---
title: "Mi-Memory：面向个人AI的生命周期记忆框架"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.18975v1"
permalink: "https://feed.astralor.com/items/itm_2b9c49f565434fedd52b9e9b"
published: "2026-07-21T11:10:38.000Z"
score: 70
channel: "tech"
tags: ["Mi-Memory","个人AI","审计合同","记忆框架","边缘部署","达尔文实验室"]
---

# Mi-Memory：面向个人AI的生命周期记忆框架

## AI 点评

> 这篇论文构建了个人AI记忆的完整生命周期蓝图，非常适合从事AI Agent、边缘计算和隐私工程的研究者。它把记忆从‘对话缓存’重新定义为可审计的治理层，弥补了当前LLM记忆系统缺乏证据链和部署意识的短板。如果你正在设计多设备联动的AI助手，Mi-Memory的架构思路值得细读。

## 速览

随着个人AI从简单聊天扩展到手机、汽车、家居、可穿戴设备等多设备场景，记忆系统需要从对话缓存升级为持久、可审计、可治理的基础设施。Mi-Memory 框架围绕结构、扩展、演进和部署四个角色，通过审计合约连接证据负载、诊断追踪、策略工件和门控回滚记录。在结构评估中，MemStack 在 LoCoMo、PersonaMem-V2 和 LongMemEval 上分别达到 93.59%、57.24% 和 87.47%。该框架为个人AI的可审计、证据门控和部署感知的记忆系统提供了重要进展。

## 关键点

- Mi-Memory 将个人AI记忆划分为结构、扩展、演进、部署四个角色，并引入审计合同以保证可追溯性。
- 通过 MemStack、MemSense/MemFuse、D²ACCI/E²MEND 和 LiteMem 四个组件分别实现各角色功能。
- 在结构评测中，MemStack 在三个基准上取得了领先结果（最高93.59%），其他模块提供初步或设计级证据。
- 框架强调证据门控和部署感知，适配延迟、成本、隐私和边缘-云约束。

## 全文

Personal AI is moving beyond chat-only interaction toward continuous services that span phones, cars, homes, wearables, cameras, and tools. In this setting, memory cannot remain a cache of prior conversations. It should serve as a continuity and governance substrate: preserving durable user state, grounding answers in multimodal and device evidence, supporting correction and forgetting, bounding policy evolution, and remaining deployable under latency, cost, privacy, and edge-cloud constraints. This technical report presents Mi-Memory, a lifecycle memory framework for Personal AI organized around four roles: Structure, Expansion, Evolution, and Deployment. A shared audit contract links these roles through four recurring artifact families: typed evidence payloads preserve source identity and provenance, diagnostic traces localize evidence loss across the serving pipeline, strategy artifacts make memory-policy changes explicit, and gate/rollback records bound accepted evolution. MiMemory instantiates the roles through MemStack, MemSense/MemFuse, D$^{2}$ACCI/E$^{2}$MEND, and LiteMem. In controlled-reference Structure evaluations, MemStack reaches 93.59%, 57.24%, and 87.47% on LoCoMo, PersonaMem-V2, and LongMemEval, respectively; other tracks report module-level, preliminary/internal, transfer-feasibility, or design-only evidence with explicit boundaries. MiMemory is a step toward auditable, evidence-gated, and deployment-aware memory systems for Personal AI. Project homepage: https://darwin-agent.github.io/Mi-Memory/ .

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_2b9c49f565434fedd52b9e9b