Selective Forgetting for Large Reasoning Models: Precisely Erasing Sensitive Knowledge While Preserving Reasoning Ability

Introduction / Main Floor: Selective Forgetting for Large Reasoning Models: Precisely Erasing Sensitive Knowledge While Preserving Reasoning Ability

This article introduces a new selective forgetting framework for Large Reasoning Models (LRMs) that can remove sensitive reasoning content while preserving general reasoning ability, addressing the limitations of traditional machine forgetting methods on reasoning models.