Unlearning Isn't Invisible: Detecting Unlearning Traces in LLMs from Model Outputs

Open in new window