opendatalab/SlimPajama-Meta-rater
收藏官方服务：
资源简介：
这个数据集包含了大约5800亿个标记，并跨越了25个不同的质量维度。数据集旨在为数据为中心的大型语言模型研究提供支持，并包括自然语言质量信号、数据重要性分数和基于模型的质量评级等指标。数据集被划分为不同的领域，并完全注释，为研究人员和开发人员提供了一个全面的资源。
This dataset contains approximately 580 billion tokens annotated across 25 different quality dimensions. It is designed for data-centric large language model research and includes metrics such as natural language quality signals, data importance scores, and model-based quality ratings. The dataset is split into different domains and is fully annotated, providing a comprehensive resource for researchers and developers.
提供机构：
opendatalab
