ER-RAG introduces a novel unified modeling approach to handle multiple heterogeneous data sources in retrieval-augmented generation systems, simplifying complex LLM architectures.

• Creates a unified framework that processes web pages, databases, and knowledge graphs through a single model
• Eliminates the need for source-specific agents in low-resource or black-box environments
• Enables efficient evidence retrieval when information is fragmented across different sources
• Particularly valuable for security applications requiring precise verification from multiple data repositories

This research significantly improves security operations by creating a more robust and streamlined information retrieval system that maintains accuracy while reducing complexity and potential security gaps when accessing diverse data sources.

ER-RAG: Enhance RAG with ER-Based Unified Modeling of Heterogeneous Data Sources