发表机构
University of Oulu; University of Helsinki（奥卢大学; 赫尔辛基大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出MELD协议，实现分布式智能体记忆的知识合并，经实验验证其在存储、召回率、一致性及消息量上均优于基线，可支撑自治智能体联邦的知识协同。
AI 中文摘要
自主智能体共享传输层并可互相调用工具，但无法共享自身知识：尚无协议能让两个智能体的记忆协调以两种方式表述的事实、关联分开存储的相关事实，或协调矛盾知识而不默默丢弃任一主张。本文提出MELD，一种针对智能体记忆联邦的自管理一致性机制，其运行时模型为知识图本身。每个“大脑”通过包含五种结果（插入、合并、关联、冲突、拒绝）的流程接纳每一条传入主张，该流程由三个信号（限定范围的主张-密钥标识、嵌入相似度、自然语言推理裁决）在上下文和新鲜度门限下判定，且仅通过一个可审计、可认证的Patch（唯一能改变状态的对象）执行。绑定到标准发布/订阅传输层，搭配每条主张状态的CRDT，使自治“大脑”无需协调器即可在主张状态上保持一致：在分区及有损路由下自修复，在良性故障模型下抵御对等节点的默默重写。MELD不裁决真相；检测到的矛盾会被保留供后续裁决，绝不默默解决。在HotpotQA干扰项数据集上，分布式合并在预指定等价性测试下的召回率不劣于集中式存储，且比朴素联合的召回率高，同时活动存储减少约11%；合并分类器在裁决后的候选对上以0.013的错误合并率达到AUC 0.968；状态CRDT在30次真实分区-愈合试验中均实现重收敛，而最后写入者优先机制仅实现11/30；语义路由在匹配召回率下消息减少约3倍。我们在涵盖运营商级5G边缘、国家HPC和本地层级的真实计算连续体上，采用经经验校准的阈值进行评估。
英文摘要
Autonomous agents share a transport and can call each other's tools, but they cannot share what they know: no protocol lets two agents' memories reconcile a fact phrased two ways, link related facts held apart, or reconcile contradictory knowledge without silently discarding either claim. We present MELD, a self-managing coherence mechanism for a federation of agent memories whose run-time model is the knowledge graph itself. Each brain admits every incoming claim through a five-outcome procedure (insert, merge, relate, conflict, or reject), decided from three signals (scoped claim-key identity, embedding similarity, and a natural-language-inference verdict) under context and freshness gates, and acting through exactly one auditable, authenticated Patch, the only object that mutates state. A binding onto standard publish/subscribe transport with a per-claim status CRDT keeps sovereign brains coherent in claim status without a coordinator: self-healing after partitions and under lossy routing, and self-protecting against silent rewrite by a peer, under a benign-fault model. MELD does not adjudicate truth; a detected contradiction is preserved for later adjudication, never silently resolved. On HotpotQA distractor, distributed merge is recall-non-inferior to a centralized store under a pre-specified equivalence test and recall-superior to naive union at about 11% less live storage; the merge classifier separates at AUC 0.968 with a 0.013 false-merge rate on adjudicated candidate pairs; the status CRDT reconverges in 30/30 real partition-heal trials where last-writer-wins manages 11/30; and semantic routing delivers about 3x fewer messages at matched recall. We evaluate on a real computing continuum spanning an operator-grade 5G edge, national HPC, and a local tier, with empirically calibrated thresholds.
Comments30 pages, 3 figures, 1 table, plus an 11-page appendix (A-N). Code and experiment data: https://doi.org/10.5281/zenodo.21878274
