MobileMem

MobileMem 被设计为一个统一基准，用于评估移动 AI 助手中的长期记忆，而不是一个独立的记忆系统。其模块化架构可无缝集成到现有移动智能体生态中，包括任务理解、记忆检索、问答、网页搜索、回复安全和执行模块。因此，MobileMem 支持不同记忆架构之间的可复现比较，并促进长期记忆评测接入真实开发管线。

MobileMem is designed as a unified benchmark for evaluating long-term memory in mobile AI assistants rather than a standalone memory system. Its modular architecture seamlessly integrates with the existing mobile agent ecosystem, including task understanding, memory retrieval, question answering, web search, response safety, and execution modules. Consequently, MobileMem supports reproducible comparison across different memory architectures and facilitates the integration of long-term memory evaluation into real-world development pipelines.

MobileMem 的一个关键特征是，它由真实智能手机记忆构建，而不是合成交互日志或人工整理的用户画像。该基准在单一评测框架中统一了基础记忆、认知记忆、偏好记忆、时间推理和视觉推理，同时将评估建立在约一年真实用户活动之上。MobileMem 建立在原生移动应用生成的异构记忆源之上，包括日历、相册、笔记、文档、待办事项、账单、语音备忘录、屏幕记忆、视频记忆以及其他用户生成记录。

A key characteristic of MobileMem is that it is constructed from authentic smartphone memory rather than synthetic interaction logs or manually curated user profiles. The benchmark unifies foundational memory, cognitive memory, preference memory, temporal reasoning, and visual reasoning within a single evaluation framework, while grounding assessment in approximately one year of real user activities. MobileMem is built upon heterogeneous memory sources generated by native mobile applications, including calendars, albums, notes, documents, to-do lists, bills, voice memos, screen memories, video memories, and other user-generated records.

MobileMem 将长期记忆评测从定性能力展示转向可度量、可比较且可持续改进的评估。该基准为检索准确性、时间一致性、偏好理解、多模态推理和长程知识整合提供综合指标，从而能够客观比较不同记忆架构和检索策略。随着基础模型日益从云服务迁移到边缘设备，MobileMem 为评估可靠、高效、隐私感知且能通过长期用户交互持续演进的记忆系统提供了有效框架。

MobileMem shifts long-term memory evaluation from qualitative capability demonstrations to measurable, comparable, and continuously improvable assessment. The benchmark provides comprehensive metrics for retrieval accuracy, temporal consistency, preference understanding, multimodal reasoning, and long-horizon knowledge integration, enabling objective comparison among different memory architectures and retrieval strategies. As foundation models increasingly migrate from cloud services to edge devices, MobileMem offers an effective framework for evaluating reliable, efficient, and privacy-aware memory systems that continuously evolve through long-term user interactions.

从核心上看，MobileMem 将长期多模态用户记忆视为个性化的基本单元。不同于主要评估孤立事实回忆的传统问答基准，MobileMem 评估 AI 助手是否能够在扩展时间跨度内持续积累、检索、整合并推理不断演化的用户记忆。通过建模真实用户轨迹、多语言交互、动态偏好和真实世界事件，该基准鼓励开发能够提供上下文感知、主动且持续自适应辅助能力的 AI 助手。

At its core, MobileMem treats long-term multimodal user memory as the fundamental unit of personalization. Unlike conventional question answering benchmarks that primarily evaluate isolated factual recall, MobileMem evaluates whether AI assistants can continuously accumulate, retrieve, integrate, and reason over evolving user memories across extended temporal horizons. By modeling authentic user trajectories, multilingual interactions, dynamic preferences, and real-world events, the benchmark encourages the development of AI assistants capable of delivering context-aware, proactive, and continuously adaptive assistance.

健康管理与用药提醒：助手结合日历、笔记、浏览记录、体检报告和处方，核对用户的药物过敏史与当前用药，并及时给出风险提醒。

旅行行程规划：助手整合用户收藏的游记、交通方式和住宿偏好，生成符合个人兴趣的具体行程。

个人影视清单与追剧记录：助手从长期积累的观剧笔记和截图中识别相似作品，进行跨时间比较并生成个性化长评。

工作复盘与年中总结：助手整合日历、文档、笔记和屏幕记忆，梳理学习轨迹、工作产出与长期存在的瓶颈。

Health Management and Medication Reminder: The assistant matches the user's allergy history with the current medication list and provides a clear risk alert.

Itinerary Planning for a Foreign Trip: The assistant integrates bookmarked guides with transportation and accommodation preferences to produce a personalized itinerary.

Personal Watchlist and Drama Tracking: The assistant identifies works with similar styles or themes and generates a comparative review grounded in long-term personal memory.

Work Review and Mid-Year Summary: The assistant integrates calendars, documents, notes, and screen memories to summarize learning trajectories, work output, and recurring bottlenecks.