A Business Process-Centric Approach for LLM-Driven Functional Test Generation

Key Points

• The research aims to enhance software testing by integrating large language models with rigorous testing methodologies.
• Developed a hybrid framework combining LLMs and graph-based navigation.
• Utilized combinatorial optimization techniques such as Pairwise and Boundary Value Analysis (BVA).
• Focused on generating tests for long-horizon business processes while preserving context.
• Achieved high-fidelity test generation for complex systems.
• Maintained state continuity during the testing process.
• Improved the effectiveness of functional testing in industrial applications.