ngram/medchat-qa

MedChatQA数据集旨在为测试LLMs在真实世界医学信息和医学交流主题上的准确问答提供基准。该数据集包含约30000个问题，覆盖约1000种FDA批准的人类处方药物。数据集由专业人士策划，语言为英语，采用MIT许可证。数据格式为JSON对象，由换行符分隔。数据集是合成的，由专家生成，但可能存在错误，使用时需谨慎。

MedChatQA Dataset is intended as a benchmark for testing the accurate question-answering performance of Large Language Models (LLMs) on real-world medical information and medical communication-related topics. This dataset comprises approximately 30,000 questions, covering around 1,000 FDA-approved human prescription drugs. Curated by medical professionals, the dataset is in English and licensed under the MIT License. Its data format is newline-delimited JSON objects. The dataset is synthetic, generated by domain experts, but may contain errors, so caution should be taken when utilizing it.

原始信息汇总

MedChatQA数据集概述

目标

• 作为测试大型语言模型（LLMs）在真实世界医疗信息和医疗沟通主题上准确问答能力的基准。

内容

• 包含医疗领域的专业人士与患者及其他专业人士之间的沟通记录。

质量要求

• 所有沟通内容要求100%事实准确，无错误。