Optimizing Context and Cost in LLM ‐Based Unit Test Generation: A Study on External Dependency Retrieval Strategies
A systematic empirical study of multiple strategies for context enrichment and optimization in LLM‐based unit test generation, conducted on seven diverse projects (three open‐source and four proprietary industrial systems), encompassing 261 distinct methods establish this optimized context strategy as a cost‐effective solution for scalable, industrial‐grade automated test generation.