Texts or Images? A Fine-grained Analysis on the Effectiveness of Input Representations and Models for Table Question Answering

TL;DR

This paper systematically compares text and image representations of tables for question answering, revealing that the optimal approach varies with question complexity and table size, and introduces FRES for dynamic representation selection.

Contribution

It provides the first controlled study on table representation effectiveness in TQA and proposes FRES, a method for dynamic selection of representations, improving performance.

Findings

Best representation-model combinations vary across setups.

FRES improves performance by 10% on average.

Controlled analysis clarifies when each representation is most effective.

Abstract

In table question answering (TQA), tables are encoded as either texts or images. Prior work suggests that passing images of tables to multi-modal large language models (MLLMs) performs comparably to or even better than using textual input with large language models (LLMs). However, the lack of controlled setups limits fine-grained distinctions between these approaches. In this paper, we conduct the first controlled study on the effectiveness of several combinations of table representations and models from two perspectives: question complexity and table size. We build a new benchmark based on existing TQA datasets. In a systematic analysis of seven pairs of MLLMs and LLMs, we find that the best combination of table representation and model varies across setups. We propose FRES, a method selecting table representations dynamically, and observe a 10% average performance improvement compared to using both representations indiscriminately.