MetaQA introduces a groundbreaking approach to detect hallucinations in Large Language Models without relying on external knowledge sources.

• Uses metamorphic relations by comparing responses to semantically equivalent questions
• Achieves 91.3% accuracy in detecting factual errors across different LLMs
• Requires zero external data sources while maintaining high reliability
• Adapts automatically to new domains and knowledge areas

This research is critical for security applications where factual accuracy is paramount, helping protect against misinformation and ensuring trustworthy AI deployments in sensitive environments.

Hallucination Detection in Large Language Models with Metamorphic Relations