中文开放指令通用程序员：初步发布（英文版）

英文标题：Chinese Open Instruction Generalist: A Preliminary Release中文摘要：通过多种方法适应 4 个子任务的内在特征，我们提出了一个项目来创建中文指令数据集，收集了约 20 万个中文指令调整样本，并总结了现有的英文和中文指令语料库以及新构建的中文指令语料库的潜在应用。英文摘要：Instruction tuning is widely recognized as a key technique for buildinggeneralist language models, which comes to the attention of researchers and thepublic with the release of InstructGPT \cite{ouyang2022training} and ChatGPT [this https URL ]. Despite impressive progress in English-orientedlarge-scale language models (\textbf{L