e-CARE

e-CARE数据集是由哈尔滨工业大学社会计算与信息检索研究中心创建，包含超过21,000个因果推理问题，是目前已知最大的人工标注常识因果推理数据集。每个问题不仅包含因果推理本身，还提供了一个自由文本形式的概念解释，解释为什么这种因果关系存在。数据集通过众包方式构建，旨在通过提供概念解释来深入理解因果事实，促进因果推理过程。e-CARE数据集的应用领域广泛，旨在解决因果推理模型在不同数据集上表现不稳定以及易受对抗攻击的问题，通过提供概念解释来提升模型的理解能力和推理稳定性。

The e-CARE dataset was created by the Research Center for Social Computing and Information Retrieval at Harbin Institute of Technology. Containing over 21,000 causal reasoning questions, it is the largest manually annotated commonsense causal reasoning dataset known to date. Each question not only includes the causal reasoning task itself but also provides a free-text conceptual explanation that elaborates on why the given causal relationship holds. Constructed via crowdsourcing, the dataset aims to deepen the understanding of causal facts and facilitate the causal reasoning process by offering conceptual explanations. The e-CARE dataset has a wide range of application scenarios, targeting the issues of unstable performance and vulnerability to adversarial attacks of causal reasoning models across different datasets, and it seeks to enhance models' comprehension ability and reasoning stability by providing conceptual explanations.