#artificial intelligence
Dec 2025
SGM: Safety Glasses for Multimodal Large Language Models via Neuron-Level Detoxification
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.
Hongbo Wang, Maungmaung Aprilpyone, Isao Echizen
· arXiv.org · 1 citation