RxnBench is a visual question answering dataset for multimodal chemical reaction image understanding, designed to evaluate the capabilities of visual language models in tasks such as chemical reaction image understanding, multimodal reasoning, and scientific question answering.

This dataset contains 1,525 multiple-choice questions on organic chemistry reaction comprehension, available in both Chinese and English. These questions are based on 305 research figures from high-impact OpenAssess journals. Each figure is accompanied by five multiple-choice visual quiz questions on interpreting organic reaction diagrams, covering areas such as image description, reaction content extraction, molecule/structure identification, and reaction mechanism assessment.