AnalogAgent AI Framework Achieves 97.4% Success Rate in Automated Analog Circuit Design

Researchers developed AnalogAgent, an LLM-based multi-agent system that automates analog circuit design with 97.4% Pass@1 accuracy using GPT-5 and 92% with Gemini. The framework uses self-evolving memory to retain technical details and improve across tasks without requiring additional expert input, significantly boosting performance in smaller models like Qwen-8B.