KGCache：基于大型语言模型的知识图谱推理的摊销子图检索
KGCache: Amortized Subgraph Retrieval for KG Reasoning with LLMs
浏览论文内容
中文总结 AI 辅助
本研究针对KGQA中重复检索知识图谱邻域的问题，提出兼容ToG与RoG范式的KGCache，经在WebQSP和CWQ上评估，可显著加速知识图谱检索。
中文摘要 AI 辅助
大型语言模型在结合知识图谱时能更可靠地回答知识密集型问题，但Think-on-Graph（ToG）、Reasoning-on-Graph（RoG）等系统会在不同问题中重复查询知识图谱的同一邻域。本研究针对知识图谱问答（KGQA）工作负载中的这种重复检索问题，提出KGCache，这是一种用于存储知识图谱一跳邻域的内存缓存。KGCache设计为兼容迭代遍历（ToG）和一次性规划（RoG）两种KGQA范式，部署在KGQA引擎与知识图谱后端之间，可将重复的实体请求从缓存中提供，而非发起新的知识图谱查询。我们在WebQSP和CWQ数据集上，使用LRU、LFU及感知轨迹的Oracle策略评估KGCache。分析显示，两个数据集在起始实体与遍历过程中到达的实体间存在大量实体复用。我们还探索了针对相似查询的语义缓存，其在WebQSP上可进一步提升命中率，但在CWQ上需进一步测试准确性。实体缓存可将知识图谱检索加速最高1.91倍，语义上下文缓存在评估的WebQSP配置中可实现最高1.06倍的全系统加速，每次命中的速度最高可达3.73倍。
英文摘要
Large language models can answer knowledge-intensive questions more reliably when they are grounded with knowledge graphs, but systems such as Think-on-Graph and Reasoning-on-Graph repeatedly query the same graph neighborhoods across different questions. In this work, we study this repeated retrieval in Knowledge Graph Question Answering~(KGQA) workloads and propose KGCache, an in-memory cache for one-hop knowledge graph neighborhoods. KGCache is designed to be compatible with both iterative traversal (ToG) and one shot planning (RoG) KGQA paradigms. KGCache is placed between the KGQA engine and the backend serving the KG, so repeated entity requests can be served from cache instead of issuing new KG queries. We evaluate KGCache on WebQSP and CWQ using LRU, LFU, and a trace-aware Oracle policy. Our analysis shows that both datasets contain substantial entity reuse among starting entities and entities reached during traversal. We also explore semantic caching for similar queries, which shows additional hit-rate gains on WebQSP and needs further accuracy testing on CWQ. Entity caching accelerates KG retrieval by up to $1.91\times$, while semantic-context caching achieves up to $1.06\times$ full-system speedup in the evaluated WebQSP configurations, with each hit being up to $3.73\times$ faster.
发表机构
- Texas A&M University（得克萨斯农工大学）
机构由 AI 辅助整理，请以论文原文为准。
