The Planning-Execution Module (PEM) is a key component of GUITester, a multi-agent framework designed for autonomous exploratory GUI testing. Its primary function is to decouple the navigation tasks of an agent from the verification process. PEM achieves this by proactively probing for potential software defects through the integration of 'testing intents' directly into its planning and execution strategy. This mechanism directly addresses the challenge of 'Goal-Oriented Masking,' where traditional agents might overlook anomalies by prioritizing the completion of their primary task. By embedding defect-seeking behaviors, PEM significantly enhances an agent's ability to autonomously discover and report defects, making it crucial for improving software quality assurance in complex graphical user interfaces, particularly for MLLM agents.
The Planning-Execution Module (PEM) helps AI agents find bugs in software by actively looking for problems instead of just completing tasks. It's part of a system called GUITester that makes AI better at testing apps by separating the job of moving around from the job of checking for errors.
PEM
Was this definition helpful?