FairytaleQA is a dataset of fairytale excerpts with human-generated QA pairs. Each question is categorized as falling into one of seven narrative elements and classified as an explicit or implicit answer.

This dataset is useful for assessing the reasoning ability of LLMs to answer questions based on text.

Permissions and licensing

License

Apache 2.0

Key restrictions

No restrictions.

How to access the data

This dataset is available to anyone. To access the data, head to the Data download tab and follow the instructions provided.