AaLLM: An End-to-End Analog Circuit Design Framework from Topology Generation to Sizing Using Large Language Models
AaLLM claims to take circuit specs straight through to a netlist, covering both topology choice and sizing.
The paper describes an open-source multi-agent LLM workflow for analog circuit design. It builds a knowledge base from papers and textbooks, then uses RAG to support design decisions. Its Designer, Critic, and Evaluator agents are meant to reduce sizing back-and-forth and SPICE calls. The authors report comparable or better figures of merit for generated topologies, 3x-4.5x fewer SPICE calls than SOTA multi-agent LLM pipelines, and a 40x wall-clock time reduction. Source: ArXiv · AI/CL/LG's note.
The paper describes an open-source multi-agent LLM workflow for analog circuit design. It builds a knowledge base from papers and textbooks, then uses RAG to support design decisions. Its Designer, Critic, and Evaluator agents are meant to reduce sizing back-and-forth and SPICE calls. The authors report comparable or better figures of merit for generated topologies, 3x-4.5x fewer SPICE calls than SOTA multi-agent LLM pipelines, and a 40x wall-clock time reduction. Source: ArXiv · AI/CL/LG's note.
score 5