PRISM: Position-encoded Regressive Inverse Spectral Model for Multilayer Thin-Film Design

Runtian Wang, Renhao Xue, Baige Chen, Hao Wu

arXiv:2605.26502 · 2026-05-27 공개 · arXiv · PDF

transformer autoregressive regression position-encoding inverse-problem thin-film optical-coatings prefix-tokens

Abstract

The inverse problem of multilayer thin-film optical coatings design represents a complex combinatorial-continuous optimization challenge. We present PRISM (Position-encoded Regressive Inverse Spectral Model), a unified decoder-only autoregressive transformer that streamlines this process by jointly predicting discrete material selection and continuous thickness regression within a single backbone. PRISM introduces two primary architectural innovations: (1) spectrum prefix conditioning, which utilizes standard prefix tokens for in-context target injection, and (2) cumulative-depth Rotary Position Embeddings, which encode continuous thickness directly into the positional representation to preserve the physical spatial relationships of the stack. Our benchmarks demonstrate that a PRISM-13M model reduces MAE by over 50\% compared to other transformer baselines while utilizing only one-fifth of the parameters. Furthermore, a 44M-parameter variant achieves state-of-the-art performance (MAE = 0.010) on our in-distribution validation benchmark and operates significantly faster than simulated annealing, offering a highly efficient alternative to classical optimization methods.

한국어 요약

📋 한 줄 요약

**[Thin-Film 광학 설계 / Autoregressive Transformer]** PRISM이 decoder-only autoregressive transformer로 multilayer thin-film의 discrete 재료·continuous 두께 통합 예측 — spectrum prefix conditioning + cumulative-depth RoPE로 13M 모델이 MAE 50% 감소, 44M은 MAE 0.010 SOTA·simulated annealing보다 significantly 빠름.

🎯 핵심 기여도

💡 핵심 아이디어

Thin-film 광학 설계의 discrete-continuous joint 최적화는 decoder-only autoregressive transformer로 통합 가능하며, 핵심은 spectrum target을 prefix token으로 in-context 주입하고 cumulative depth를 RoPE에 직접 인코딩해 stack의 physical spatial relationship을 positional 표현으로 보존하는 것이다 — classical simulated annealing 대비 SOTA·고속.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: Thin-film 광학 설계의 transformer 기반 unified solution, RoPE로 continuous 두께 인코딩한다는 새 positional embedding 활용, classical optimization 대비 양적·속도적 우월성, parameter-efficient(13M으로 SOTA-급) 결과. **한계**: In-distribution 검증 위주로 out-of-distribution 일반화 추가 검증, multilayer thin-film 도메인 특화로 다른 광학 inverse problem 적용성, MAE 0.010의 실제 제작 tolerance 적합성.

🚀 실용적 활용