HC-RAG: Evidence-Centric Retrieval-Augmented Generation over Heterogeneous Financial Filings

The episode discusses 'HC-RAG,' a system for answering questions over complex financial filings. The hosts detail how HC-RAG uses a graph structure to preserve document hierarchy, aligns text and tables, and routes queries by intent. It achieves high accuracy and efficiency on specialized benchmarks.

Open the full episode

More from AI Radio