RAIA-BRASIL/agent_eval_questions

数据集名称为Questões geradas com Agents em Português（葡萄牙语生成的代理问题），由 Álvaro Lopes、Artur de Vlieger、Fabrício Salomon、Leticia Bossatto Marchezi、Luis Felipe Jorge和Otávio Coletti共同开发，赞助商为Pico。数据集包含的问题少于1000个，标签包括代理、问题、 vestibular、葡萄牙语等。该数据集用于比较根据代理策略和提示工程生成的LLM问题的差异。数据集结构包括用户请求、基线模型响应、真实ENEM问题、few-shot代理响应和配对代理响应等字段。数据集的创建是为了改善使用人工智能创建问题的过程。

The dataset is named Questões geradas com Agents em Português (Portuguese-generated Agent Questions), developed by Álvaro Lopes, Artur de Vlieger, Fabrício Salomon, Leticia Bossatto Marchezi, Luis Felipe Jorge, and Otávio Coletti, sponsored by Pico. The dataset contains less than 1000 questions, with tags including agents, questions, vestibular, Portuguese, etc. It is used to compare the differences between LLM questions generated according to agent strategies and prompt engineering. The dataset structure includes fields such as user request, baseline model response, real ENEM questions, few-shot agent response, and paired agent response. The dataset was created to improve the process of creating questions using artificial intelligence.