CURE:Circuit-Aware Unlearning for LLM-based Recommendation

TL;DR

CURE introduces a circuit-aware unlearning framework for LLM-based recommendation systems, disentangling model components to improve privacy-preserving unlearning without sacrificing utility.

Contribution

The paper proposes a novel circuit-based method to selectively update model components, reducing gradient conflicts and enhancing unlearning effectiveness in LLMRec.

Findings

Achieves more effective unlearning than existing methods.

Reduces gradient conflicts during unlearning.

Maintains model utility while removing user data.

Abstract

Recent advances in large language models (LLMs) have opened new opportunities for recommender systems by enabling rich semantic understanding and reasoning about user interests and item attributes. However, as privacy regulations tighten, incorporating user data into LLM-based recommendation (LLMRec) introduces significant privacy risks, making unlearning algorithms increasingly crucial for practical deployment. Despite growing interest in LLMRec unlearning, most existing approaches formulate unlearning as a weighted combination of forgetting and retaining objectives while updating model parameters in a uniform manner. Such formulations inevitably induce gradient conflicts between the two objectives, leading to unstable optimization and resulting in either ineffective unlearning or severe degradation of model utility. Moreover, the unlearning procedure remains largely black-box, undermining its transparency and trustworthiness. To tackle these challenges, we propose CURE, a circuit-aware unlearning framework that disentangles model components into functionally distinct subsets and selectively updates them. Here, a circuit refers to a computational subgraph that is causally responsible for task-specific behaviors. Specifically, we extract the core circuits underlying item recommendation and analyze how individual modules within these circuits contribute to the forget and retain objectives. Based on this analysis, these modules are categorized into forget-specific, retain-specific, and task-shared groups, each subject to function-specific update rules to mitigate gradient conflicts during unlearning. Experiments on real-world datasets show that our approach achieves more effective unlearning than existing baselines.