FineRob
收藏资源简介:
FineRob数据集由中国科学院信息工程研究所创建,专注于社交媒体上的细粒度用户行为模拟。该数据集收集了来自Twitter、Reddit和Zhihu三个平台的1866名用户的完整行为历史,包含78600条细粒度行为模拟问答记录。数据集的创建过程包括从这些平台收集用户行为数据,并将其分解为对象、类型和内容三个细粒度元素。FineRob数据集主要用于评估大型语言模型在模拟角色特定行为方面的能力,旨在解决在复杂现实场景中准确模拟用户行为的问题。
FineRob Dataset was created by the Institute of Information Engineering, Chinese Academy of Sciences, focusing on fine-grained user behavior simulation on social media. It collects complete behavioral histories of 1866 users from three platforms: Twitter, Reddit and Zhihu, containing 78,600 fine-grained behavioral simulation question-and-answer records. The dataset creation process includes collecting user behavioral data from these platforms and decomposing it into three fine-grained elements: object, type and content. The FineRob Dataset is primarily used to evaluate the capabilities of large language models in simulating role-specific behaviors, aiming to address the challenge of accurately simulating user behaviors in complex real-world scenarios.




