CEGen: Cause-Effect Graph Generation Using Large Language Models

Black-box testing is vital for software quality assurance, focusing on system behavior without internal details. Cause-Effect Graphs (CEGs) visualize relationships between inputs and outputs, aiding in complex logical scenario testing. Despite their usefulness, creating CEGs is labor-intensive and requires expertise. This paper introduces CEGen, a novel method using Large Language Models (LLMs) to generate CEGs from natural language. CEGen employs LLMs to create truth tables from specifications and constructs CEGs algorithmically, making the process accessible to non-experts. Evaluation on ten problems from a tester training book shows that CEGen successfully generated error-free CEGs in 52% of attempts.