TableBench

In order to better evaluate the ability of LLM in table question answering, the research teams of Beihang University, University of Waterloo, Fudan University and Beijing Information Science and Technology University jointly released the TableBench table question answering benchmark. The related paper results are "TableBench: A Comprehensive and Complex Benchmark for Table Question Answering".

TableBench is a comprehensive and complex TableQA benchmark with human annotations, containing 886 samples from 18 domains, designed to facilitate fact-checking, numerical reasoning, data analysis, and visualization tasks. In addition, the team has open-sourced the TableInstruct large-scale instruction corpus, which is designed to guide LLM with various reasoning methods.