BEHAVIORCHAIN
收藏资源简介:
BEHAVIORCHAIN是由北京大学等机构创建的高质量、多样化的人格化行为链数据集,包含101个独立人物的行为链,共15846个独立行为。数据集通过从小说和传记文学中自动提取行为链构建而成,旨在为研究连续人类行为提供有价值的模拟环境。该数据集适用于评估大型语言模型模拟连续人类行为的能力,特别是在动态场景下的行为预测和生成任务。
BEHAVIORCHAIN is a high-quality, diverse personalized behavior chain dataset created by Peking University and other institutions, which includes behavior chains of 101 distinct individuals with a total of 15,846 independent behaviors. The dataset is constructed by automatically extracting behavior chains from novels and biographical literature, aiming to provide a valuable simulation environment for research on continuous human behavior. This dataset is suitable for evaluating the ability of large language models (LLMs) to simulate continuous human behavior, particularly for behavior prediction and generation tasks in dynamic scenarios.




