IndicInstruct
收藏arXiv2025-09-30 收录
下载链接：
https://huggingface.co/datasets/ai4bharat/indic-instruct-data-v0.1
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一组多样化的针对印地语的指令调优数据集，通过将高质量的英语监督指令调优数据集翻译成印地语而创建。为了确保高质量，翻译后的印地语示例经过了筛选，要求其chrF++分数达到50分以上。数据集规模包含38.5万个示例，旨在为印地语语言模型进行指令调优任务。
This dataset is a diverse instruction tuning dataset tailored for Hindi, constructed by translating high-quality English supervised instruction tuning datasets into Hindi. To ensure high data quality, the translated Hindi examples were filtered with a minimum chrF++ score of 50. Containing 385,000 instances in total, this dataset is designed for instruction tuning tasks of Hindi language models.
提供机构：
ai4bharat
