Protecting multimodal large language models against misleading visualizations

Open in new window