ozoneproject/news-prompts

一个精心策划的数据集，包含800个真实的用户提示，用于评估AI系统在发布者内容上的表现。每个提示代表一个真实用户在寻求新闻或信息时可能提出的问题。这些提示旨在测试RAG（检索增强生成）系统检索和利用发布者文章的能力。数据集包含68个类别和8种文章类型，每种类型有100个提示。文章类型包括硬新闻、软新闻、分析、观点、数据研究、产品、操作指南和聚合内容。

A curated dataset of 800 realistic user prompts for evaluating AI systems against publisher content. Each prompt represents a question a real user might ask when seeking news or information. The prompts are designed to test how well RAG (Retrieval-Augmented Generation) systems can retrieve and utilize publisher articles. The dataset includes 68 categories and 8 article types, with 100 prompts each. Article types include hard_news, soft_news, analysis, opinion, data_research, product, how_to, and aggregation.