japan-ai-official/igakuqa-subset-curated

IgakuQA精选子集是一个高质量的医疗考试问题数据集，它从IgakuQA日本医疗考试数据集中精心挑选了66个纯文本样本。这些样本基于难度评分、一致性评分、模型共识、响应质量和文本约束等标准进行选择。该数据集适合用于文本模型的医学AI评估、微调、代理服务评估、医学知识能力研究以及了解模型在医学领域的局限性。

The IgakuQA Curated Subset is a high-quality medical examination question dataset consisting of 66 carefully selected text-only samples from the IgakuQA Japanese medical examination dataset. These samples were chosen based on criteria such as difficulty score, consistency score, model agreement, response quality, and text-only constraint. The dataset is suitable for medical AI evaluation with text-only models, fine-tuning on challenging Japanese medical questions, agent service evaluation without image support, research on medical knowledge capabilities without visual components, and understanding model limitations in medical domains.