AutoMS: Multi-Agent Evolutionary Search for Cross-Physics Inverse Microstructure Design

Designing microstructures with coupled cross-physics objectives is a fundamental challenge where traditional topology optimization is often computationally prohibitive and deep generative models frequently suffer from physical hallucinations. We introduce AutoMS, a multi-agent neuro-symbolic framework that reformulates inverse design as an LLM-driven evolutionary search. AutoMS leverages LLMs as semantic navigators to decompose complex requirements and coordinate agent workflows, while a novel Simulation-Aware Evolutionary Search (SAES) mechanism handles low-level numerical optimization via local gradient approximation and directed parameter updates. This architecture achieves a state-of-the-art 83.8\% success rate on 17 diverse cross-physics tasks, significantly outperforming both traditional evolutionary algorithms and existing agentic baselines. By decoupling open-ended semantic orchestration from simulation-grounded numerical search, AutoMS provides a robust pathway for navigating complex physical landscapes that remain intractable for standard generative or purely linguistic approaches.

Designing microstructure often depends on finding tiny internal structures that can satisfy several physical needs at once, such as being strong, flexible, and good at conducting heat or electricity. This is difficult because improving one property may hurt another, and testing many possible designs can be very expensive. This paper presents AutoMS, an AI system that helps search for better microstructure designs. Instead of making a single guess, AutoMS breaks down the design goal, generates candidate structures, tests them with physics simulations, and uses the results to guide the next round of search. Across 17 challenging design tasks, AutoMS successfully finds valid designs more often than existing search-based and AI-based methods.