AI RESEARCH

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

arXiv CS.LG

ArXi:2506.14003v5 Announce Type: replace Machine unlearning (MU) for large language models (LLMs), commonly referred to as LLM unlearning, seeks to remove specific undesirable data or knowledge from a trained model, while maintaining its performance on standard tasks. While unlearning plays a vital role in protecting data privacy, enforcing