2025 05 Blur Unlearning

A new paper BLUR: A Bi-Level Optimization Approach for LLM Unlearning, joint work with Hadi, Sijia and Amazon colleagues is available. In this work, we propose a new formulation of the unlearning problem, based on a (simple) bilevel optimization, which can prioritize the unlearning capabilities while maintaining the desirable content from the LLM output. Code available here.