NurValues
收藏资源简介:
NurValues是一个基于真实世界护理行为的评估基准,涵盖了五个核心价值维度:利他主义、人性尊严、诚信、公正和专业精神。数据集由1100个真实世界的护理行为实例组成,这些实例通过五个多月的纵向田野研究在三个不同级别的医院中收集。这些实例由五名临床护士进行标注,并使用LLM生成的反事实版本进行增强,每个原始案例都与一个价值对齐和一个价值违反版本配对,形成了2200个标签实例的Easy-Level数据集。为了增加对抗性复杂性,每个实例进一步转化为基于对话的格式,其中嵌入上下文线索和细微的误导信号,产生了Hard-Level数据集。NurValues旨在为临床环境中开发具有价值敏感性的LLM提供基础。
NurValues is an evaluation benchmark based on real-world nursing behaviors, covering five core value dimensions: altruism, human dignity, integrity, justice, and professionalism. The dataset consists of 1,100 real-world nursing behavior instances, which were collected across three hospitals of different tiers via a longitudinal field study spanning over five months. These instances were annotated by five clinical nurses and augmented with counterfactual versions generated by LLMs. Each original case is paired with one value-aligned and one value-violating variant, forming the Easy-Level dataset with 2,200 labeled instances. To increase adversarial complexity, each instance was further converted into a dialogue-based format embedded with contextual cues and subtle misleading signals, resulting in the Hard-Level dataset. NurValues aims to provide a foundational resource for developing value-sensitive LLMs in clinical settings.
NurValues Benchmark 数据集概述
基本信息
- 许可证: CC-BY-NC-4.0
- 任务类别: 文本分类
- 语言: 英语 (en)、中文 (zh)
- 标签: 医学 (medical)
- 数据规模: 1K<n<10K
数据集内容
数据集级别
- Easy-Level 数据集:
text_CN: 简单案例(中文)text_EN: 简单案例(英文)
- Hard-Level 数据集:
complicated_text_CN: 复杂对话(中文)complicated_text_EN: 复杂对话(英文)
通用列
index: 样本索引Nursing_Value: 相关护理价值观(包括 Altruism、Human_Dignity、Integrity、Justice 和 Professionalism)Alignment: 护士行为是否与护理价值观对齐0: 未对齐1: 对齐
使用方式
Python import pandas as pd df = pd.read_csv("nursing_value_CN+EN.csv")
相关文献与引用
- 标题: NurValues: Real-World Nursing Values Evaluation for Large Language Models in Clinical Context
- arXiv链接: https://arxiv.org/abs/2505.08734
- 引用格式: bibtex @misc{yao2025nurvaluesrealworldnursingvalues, title={NurValues: Real-World Nursing Values Evaluation for Large Language Models in Clinical Context}, author={Ben Yao and Qiuchi Li and Yazhou Zhang and Siyu Yang and Bohan Zhang and Prayag Tiwari and Jing Qin}, year={2025}, eprint={2505.08734}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2505.08734}, }
其他资源
- GitHub仓库: https://github.com/BenYyyyyy/NurValues

- 1NurValues: Real-World Nursing Values Evaluation for Large Language Models in Clinical Context香港理工大学; 哥本哈根大学; 天津大学; 华中科技大学同济医学院附属同济医院; 哈尔姆斯塔德大学 · 2025年



