Asymmetric dynamic routing improves efficiency in knowledge graph queries
Asymmetric Dynamic Routing: Balancing Reasoning Depth and Computational Efficiency in Hypergraph RAG
Information Retrieval
Summary
Large language models sometimes get confused or waste effort when answering different kinds of questions because they treat all queries the same way. The authors found that using a flexible approach that changes how it searches a knowledge graph based on the question’s complexity helps fix this. Their method, called asymmetric dynamic routing, decides between three ways of searching to balance accuracy and speed. Tests show it keeps answers good while using less computing time and fewer data tokens.
What this means in practice
- •For enterprise ai teams: Reduce computing costs and speed up answer retrieval for complex knowledge-base queries by adapting search strategies based on query difficulty.
- •For robotics control engineers: Improve real-time decision making by efficiently accessing hierarchical knowledge graphs with adaptive traversal for different reasoning needs.
Authors
Qi Sun, Yijia Zhang, Xingliang Hou, Caibo Li, Qiang Li, Yu Guo
Abstract
While graph-based and hypergraph-based Retrieval-Augmented Generation (RAG) significantly mitigate hallucinations in Large Language Models (LLMs), existing structure-based RAG systems typically adopt static traversal strategies regardless of the query complexity. We identify this ``static retrieval fallacy'' as a primary source of computational redundancy for simple queries and cognitive context gaps for complex reasoning tasks. To balance reasoning quality and inference efficiency, we propose Asymmetric Dynamic Routing (ADR), an intent-conditioned retrieval framework operating over hierarchical knowledge graphs. ADR employs a lightweight structured classifier to dynamically dispatch queries among three asymmetric topological traversal operators: localized fact anchoring, bottom-up adjacency diffusion, and top-down insight grounding, which collectively enable bidirectional information flow across hierarchical knowledge layers. Extensive empirical evaluations across five domain-specific corpora demonstrate that ADR maintains strong reasoning performance while reducing prompt token consumption by up to 48.7\% and end-to-end query latency by 45.3\%, yielding a favorable quality--efficiency trade-off for query-adaptive Hypergraph RAG.