[论文解读] SuperLocalMemory V3: Information-Geometric Foundations for Zero-LLM Enterprise Agent Memory

Varun Pratap Bhardwaj|arXiv (Cornell University)|Mar 15, 2026

Topological and Geometric Data Analysis被引用 0

一句话总结

SLM-V3 用费舍-信息加权检索替代余弦相似性，增加层厚一致性、使用黎曼 Langevin 动力学管理内存生命周期，实现零-LLM 检索，在 LoCoMo 上呈现强劲的经验提升。

ABSTRACT

Persistent memory is a central capability for AI agents, yet the mathematical foundations of memory retrieval, lifecycle management, and consistency remain unexplored. Current systems employ cosine similarity for retrieval, heuristic decay for salience, and provide no formal contradiction detection. We establish information-geometric foundations through three contributions. First, a retrieval metric derived from the Fisher information structure of diagonal Gaussian families, satisfying Riemannian metric axioms, invariant under sufficient statistics, and computable in O(d) time. Second, memory lifecycle formulated as Riemannian Langevin dynamics with proven existence and uniqueness of the stationary distribution via the Fokker-Planck equation, replacing hand-tuned decay with principled convergence guarantees. Third, a cellular sheaf model where non-trivial first cohomology classes correspond precisely to irreconcilable contradictions across memory contexts. On the LoCoMo benchmark, the mathematical layers yield +12.7 percentage points over engineering baselines across six conversations, reaching +19.9 pp on the most challenging dialogues. A four-channel retrieval architecture achieves 75% accuracy without cloud dependency. Cloud-augmented results reach 87.7%. A zero-LLM configuration satisfies EU AI Act data sovereignty requirements by architectural design. To our knowledge, this is the first work establishing information-geometric, sheaf-theoretic, and stochastic-dynamical foundations for AI agent memory systems.

研究动机与目标

• 识别检索、生命周期与一致性中持久化代理记忆的数学基础.
• 用有信息几何度量代替启发式记忆检索。
• 提供一个正式机制来检测跨记忆上下文的矛盾。
• 演示可扩展的、符合数据主权要求的零-LLM 记忆操作。

提出的方法

• 引入四通道检索架构，整合 Fisher–Rao 语义检索、BM25、实体图和时间通道。
• 为具有每维不确定性的记忆嵌入开发方差加权的费舍信息度量。
• 将记忆存储建模为一个胞腔（cellular）层，使用第一上同调来检测不可调和的矛盾。
• 将记忆生命周期公式化为黎曼 Langevin 动力学，并证明其平稳分布的存在性/唯一性。
• 通过带权倒排相关性融合与神经重排序实现四通道融合以优化。
• 在 LoCoMo 上对云无检索进行零-LLM 配置评估并报告相较基线的提升。

Figure 1 : The SLM-V3 architecture. Left: Ingestion pipeline processes content through entropy gating, fact extraction, entity resolution, graph construction, and sheaf consistency checking ( $H^{1}\neq 0$ detects contradictions). Center: Four-channel retrieval with Fisher-information-weighted scori

实验结果

研究问题

• RQ1 在持久化 AI 记忆的检索、生命周期管理和一致性方面，哪些数学结构是合适的？
• RQ2 费舍信息基础的检索在高维嵌入中是否能优于余弦相似性？
• RQ3 如何在多上下文存储中形式化检测并诊断记忆矛盾？
• RQ4 在信息几何基础上实现零-LLM 部署，是否在无云访问的情况下仍能达到有竞争力的检索？

主要发现

• 在六轮 LoCoMo 对话（n=832）中，经过三层数学结构的加持，平均提升为 +12.7 个百分点，相对于工程基线。
• 在最具挑战性的对话中，获得最大提升为 +19.9 个百分点。
• 四通道检索在无需云环境的情况下实现约 75% 的检索质量。
• 零-LLM 配置达到 75% 的检索质量，证明数据主权的可行性。
• 消融实验表明跨编码器重排序是性能提升的最大贡献者。
• 跨跳推理显示来自数学层的提升为 +12 个百分点。

Figure 2 : Competitive landscape of agent memory systems (March 2026) evaluated on LoCoMo. All systems above SLM-V3 require cloud LLM dependency. SLM-V3 Mode A Retrieval (74.8%) is the highest reported score achievable without cloud dependency during retrieval. Stars ( $\star$ ) denote zero-LLM conf