KGVoyager: Knowledge Graph Agnostic Question Answering via Agentic Navigation

arXiv:2609.01780v2 Announce Type: replace Abstract: Knowledge Graph Question Answering (KGQA) over RDF graphs remains challenging in domain-specific settings, where formal ontologies and curated text-SPARQL pairs are often unavailable. We present KG Voyager, a KG-agnostic agentic architecture that generates SPARQL queries from natural language questions by dynamically discovering graph structure and semantics, requiring only a query endpoint of the underlying graph. Using a think-act-observe loop with search, exploration, and execution tools, KG Voyager maps terms to graph Internationalized Resource Identifiers (IRIs), uncovers structure, and refines queries through execution feedback -- all without pre-existing ontologies or examples. Unlike the prior state of the art, KG Voyager requires only a lightweight class index which renders it applicable for far more real-world endpoints. Across four benchmarks, KG Voyager improves F1 by $\sim$4 points on average compared to the state of the art while cutting cost by $\sim$38% and runtime by $\sim$24%.