Search papers, labs, and topics across Lattice.
PRISM, a decoder-only autoregressive transformer, tackles the inverse problem of multilayer thin-film design by jointly predicting discrete material selection and continuous thickness regression. The architecture incorporates spectrum prefix conditioning for target injection and cumulative-depth Rotary Position Embeddings to encode continuous thickness within the positional representation. Experiments show that a PRISM-13M model reduces MAE by over 50% compared to other transformer baselines, and a 44M-parameter variant achieves state-of-the-art performance with significantly faster operation than simulated annealing.
Achieve state-of-the-art performance in multilayer thin-film design with a 5x smaller model and faster runtime by encoding continuous thickness directly into the positional embeddings of an autoregressive transformer.
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.
