SGM: Safety Glasses for Multimodal Large Language Models via Neuron-Level Detoxification

ACL ARR 2026 January Submission795 Authors

Keywords: multimodal large language model, toxic language, detoxification, expert neuron, white-box intervention

Abstract:

Disclaimer: Samples in this paper may be harmful and cause discomfort.

Multimodal large language models (MLLMs) enable multimodal understanding but inherit toxic, biased, and NSFW signals from weakly curated pretraining corpora, causing safety risks, especially under adversarial triggers that late, opaque training-free detoxification methods struggle to handle. We propose SGM, a white-box neuron-level multimodal intervention that acts like safety glasses for toxic neurons: it selectively recalibrates a small set of toxic expert neurons via expertise-weighted soft suppression, neutralizing harmful cross-modal activations without any parameter updates. We establish MM-TOXIC-QA, a multimodal toxicity evaluation framework, and compare SGM with existing detoxification techniques. Experiments on open-source MLLMs show that SGM mitigates toxicity in standard and adversarial conditions, cutting harmful rates from $48.2%$ to $2.5%$ while preserving fluency and multimodal reasoning. SGM is extensible, and its combined defenses, denoted as SGM*, integrate with existing detoxification methods for stronger safety performance, providing an interpretable, low-cost solution for toxicity-controlled multimodal generation.

Paper Type: Long

Research Area: Safety and Alignment in LLMs

Research Area Keywords: safety and alignment, multimodality, bias/toxicity

Contribution Types: Model analysis & interpretability, Approaches low compute settings-efficiency, Data resources

Languages Studied: English

Submission Number: 795