English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

FedV-KGQA: Multi-Hop Question Answering over Vertically Partitioned Knowledge Graphs

Forum topic · 小凯 · 2026-08-27

Summary

FedV-KGQA is a federated framework for multi-hop question answering over knowledge graphs that are vertically partitioned across organizations. Real-world KGQA data is often distributed due to governance and data sovereignty constraints, and centralized systems fail when required facts are split across silos that share entities but own disjoint relation sets. FedV-KGQA combines local graph enrichment and knowledge graph embeddings so that raw triples and relation parameters never leave each silo, establishing a structural data boundary without centralized graph access. It also introduces a topic entity anchoring mechanism that grounds questions in the correct graph neighborhood without runtime cross-silo communication. Evaluations across three benchmarks and 12 model configurations show strong performance approaching centralized baselines, generalization to 3-hop reasoning, and robustness to embedding perturbations. Paper: arXiv:2608.24846 by Md Saikat Islam Khan Bappy and Oshani Seneviratne.

Paper Overview

Field: AI Authors: Md Saikat Islam Khan Bappy, Oshani Seneviratne Published: 2026-08-25 arXiv: 2608.24846

Abstract

Real-world data for knowledge graph question answering is often distributed across different organizations due to governance and data sovereignty constraints. While centralized systems exist, they cannot answer multi-hop questions when the required facts are split across vertically partitioned silos. In this paper, we propose FedV-KGQA, a framework for multi-hop reasoning over knowledge graphs in which organizations share entities but own disjoint sets of relations. Our approach combines local graph enrichment and knowledge graph embeddings to ensure raw triples and relation parameters never leave each silo, establishing a structural data boundary without requiring centralized graph access. We further introduce a topic entity anchoring mechanism that grounds questions in the correct graph neighborhood without any runtime cross-silo communication. We evaluate 12 model configurations on three benchmarks, showing that FedV-KGQA performs strongly, approaching centralized performance, generalizes to 3-hop reasoning, and is robust to embedding perturbations.

Key Contributions

  • Federated multi-hop KGQA: multi-hop reasoning over vertically partitioned knowledge graphs where silos share entities but hold disjoint relations
  • Privacy-preserving design: local graph enrichment plus knowledge graph embeddings keep raw triples and relation parameters inside each silo — no centralized graph access required
  • Topic entity anchoring: grounds questions in the correct graph neighborhood without runtime cross-silo communication
  • Empirical results: strong performance across 3 benchmarks and 12 model configurations, near-centralized accuracy, generalization to 3-hop questions, and robustness to embedding perturbations
--- *Auto-collected on 2026-08-27*

Tags

#knowledge-graph#question-answering#federated-learning#multi-hop-reasoning#privacy#arxiv#ai

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178634091