Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ST-Raptor: An Agentic System for Semi-Structured Table QA

Jinxiu Qu, Zirui Tang, Hongzhang Huang, Boyu Niu, Wei Zhou, Jiannan Wang, Yitong Song, Guoliang Li, Xuanhe Zhou, Fan Wu

TL;DR

The paper addresses the challenge of QA over semi-structured tables where conventional flattening and flat headers lose semantic structure. It proposes ST-Raptor, an agentic system that builds a layout-aware HO-Tree from raw tables via a Table2Tree pipeline combining multimodal extraction and embedding alignment, and supports interactive editing and multi-turn reasoning through an Orchestration Agent. A nine-core-tree-operation framework plus a column-type-aware sub-operation decomposition enable precise, faithful query execution with forward and backward verification. Empirical results on SSTQA and WikiTQ-ST show ST-Raptor outperforms strong baselines by meaningful margins (e.g., 11.2% on accuracy) and demonstrates improved usability via interactive UI and human-in-the-loop editing. The work advances practical semi-structured table QA by preserving hierarchical structure, enabling user control, and providing robust multi-turn multimodal QA.

Abstract

Semi-structured table question answering (QA) is a challenging task that requires (1) precise extraction of cell contents and positions and (2) accurate recovery of key implicit logical structures, hierarchical relationships, and semantic associations encoded in table layouts. In practice, such tables are often interpreted manually by human experts, which is labor-intensive and time-consuming. However, automating this process remains difficult. Existing Text-to-SQL methods typically require converting semi-structured tables into structured formats, inevitably leading to information loss, while approaches like Text-to-Code and multimodal LLM-based QA struggle with complex layouts and often yield inaccurate answers. To address these limitations, we present ST-Raptor, an agentic system for semi-structured table QA. ST-Raptor offers an interactive analysis environment that combines visual editing, tree-based structural modeling, and agent-driven query resolution to support accurate and user-friendly table understanding. Experimental results on both benchmark and real-world datasets demonstrate that ST-Raptor outperforms existing methods in both accuracy and usability. The code is available at https://github.com/weAIDB/ST-Raptor, and a demonstration video is available at https://youtu.be/9GDR-94Cau4.

ST-Raptor: An Agentic System for Semi-Structured Table QA

TL;DR

The paper addresses the challenge of QA over semi-structured tables where conventional flattening and flat headers lose semantic structure. It proposes ST-Raptor, an agentic system that builds a layout-aware HO-Tree from raw tables via a Table2Tree pipeline combining multimodal extraction and embedding alignment, and supports interactive editing and multi-turn reasoning through an Orchestration Agent. A nine-core-tree-operation framework plus a column-type-aware sub-operation decomposition enable precise, faithful query execution with forward and backward verification. Empirical results on SSTQA and WikiTQ-ST show ST-Raptor outperforms strong baselines by meaningful margins (e.g., 11.2% on accuracy) and demonstrates improved usability via interactive UI and human-in-the-loop editing. The work advances practical semi-structured table QA by preserving hierarchical structure, enabling user control, and providing robust multi-turn multimodal QA.

Abstract

Semi-structured table question answering (QA) is a challenging task that requires (1) precise extraction of cell contents and positions and (2) accurate recovery of key implicit logical structures, hierarchical relationships, and semantic associations encoded in table layouts. In practice, such tables are often interpreted manually by human experts, which is labor-intensive and time-consuming. However, automating this process remains difficult. Existing Text-to-SQL methods typically require converting semi-structured tables into structured formats, inevitably leading to information loss, while approaches like Text-to-Code and multimodal LLM-based QA struggle with complex layouts and often yield inaccurate answers. To address these limitations, we present ST-Raptor, an agentic system for semi-structured table QA. ST-Raptor offers an interactive analysis environment that combines visual editing, tree-based structural modeling, and agent-driven query resolution to support accurate and user-friendly table understanding. Experimental results on both benchmark and real-world datasets demonstrate that ST-Raptor outperforms existing methods in both accuracy and usability. The code is available at https://github.com/weAIDB/ST-Raptor, and a demonstration video is available at https://youtu.be/9GDR-94Cau4.
Paper Structure (6 sections, 2 figures)

This paper contains 6 sections, 2 figures.

Table of Contents

  1. INTRODUCTION
  2. ST-RAPTOR OVERVIEW
  3. DEMONSTRATION
  4. End-to-End Experience
  5. Scenario 1: Interactive Tree Editing for Query Reorientation
  6. Scenario 2: Complex Query Execution on Real-World Tables

Figures (2)

  • Figure 1: Unique Challenges of Semi-Structured Table QA -- (1) Flattening Nested Tables Causes Semantic Misalignment; (2) Dehierarchizing Multi-Level Headers Obscures Structural Relationships; (3) Fragmenting Merged Cells Breaks Original Semantics.
  • Figure 2: System Overview of ST-Raptor.