Palliative Care Adversarial Dataset (PCAD)
收藏资源简介:
Palliative Care Adversarial Dataset(PCAD)是由伦敦大学国王学院和加拿大蒙特利尔北岛大学社会卫生科学中心开发的两个对抗性数据集,旨在评估大型语言模型在缓和医疗中的偏见。该数据集包含100个对抗性问题(PCAD-Direct)和84对反事实场景(PCAD-Counterfactual),针对四个护理维度和三个身份轴设计,以揭示大型语言模型在缓和医疗中的偏见和公平性问题。
Palliative Care Adversarial Dataset (PCAD) is a pair of adversarial datasets developed by King's College London and the Centre for Social and Health Sciences at North Island University, Montreal, Canada. It is designed to evaluate biases of large language models (LLMs) in palliative care. This dataset contains 100 adversarial questions (PCAD-Direct) and 84 pairs of counterfactual scenarios (PCAD-Counterfactual), which are formulated based on four care dimensions and three identity axes to uncover biases and fairness-related issues of large language models in palliative care.




