"A vision-language model that detect toxic memes online and generates informed rebuttals." - LLMs Tools for Public Discourse