What Inputs Drive Effective Large Language Model-Based Unit Test Generation?

Large language models (LLMs) have revolutionized software engineering by automating critical tasks. We study five state-of-the-art LLMs, investigating their capabilities in generating unit test cases while focusing on how different inputs impact test correctness, bug detection capability, and code coverage.