LHMKE

LHMKE是由天津大学智能与计算学部创建的大型综合性多学科知识评估基准，用于评估中文大型语言模型的知识获取能力。该数据集包含10,465个问题，覆盖75个任务和30个学科，从小学到专业认证考试。LHMKE不仅包括客观题，还包含主观题，以更全面地评估语言模型的知识水平。数据集中的每个问题都来源于实际的标准考试，并具有特定的分数，使得每个学科的评分系统标准化。LHMKE旨在解决现有基准在评估语言模型知识获取方面的不足，特别是在问题类型的多样性方面。

LHMKE is a large-scale comprehensive multi-disciplinary knowledge evaluation benchmark developed by the College of Intelligence and Computing, Tianjin University, for assessing the knowledge acquisition capabilities of Chinese large language models. This dataset contains 10,465 questions spanning 75 tasks across 30 disciplines, ranging from primary school level to professional certification examinations. LHMKE includes both objective and subjective questions to enable a more comprehensive evaluation of the knowledge proficiency of language models. Every question in the dataset is derived from actual standardized examinations and is assigned a specific score, which standardizes the scoring system for each discipline. LHMKE is designed to address the limitations of existing benchmarks in evaluating the knowledge acquisition of language models, especially regarding the diversity of question types.

搜集汇总

背景与挑战

背景概述

LHMKE是一个用于评估中文大型语言模型的大规模、综合性、多学科知识评估基准。它旨在全面测试模型在多个学科领域的知识掌握情况，具有规模大和覆盖范围广的特点。