Metamorphic Testing

Metamorphic testing is a software testing technique that addresses the challenge of verifying program correctness when traditional "oracle" methods (comparing outputs to known correct results) are unavailable or impractical. Current research focuses on applying metamorphic testing to diverse domains, including autonomous systems (e.g., self-driving cars, drones), large language models (LLMs), and image processing (e.g., deepfake detection, hand pose estimation), often leveraging metamorphic relations to identify inconsistencies in model behavior under various input transformations. This approach is significant because it enables robust evaluation of complex systems where defining ground truth is difficult, improving the reliability and safety of AI-driven applications and software.