Causality in Script Knowledge (CSK)
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/tony-hong/causal-script
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为CSK，基于DeScript数据集构建，包含了21个描述日常活动的英文故事。每个故事都包含事件A和事件B，并设有特定条件以测试因果推断。数据集涵盖了不同的故事条件（A导致B，非A导致B，无原因导致B），旨在探究人类和大型语言模型如何处理因果关系。该数据集涉及多参与者的21个故事，其任务是研究阅读理解中的因果推断。
This dataset, named CSK, is constructed based on the DeScript dataset. It contains 21 English stories describing daily activities. Each story includes events A and B, with specific conditions designed to test causal inference. The dataset covers diverse story conditions: A causes B, non-A causes B, and B occurs without a cause. It aims to explore how humans and large language models (LLMs) handle causal relationships. Comprising 21 stories involving multiple participants, this dataset is developed to investigate causal inference in reading comprehension.
