Explore model responses to prompts and visualize correctness