Modality-Aware Neuron Pruning for Unlearning in Multimodal Large Language Models

TL;DR

This paper introduces MANU, a framework for selectively unlearning sensitive information in multimodal large language models by pruning neurons based on their importance to specific modalities, enhancing privacy without sacrificing model utility.

Contribution

The paper presents a novel modality-aware neuron unlearning method that effectively isolates and removes modality-specific knowledge in MLLMs, addressing the challenge of entangled multimodal information.

Findings

MANU achieves balanced unlearning across modalities.

Selective neuron pruning preserves overall model utility.

Effective removal of targeted sensitive information.

Abstract

Generative models such as Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) trained on massive datasets can lead them to memorize and inadvertently reveal sensitive information, raising ethical and privacy concerns. While some prior works have explored this issue in the context of LLMs, it presents a unique challenge for MLLMs due to the entangled nature of knowledge across modalities, making comprehensive unlearning more difficult. To address this challenge, we propose Modality Aware Neuron Unlearning (MANU), a novel unlearning framework for MLLMs designed to selectively clip neurons based on their relative importance to the targeted forget data, curated for different modalities. Specifically, MANU consists of two stages: important neuron selection and selective pruning. The first stage identifies and collects the most influential neurons across modalities relative to the targeted forget knowledge, while the second stage is dedicated to pruning those selected neurons. MANU effectively isolates and removes the neurons that contribute most to the forget data within each modality, while preserving the integrity of retained knowledge. Our experiments conducted across various MLLM architectures illustrate that MANU can achieve a more balanced and comprehensive unlearning in each modality without largely affecting the overall model utility.