Orthogonal Hierarchical Decomposition for Structure-Aware Table Understanding with Large Language Models

TL;DR

This paper introduces Orthogonal Hierarchical Decomposition (OHD), a novel method for representing complex tables with hierarchical structures to improve understanding and reasoning by large language models, outperforming existing approaches.

Contribution

The paper proposes the OHD framework with Orthogonal Tree Induction and a dual-pathway association protocol to explicitly capture hierarchical dependencies in complex tables for LLMs.

Findings

OHD outperforms existing methods on AITQA and HiTab benchmarks.

The orthogonal tree decomposition effectively captures hierarchical table structures.

Semantic alignment improves question answering accuracy.

Abstract

Complex tables with multi-level headers, merged cells and heterogeneous layouts pose persistent challenges for LLMs in both understanding and reasoning. Existing approaches typically rely on table linearization or normalized grid modeling. However, these representations struggle to explicitly capture hierarchical structures and cross-dimensional dependencies, which can lead to misalignment between structural semantics and textual representations for non-standard tables. To address this issue, we propose an Orthogonal Hierarchical Decomposition (OHD) framework that constructs structure-preserving input representations of complex tables for LLMs. OHD introduces an Orthogonal Tree Induction (OTI) method based on spatial--semantic co-constraints, which decomposes irregular tables into a column tree and a row tree to capture vertical and horizontal hierarchical dependencies, respectively. Building on this representation, we design a dual-pathway association protocol to symmetrically reconstruct semantic lineage of each cell, and incorporate an LLM as a semantic arbitrator to align multi-level semantic information. We evaluate OHD framework on two complex table question answering benchmarks, AITQA and HiTab. Experimental results show that OHD consistently outperforms existing representation paradigms across multiple evaluation metrics.