General purpose models for the chemical sciences

2025-07-10

AI Summary

Chemical science confronts challenges arising from heterogeneous, sparse, and semantically ambiguous data, rendering conventional machine learning—reliant on large-scale labeled datasets—ineffective for robust modeling. To address this, we propose a novel paradigm of chemical intelligence grounded in General-Purpose Models (GPMs), integrating large language models, self-supervised learning, transfer learning, and in-context learning to enable end-to-end reasoning across molecular design, reaction prediction, and other tasks—even under low-shot or zero-shot conditions. Our framework unifies multimodal, heterogeneous chemical data representations and substantially reduces dependence on human-annotated labels. We validate its prototype capabilities across multiple scientific workflows, including property prediction, retrosynthetic planning, and reaction outcome forecasting. Furthermore, we systematically survey state-of-the-art GPM applications in chemistry, establishing both a methodological foundation and a technical roadmap for advancing chemical AI from narrow, task-specific models toward broad, general-purpose intelligence.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Data-driven techniques have a large potential to transform and accelerate the chemical sciences. However, chemical sciences also pose the unique challenge of very diverse, small, fuzzy datasets that are difficult to leverage in conventional machine learning approaches completely. A new class of models, general-purpose models (GPMs) such as large language models, have shown the ability to solve tasks they have not been directly trained on, and to flexibly operate with low amounts of data in different formats. In this review, we discuss fundamental building principles of GPMs and review recent applications of those models in the chemical sciences across the entire scientific process. While many of these applications are still in the prototype phase, we expect that the increasing interest in GPMs will make many of them mature in the coming years.

Problem

• Addressing diverse small fuzzy chemical datasets
• Leveraging general-purpose models for chemical tasks
• Exploring GPM applications across chemical sciences

Innovation

• General-purpose models for diverse chemical datasets
• Leveraging large language models in chemistry
• Flexible operation with low data amounts