COIG-CQIA 高质量中文指令微调数据集

COIG-CQIA 全称为 Chinese Open Instruction Generalist – Quality is All You Need， 是一个开源的高质量指令微调数据集，旨在为中文 NLP 社区提供高质量且符合人类交互行为的指令微调数据。 COIG-CQIA 以中文互联网获取到的问答及文章作为原始数据，经过深度清洗、重构及人工审核构建而成。

COIG-CQIA, whose full name is Chinese Open Instruction Generalist – Quality is All You Need, is an open-source high-quality instruction tuning dataset designed to provide high-quality instruction tuning data that aligns with human interactive behaviors for the Chinese natural language processing (NLP) community. It is constructed using raw Q&A pairs and articles sourced from Chinese internet platforms, and undergoes thorough data cleaning, restructuring and manual review during its development.

搜集汇总

背景与挑战

背景概述

COIG-CQIA是一个开源的高质量中文指令微调数据集，全称为Chinese Open Instruction Generalist – Quality is All You Need，旨在为中文NLP社区提供符合人类交互行为的指令微调数据。该数据集从中文互联网的多个来源（如问答社区、维基平台、考试资料和现有NLP数据集）收集原始数据，经过深度清洗、重构和人工审核，注重数据的质量、多样性和真实交互模式，适用于大语言模型的训练和微调。