Walk&Retrieve: Simple Yet Effective Zero-shot Retrieval-Augmented Generation via Knowledge Graph Walks

2025-05-22

AI Summary

To address knowledge hallucination and obsolescence in large language models (LLMs), this paper proposes a zero-shot knowledge graph (KG)-enhanced generation framework. The method eliminates reliance on representation alignment or fine-tuning, introducing instead an unsupervised paradigm for constructing high-quality retrieval corpora—based on lightweight random walks over KGs and knowledge verbalization—to dynamically adapt to evolving KGs. The resulting zero-shot retrieval-augmented generation (RAG) paradigm achieves high retrieval accuracy, low latency, and scalability to billion-scale triplets. Experiments demonstrate that our approach significantly outperforms mainstream RAG baselines in response accuracy and hallucination suppression, while reducing query latency by a substantial margin.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Large Language Models (LLMs) have showcased impressive reasoning abilities, but often suffer from hallucinations or outdated knowledge. Knowledge Graph (KG)-based Retrieval-Augmented Generation (RAG) remedies these shortcomings by grounding LLM responses in structured external information from a knowledge base. However, many KG-based RAG approaches struggle with (i) aligning KG and textual representations, (ii) balancing retrieval accuracy and efficiency, and (iii) adapting to dynamically updated KGs. In this work, we introduce Walk&Retrieve, a simple yet effective KG-based framework that leverages walk-based graph traversal and knowledge verbalization for corpus generation for zero-shot RAG. Built around efficient KG walks, our method does not require fine-tuning on domain-specific data, enabling seamless adaptation to KG updates, reducing computational overhead, and allowing integration with any off-the-shelf backbone LLM. Despite its simplicity, Walk&Retrieve performs competitively, often outperforming existing RAG systems in response accuracy and hallucination reduction. Moreover, it demonstrates lower query latency and robust scalability to large KGs, highlighting the potential of lightweight retrieval strategies as strong baselines for future RAG research.

Problem

• Aligning knowledge graph and textual representations effectively
• Balancing retrieval accuracy and efficiency in RAG
• Adapting to dynamically updated knowledge graphs seamlessly

Innovation

• Leverages walk-based graph traversal for retrieval
• Uses knowledge verbalization for corpus generation
• Requires no fine-tuning for domain adaptation