AutoMS: Multi-Agent Evolutionary Search for Cross-Physics Inverse Microstructure Design
Summary: arXiv:2603.27195v1 Announce Type: new
Abstract
Designing microstructures that satisfy coupled cross-physics objectives is a fundamental challenge in material science. This inverse design problem involves a vast, discontinuous search space where traditional topology optimization is computationally prohibitive, and deep generative models often suffer from “physical hallucinations,” lacking the capability to ensure rigorous validity.
Introduction to AutoMS
To address these limitations, we introduce AutoMS, a multi-agent neuro-symbolic framework that reformulates inverse design as an LLM-driven evolutionary search. Unlike methods that treat LLMs merely as interfaces, AutoMS integrates them as “semantic navigators” to initialize search spaces and break local optima.
Key Features of AutoMS
- Simulation-Aware Evolutionary Search (SAES): This novel approach addresses the “blindness” of traditional evolutionary strategies by utilizing simulation feedback to perform local gradient approximation and directed parameter updates.
- Specialized Agents: AutoMS orchestrates specialized agents including Manager, Parser, Generator, and Simulator to effectively navigate complex physical landscapes.
- Performance Metrics: AutoMS achieves a state-of-the-art 83.8% success rate on 17 diverse cross-physics tasks, nearly doubling the performance of traditional NSGA-II (43.7%) and significantly outperforming ReAct-based LLM baselines (53.3%).
- Efficiency Improvement: Our hierarchical architecture reduces total execution time by 23.3%, showcasing the efficiency of the multi-agent system.
Impact and Applications
AutoMS demonstrates that autonomous agent systems can effectively bridge the gap between semantic design intent and rigorous physical validity. By integrating LLMs into the design process, AutoMS offers a promising solution to the challenges posed by complex material design tasks.
Conclusion
In summary, AutoMS represents a significant advancement in the field of material science, providing a robust framework for the inverse design problem. By leveraging multi-agent systems and simulation-aware strategies, AutoMS enhances the capability to explore vast search spaces effectively, making it a valuable tool for researchers and engineers in the quest for innovative microstructure designs.
Future Directions
As we continue to refine and develop AutoMS, we anticipate further enhancements in both efficiency and accuracy. Future work will focus on:
- Expanding the range of materials and physics involved in the design process.
- Improving the scalability of the framework for larger and more complex design problems.
- Integrating additional machine learning techniques to further enhance the performance of the evolutionary search.