向大规模的化学反应图像解析通过一个多式联络式的大型语言模型
Yufan Chen1, Ching Ting Leung1, Jianwei Sun1,2
1Department of Chemical and Biological Engineering, Hong Kong University of Science and Technology Hong Kong SAR China hanyugao@ust.hk.
Chemical science
|October 17, 2025
概括
我们开发了RxnIM,一个多式大型语言模型,自动从图像中提取化学反应数据. 这通过从已发表的文献中创建机器可读的反应数据库,在有机化学中推进了人工智能.
科学领域:
- 有机化学 有机化学
- 人工智能的人工智能
- 计算化学的计算化学
背景情况:
- 人工智能 (AI) 对有机化学有希望,但需要高质量,机器可读的反应数据.
- 目前从文献中提取反应数据的方法是手动的或与图像解析作斗争,阻碍AI应用.
- 缺乏结构化,机器可读的反应数据限制了AI在化学研究中的潜力.
研究的目的:
- 介绍反应图像多模式大语言模型 (RxnIM) 用于解析化学反应图像.
- 允许从图像中自动提取反应组件和条件,将其转化为机器可读的格式.
- 促进从化学文献中创建大规模,机器可读的反应数据库.
主要方法:
- 开发了RxnIM,这是一个针对化学反应图像解析的新型多式联络大型语言模型.
- 实施了一种专门的大规模数据集生成方法来训练RxnIM模型.
- 评估了RxnIM在提取化学反应数据的各种基准上的表现.
主要成果:
- RxnIM成功地分析化学反应图像,提取关键组件并解释文本条件.
- 在基准数据集上获得了88%的F1平均得分,比现有方法高出5%.
- 在将视觉反应信息转换为结构化,机器可读数据方面表现出强大的性能.
结论:
- RxnIM代表了从图像中自动生成机器可读的反应数据的重大进步.
- 这项工作为人工智能驱动的有机化学研究提供了必要的数据资源.
- 开发的模型,数据集和代码被公开发布,以支持研究社区.
相关概念视频
Molecular Models
43.5K
Physical models representing molecular architectures of chemical compounds play essential roles in understanding chemistry. The use of molecular models makes it easier to visualize the structures and shapes of atoms and molecules.
43.5K
Chemical Reactions
95.0K
A chemical reaction is a process by which the bonds in the atoms of substances are rearranged to generate new substances. Matter cannot be created or destroyed in a chemical reaction—the same type and number of atoms that make up the reactants are still present in the products. Merely, the rearrangement of chemical bonds produces new compounds.
Chemical Reactions Rearrange Atoms into New Substances
A chemical reaction takes starting materials—the reactants—and changes them...
Chemical Reactions Rearrange Atoms into New Substances
A chemical reaction takes starting materials—the reactants—and changes them...
95.0K
Chemical Reactions
13.3K
A balanced chemical equation provides the information of chemical formulas of the reactants and products involved in the chemical change. A reaction’s stoichiometry helps predict how much of the reactant is needed to produce the desired amount of product, or in some cases, how much product will be formed from a specific amount of the reactant.
The relative amounts of reactants and products represented in a balanced chemical equation are often referred to as stoichiometric amounts. However, in...
The relative amounts of reactants and products represented in a balanced chemical equation are often referred to as stoichiometric amounts. However, in...
13.3K
Multi-Step Reactions
8.6K
Chemical reactions often occur in a stepwise fashion involving two or more distinct reactions taking place in a sequence. A balanced equation indicates the reacting species and the product species, but it reveals no details about how the reaction occurs at the molecular level. The reaction mechanism (or reaction path) provides details regarding the precise, step-by-step process by which a reaction occurs. Each of the steps in a reaction mechanism is called an elementary reaction. These...
8.6K
The Photochemical Reaction Center
5.2K
Reaction centers are pigment-protein complexes that initiate energy conversion from photons to chemical entities. Therefore, photochemical reaction center is a more appropriate term that describes these complexes. The Nobel laureates Robert Emerson and William Arnold provided the first experimental evidence of photochemical reaction centers by demonstrating the participation of nearly 2,500 chlorophyll molecules for the release of just one molecule of oxygen. Despite thousands of photosynthetic...
5.2K
Predicting Reaction Outcomes
10.0K
Kinetics describes the rate and path by which a reaction occurs. In contrast, thermodynamics deals with state functions and describes the properties, behavior, and components of a system. It is not concerned with the path taken by the process and cannot address the rate at which a reaction occurs. Although it does provide information about what can happen during a reaction process, it does not describe the detailed steps of what appears on an atomic or a molecular level. On the other hand,...
10.0K


