allganize/IFEval-Ko
收藏Hugging Face2025-06-09 更新2025-05-31 收录
下载链接:
https://hf-mirror.com/datasets/allganize/IFEval-Ko
下载链接
链接失效反馈官方服务:
资源简介:
IFEval-Ko是一个韩语版本的指令遵循基准数据集,用于评估大型语言模型(LLMs)在韩语中的指令遵循能力。该数据集基于Google的开源IFEval基准,并使用lm-evaluation-harness框架进行评估。数据集由Allganize Inc. LLM团队进行改编,移除了一些不合适的任务,并对单位进行了转换和标准化处理。
IFEval-Ko is a Korean adaptation of the instruction-following benchmark for evaluating the instruction-following capabilities of large language models (LLMs) in the Korean language, based on Googles open-source IFEval benchmark and utilizing the lm-evaluation-harness framework. The dataset has been adapted by the LLM TEAM at Allganize Inc., with removal of inappropriate tasks, unit conversions, and standardizations.
提供机构:
allganize



