The CRAG-MoW framework introduces an agentic workflow system that combines Large Language Models with domain-specific tools to revolutionize materials discovery and optimization.

• Integrates multi-modal data across chemistry and materials science
• Employs automated reasoning and self-correction mechanisms
• Optimizes search across vast chemical design spaces
• Provides a benchmarking framework for materials science AI applications

This research bridges critical gaps in engineering by creating practical automation tools for materials scientists, potentially accelerating discovery cycles and reducing development costs in advanced materials engineering.

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