Methodology
How Humansplain benchmarks vision-language models on explaining humor.
Humansplain v1
Tell me why this is funny in one sentence. Use simple and easy words. Get directly into the main reason. Drop any prefacing like "this is funny because" or "the joke is" or "the punchline is". Use plain text only.
All models receive the same image and this prompt. Responses are shown as options A/B/C/D (randomly assigned). Users multiselect the answers closest to why it's funny, or choose "None" and humansplain their own explanation.