CITYWALK is an innovative approach that significantly improves automated unit test generation for C++ code using large language models by incorporating project-specific context and language knowledge.

• Addresses the unique challenges of compiled languages like C++ that previous LLM test generators overlooked
• Leverages dependency awareness to understand project relationships and structures
• Incorporates C++-specific knowledge to handle complex language features
• Demonstrates superior test quality compared to vanilla LLM approaches

This research matters for Engineering teams by reducing the time-consuming burden of manual unit test creation while improving overall code quality and reliability in C++ codebases.

CITYWALK: Enhancing LLM-Based C++ Unit Test Generation via Project-Dependency Awareness and Language-Specific Knowledge