With the rapid emergence of personal AI agents based on Large Language Models (LLMs), implementing them on-device has become essential for privacy and responsiveness. To handle the inherently personal and context-dependent nature of real-world requests, such agents must ground their generation in device-resident personal context. However, under tight memory budgets, the core bottleneck is what to store so that retrieval remains aligned with the user. We propose EPIC (Efficient Preference-aligned Index Construction), which focuses on user preferences as a compact and stable form of personal context and integrates them throughout the RAG pipeline. EPIC selectively retains preference-relevant information from raw data and aligns retrieval toward preference-aligned contexts. Across four benchmarks covering conversations, debates, explanations, and recommendations, EPIC reduces indexing memory by 2,404 times, improves preference-following accuracy by 18.79 %p, and achieves 32.17 times lower retrieval latency over the best-performing baseline. In on-device experiments, EPIC maintains under 1 MB memory and achieves 5.21 to 29.35 ms/query latency across three platforms, while supporting streaming updates under preference drift. Our code and data are available at https://github.com/UbiquitousAILab/EPIC.

翻译：随着基于大语言模型（LLMs）的个人AI代理迅速涌现，将其部署在设备端已成为保障隐私和响应速度的关键。为应对现实请求中固有的个性化与上下文依赖性，此类代理必须基于设备本地存储的个人上下文生成结果。然而，在紧张的内存预算约束下，核心瓶颈在于存储什么内容，使得检索始终与用户偏好对齐。我们提出EPIC（高效偏好对齐索引构建），该方法将用户偏好作为紧凑且稳定的个人上下文形式，并将其整合到RAG流水线的全过程。EPIC从原始数据中选择性保留偏好相关信息，并引导检索朝向偏好对齐的上下文。在涵盖对话、辩论、解释和推荐的四个基准测试中，EPIC将索引内存减少了2,404倍，偏好跟随准确率提升了18.79个百分点，并在最佳性能基线方案上实现了32.17倍的检索延迟降低。在设备端实验中，EPIC在保持内存低于1 MB的条件下，在三个平台上实现了5.21至29.35毫秒/查询的延迟，同时支持偏好漂移下的流式更新。我们的代码和数据已开源：https://github.com/UbiquitousAILab/EPIC。