English(EN) Privacy Without Regret: Differentially Private Inference-Time Alignment

新方法在不损失性能的情况下增强了大型语言模型的隐私和对齐能力

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

研究人员开发了新的方法 PrivBoN 和 PrivITP，以增强大型语言模型在推理过程中的差分隐私。这些技术解决了奖励破解问题以及敏感训练数据缺乏隐私保护的问题。PrivBoN 使用 Gumbel 噪声来实现差分隐私和 KL 正则化对齐，在特定隐私预算下匹配理论性能。PrivITP 进一步通过结合 $\chi^2$-正则化拒绝采样和高斯机制来改进这一点，提供与响应数量无关的事后 $(\epsilon,\delta)$-DP，并将正则化与隐私参数解耦。 AI

影响这些方法可以通过在推理过程中保护敏感用户数据，从而实现更安全的 LLM 部署。

排序理由详细介绍大型语言模型差分隐私新方法的学术论文。[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; 详细介绍大型语言模型差分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]

arXiv cs.LG TIER_1 English(EN) · Ishi Jain, Nandini Bhattad, Sayak Ray Chowdhury · 2026-08-28 04:00

无悔隐私：差分隐私的推理时对齐

arXiv:2608.26324v1 Announce Type: new Abstract: Best-of-N (BoN) sampling is the simplest and most widely deployed inference-time alignment strategy, but it suffers from two distinct problems: reward hacking, in which the selected response exploits errors in the proxy reward model…

本文如何被排名

报道来源 [1]

无悔隐私：差分隐私的推理时对齐

相关实体

相关话题