CoSP: Reconfigurable Metamaterial Inverse Design via Contrastive Pretrained Large Language Model
Advanced Science · 2026-11 · 13 (61)
DOI: 10.1002/advs.76917 ↗At a glance
Journal article in Advanced Science, published online on 2026-08-11. Written by Shujie Yang and 4 co-authors. Institution: Institute of Data and Information, Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, Guangdong, China. 63 references. The full text is freely available (open access, CC BY 4.0). Funded by: National Key Research and Development Program of China, National Natural Science Foundation of China, Shenzhen Science and Technology Innovation Program.
Abstract (original language)
Metamaterials for light manipulation at subwavelength scales face significant design challenges due to their complex and sophisticated structures, leading to the emergence of deep learning as a powerful tool to streamline their design process. However, existing deep learning‐based inverse design methods fall short in the design of reconfigurable metamaterials (RMMs), whose optical characteristics switch between different states upon external stimuli. To address this challenge, CoSP, an intelligent inverse design method for RMMs based on a contrastive pretrained large language model (LLM), is proposed. By performing contrastive pretraining on multi‐state spectra, a well‐trained spectrum encoder is obtained and coupled to a GPT‐style decoder trained end‐to‐end from scratch. Equipped with the preservation of linguistic capabilities, CoSP is capable of describing material structures with target optical properties in natural language. Numerical experiments demonstrate that CoSP can design RMM structures for multi‐state, multi‐band optical responses, showing great potential in versatile applications such as thermal management, optical computation, and telecommunications.
Subjects
Author keywords
Authors and institutions
- Institute of Data and Information, Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, Guangdong, China.
- Center of Double Helix, Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, Guangdong, China.
- Intelligent Passive Thermal Control Center, Research Institute of Tsinghua University in Shenzhen, Shenzhen, Guangdong, China.
ORCID profiles: Xuzhe Zhao, Kaichen Dong
Funding
- National Key Research and Development Program of China
- National Natural Science Foundation of China
- Shenzhen Science and Technology Innovation Program
Cite
APA
Yang, S., Zhang, Y., Zhao, X., Tang, Y., & Dong, K. (2026). CoSP: Reconfigurable Metamaterial Inverse Design via Contrastive Pretrained Large Language Model. Advanced Science, 13(61). https://doi.org/10.1002/advs.76917
Vancouver
Yang S, Zhang Y, Zhao X, Tang Y, Dong K. CoSP: Reconfigurable Metamaterial Inverse Design via Contrastive Pretrained Large Language Model. Advanced Science. 2026;13(61). doi:10.1002/advs.76917
BibTeX
@article{Yang2026,
author = {Yang, Shujie and Zhang, Yuqi and Zhao, Xuzhe and Tang, Yansong and Dong, Kaichen},
title = {CoSP: Reconfigurable Metamaterial Inverse Design via Contrastive Pretrained Large Language Model},
journal = {Advanced Science},
year = {2026},
volume = {13},
number = {61},
doi = {10.1002/advs.76917}
}
EUVIMED