BioMol-MQA

BioMol-MQA是一个多模态问答数据集，旨在测试大型语言模型在多模态生物分子相互作用推理方面的能力。该数据集由两部分组成：一部分是多模态知识图谱，包含文本和分子结构用于信息检索；另一部分是设计用于测试大型语言模型在多模态知识图谱上检索和推理能力的问题。数据集的构建过程包括数据收集、文本后处理、分子相互作用提取、问题生成和问题验证。

BioMol-MQA is a multimodal question answering dataset intended to evaluate the reasoning capabilities of large language models (LLMs) regarding multimodal biomolecular interactions. This dataset consists of two components: one is a multimodal knowledge graph that incorporates textual content and molecular structures for information retrieval, and the other is a set of questions designed to test the retrieval and reasoning abilities of LLMs over the multimodal knowledge graph. The construction pipeline of the dataset encompasses data collection, text post-processing, molecular interaction extraction, question generation, and question validation.