PankRAG: Enhancing Graph Retrieval via Globally Aware Query Resolution and Dependency-Aware Reranking Mechanism
Ningyuan Li, Junrui Liu, Yi Shan, Minghui Huang, Ziren Gong, Tong Li
TL;DR
PankRAG tackles the bottleneck in graph-based retrieval-augmented generation where latent relations are missed and hallucinations arise from single-phase retrieval. It combines a globally-aware hierarchical planning pathway that reveals parallel and progress relationships with a dependency-aware reranking mechanism that validates retrieved content against resolved sub-question dependencies, employing the combined score $\alpha \times \mathcal{R}_i + \beta \times \mathcal{M}_i$ where $\mathcal{M}_i = \text{Cos}(\text{ori}_i + \text{sub}_i)$ and $\alpha+\beta=1$. Empirical results on ACQ and SCQ benchmarks show consistent improvements over strong baselines, with ablations confirming the contributions of the Path and Ank components. The approach yields better factual grounding, reduced hallucinations, and strong generalization across domains, making it practically impactful for complex retrieval-augmented generation tasks.
Abstract
Recent graph-based RAG approaches leverage knowledge graphs by extracting entities from a query to fetch their associated relationships and metadata. However, relying solely on entity extraction often results in the misinterpretation or omission of latent critical information and relationships. This can lead to the retrieval of irrelevant or contradictory content, as well as the exclusion of essential information, thereby increasing hallucination risks and undermining the quality of generated responses. In this paper, we propose PankRAG, a framework designed to capture and resolve the latent relationships within complex queries that prior methods overlook. It achieves this through a synergistic combination of a globally-aware hierarchical resolution pathway and a dependency-aware reranking mechanism. PankRAG first generates a globally aware resolution pathway that captures parallel and progress relationships, guiding LLMs to resolve queries through a hierarchical reasoning path. Additionally, its dependency-aware reranking mechanism utilizes resolved sub-question dependencies to augment and validate the retrieved content of the current unresolved sub-question. Experimental results demonstrate that PankRAG consistently outperforms existing state-of-the-art methods, underscoring its generalizability.