Augmenting large language models with chemistry tools

By Andrés M Bran, Sam Cox, Oliver Schilter, Carlo Baldassari, Andrew D. White, Philippe Schwaller. Nature Machine Intelligence, 2023

1,063 citations recorded by the source corpus.

Large language models (LLMs) have shown strong performance in tasks across domains but struggle with chemistry-related problems. These models also lack access to external knowledge sources, limiting their usefulness in scientific applications. We introduce ChemCrow, an LLM chemistry agent designed to accomplish tasks across organic synthesis, drug discovery and materials design. By integrating 18 expert-designed tools and using GPT-4 as the LLM, ChemCrow augments the LLM performance in chemistry, and new capabilities emerge. Our agent autonomously planned and executed the syntheses of an insect repellent and three organocatalysts and guided the discovery of a novel chromophore. Our evaluation, including both LLM and expert assessments, demonstrates ChemCrow’s effectiveness in automating a diverse set of chemical tasks. Our work not only aids expert chemists and lowers barriers for non-experts but also fosters scientific advancement by bridging the gap between experimental and computational chemistry. Large language models can be queried to perform chain-of-thought reasoning on text descriptions of data or computational tools, which can enable flexible and autonomous workflows. Bran et al. developed ChemCrow, a GPT-4-based agent that has access to computational chemistry tools and a robotic chemistry platform, which can autonomously solve tasks for designing or synthesizing chemicals such as drugs or materials.

Subfields: Artificial Intelligence, Information Systems, Materials Chemistry

View Paper: https://www.semanticscholar.org/paper/354dcdebf3f8b5feeed5c62090e0bc1f0c28db06

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Sebastian Nowozin
2. Heng Ji
3. Jiawei Han
4. Pushmeet Kohli
5. Quoc V. Le
6. Gao Huang
7. Meng Jiang
8. Percy Liang

In Information Systems

1. Sebastian Nowozin
2. Pushmeet Kohli
3. Heng Ji
4. Gao Huang
5. Meng Jiang
6. Jiawei Han
7. Bowen Jin
8. Marinka Zitnik

In Materials Chemistry

1. Pushmeet Kohli
2. Gao Huang
3. Heng Ji
4. Meng Jiang
5. Y. Liu
6. Marinka Zitnik

Citations in the corpus

• ExpeL: LLM Agents Are Experiential Learners (2023, background): “…fields like robotics, natural sciences, game-playing, and workflows has surged, with emphasis on their world knowledge in fewshot settings (Ha, Florence, and Song 2023; Mu et al. 2023; Bran et al. 2023; Boiko, MacKnight, and Gomes 2023; Yang et al. 2023b; Lin et al. 2023a; Nakano et al. 2021; Wang…”
• Competition-Level Problems are Effective LLM Evaluators (2023, background): “LLMs are also widely used in scientific research in other fields (Wang et al., 2023), such as physics (Yeadon and Halliday, 2023), chemistry (Castro Nascimento and Pimentel, 2023; Bran et al., 2023), etc.”
• Empowering Biomedical Discovery with AI Agents (2024, positive): “For instance, ChemCrow [2] combines chain-of-thought reasoning [80] with ML tools to support tasks in organic chemistry, identifying and summarizing literature to inform experiments.”
• Large Language Models on Graphs: A Comprehensive Survey (2023, methodology): “Since string representations of molecular graphs usually carry new and in-depth domain knowledge, which is beyond the knowledge of LLMs, recent work [146], [157], [165] also attempts to utilize the reasoning ability of LLMs, instead of using them as a knowledge source for predicting the property of…”
• AlphaEvolve: A coding agent for scientific and algorithmic discovery (2025, positive): “Other works use LLMs to optimize experimental planning and design [7, 10, 43, 75] or experiment execution and workflow [28, 62, 82, 105, 116].”
• Tool learning with large language models: a survey (2024, positive): “…the employment of search engine tool can enable LLMs to access contemporary information [17, 28, 47–51], while the integration of database tool allows LLMs to access structured databases to retrieve specific information or execute complex queries, thus expanding their knowledge base [52–57].”
• PDE-Agents: An LLM-Orchestrated Multi-Agent Framework for Automated Finite Element Simulations with Knowledge Graph-Augmented Reasoning (2026, unknown): “ChemCrow [12] demonstrated that equipping an LLM with chemistry tools (RDKit, reaction planners) enables autonomous laboratory-scale reasoning.”