MARS: Multi-hop Adaptive Retrieval and SPARQL Generation for KGQA

2026-07-16

AI Summary

This work addresses the susceptibility of large language models to hallucination and insufficient reliability in knowledge-intensive tasks by proposing a fine-tuning-free knowledge graph question answering (KGQA) approach. The method employs structured multi-hop adaptive retrieval to dynamically link question entities to a knowledge graph and adaptively determines either the retrieval depth or the final SPARQL query based on question complexity. By integrating multi-hop entity linking, iterative graph traversal, and query generation, the system supports multilingual inputs without requiring model fine-tuning. It achieves state-of-the-art performance on three widely used KGQA benchmarks while maintaining high efficiency and scalability. The associated code and resources have been publicly released.

Abstract

Large language models (LLMs) have demonstrated strong reasoning performance, but their tendency to hallucinate limits their reliability in knowledge-intensive tasks requiring up-to-date and grounded information. Combining knowledge graphs (KGs) with LLMs facilitates the use of explicit symbolic knowledge that can be continuously updated without costly fine-tuning, while benefiting from rapidly advancing LLM reasoning. We propose MARS, a scalable knowledge graph question answering (KGQA) approach that requires no model fine-tuning. Rather than relying on open-ended agentic exploration, MARS performs a structured retrieval procedure that links question entities to the KG and iteratively retrieves relevant next-hop information. At each step, MARS decides whether to continue graph traversal or to generate the final SPARQL query, allowing the model to adapt the retrieval depth to the question while keeping the overall pipeline more predictable than fully agentic approaches. We evaluate MARS on three established KGQA benchmarks across several LLMs and settings, including multilingual evaluation, and provide insights through ablation studies and error analysis. Our approach achieves competitive performance relative to state-of-the-art methods while remaining efficient and scalable. The evaluation results, code and resources are publicly available: https://github.com/dice-group/mars-kgqa.

Problem

• knowledge graph question answering
• large language models
• hallucination
• SPARQL generation
• retrieval

Innovation

• Multi-hop Retrieval
• Adaptive SPARQL Generation
• Knowledge Graph Question Answering
• LLM-KG Integration
• Zero-shot KGQA