Pushing the Limits of Inverse Lithography with Generative Reinforcement Learning

2026-02-21

AI Summary

This work addresses the challenge of inverse lithography technology (ILT), where highly non-convex optimization often leads to suboptimal local minima that conventional and existing generative methods struggle to overcome. To this end, we propose the first application of generative reinforcement learning to ILT, introducing a fine-tuning mechanism based on Group Relative Policy Optimization (GRPO) and a multi-candidate sampling-and-screening pipeline. Specifically, a WGAN-based generator is first pre-trained to learn the mapping from target patterns to mask distributions, then fine-tuned using a combination of reconstruction loss and fast batch-wise ILT optimization. The proposed approach effectively mitigates non-convexity issues, achieving significant reductions in edge placement error (EPE) violations under 3nm tolerance on LithoBench while doubling throughput. On the ICCAD13 benchmark, it demonstrates over 20% EPE improvement and a 3× speedup, outperforming both traditional solvers and purely generative methods.

Technology Category

• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Natural Language Processing: Generation
• Search and Optimization: Learning to Search

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Inverse lithography (ILT) is critical for modern semiconductor manufacturing but suffers from highly non-convex objectives that often trap optimization in poor local minima. Generative AI has been explored to warm-start ILT, yet most approaches train deterministic image-to-image translators to mimic sub-optimal datasets, providing limited guidance for escaping non-convex traps during refinement. We reformulate mask synthesis as conditional sampling: a generator learns a distribution over masks conditioned on the design and proposes multiple candidates. The generator is first pretrained with WGAN plus a reconstruction loss, then fine-tuned using Group Relative Policy Optimization (GRPO) with an ILT-guided imitation loss. At inference, we sample a small batch of masks, run fast batched ILT refinement, evaluate lithography metrics (e.g., EPE, process window), and select the best candidate. On \texttt{LithoBench} dataset, the proposed hybrid framework reduces EPE violations under a 3\,nm tolerance and roughly doubles throughput versus a strong numerical ILT baseline, while improving final mask quality. We also present over 20\% EPE improvement on \texttt{ICCAD13} contest cases with 3$\times$ speedup over the SOTA numerical ILT solver. By learning to propose ILT-friendly initializations, our approach mitigates non-convexity and advances beyond what traditional solvers or GenAI can achieve.

Problem

• inverse lithography
• non-convex optimization
• mask synthesis
• local minima
• semiconductor manufacturing

Innovation

• Generative Reinforcement Learning
• Inverse Lithography
• Conditional Sampling
• Group Relative Policy Optimization
• Mask Synthesis