WikiHint is a human-annotated dataset designed for automatic hint generation and ranking for factoid questions. This dataset, based on Wikipedia, contains 5,000 hints for 1,000 questions and supports research in hint evaluation, ranking, and generation.
1,000 questions with 5,000 manually created hints.
Hints ranked by human annotators based on helpfulness.
Evaluated using LLMs (LLaMA, GPT-4) and human performance studies.
Supports hint ranking and automatic hint… See the full description on the dataset page:
https://huggingface.co/datasets/JamshidJDMY/WikiHint.
