English(EN) When Privacy Hurts Mergeability: Geometry-Aware Model Merging under Differential Privacy

新的DP-Merging框架改进了私有模型组合

作者 PulseAugur 编辑部 · [1 个来源] · 2026-08-28 04:00

研究人员开发了DP-Merging，一个旨在改进差分隐私任务模型可合并性的新框架。该方法解决了两个关键的几何障碍：局部尖锐性和参考漂移，这阻碍了私有模型的组合。DP-Merging引导私有任务模型趋向更平坦的损失区域，并将其与共享的预训练初始化对齐，从而减少了与合并相关的损失增加。实验表明，DP-Merging在保持差分隐私保证的同时，提高了私有合并模型在视觉和语言任务上的性能。 AI

影响增强了在不损害数据隐私的情况下组合私有AI模型的能力，可能促进更具协作性的AI开发。

排序理由关于差分隐私下模型合并新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 22 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 关于差分隐私下模型合并新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, safety; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Jin Liu, Junkang Liu, Ning Xi, Yinbin Miao, Dawei Wei, Ke Cheng, Jianfeng Ma · 2026-08-28 04:00

当隐私损害可合并性：差分隐私下的几何感知模型合并

arXiv:2608.26655v1 Announce Type: new Abstract: Model merging promises to construct a single multi-task model from independently fine-tuned task models without accessing the original task data. This makes it attractive when task data cannot be centralized, but released task model…

本文如何被排名

报道来源 [1]

当隐私损害可合并性：差分隐私下的几何感知模型合并

相关实体

相关话题