Research on trustworthy Software Testing Techniques Based on Large Models

This paper focuses on trustworthy software testing techniques enabled by large language models. We first give an overview of large language models and their applications in software engineering. Then we analyze the challenges of applying these models to software testing, which include overfitting, bias and trust issues. We evaluate LMTester on both synthetic and real-world programs compared with baseline testing tools. The results demonstrate its advantages in improving test coverage, detecting bugs, and assuring trustworthiness. We also discuss the limitations of current techniques and point out several future directions. Our work is expected to facilitate trustworthy applications of large language models in software testing and quality assurance.