Tables as Texts or Images: Evaluating the Table Reasoning Ability of LLMs and MLLMs
This work addresses the challenge of optimizing table reasoning for AI researchers and practitioners, but it is incremental as it focuses on benchmarking and comparing existing methods without introducing new models.
The paper tackled the problem of evaluating how well large language models (LLMs) and multimodal LLMs (MLLMs) reason with tables by testing them across six benchmarks for tasks like question-answering and fact-checking, and found that performance varies significantly based on whether tables are represented as text or images and the prompting strategies used.
In this paper, we investigate the effectiveness of various LLMs in interpreting tabular data through different prompting strategies and data formats. Our analyses extend across six benchmarks for table-related tasks such as question-answering and fact-checking. We introduce for the first time the assessment of LLMs' performance on image-based table representations. Specifically, we compare five text-based and three image-based table representations, demonstrating the role of representation and prompting on LLM performance. Our study provides insights into the effective use of LLMs on table-related tasks.