Lee receives competitive Amazon Research Award

Department of Electrical and Computer Engineering Assistant Professor Kangwook Lee recently received a globally-recognized Amazon Research Award. Lee is one of 98 recipients who represent 51 universities across 15 countries. In addition to funding, Lee will have access to over 300 Amazon public datasets and AWS AI/ML services and tools for his research: Information and Coding Theory-Based Framework for Prompt Engineering.

“Many people are now familiar with large language models (LLMs) like ChatGPT and how slight changes in the design of ‘prompts’ can significantly affect the quality of responses. While there are many prompting tips and research papers available, most existing methods are empirically designed and lack strong theoretical foundations. In my recent work, I have been exploring whether we can understand prompting through the lens of information and coding theory, a classical field that has revolutionized communication systems. This project aims to develop more theoretically principled methods for prompt engineering, ” Lee explains.

Amazon Research Awards offer unrestricted funds and AWS Promotional Credits to academic researchers exploring diverse research topics across multiple disciplines. Lee explains why these funds are important to his research in particular, “To validate our ideas and develop new theories, we need to run many experiments with LLMs. These experiments are highly costly and difficult to run on a lab-scale compute cluster. Amazon’s support provides my research group with the necessary computing resources and datasets, which is incredibly helpful.”

Proposals within six categories (AI for Information Security, Automated Reasoning, AWS AI, AWS Cryptography and Privacy, AWS Database Services, and Sustainability) were reviewed for the quality of their scientific content and their potential to impact both the research community and society. Lee believes his research in the AWS AI category has future implications, “LLMs will become increasingly prevalent in the coming years, and prompt engineering will continue to play an important role. I am excited to develop theoretically principled prompt engineering methods through this project.”