Topo-RAG: Topology-aware retrieval for hybrid text-table documents

TL;DR

Topo-RAG introduces a topology-aware retrieval framework that preserves the structural relationships in hybrid text-table documents, significantly improving retrieval performance over linearization methods.

Contribution

It proposes a dual architecture that separately processes narrative and tabular data, maintaining their topology for more effective retrieval in complex enterprise datasets.

Findings

Achieves 18.4% improvement in nDCG@10 on hybrid queries

Demonstrates the importance of preserving data topology in retrieval

Outperforms linearization-based retrieval methods

Abstract

In enterprise datasets, documents are rarely pure. They are not just text, nor just numbers; they are a complex amalgam of narrative and structure. Current Retrieval-Augmented Generation (RAG) systems have attempted to address this complexity with a blunt tool: linearization. We convert rich, multidimensional tables into simple Markdown-style text strings, hoping that an embedding model will capture the geometry of a spreadsheet in a single vector. But it has already been shown that this is mathematically insufficient. This work presents Topo-RAG, a framework that challenges the assumption that "everything is text". We propose a dual architecture that respects the topology of the data: we route fluid narrative through traditional dense retrievers, while tabular structures are processed by a Cell-Aware Late Interaction mechanism, preserving their spatial relationships. Evaluated on SEC-25, a synthetic enterprise corpus that mimics real-world complexity, Topo-RAG demonstrates an 18.4% improvement in nDCG@10 on hybrid queries compared to standard linearization approaches. It's not just about searching better; it's about understanding the shape of information.