Korean National Educational Test Benchmark (KoNET)
收藏资源简介:
KoNET是一个基于韩国国家教育考试的新基准数据集,旨在评估多模态生成AI系统。该数据集包括韩国小学、初中、高中和大学入学考试四种考试,提供了详细的题目难度分析,能够对AI在不同教育水平上的能力进行细微评估。KoNET的独特之处在于,它专注于韩国语言,提供了对非英语和东亚语言环境中AI性能的见解。数据集的构建过程包括从韩国课程评价院公开发布的官方PDF中解析问题。KoNET的应用领域主要是教育技术,特别是在AI辅导市场,它可以帮助评估和改进多模态生成AI在教育场景中的表现。
KoNET is a novel benchmark dataset based on the Korean National Education Examinations, designed to evaluate multimodal generative AI systems. This dataset covers four types of examinations in the Korean education system: primary school, middle school, high school, and college entrance examinations, and provides detailed question difficulty analysis to enable fine-grained assessment of AI capabilities across different educational levels. What distinguishes KoNET is its exclusive focus on the Korean language, offering insights into AI performance in non-English and East Asian language contexts. The dataset’s construction process involves extracting questions from official publicly released PDFs issued by the Korea Institute for Curriculum and Evaluation. The main application domains of KoNET are educational technology, particularly the AI tutoring market, where it can assist in evaluating and enhancing the performance of multimodal generative AI in educational scenarios.

- 1Evaluating Multimodal Generative AI with Korean Educational StandardsNAVER Cloud AI · 2025年



