RKLD: Reverse KL-Divergence-based Knowledge Distillation for Unlearning Personal Information in Large Language Models
This addresses the challenge of balancing forget quality and model utility in LLMs for personal data unlearning, which is crucial for regulatory compliance, though it appears incremental as it builds on existing unlearning methods.
The paper tackles the problem of unlearning personal information in large language models (LLMs) to comply with regulations like the Right to Be Forgotten, proposing RKLD, a reverse KL-divergence-based knowledge distillation method that achieves significant forget quality while effectively maintaining model utility in experiments.
With the passage of the Right to Be Forgotten (RTBF) regulations and the scaling up of language model training datasets, research on model unlearning in large language models (LLMs) has become more crucial. Before the era of LLMs, machine unlearning research focused mainly on classification tasks in models with small parameters. In these tasks, the content to be forgotten or retained is clear and straightforward. However, as parameter sizes have grown and tasks have become more complex, balancing forget quality and model utility has become more challenging, especially in scenarios involving personal data instead of classification results. Existing methods based on gradient ascent and its variants often struggle with this balance, leading to unintended information loss or partial forgetting. To address this challenge, we propose RKLD, a novel \textbf{R}everse \textbf{KL}-Divergence-based Knowledge \textbf{D}istillation unlearning algorithm for LLMs targeting the unlearning of personal information. Through RKLD, we achieve significant forget quality and effectively maintain the model utility in our experiments.