Retrieval-Augmented Generation (RAG) has become a key paradigm to enhance Large Language Models (LLMs) with external knowledge. While most RAG systems focus on text corpora, real-world information is often stored in tables across web pages, Wikipedia, and relational databases. Existing methods struggle to retrieve and reason across multiple heterogeneous tables.
For MultiTableQA, we release a comprehensive benchmark, including five different… See the full description on the dataset page:
https://huggingface.co/datasets/jiaruz2/MultiTableQA_TabFact.
