CAMEL

The CAMEL dataset features an extensive collection of around 584K instructions, among which 107K have been translated into multiple languages. This dataset provides a wide array of dialogue resources covering both multilingual and code domains. The datasets introduces a communication agent framework called “role-playing” generated through three types of prompts and involving two contexts: “AI Society” and “Programming.

CAMEL数据集（CAMEL dataset）拥有规模庞大的指令集合，共计约58.4万条，其中10.7万条已被翻译为多语言版本。该数据集提供覆盖多语言与代码领域的丰富对话资源。本数据集引入了一种名为「角色扮演（role-playing）」的通信智能体框架，该框架通过三类提示词生成，涵盖"AI Society"与"Programming"两类上下文。