CLIcK 韩语文化与语言智能数据集

CLIcK 数据集由韩国科学技术院创建，旨在填补韩语大模型在文化与语言知识评估方面的空白。该数据集包含 1995 对问答样本，源自官方韩国考试与教科书，覆盖语言和文化两大类，细分为 11 个子类别，每个样本都提供了细粒度的注释，指明回答问题所需的文化和语言知识。

The CLIcK dataset, developed by the Korea Advanced Institute of Science and Technology (KAIST), is designed to address the gap in cultural and linguistic knowledge evaluation for Korean large language models (LLMs). This dataset contains 1995 question-answer pairs derived from official Korean examinations and textbooks, covering two main categories: language and culture, which are further subdivided into 11 subcategories. Each sample provides fine-grained annotations indicating the cultural and linguistic knowledge required to answer the corresponding question.

搜集汇总

背景与挑战

背景概述

CLIcK 韩语文化与语言智能数据集由韩国科学技术院创建，包含1995对源自韩国考试、教科书及GPT-4生成的问答样本，覆盖语言与文化两大类的11个子类别，并提供细粒度注释。该数据集经过严格人工审核，旨在评估韩语大模型的文化与语言理解能力，为相关研究提供高质量基准。