为大规模企业数据设计动态内存管理，提升检索效率与准确性。
MEMONDEMAND: A Memory Management System for Large-Scale Enterprise Data
- 构建动态多层层级结构，按数据集特性自适应抽象深度。
- 双层内存机制分离路由与证据加载，10M到618M token均显著提效。
- 按需提升节点优先级，有限内存预算下实现高效更新。
企业数据仓库规模庞大、异构且持续更新，同时支持高效访问、源可信证据和跨查询适配时，检索难度高。企业记忆扩展了模型上下文外的检索能力，但现有系统未能协同解决集合特异性层次构建、低成本路由、细粒度证据加载及工作负载感知的记忆更新问题。本文提出MEMONDEMAND（按需记忆），包含三个协调机制：动态多层层次结构，为每个数据集确定抽象层级与深度；每层双内存设计，分离精炼路由与详细证据；按需记忆提升机制，在有限活跃状态预算下动态调整节点优先级。在EnterpriseRAG-Bench上，MEMONDEMAND在10M至618M token所有规模下均超越最强基线（LB#1），分别取得12.23%和4.66%的性能提升。FinanceBench、HotpotQA和FRAMES上的结果进一步验证其在金融、多跳和事实检索场景下的优异表现。综合表明，MEMONDEMAND是适用于大规模企业数据仓库的精准、高效、可扩展记忆解决方案。
原文摘要 · Abstract (English)
Enterprise repositories are large, heteroge- neous, and continuously updated, making re- trieval difficult when efficient access, source- faithful evidence, and cross-query adaptation must be supported together. Enterprise mem- ory extends retrieval beyond the model con- text, but existing systems do not jointly address collection-specific hierarchy construction, low- cost routing, detailed evidence loading, and workload-aware memory updates at this scale. We introduce MEMONDEMAND, short for On- Demand Memory, a memory management sys- tem with three coordinated mechanisms: a dy- namic multi-level hierarchy that determines the abstraction structure and depth for each col- lection, dual memory at every hierarchy level that separates distilled routing from detailed evidence, and on-demand memory promotion that updates node priority under a bounded active-state budget. On EnterpriseRAG-Bench, MEMONDEMAND outperforms the strongest published LB#1 result at every evaluated scale from 10M tokens through the complete 618M- token collection, with gains of 12.23% at 10M and 4.66% at 618M. Results on FinanceBench, HotpotQA, and FRAMES further show strong performance across financial, multi-hop, and fact-retrieval settings. Together, these results establish MEMONDEMAND as an accurate, ef- ficient, and scalable memory solution for very large enterprise repositories across data scales, domains, and evidence requirements. Our code is available at https://github.com/ xfab-xinyuansong/MemOnDemand.git.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。