N2N-GQA: Noise-to-Narrative for Graph-Based Table-Text Question Answering Using LLMs

TL;DR

N2N-GQA addresses the challenge of multi-hop QA over hybrid table-text data by transforming noisy, retrieved documents into a dynamic evidence graph. The approach introduces GraphRank to blend semantic relevance with graph centrality, and uses structured query planning, iterative graph-based evidence gathering, and a Bridge-Aware Hybrid Selector to curate a coherent final context for answer synthesis. Empirically, the zero-shot framework achieves $EM = 48.80$ on OTT-QA, closely approaching fine-tuned methods like CORE and COS, and exhibits substantial gains over vanilla RAG, demonstrating that graph-based evidence organization provides a scalable, interpretable foundation for robust multi-hop QA without task-specific training. The work highlights graph-structured reasoning as a core mechanism for improving retrieval-augmented generation, with practical implications for deploying adaptable QA systems across domains and encouraging future work on richer graph representations and broader applicability.

Abstract

Multi-hop question answering over hybrid table-text data requires retrieving and reasoning across multiple evidence pieces from large corpora, but standard Retrieval-Augmented Generation (RAG) pipelines process documents as flat ranked lists, causing retrieval noise to obscure reasoning chains. We introduce N2N-GQA. To our knowledge, it is the first zeroshot framework for open-domain hybrid table-text QA that constructs dynamic evidence graphs from noisy retrieval outputs. Our key insight is that multi-hop reasoning requires understanding relationships between evidence pieces: by modeling documents as graph nodes with semantic relationships as edges, we identify bridge documents connecting reasoning steps, a capability absent in list-based retrieval. On OTT-QA, graph-based evidence curation provides a 19.9-point EM improvement over strong baselines, demonstrating that organizing retrieval results as structured graphs is critical for multihop reasoning. N2N-GQA achieves 48.80 EM, matching finetuned retrieval models (CORE: 49.0 EM) and approaching heavily optimized systems (COS: 56.9 EM) without any task specific training. This establishes graph-structured evidence organization as essential for scalable, zero-shot multi-hop QA systems and demonstrates that simple, interpretable graph construction can rival sophisticated fine-tuned approaches.

Figures (1)

• Figure 1: Overview of N2N-GQA. The pipeline begins with structured query planning, followed by iterative evidence gathering where documents are modeled as a dynamic graph. The "noise-to-narrative" process uses centrality-based pruning to identify bridge documents, resulting in a curated evidence graph for final answer synthesis.