Abstract: Large Language Models (LLMs), such as ChatGPT, have shown impressive performance in various downstream tasks spanning diverse fields. In this talk, I will present our recent work on leveraging LLMs for quality assurance of real-world software systems, encompassing software testing, program repair, and program synthesis. More specifically, I will first talk about how LLMs can be directly applied for both generation- and mutation-based fuzz testing studied for decades, while being fully automated, generalizable, and applicable to challenging domains (including quantum computing systems). Next, I will talk about AlphaRepair, which reformulates the automated program repair (APR) problem as an infilling (or cloze) task and demonstrates that LLMs can directly outperform traditional APR techniques studied for over a decade. Lastly, I will also briefly talk about our recent work on LLM-based program synthesis, including Magicoder and EvalPlus.

Biography: Lingming Zhang is an Associate Professor in the Department of Computer Science at the University of Illinois Urbana-Champaign (UIUC). His main research interests lie in Software Engineering and Programming Languages, as well as their synergy with Machine Learning, including Large Language Models (LLMs) for Code and ML System Reliability. To date, his work has found 1000+ new bugs and vulnerabilities in real-world software systems, including deep learning compilers/libraries, C/C++ compilers, Java virtual machines, operating systems, and even quantum computing systems. He is the recipient of ACM SIGSOFT Early Career Researcher Award, NSF CAREER Award, UIUC Dean’s Award for Excellence in Research, UIUC List of Teachers Ranked as Outstanding, multiple ACM SIGSOFT Distinguished Paper Awards, and various awards/grants from Alibaba, Amazon, Google, Kwai Inc., Meta, NVIDIA, and Samsung. He currently serves as program co-chair for ASE 2025 and LLM4Code 2024, and associate chair for OOPSLA 2024. For more details, please visit: http://lingming.cs.illinois.edu/