Leveraging Chemistry Foundation Models to Facilitate Structure Focused Retrieval Augmented Generation in Multi-Agent Workflows for Catalyst and Materials Design

TL;DR

This paper explores how large pre-trained chemistry models can enhance structure-focused retrieval and generation in multi-agent systems for materials and catalyst design, improving information access across diverse data types.

Contribution

It introduces the use of chemistry foundation models for semantic retrieval in multi-agent workflows, enabling cross-modal and structure-based queries in materials research.

Findings

Chemistry foundation models enable structure-focused retrieval for molecules and polymers.

Integration with multi-modal models allows cross-domain information retrieval.

Multi-agent systems benefit from these models for natural language and topological queries.

Abstract

Molecular property prediction and generative design via deep learning models has been the subject of intense research given its potential to accelerate development of new, high-performance materials. More recently, these workflows have been significantly augmented with the advent of large language models (LLMs) and systems of autonomous agents capable of utilizing pre-trained models to make predictions in the context of more complex research tasks. While effective, there is still room for substantial improvement within agentic systems on the retrieval of salient information for material design tasks. Within this context, alternative uses of predictive deep learning models, such as leveraging their latent representations to facilitate cross-modal retrieval augmented generation within agentic systems for task-specific materials design, has remained unexplored. Herein, we demonstrate that large, pre-trained chemistry foundation models can serve as a basis for enabling structure-focused, semantic chemistry information retrieval for both small-molecules, complex polymeric materials, and reactions. Additionally, we show the use of chemistry foundation models in conjunction with multi-modal models such as OpenCLIP facilitate unprecedented queries and information retrieval across multiple characterization data domains. Finally, we demonstrate the integration of these models within multi-agent systems to facilitate structure and topological-based natural language queries and information retrieval for different research tasks.