PubMedQA

该数据集名为PubMedQA，旨在评估基于PubMed摘要中提取问题而构建的医学问答系统的性能。此外，采用自一致性提示的Med-PaLM 2模型在此数据集上取得了81.8%的准确率，尽管测试集规模较小，但这一成绩已达到当前最高水平。该数据集包含500个示例，其任务是医疗问题解答。

This dataset, named PubMedQA, is designed to evaluate the performance of medical question answering systems built using questions extracted from PubMed abstracts. Furthermore, the Med-PaLM 2 model adopting self-consistency prompting achieved an accuracy of 81.8% on this dataset. Although the test set is small in scale, this result has reached the current state-of-the-art level. This dataset contains 500 examples, with its task focused on medical question answering.

搜集汇总

背景与挑战

背景概述

PubMedQA是一个专注于生物医学研究问答的数据集，包含专家标注、未标注和人工生成的QA实例，总计超过27万条数据，旨在通过研究摘要回答是/否/可能类型的问题。