McGill-NLP/CHASE-QA
收藏Hugging Face2025-02-21 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/McGill-NLP/CHASE-QA
下载链接
链接失效反馈官方服务：
资源简介：
CHASE数据集是一个创新的评估框架，旨在通过合成的方式生成挑战性问题，以便对大型语言模型进行严格的评估。该数据集涵盖了文档问答、代码补全和数学推理三个领域，能够自动生成具有挑战性的问题，无需人工参与，从而提高了评估的效率和准确性。
CHASE is an innovative evaluation framework designed to generate challenging problems synthetically for rigorous assessment of large language models. The dataset covers three domains: document-based question answering, code completion at the repository level, and math reasoning, automating the creation of challenging problems without human involvement, thus enhancing the efficiency and accuracy of evaluation.
提供机构：
McGill-NLP
