This research demonstrates how goal-driven and constraint-guided large language model agents can generate viable hypotheses for materials discovery and design.

• Combines LLMs with expert knowledge to accelerate the materials discovery pipeline
• Creates a novel dataset from recent publications featuring real-world materials science goals
• Establishes a framework where AI agents operate under specific constraints to generate testable hypotheses
• Demonstrates potential to significantly reduce time and resources needed for materials innovation

For the engineering sector, this approach offers a transformative method to rapidly discover application-specific materials, potentially revolutionizing product development cycles and enabling more sustainable manufacturing processes.

Hypothesis Generation for Materials Discovery and Design Using Goal-Driven and Constraint-Guided LLM Agents