PulseAugur
中
实时 09:35:41
English(EN) Prune the Leaky Neurons: Mitigating Private Data Leakage in Fine-Tuned LLMs via Joint-Embedding-Guided Pruning

新的JEPA-Prune方法以最小的效用损失移除了微调LLM中的PII

研究人员开发了一种名为JEPA-Prune的新方法,以解决微调后大型语言模型(LLM)中的隐私问题。该技术旨在移除个人身份信息(PII),同时不显著降低模型的性能。JEPA-Prune使用联合嵌入预测架构(JEPA)在无PII的数据上训练一个参考模型,使其能够识别并修剪微调模型中的PII特定神经元。实验表明,修剪少量神经元可以显著降低PII推理准确性和提取召回率,同时保持效用。 AI

影响 这项研究为减轻LLM中的隐私风险提供了一种新颖的方法,有可能在敏感数据上更安全地部署微调模型。

排序理由 该集群包含一篇详细介绍LLM隐私新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

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

新的JEPA-Prune方法以最小的效用损失移除了微调LLM中的PII

本文如何被排名

Signal score
13 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
该集群包含一篇详细介绍LLM隐私新方法的学术论文。[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]

  1. arXiv cs.AI TIER_1 English(EN) · Mohamed Shaaban, Mohamed Elmahallawy ·

    剪除泄露的神经元:通过联合嵌入引导的剪枝来减轻微调 LLM 中的私有数据泄露

    arXiv:2610.22792v1 Announce Type: cross Abstract: Fine-tuning large language models (LLMs) on proprietary records enables domain adaptation but introduces two challenges: (i) privacy leakage of memorized private data, such as personally identifiable information (PII), and (ii) ba…