Enhancing Test Generation for Hard-to-Cover Branches via Program Analysis and Large Language Models

TELPA, a novel LLM-based test generation technique, leverages program analysis to enhance the coverage of hard-to-cover branches by extracting real usage scenarios, understanding inter-procedural dependencies, and guiding LLMs with counter-examples.