LLMs

Red Teaming Multimodal Language Models: Evaluating Harm Across Prompt Modalities and Models

Evaluation of harmlessness of four leading Multimodal LLMs. Preprint. September 2025.