AUTOCIRCUIT-RL: Reinforcement Learning-Driven LLM for Automated Circuit Topology Generation

Prashanth Vijayaraghavan, Luyao Shi, Ehsan Degan, Vandana Mukherjee, Xin Zhang

TL;DR: We introduce AUTOCIRCUIT-RL, an AI system that uses reinforcement learning and large language models to automatically generate efficient and valid analog circuit designs from constraints.

Abstract: Analog circuit topology synthesis is integral to Electronic Design Automation (EDA), enabling the automated creation of circuit structures tailored to specific design requirements. However, the vast design search space and strict constraint adherence make efficient synthesis challenging. Leveraging the versatility of Large Language Models (LLMs), we propose AUTOCIRCUIT-RL, a novel reinforcement learning (RL)-based framework for automated analog circuit synthesis. The framework operates in two phases: instruction tuning, where an LLM learns to generate circuit topologies from structured prompts encoding design constraints, and RL refinement, which further improves the instruction-tuned model using reward models that evaluate validity, efficiency, and output voltage. The refined model is then used directly to generate topologies that satisfy the design constraints. Empirical results show that AUTOCIRCUIT-RL generates ~12% more valid circuits and improves efficiency by ~14% compared to the best baselines, while reducing duplicate generation rates by ~38%. It achieves over 60% success in synthesizing valid circuits with limited training data, demonstrating strong generalization. These findings highlight the framework's effectiveness in scaling to complex circuits while maintaining efficiency and constraint adherence, marking a significant advancement in AI-driven circuit design.

Lay Summary:

Designing analog circuits — the basic building blocks inside electronic devices — usually takes a lot of time and expert knowledge. It’s hard to automate because there are so many ways to design a circuit, and each one has to follow strict performance rules.
In this work, we created a tool called AUTOCIRCUIT-RL that uses AI to help design these circuits automatically. It works in two steps. First, we use a large language model (LLM) — similar to the ones used in chatbots — and teach it to generate circuit designs based on given instructions, like what the circuit should do. Then, we improve the model's designing ability by giving it feedback from AI. This feedback helps the model learn which designs work well — based on how efficient they are, if they produce the right output voltage, and whether the design actually works.
Compared to older methods, our tool makes more valid and efficient circuits, even when it doesn't get much training data. It also avoids generating the same design over and over. This could make circuit design faster and easier for engineers

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Primary Area: Applications->Everything Else

Keywords: Analog circuit, circuit topology, reinforcement learning, instruction tuning, RL refinement

Submission Number: 13250