SCITAB
收藏arXiv2023-10-23 更新2024-06-21 收录
下载链接：
https://github.com/XinyuanLu00/SciTab
下载链接
链接失效反馈官方服务：
资源简介：
SCITAB是一个专为科学事实验证设计的挑战性评估数据集，包含1225条由专家验证的科学声明，这些声明源自真实的科学出版物，并需要通过组合推理进行验证。数据集中的声明与包含证据的科学表格配对，并进行了标注。SCITAB数据集通过模拟真实科学事实验证的需求，提供了更全面和细致的科学推理表示。该数据集主要用于评估和推动科学事实验证领域的发展，特别是在处理复杂声明和基于表格的证据方面。
SCITAB is a challenging evaluation dataset designed specifically for scientific fact verification. It contains 1,225 expert-validated scientific claims derived from real scientific publications, which require verification through compositional reasoning. The claims in the dataset are paired with annotated scientific tables that contain supporting evidence. The SCITAB dataset provides more comprehensive and fine-grained representations of scientific reasoning by simulating the real-world demands of scientific fact verification. This dataset is primarily used to evaluate and advance the development of the scientific fact verification field, particularly for tasks involving complex claims and table-based evidence.
提供机构：
新加坡国立大学
创建时间：
2023-05-23
