Answers are grounded in this paper and cite their sources.
Large language models (LLMs) have shown promise in science, such as structure-property prediction and acting as AI agents, yet their intrinsic knowledge and reasoning capability for scientific discovery remains underexplored. We introduce LLM-EO, an integration of LLMs into evolutionary optimization, and demonstrate its success in optimizing transition metal complexes (TMCs). LLM-EO demonstrates advantages in few-sample learning due to the intrinsic chemical knowledge embedded within LLMs and their ability to leverage entire historical data collected during optimizations. Through natural language instructions, LLM-EO offers enhanced accessibility for multiobjective optimizations, potentially lowering barriers for experimental chemists without extensive programming expertise. As generative models, LLM-EO possesses the capability to propose novel ligands and TMCs with unique chemical properties by amalgamating both internal knowledge and external chemistry data, thus combining the benefits of efficient optimization and generation. With advancements in LLMs, both in their capacity as pretrained foundational models and new strategies in post-training inference, we anticipate broad applications of LLM-EO in chemistry and materials design.
No takes yet. Share an insight, caveat, or question.
Lü et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: