<ul><li>Recent advancements in language models have enhanced their ability to reason with tabular data.</li><li>TableRAG is a Retrieval-Augmented Generation (RAG) framework designed for LM-based table understanding.</li><li>TableRAG leverages query expansion and schema/cell retrieval for efficient data encoding and precise retrieval.</li><li>TableRAG achieves the highest retrieval quality and state-of-the-art performance on large-scale table understanding.</li></ul>

TableRAG: Million-Token Table Understanding with Language Models

Discover more