Yiting Qu, Michael Backes, Yang Zhang: Bridging the Gap in Vision Language Models in Identifying Unsafe Concepts Across Modalities. Zenodo 2025