arXiv cs.CL· Zhuomin Chen, Jingchao Ni, Qingkai Dong, Xu Zheng, Gabriel H\"ubner Lucchesi, Mo Sha, Wei Cheng, Dongsheng Luo·· 12 小时前AI 评分34
Heirloom：在手机端将个性化适配器迁移至新一代基座模型
Heirloom: Carrying a Phone's Personalization Across Base-Model Generations
AI 导读
Heirloom 可在不保留用户标注历史的情况下，把手机端个性化适配器迁移到新基座模型：以旧个性化模型为教师，在存储的无标注输入、厂商提示词或新基座生成的提示词上收集预测，再卸载旧基座、训练新适配器对齐这些预测，内存中始终只保留一个基座模型。
正文 · 原文
Abstract:On-device language models can learn a user's preferences through adapters trained on private data. However, a base-model upgrade may change the layers or vocabulary, leaving the personal adapter incompatible. Replacing it with a non-personal adapter loses much of the personalization, while retraining requires retaining the user's labeled history. We present Heirloom, a system that transfers a personal adapter to a new base model on the phone without retaining labels. Heirloom uses the old personal model as a teacher, collecting its predictions on stored unlabeled inputs, vendor-provided prompts, or prompts generated by the new base. It then unloads the old base and trains an adapter for the new model to match those predictions, keeping only one base model in memory at a time. We evaluate Heirloom across three generation upgrades from two model families, an architecture-preserving refresh, and three personalization tasks. On movie tagging, migration from stored inputs maintains pre-upgrade accuracy and differs from labeled retraining by at most 1.7 percentage points across the three upgrades. On a phone's neural processing unit (NPU), a fixed-length migration for the 4B upgrade takes 38 minutes and consumes 20% of the battery end to end, making the stored-input setting feasible during overnight charging.
| Comments: | 35 pages, 12 figures, 20 tables |
| Subjects: | Machine Learning (cs.LG); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.00650 [cs.LG] |
| (or arXiv:2611.00650v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.00650 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Dongsheng Luo [view email]
[v1]
Sat, 31 Oct 2026 05:26:50 UTC (420 KB)
来源：arXiv cs.CL · arxiv.org