BAAI/IndustryCorpus_finance

该数据集是一个高质量的多行业分类中英文预训练数据集，通过22个行业数据处理操作符从超过100TB的开源数据集中清洗和过滤出3.4TB的高质量数据，包括1TB的中文数据和2.4TB的英文数据。中文数据被标注了12种标签，如字母数字比例、平均行长度、语言置信度分数等。数据集涵盖了18个行业类别，如医疗、教育、文学、金融等。为了验证数据集的性能，作者在医疗行业示范模型上进行了持续预训练、SFT和DPO训练，结果显示客观性能提高了20%，主观胜率为82%。数据集还通过基于规则的过滤、基于模型的过滤和数据去重等步骤进一步优化。

The dataset contains 3.4TB of high-quality multi-industry classified Chinese and English pre-training data, which has been cleaned and filtered through 22 industry data processing operators, resulting in 1TB of Chinese data and 2.4TB of English data. The dataset covers 18 industry categories and has annotated the Chinese data with 12 types of labels, including alphanumeric ratio, average line length, language confidence score, maximum line length, and perplexity. Additionally, the dataset has undergone model validation, showing a 20% performance improvement on a medical industry demonstration model.