DataSet

A benchmark for multi-dimensional question generation evaluation, which consists of 200 instances from SQuAD and HotpotQA, each instance contains 15 questions generated by 15 different QG models.

Evalutaion dimensions:

• fluency
• clarity
• conciseness
• relevance
• consistency
• answerability
• answer consistency

Models

Trained QG models used for generating questions to be evaluated.