ECQA

该数据集（ECQA）为用户提供了针对CSQA数据集中未选中选项的反事实解释。此外，ECQA被用于对大型语言模型生成的理由和自由形式的人类生成的理由进行直接比较。该数据集通过众包方式为CSQA中的实例提供了解释，其任务是评估理由生成的效果。

This dataset (ECQA) provides users with counterfactual explanations for the unselected options within the CSQA dataset. Additionally, ECQA is utilized to directly compare rationales generated by large language models (LLMs) with free-form human-generated rationales. Developed via crowdsourcing to deliver explanations for instances in CSQA, this dataset is intended to evaluate the effectiveness of rationale generation.