DataMosaic introduces a novel Extract-Reason-Verify framework that addresses key limitations in LLM-based data analytics by making results both explainable and verifiable.

• Employs a multi-agent architecture to extract data chunks, reason with them, and verify outputs
• Overcomes limitations of traditional RAG approaches when dealing with noisy or multi-modal data
• Provides transparent reasoning paths that users can inspect and validate
• Significantly reduces hallucinations through explicit verification mechanisms

This research enhances security posture by ensuring trustworthiness in AI-driven analytics, reducing the risk of making decisions based on hallucinated or erroneous outputs—critical for sensitive data environments.

DataMosaic: Explainable and Verifiable Multi-Modal Data Analytics through Extract-Reason-Verify