Content-Specific Humorous Image Captioning Using Incongruity Resolution Chain-of-Thought

Download paper here

This paper proposed IRCoT, a new prompting framework that enables multi-modal large language models to generate humorous captions that are unique to each image.