A Survey on Advances in Retrieval-Augmented Generation over Tabular Data and Table QA

Hassan Soliman, Iryna Gurevych

Keywords: TableQA, Retrieval-Augmented Generation, Multi-Table Retrieval, Multimodal Table Retrieval, Generative Information Retrieval

TL;DR: The paper surveys progress in retrieval-augmented generation (RAG) and question answering (QA) over tables, highlighting trends and limitations in these areas.

Abstract: Recent advancements in retrieval-augmented generation (RAG) and question answering (QA) over tabular data have demonstrated significant potential in addressing challenges in data retrieval, semantic understanding, and complex reasoning. This work reviews key trends, insights, and limitations across various domains, emphasizing advancements in TableQA, multi-table retrieval, multimodal table retrieval, and generative information retrieval (GenIR). These developments are critical for improving machine interaction with structured datasets, paving the way for scalable and accurate decision-making tools in real-world applications.

Submission Number: 6