Protecting multimodal large language models against misleading visualizations

Re-assign community