牛哥精选 · 所有

📋 全部 ☁️ 云服务 🤖 AI 平台 🔗 API 中转 🔐 安全/认证 💳 支付 📧 通讯 📊 数据分析 🖼 媒体处理 🌐 域名/DNS

📝 深度技术 arXiv NLP 2026-07-22

Guard Vector: Beyond English LLM Guardrails with Task-Vector Composition and Streaming-Aware Prefix SFT

Guard Vector通过任务向量组合与流式感知前缀SFT，突破英语限制实现多语言LLM安全护栏，提供可扩展的防护新范式。

arXiv:2509.23381v2 Announce Type: replace Abstract: We introduce Guard Vector, a safety task vector computed as the parameter difference between a gua…

guard vect llm安全护栏任务向量前缀微调流式感知

🤖 AI·大模型 arXiv 机器学习 2026-07-10

Functional and Secure Code Generation with Task Vectors

用任务向量让大模型生成既功能正确又安全无漏洞的代码，新方法直击LLM代码生成痛点

arXiv:2607.07881v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used for code generation, but they struggle to generat…

任务向量代码生成安全漏洞 llm 功能正确性

📝 深度技术 arXiv 机器学习 2026-06-03

RogueMerge: Robust and Unified Attacks against LLM Model Merging

针对LLM模型合并的供应链漏洞，提出统一鲁棒的攻击方法RogueMerge，揭示第三方任务向量的安全威胁。

arXiv:2606.03344v1 Announce Type: cross Abstract: Model merging composes specialized capabilities into a single LLM by aggregating task vectors source…

llm安全模型合并供应链攻击任务向量 roguemerge

📅 日期

2026-05-20 2026-05-19

🐂 牛哥精选

Guard Vector: Beyond English LLM Guardrails with Task-Vector Composition and Streaming-Aware Prefix SFT

Functional and Secure Code Generation with Task Vectors

RogueMerge: Robust and Unified Attacks against LLM Model Merging

📅 日期