CSQ: A Chinese Elementary Science Question Dataset with Rich Discipline Properties in Adaptive Problem-Solving Process Generation

This is currently the world's largest Chinese Science Question (CSQ) dataset, which includes benchmarks and training sets and is designed to evaluate and improve the scientific problem-solving ability of LLMs. CSQ consists of 12,000 high-quality samples with a variety of question types and different subject attributes, covering four subjects and multiple topics in Chinese primary schools. It is deeply coupled with the Science Curriculum Standards for Compulsory Education of China (2022), providing a new way for large language models to empower science education, and also providing a research foundation for science curriculum ITS based on LLMs.

本数据集为目前全球规模最大的中文科学试题（Chinese Science Question，CSQ）数据集，涵盖基准测试集与训练集，旨在评估并提升大语言模型（Large Language Model，LLM）的科学解题能力。该数据集包含12000条高质量样本，具备多样题型与多元学科属性，覆盖中国小学阶段的四门学科及多个知识点主题。其深度契合《中国义务教育科学课程标准（2022年版）》，既为大语言模型赋能科学教育提供了全新路径，也为基于大语言模型的科学课程ITS研究奠定了科研基础。

搜集汇总

背景与挑战

背景概述

CSQ是一个包含12,000个高质量样本的中文小学科学问题数据集，涵盖四个学科和多个主题，旨在评估和提升大型语言模型（LLMs）的科学问题解决能力。该数据集与《中国义务教育科学课程标准（2022年版）》深度结合，为科学教育提供了新的研究基础。