Dataset · jmhessel
New Yorker humor understanding benchmark
jmhessel/newyorker_caption_contest
Cartoons and captions from the New Yorker contest for evaluating humor understanding in models.
What's inside
Data from The New Yorker caption contest: English images and text captions, between 100k and 1M examples. Tasks include image-to-text, multiple-choice, classification, text generation, and VQA. These are benchmarks from the paper "Do Androids Laugh at Electric Sheep?"; details at capcon.dev.
What you can do with it
- Evaluate humor understanding in multimodal models
- Train fun-caption selection for images
- Test caption generation
- Tasks
- Image → text, Multiple choice, Text classification, Text generation, Visual question answering, Other
- Languages
- English
- Size
- 100K–1M rows
- Downloads this month
- 7.2K
- Likes
- 78
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("jmhessel/newyorker_caption_contest")