IMB-QA, IMB-MCQA
收藏资源简介:
IMB数据集由两个子集组成:IMB-QA,包含从意大利在线医疗论坛MedicItalia和Dica33收集的782,644个医患对话,跨越77个医疗类别;IMB-MCQA包含从医学专业入学考试模拟器CompitoInClasse.org收集的25,862个多项选择题。这些数据集通过使用大型语言模型(LLM)进行数据清理和匿名化,以提高数据质量和隐私保护。IMB数据集旨在支持多语言医疗问答的研究,并为评估不同类型医疗沟通中的模型性能提供独特的机遇。
The IMB dataset comprises two subsets: IMB-QA and IMB-MCQA. IMB-QA contains 782,644 doctor-patient dialogues collected from two Italian online medical forums, MedicItalia and Dica33, spanning 77 medical categories. IMB-MCQA includes 25,862 multiple-choice questions collected from the medical professional entrance exam simulator platform CompitoInClasse.org. Both subsets have undergone data cleaning and anonymization using Large Language Models (LLMs) to enhance data quality and privacy protection. The IMB dataset aims to support research on multilingual medical question answering and provides a unique opportunity to evaluate model performance across different types of medical communication.

- 1IMB: An Italian Medical Benchmark for Question AnsweringUniversity of Naples Federico II, Department of Electrical Engineering and Information Technology (DIETI), Italy and Consorzio Interuniversitario Nazionale per l'Informatica (CINI)-ITEM National Lab, Naples, Italy · 2025年



