locuslab/fineweb_annotated

这是一个包含文本数据及其安全相关元数据的大型数据集，适用于训练文本安全评估模型。数据集分为多个配置，每个配置都包含训练集。每个样本都包括一个唯一标识符、文本内容以及安全元数据，其中安全元数据提供了不同维度的安全评分和原因。

This is a large dataset of text data and its safety-related metadata, suitable for training text safety assessment models. The dataset is divided into multiple configurations, each containing a training set. Each sample includes a unique identifier, text content, and safety metadata, which provides safety scores and reasons from different dimensions.