CIRCUIT: A Benchmark for Circuit Interpretation and Reasoning Capabilities of LLMs

The application of Large Language Models (LLMs) in analog integrated circuit design could pioneer a new era of AI applications in domains traditionally dominated by human expertise. Analog semiconductor chips are the core building blocks in sensing and communication systems. Contrary to digital chip development, where computer-aided design automation has been widely adopted for a few decades, analog design, often perceived more as a craftsmanship than a well-established engineering procedure, relies heavily on the designer's experience and intuition to navigate in the trade space of efficiency, noise, linearity, and speed to meet certain specifications. This domain's depth, requiring a blend of acumen and creativity, underscores the high barriers to entry and the extensive training required to master its intricacies, which exacerbated the critical labor shortfall of the semiconductor industry in this decade [Ravi, 2023]. The advent of AI-assisted design automation in analog circuit design holds considerable promise to tackle the aforementioned challenge. It offers the potential to significantly streamline design cycles, enabling engineers to focus more on strategic, high-level design considerations and the exploration of novel ideas and applications.