TableEval: A Real-World Benchmark for Complex, Multilingual, and Multi-Structured Table Question Answering

TL;DR

TableEval is a comprehensive benchmark designed to evaluate large language models on complex, multilingual, and multi-structured table question answering tasks, addressing real-world challenges and providing a new evaluation framework.

Contribution

The paper introduces TableEval, a realistic, multi-domain, multilingual TableQA benchmark with a novel semantic accuracy metric, SEAT, to better assess LLM performance on complex table reasoning.

Findings

State-of-the-art LLMs show significant gaps in complex TableQA tasks.

SEAT correlates highly with human judgment, improving evaluation accuracy.

Tables from diverse domains and languages reveal limitations of current models.

Abstract

LLMs have shown impressive progress in natural language processing. However, they still face significant challenges in TableQA, where real-world complexities such as diverse table structures, multilingual data, and domain-specific reasoning are crucial. Existing TableQA benchmarks are often limited by their focus on simple flat tables and suffer from data leakage. Furthermore, most benchmarks are monolingual and fail to capture the cross-lingual and cross-domain variability in practical applications. To address these limitations, we introduce TableEval, a new benchmark designed to evaluate LLMs on realistic TableQA tasks. Specifically, TableEval includes tables with various structures (such as concise, hierarchical, and nested tables) collected from four domains (including government, finance, academia, and industry reports). Besides, TableEval features cross-lingual scenarios with tables in Simplified Chinese, Traditional Chinese, and English. To minimize the risk of data leakage, we collect all data from recent real-world documents. Considering that existing TableQA metrics fail to capture semantic accuracy, we further propose SEAT, a new evaluation framework that assesses the alignment between model responses and reference answers at the sub-question level. Experimental results have shown that SEAT achieves high agreement with human judgment. Extensive experiments on TableEval reveal critical gaps in the ability of state-of-the-art LLMs to handle these complex, real-world TableQA tasks, offering insights for future improvements. We make our dataset available here: https://github.com/wenge-research/TableEval.