Designing multiple metamaterials inversely with generative AI, including fabrication structures

1/ AI 활용 방법: Structuring language instructions and combining them with a diffusion model for generative inverse design
2/ 적용 과제: Inverse design of metamaterial candidate structures that satisfy target properties
3/ 보고된 결과: The design conditions were structured, candidate structures were generated and screened, and actually fabricated metamaterials were presented
4/ This study, announced on 2026-11-05, used MetaGen-LLM, which structures language instructions and then combines a diffusion model with a generative model, to perform inverse design of multi-class metamaterials.
5/ The system organized design conditions with an LLM parser, and connected a diffusion-based parameter generator, geometry generator, and surrogate predictor to create and filter candidate structures.
6/ The study also presented metamaterials that were actually fabricated, showing a workflow in which generative AI produces concrete structural candidates at the design stage.
7/ From a materials research perspective, the key point is the pipeline that converts requirements described in natural language into structural parameters and geometric candidates.
8/ However, this result is a case study for a specific metamaterial problem, and it cannot be assumed that the same quality of candidates will be generated automatically for other material systems or constraint conditions.
9/ Candidate performance requires separate confirmation through actual fabrication and measurement. [1]