chatelet/distil-10k

Distil 10k是一个包含10k条英文自然语言提示的数据集，覆盖了广泛的领域，这些提示是由GPT-5生成的，并由人类进行了审核。该数据集主要用于将大型模型压缩成更小的模型。数据集包含的领域有创意写作、代码生成、数学问题解决、翻译、推理与逻辑、科学解释、历史分析、角色扮演与对话模拟、观点与辩论、指导手册、伦理与道德、商业与战略、技术文档、教育评估和幽默与娱乐等。

Distil 10k is a dataset of 10k English natural language prompts across a wide range of domains, synthetically generated by GPT-5 and reviewed by humans, primarily intended for distillation of large models into smaller ones. The dataset covers domains such as Creative Writing, Code Generation, Mathematical Problem Solving, Translation, Reasoning & Logic, Scientific Explanation, Historical Analysis, Roleplay & Dialogue Simulation, Opinion & Debate, Instructional Guides, Ethics & Morality, Business & Strategy, Technical Documentation, Educational Assessment, and Humour & Entertainment.