Norwegian Medical Question Answering Dataset - NorMedQA

This benchmark dataset consists of 1401 medical question-and-answer pairs primarily in Norwegian (Bokmål and Nynorsk), designed for evaluating Large Language Models (LLMs). The content originates from publicly available sources containing medical exam questions and has undergone cleaning and preprocessing. The dataset is structured in JSON format, with each record containing the source document name, question number (where available), the question text, and the reference answer text and the wrong answers text if the answer was multiple choice. It is suitable for use within evaluation frameworks such as lm-evaluation-harness (Github with config and code example: https://github.com/kelkalot/normedqa)to assess model capabilities in medical knowledge retrieval and reasoning specific to the Norwegian context.

本基准数据集共计收录1401组医疗问答对，主要采用挪威语（书面挪威语（Bokmål）与新挪威语（Nynorsk）），专为评估大语言模型（Large Language Models，LLMs）而设计。其内容源自公开可获取的医疗考试试题资源，已完成清洗与预处理流程。数据集以JSON格式构建，每条记录包含来源文档名称、（如有提供的）问题编号、问题文本与参考答案文本；若题型为选择题，则额外包含错误选项文本。该数据集可适配lm-evaluation-harness（附带配置与代码示例的GitHub仓库：https://github.com/kelkalot/normedqa）等评估框架，用于测评模型在挪威语语境下的医疗知识检索与推理能力。