ellamind/summary_texts_new_extract_instructs

该数据集是基于ellamind/summary_texts_new_extract数据集扩展而来，为每篇文本添加了3条使用OpenAI的GPT-4.1-nano模型生成的自定义指令。数据集包含原始字段（文本、令牌数、日期、哈希和其他元数据）以及新增字段（3条自定义指令）。指令生成考虑了内容相关性、多样性、实用性、变化性，并支持多种语言。数据集可用于微调指令跟随模型、评估模型遵循多样化指令的能力、数据增强以及指令分析。

This dataset is an extension of ellamind/summary_texts_new_extract, adding 3 custom instructions per text generated using OpenAIs GPT-4.1-nano model. It includes original fields (text, tokens, date, hash, and other metadata) and new fields (3 custom instructions). The instruction generation considers content-awareness, diversity, practicality, and variability, and supports multiple languages. The dataset can be used for fine-tuning instruction-following models, testing a models ability to follow varied instructions, data augmentation, and instruction analysis.