Table Question Answering and Retrieval-Augmented Generation

The field of table question answering and retrieval-augmented generation is moving towards more advanced and adaptable methods. Recent developments have focused on improving the ability of large language models to understand and reason about tabular data, with a particular emphasis on multi-modal and domain-specific approaches. This includes the development of new frameworks and techniques for generating domain-grounded question-answer-context triples, as well as methods for detecting errors and inconsistencies in tabular data. Notable papers in this area include: VeritasFi, which presents a groundbreaking framework for multi-modal financial question answering. SG-XDEAT, which proposes a novel framework for supervised learning on tabular data that integrates sparsity-guided cross-dimensional and cross-encoding attention with target-aware conditioning.

Sources

Table Question Answering in the Era of Large Language Models: A Comprehensive Survey of Tasks, Methods, and Evaluation

VeritasFi: An Adaptable, Multi-tiered RAG Framework for Multi-modal Financial Question Answering

Domain-Specific Data Generation Framework for RAG Adaptation

Towards Cross-Modal Error Detection with Tables and Images

SG-XDEAT: Sparsity-Guided Cross-Dimensional and Cross-Encoding Attention with Target-Aware Conditioning in Tabular Learning

FreshTab: Sourcing Fresh Data for Table-to-Text Generation Evaluation

Built with on top of