Available at: https://digitalcommons.calpoly.edu/theses/2770
Date of Award
1-2024
Degree Name
MS in Electrical Engineering
Department/Program
Electrical Engineering
College
College of Engineering
Advisor
Andrew Danowitz
Advisor Department
Electrical Engineering
Advisor College
College of Engineering
Abstract
In recent years, there has been an explosion of advancements in artificial intelligence, especially in language models. These models have become essential in aiding and providing information for various tasks. This study explores five proprietary and open-source large language models (LLMs) and examines their reliability and accuracy in selecting parts and constructing connections of ten circuit design tasks from our benchmark. During our investigations, we assessed that the default textual outputs from these LLMs could lead to ambiguous responses that are either too general or open to multiple interpretations. To enhance clarity, we developed an artificial intelligence (AI)-based pipeline that translates responses from LLMs into netlists, eliminating the need for further training or fine-tuning. Our study aims to highlight the reliability and accuracy of the default responses, develop a solution that provides a more explicit netlist description, and compare default and netlist outputs.
Included in
Digital Circuits Commons, Electronic Devices and Semiconductor Manufacturing Commons, Hardware Systems Commons