Skip to content
Hot storiesContinuously updated

研究人员提出端侧模型个性化迁移系统Heirloom

1 reports1 reporting sourcesUpdated 19 hours ago

Start here

AI overview

2026年11月,相关论文介绍了可在手机端进行模型个性化迁移的系统Heirloom。该系统能跨代际将个性化adapter迁移至新基座模型,运行中无需保留带标签的历史数据,且内存中始终仅保留一个基座模型。实验显示,其在电影打标任务中保持了升级前准确率,与带标签重新训练的差距最多仅1.7个百分点;在手机NPU上执行4B模型升级迁移耗时38分钟、耗电20%,适合在夜间充电时完成。

AI-generated from reports · Updated 15 hours ago

Report timeline

Follow the reports to see different sides of the story.

Nov 3
  1. arXiv 机器学习
    Heirloom:跨基座模型代际迁移手机个性化

    Heirloom 是一款可在手机端将个性化 adapter 迁移至新基座模型的系统,无需保留带标签的历史数据,且内存中始终仅保留一个基座模型。该系统在电影打标任务中保持了升级前准确率,与带标签重新训练的差距最多仅 1.7 个百分点。在手机 NPU 上,4B 模型升级的迁移耗时 38 分钟并耗电 20%,适合在夜间充电时完成。

Story heat trend

Current heat 6·Comparable peak 9(Nov 3 18:00)·Comparable change, last 24 hours–

02.557.510Nov3Nov3Nov4Nov4

Trends compare only the same parties observed continuously, so the scope can be narrower than current heat counts. Move the pointer or tap the chart to see hourly heat; use the left and right arrow keys on a keyboard.