M3KE
收藏资源简介:
M3KE是一个针对中文大型语言模型的知识评估基准,包含20,477个问题,覆盖从小学到大学的所有主要教育水平,涵盖人文、历史、政治、法律、教育、心理学、科学、技术、艺术和宗教等多个学科。该数据集旨在通过零样本和少样本设置测试中文大型语言模型的多任务准确性,以衡量其知识获取能力。数据集中的所有问题均为四选项选择题,确保评估过程的标准化和统一。
M3KE is a knowledge evaluation benchmark tailored for Chinese large language models, which consists of 20,477 questions covering all major educational levels from primary school to university and spanning multiple disciplines including humanities, history, politics, law, education, psychology, science, technology, art, and religion. This benchmark is designed to test the multi-task accuracy of Chinese large language models under zero-shot and few-shot settings, so as to measure their knowledge acquisition capabilities. All questions in this dataset are four-option multiple-choice questions, ensuring the standardization and consistency of the evaluation process.

- 1M3KE: A Massive Multi-Level Multi-Subject Knowledge Evaluation Benchmark for Chinese Large Language Models天津大学智能与计算学部 · 2023年



