Postdoc, University of Cambridge
2 papers at NeurIPS 2025
We propose LUNAR, an novel LLM unlearning method by redirecting the representation of unlearned data for effective, controlled and efficient unlearning.
