在多式化学奥林匹克考试中评估大型语言模型
Yiming Cui1,2, Xin Yao3,4, Yuxuan Qin3,4
1State Key Laboratory of Cognitive Intelligence, Hefei, China. ymcui@ieee.org.
Communications chemistry
|December 14, 2025
概括
多模式大语言模型 (LLM) 难以进行化学推理,往往无法有效地整合视觉和文本数据. 思想链提示提高了性能,突出了在科学领域需要更好的多式联络人工智能.
科学领域:
- 人工智能的人工智能
- 化学教育 化学教育
背景情况:
- 多模式大语言模型 (LLM) 在科学推理方面面临挑战,特别是在化学领域,其中涉及复杂的视觉和符号数据.
- 视觉和文本信息的有效整合对于化学中高级问题解决至关重要.
研究的目的:
- 系统地评估40个多模式LLM在化学特定推理任务上的表现.
- 确定当前模型能够从不同方式融合信息的能力的局限性.
- 评估思维链提示对多式模式科学推理的影响.
主要方法:
- 美国国家化学奥林匹克赛 (USNCO) 试题的精选基准,涵盖了20多年的时间.
- 40个专有和开源多式联络法学士的系统评估.
- 除研究和基于封闭的可解释性,以分析模型行为和视觉接地.
- 使用和不使用思维链提示的性能比较.
主要成果:
- 许多评估的多模式LLM在模式融合方面表现出显著的困难,有时在视觉输入方面表现更差.
- 连锁思维提示明显提高了多式化学问题解决的准确性和视觉接地.
- 性能变化凸显了当前多式联络LLM科学推理能力的关键局限性.
结论:
- 目前的多式联络LLM需要大幅度改进,以便在化学中进行强有力的科学推理.
- 思想链提示为增强多式人工智能性能和可解释性提供了一个有希望的策略.
- 开发的基准是衡量领域特定多式联络人工智能研究进展的关键工具.
相关概念视频
Molecular Models
43.3K
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.
43.3K
The Small x Assumption
49.3K
If a reaction has a small equilibrium constant, the equilibrium position favors the reactants. In such reactions, a negligible change in concentration may occur if the initial concentrations of reactants are high and the Kc value is small. In such circumstances, the equilibrium concentration is approximately equal to its initial concentration. This estimation can be used to simplify the equilibrium calculations by assuming that some equilibrium concentrations are equal to the initial...
49.3K
Chemical and Solubility Equilibria
4.8K
The free energy change associated with dissolving a solute in a liter of solvent is called the free energy of a solution, ΔGsolution. The overall ΔGsolution is expressed as the balance of ΔGinteraction against the always-favorable free-energy of mixing, ΔGmixing. Solution formation is favorable if ΔGsolution is less than zero, whereas it is unfavorable if ΔGsolution is greater than zero. In short, for a solution to form and complete dissolution to take place,...
4.8K
Predicting Molecular Geometry
44.4K
VSEPR Theory for Determination of Electron Pair Geometries
44.4K
Classification of Elements and Compounds
72.5K
Pure substances consist of only one type of matter. A pure substance can be an element or a compound. An element consists of only one type of atom, while a compound consists of two or more types of atoms held together by a chemical bond. Elements are classified as atomic or molecular based on the nature of their basic units.
Compounds are pure substances composed of two or more elements in fixed, definite proportions. Compounds are classified as ionic or molecular (covalent) based on the bonds...
Compounds are pure substances composed of two or more elements in fixed, definite proportions. Compounds are classified as ionic or molecular (covalent) based on the bonds...
72.5K
Experimental Determination of Chemical Formula
46.2K
The elemental makeup of a compound defines its chemical identity, and chemical formulas are the most concise way of representing this elemental makeup. When a compound’s formula is unknown, measuring the mass of its constituent elements is often the first step in determining the formula experimentally.
46.2K


