AI-Driven Testing: Unleashing Autonomous Systems for Superior Software Quality Using Generative AI

In software development, achieving comprehensive code coverage is imperative for ensuring the reliability and robustness of programs. This paper presents a novel approach to attaining between 80% and 100% code coverage in C programs by leveraging Large Language Models (LLMS) for test case generation and subsequently measuring the coverage using the GCOV tool. The methodology outlined in this research harnesses the capabilities of LLMS, such as GPT-3.5-turbo, Code llama, and llama-2 to autonomously generate diverse and complex test cases tailored to the intricacies of C code. Through empirical experimentation and case studies, we demonstrate the feasibility and efficacy of our approach in achieving from 80%-100% code coverage in a variety of C programs across different domains. Furthermore, we discuss the implications of our methodology in enhancing software quality, reducing the likelihood of undiscovered defects, and ultimately improving the reliability and maintainability of C codebases.