Towards a universal meta-optics solver via large language models
Abstract
Metasurface design increasingly requires fast models that can operate across structurally distinct device families, rather than retraining a separate surrogate for every geometry class. Conventional neural network surrogates often depend on fixed-dimensional descriptors, family-specific output formats, and repeated architecture tuning, which limits their scalability across heterogeneous meta-atoms. Here, we present a unified large language model (LLM) workflow for multi-family metasurface modeling and inverse-design. Geometries, design parameters, and optical response channels were converted into a shared instruction-following text format and used to fine-tune Gemma-2-9B across 8 metasurface families. Compared with single-family baselines, the joint model simultaneously predicted the optical responses of all metasurface families while reducing the MSE for each family by an average of 56.5%. The same representation was also used for inverse design. These results show that a shared sequence-based LLM interface can provide a practical route to cross-family metasurface design while reducing the need for task-specific surrogate architectures.
Versions
Related
- Inverse design of large-scale freeform meta-optics by breaking the memory wall of full-wave simulation
- Integrated robust universal linear optics based on white-boxed multi-plane light conversion
- Vision-Language Agents for Active Perception in Optics Laboratories
- Meta-Optics-Based Diffractive Learning Machine for Angle-Multiplexed Multi-Task Processing
- Universal Synthesis of Large-Area Hybrid Perovskite Monolayers Enabling Lithography-Free, Low-Static-Power Optoelectronics
BibTeX
@misc{zhang2026towards,
title = {{Towards a universal meta-optics solver via large language models}},
author = {Huanshu Zhang and Lei Kang and Yuyan Chen and Luxiang Wang and Zhaolong Cao and Douglas H. Werner},
howpublished = {arXiv},
year = {2026},
month = aug,
doi = {10.48550/arxiv.2608.26417},
eprint = {2608.26417},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.26417},
}
RIS
TY - UNPB
TI - Towards a universal meta-optics solver via large language models
AU - Huanshu Zhang
AU - Lei Kang
AU - Yuyan Chen
AU - Luxiang Wang
AU - Zhaolong Cao
AU - Douglas H. Werner
PY - 2026
DA - 2026/08/26
PB - arXiv
DO - 10.48550/arxiv.2608.26417
UR - https://arxiv.org/abs/2608.26417
AB - Metasurface design increasingly requires fast models that can operate across structurally distinct device families, rather than retraining a separate surrogate for every geometry class. Conventional neural network surrogates often depend on fixed-dimensional descriptors, family-specific output formats, and repeated architecture tuning, which limits their scalability across heterogeneous meta-atoms. Here, we present a unified large language model (LLM) workflow for multi-family metasurface modeling and inverse-design. Geometries, design parameters, and optical response channels were converted into a shared instruction-following text format and used to fine-tune Gemma-2-9B across 8 metasurface families. Compared with single-family baselines, the joint model simultaneously predicted the optical responses of all metasurface families while reducing the MSE for each family by an average of 56.5%. The same representation was also used for inverse design. These results show that a shared sequence-based LLM interface can provide a practical route to cross-family metasurface design while reducing the need for task-specific surrogate architectures.
ER -