CMoralEval
收藏资源简介:
CMoralEval数据集由天津大学智能与计算学部等机构创建,旨在评估中文大型语言模型的道德表现。该数据集包含30,388条数据,来源于中国法律与伦理电视节目和一系列道德异常案例,覆盖家庭道德、社会道德、职业伦理、网络伦理和个人道德五个类别。数据集通过人工标注和AI辅助生成,确保了数据的多样性和真实性,适用于研究模型在道德决策中的表现。
The CMoralEval dataset was developed by institutions including the College of Intelligence and Computing at Tianjin University, with the objective of evaluating the moral performance of Chinese large language models. This dataset comprises 30,388 instances, sourced from Chinese legal and ethical television programs and a collection of moral anomaly cases, covering five categories: family morality, social morality, professional ethics, cyber ethics, and personal ethics. The dataset was constructed through manual annotation and AI-assisted generation to ensure its diversity and authenticity, making it suitable for investigating models' performance in moral decision-making scenarios.




