English(EN) Follow the Gradient: Tracing Safety Degradation Back to Training Data

新方法将大型语言模型安全退化追溯至训练数据

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-28 04:00

研究人员开发了一种名为SimpleLoRIF的新方法，用于识别和移除大型语言模型训练集中的潜在有害数据。该技术利用梯度嵌入来精确定位那些看似良性但可能导致不安全行为的微调示例。通过将SimpleLoRIF应用于Tulu 3模型，他们创建了SafeTulu，该模型在抵御攻击的脆弱性方面显著降低，并保持或提高了其能力。 AI

影响提供了一种通过过滤训练数据来提高大型语言模型安全性的新技术，有望带来更强大、更安全的AI系统。

排序理由详细介绍大型语言模型安全新方法的学术论文。[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 →

本文如何被排名

Signal score: 12 / 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安全新方法的学术论文。[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: Same-day; Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.AI TIER_1 English(EN) · Saket Reddy, Amro Abbas, Anshuman Suri, Ari Morcos, David Schwab, Pratyush Maini · 2026-10-28 04:00

循梯度而行：追溯安全退化至训练数据

arXiv:2610.37481v1 Announce Type: cross Abstract: Fine-tuning Large Language Models (LLMs) on instruction data can quietly erode their safety, even when no training example looks harmful. Because this safety-degrading data looks benign, standard content filters struggle to catch …

本文如何被排名

报道来源 [1]

循梯度而行：追溯安全退化至训练数据

相关实体

相关话题