TRU: Targeted Reverse Update for Efficient Multimodal Recommendation Unlearning

TL;DR

This paper introduces TRU, a targeted reverse update framework for efficient unlearning in multimodal recommendation systems, addressing non-uniform data influence across model components.

Contribution

TRU provides a novel, targeted unlearning method that improves forgetting efficiency and effectiveness by focusing on model hierarchy and influence distribution.

Findings

TRU outperforms prior methods in retain-forget trade-offs.

Experiments show deeper forgetting and behavior closer to full retraining.

TRU is effective across multiple datasets and model backbones.

Abstract

Multimodal recommendation systems (MRS) jointly model user-item interaction graphs and rich item content, but this tight coupling makes user data difficult to remove once learned. Approximate machine unlearning offers an efficient alternative to full retraining, yet existing methods for MRS mainly rely on a largely uniform reverse update across the model. We show that this assumption is fundamentally mismatched to modern MRS: deleted-data influence is not uniformly distributed, but concentrated unevenly across \textit{ranking behavior}, \textit{modality branches}, and \textit{network layers}. This non-uniformity gives rise to three bottlenecks in MRS unlearning: target-item persistence in the collaborative graph, modality imbalance across feature branches, and layer-wise sensitivity in the parameter space. To address this mismatch, we propose \textbf{targeted reverse update} (TRU), a plug-and-play unlearning framework for MRS. Instead of applying a blind global reversal, TRU performs three coordinated interventions across the model hierarchy: a ranking fusion gate to suppress residual target-item influence in ranking, branch-wise modality scaling to preserve retained multimodal representations, and capacity-aware layer isolation to localize reverse updates to deletion-sensitive modules. Experiments across two representative backbones, three datasets, and three unlearning regimes show that TRU consistently achieves a better retain-forget trade-off than prior approximate baselines, while security audits further confirm deeper forgetting and behavior closer to a full retraining on the retained data.