Instruction-following data
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/OFA-Sys/ExpertLLaMA
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是由GPT-3.5生成的一套新的遵循指令数据，旨在提升对齐大型语言模型在回答问题时的质量。此外，该数据集采用了基于GPT4的评估方法来评估所生成答案的质量。该数据集的规模为52k条指令，其任务是对指令遵循。
This dataset is a new set of instruction-following data generated by GPT-3.5, designed to improve the response quality of aligned large language models. Additionally, it adopts a GPT-4-based evaluation method to assess the quality of the generated answers. This dataset contains 52k instruction instances, with its core task being instruction following.
提供机构：
OFA-Sys
