▲ | pera 6 days ago | |
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 |