Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Survey of Table Reasoning with Large Language Models

Xuanliang Zhang, Dingzirui Wang, Longxu Dou, Qingfu Zhu, Wanxiang Che

TL;DR

The survey addresses how Large Language Models (LLMs) enable table reasoning across tasks such as table QA, table fact verification, table-to-text, and text-to-SQL, contrasting them with the pre-LLM era. It categorizes techniques into two classes: those that extend pre-LLM methods (supervised fine-tuning and result ensemble) and those unique to LLMs (in-context learning, instruction design, and step-by-step reasoning), and analyzes their effectiveness and prevalence across benchmarks. The authors explain why LLMs excel, highlighting improved structure understanding from instruction following and enhanced schema linking via decomposed reasoning, and provide a roadmap for future work, including data diversification, automatic prompt optimization, tool use, RAG, and multimodal/dialogue-enabled extensions. The paper contributes a structured synthesis, comparative insights across benchmarks, and a resource repository to guide ongoing research and practical deployment in table reasoning with LLMs.

Abstract

Table reasoning, which aims to generate the corresponding answer to the question following the user requirement according to the provided table, and optionally a text description of the table, effectively improving the efficiency of obtaining information. Recently, using Large Language Models (LLMs) has become the mainstream method for table reasoning, because it not only significantly reduces the annotation cost but also exceeds the performance of previous methods. However, existing research still lacks a summary of LLM-based table reasoning works. Due to the existing lack of research, questions about which techniques can improve table reasoning performance in the era of LLMs, why LLMs excel at table reasoning, and how to enhance table reasoning abilities in the future, remain largely unexplored. This gap significantly limits progress in research. To answer the above questions and advance table reasoning research with LLMs, we present this survey to analyze existing research, inspiring future work. In this paper, we analyze the mainstream techniques used to improve table reasoning performance in the LLM era, and the advantages of LLMs compared to pre-LLMs for solving table reasoning. We provide research directions from both the improvement of existing methods and the expansion of practical applications to inspire future research.

A Survey of Table Reasoning with Large Language Models

TL;DR

The survey addresses how Large Language Models (LLMs) enable table reasoning across tasks such as table QA, table fact verification, table-to-text, and text-to-SQL, contrasting them with the pre-LLM era. It categorizes techniques into two classes: those that extend pre-LLM methods (supervised fine-tuning and result ensemble) and those unique to LLMs (in-context learning, instruction design, and step-by-step reasoning), and analyzes their effectiveness and prevalence across benchmarks. The authors explain why LLMs excel, highlighting improved structure understanding from instruction following and enhanced schema linking via decomposed reasoning, and provide a roadmap for future work, including data diversification, automatic prompt optimization, tool use, RAG, and multimodal/dialogue-enabled extensions. The paper contributes a structured synthesis, comparative insights across benchmarks, and a resource repository to guide ongoing research and practical deployment in table reasoning with LLMs.

Abstract

Table reasoning, which aims to generate the corresponding answer to the question following the user requirement according to the provided table, and optionally a text description of the table, effectively improving the efficiency of obtaining information. Recently, using Large Language Models (LLMs) has become the mainstream method for table reasoning, because it not only significantly reduces the annotation cost but also exceeds the performance of previous methods. However, existing research still lacks a summary of LLM-based table reasoning works. Due to the existing lack of research, questions about which techniques can improve table reasoning performance in the era of LLMs, why LLMs excel at table reasoning, and how to enhance table reasoning abilities in the future, remain largely unexplored. This gap significantly limits progress in research. To answer the above questions and advance table reasoning research with LLMs, we present this survey to analyze existing research, inspiring future work. In this paper, we analyze the mainstream techniques used to improve table reasoning performance in the LLM era, and the advantages of LLMs compared to pre-LLMs for solving table reasoning. We provide research directions from both the improvement of existing methods and the expansion of practical applications to inspire future research.
Paper Structure (37 sections, 4 figures, 2 tables)

This paper contains 37 sections, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Paper Selection Criteria
  4. Task Definition
  5. Benchmarks
  6. What Techniques Can Improve Table Reasoning Performance in the LLM Era
  7. Mainstream Techniques Following pre-LLMs
  8. Supervised Fine-Tuning
  9. Highlight
  10. Result Ensemble
  11. Highlight
  12. Mainstream Techniques Unique to LLMs
  13. In-context Learning
  14. Highlight
  15. Instruction Design
  16. ...and 22 more sections

Figures (4)

  • Figure 1: The illustration of various table reasoning tasks.
  • Figure 2: The structure overview of our paper, taking the most representative works as an example.
  • Figure 3: The mainstream techniques that can be utilized to improve table reasoning performance in the LLM era.
  • Figure 4: The research trend using different techniques over months. #Paper denotes the number of papers.