arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.18271cs.CLcs.AIcs.IRcs.LG
收藏

从体积到价值:面向设备端RAG的偏好对齐记忆构建

From Volume to Value: Preference-Aligned Memory Construction for On-Device RAG

Department of Computer Science and Engineering, Ulsan National Institute of Science and Technology (UNIST), Ulsan, Republic of Korea(计算机科学与工程系,全州国立科学与技术研究所(UNIST),全州,韩国)

Changmin Lee, Jaemin Kim, Taesik Gong

AI总结:

本文提出EPIC方法,通过将用户偏好作为紧凑且稳定的个人上下文形式,整合到RAG流程中,以在有限内存下提高检索与用户偏好的对齐度,从而减少内存使用并提升准确性。

摘要:

随着基于大型语言模型(LLMs)的个人AI代理的迅速发展,将其部署到设备上已成为隐私和响应性的重要需求。为了处理现实世界请求中固有的个人和上下文依赖性,这些代理必须基于设备上存储的个人上下文进行生成。然而,在内存预算紧张的情况下,核心瓶颈是存储什么内容以确保检索与用户保持一致。我们提出EPIC(高效偏好对齐索引构建),专注于用户偏好作为紧凑且稳定的个人上下文形式,并在整个RAG流程中整合它们。EPIC会选择性地保留与偏好相关的信息,并将检索对准偏好对齐的上下文。在四个涵盖对话、辩论、解释和推荐的基准测试中,EPIC将索引内存减少了2,404倍,提高了偏好遵循的准确性20.17个百分点,并在最佳表现基线之上实现了33.33倍更低的检索延迟。在我们的设备端实验中,EPIC在29.35毫秒/查询的流式更新下保持内存占用低于1 MB。

英文摘要:

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.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑