QuanTemp

该数据集是一个专注于数值性声明的多样化多领域集合，它包含了时间序列、统计以及其他多种细粒度方面的数据，配备了详尽的元数据和无泄露的证据收集。此外，该数据集在声明分布上并不均衡，真实声明占18.79%，虚假声明占57.93%，冲突声明占23.27%。数据集还包括了由423,320个片段组成的全面证据收集。在规模上，该数据集包含了15,514个声明，分为训练集（9935个）、验证集（3084个）和测试集（2495个）。其任务是验证数值性声明。

This dataset is a diverse multi-domain collection focused on numerical claims. It incorporates data from time series, statistics, and various other fine-grained aspects, and features comprehensive metadata and evidence collection free of data leakage. Moreover, the distribution of claims in this dataset is imbalanced: true claims account for 18.79%, false claims account for 57.93%, and conflicting claims account for 23.27%. The dataset also includes a comprehensive evidence collection composed of 423,320 snippets. In terms of scale, this dataset contains 15,514 claims, which are split into the training set (9,935 samples), validation set (3,084 samples), and test set (2,495 samples). The core task of this dataset is to verify numerical claims.