This research introduces CRAG-MoW, a novel framework that leverages large language models to optimize and accelerate materials discovery across complex design spaces.

• Combines multiple specialized workflows to integrate multidisciplinary scientific knowledge
• Enables self-correction capabilities in AI systems for materials science applications
• Creates a more efficient search process for identifying promising chemical compounds
• Addresses critical gaps in applying LLMs to materials science through practical implementation frameworks

This advancement significantly impacts engineering by providing automated tools for exploring vast chemical spaces, potentially reducing development cycles for new materials and enabling discoveries that would be impractical through traditional methods.

Agentic Mixture-of-Workflows for Multi-Modal Chemical Search