Answers are grounded in this paper and cite their sources.
Inverse design of electromagnetic metasurface plays a pivotal role in tailoring subwavelength structures to achieve specific wavefront manipulations and exotic functionalities. The design of metasurface typically begins with finding a proper structure prototype according to user demand, requiring domain experts to take considerable time for literature retrieval and filtering. Recently, large language models (LLMs) pave a promising way for research automation, owing to its capability of knowledge query, step‐by‐step reasoning, and on‐demand tool calling. Here, we propose MetaChatAgent, an LLM‐assisted toolbox empowered by multi‐agent collaboration and retrieval‐augmented generation. It integrates three specialized agents to replace labor‐intensive research processes. A data‐analysis agent is firstly developed to execute the holistic literature mining from semantic knowledge databases, enabling human‐free data mining and feature analysis. Next, a retrieval‐reasoning agent is presented to retrieve the relevant knowledge in the vast literature according to user demand, whose effectiveness is validated by a query‐answer benchmark of metasurfaces across various scenarios. Finally, a reproduction agent is proposed to automatically reconstruct the parametric structure of metasurface and conduct the property reproduction facilitated by specialized simulation tools. This work establishes a knowledge‐guided multi‐agent framework for AI‐in‐the‐loop metasurface research, automating tasks including literature mining, evidence‐grounded analysis, structural reconstruction, and simulation‐based validation.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: