KokushiMD-10

该数据集名为KokushiMD-10，是一个多模态的基准数据集，它由日本的十项国家医疗执照考试构建而成，覆盖了医学、牙医学、护理学、药学以及相关健康专业等多个领域。该数据集包含了超过11,588个真实的考试题目，这些题目配有临床图片和专家标注的解题依据。此外，该数据集还评估了超过30个最先进的LLM模型，这些模型在文本和基于图像的设置下进行评估，并被分为四种类型的题目：单选题、多选题、数值计算题和填空题。其规模之大，涵盖了十个职业领域及多个专业，共有超过11,588个考试题目。该数据集的任务是评估大型语言模型在医疗执照考试中的表现。

The dataset named KokushiMD-10 is a multimodal benchmark dataset constructed from ten national medical licensing examinations in Japan, covering multiple domains including medicine, dentistry, nursing, pharmacy, and related health professions. It contains over 11,588 real exam questions paired with clinical images and expert-annotated solution rationales. Additionally, this dataset has evaluated more than 30 state-of-the-art LLM models, with assessments conducted under both text-only and image-based settings. The included exam questions are categorized into four types: single-choice questions, multiple-choice questions, numerical calculation questions, and fill-in-the-blank questions. With its large scale, the dataset spans ten professional fields and multiple specialties, totaling over 11,588 exam questions. The core task of this dataset is to evaluate the performance of large language models in medical licensing examinations.