arxiv.org/html/2402.08787v6 https://arxiv.org/html/2402.08787v6