BRIDGE
收藏资源简介:
BRIDGE是一个多语言的大语言模型基准,包含来自现实世界临床数据源的87个任务,涵盖了9种语言和超过一百万个样本。该数据集旨在评估大语言模型在多语言、真实世界临床文本理解方面的性能。数据集内容来源于电子健康记录(EHR)笔记或临床案例报告,以及真实世界的在线医患咨询记录。数据集创建过程基于对全球临床文本资源的系统审查。该数据集的应用领域包括临床决策支持、患者信息解读、个性化预防建议等,旨在解决医疗领域文本理解和处理的问题。
BRIDGE is a multilingual benchmark for large language models (LLMs) that encompasses 87 tasks derived from real-world clinical data sources, covering 9 languages and over one million samples. This benchmark is designed to evaluate the performance of LLMs in multilingual, real-world clinical text understanding. The dataset content is sourced from electronic health record (EHR) notes, clinical case reports, and real-world online patient-physician consultation records. The development of this benchmark was based on a systematic review of global clinical text resources. Its application areas include clinical decision support, patient information interpretation, personalized preventive care recommendations, and more, aiming to address the challenges of text understanding and processing in the healthcare domain.




