UniParser/RxnBench

RxnBench（SF-QA）是一个视觉问答（VQA）基准数据集，包含1,525个博士级别的有机化学反应理解多选题（MCQ）。该基准数据集基于305个来自高影响力OpenAssess期刊的科学图表构建。每个图表由领域专家精心设计了五个多选题，旨在解释有机反应图。这些问题经过多轮严格的审查和修订，以确保其清晰性和科学准确性。问题涵盖多种类型，包括化学反应图像的描述、反应内容的提取、分子或Markush结构的识别以及反应机制的确定。该基准数据集挑战视觉语言模型在有机化学基础知识、多模态上下文推理和化学推理技能方面的能力。数据集提供英文和中文两个版本。

RxnBench (SF-QA) is a visual question answering (VQA) benchmark comprising 1,525 multiple-choice questions (MCQs) at the PhD-level of organic chemistry reaction understanding. The benchmark is built from 305 scientific figures drawn from high-impact OpenAssess journals. For each figure, domain experts carefully designed five multiple-choice VQA questions targeting the interpretation of organic reaction diagrams. These questions were further refined through multiple rounds of rigorous review and revision to ensure both clarity and scientific accuracy. The questions cover a variety of types, including the description of chemical reaction images, extraction of reaction content, recognition of molecules or Markush structures, and determination of mechanisms. This benchmark challenges visual-language models on their foundational knowledge of organic chemistry, multimodal contextual reasoning, and chemical reasoning skills. The benchmark is released in both English and Chinese versions.