牛哥精选 · 半年

📋 全部 🤖 AI·大模型 ⚡ 效率工具 📝 深度技术 🚀 产品观察 💰 商业科技 🔓 开源项目 🎨 设计创意 📖 阅读推荐 🏷 资源合集 🌱 成长效率

📝 深度技术 arXiv NLP 2026-06-25

Detect, Unlearn, Restore: Defending Text Summarization Models Against Data Poisoning

针对摘要模型微调时的数据投毒攻击，提出检测-遗忘-恢复三步防御框架，安全研究必备。

arXiv:2606.26036v1 Announce Type: new Abstract: Training-time data poisoning during fine-tuning poses a significant threat to large language models (L…

数据投毒文本摘要大模型安全防御机制微调安全

🤖 AI·大模型 arXiv AI 2026-05-25

TCAP: Tri-Component Attention Profiling for Unsupervised Backdoor Detection in MLLM Fine-Tuning

MLLM微调时隐藏的后门如何绕开？TCAP用三组件注意力分析实现无监督检测，无需标签即可揪出异常。

arXiv:2601.21692v2 Announce Type: replace Abstract: Fine-Tuning-as-a-Service (FTaaS) facilitates the customization of Multimodal Large Language Models…

多模态大模型后门检测注意力机制无监督学习微调安全

📅 日期

2026-05-20 2026-05-19

🐂 牛哥精选

Detect, Unlearn, Restore: Defending Text Summarization Models Against Data Poisoning

TCAP: Tri-Component Attention Profiling for Unsupervised Backdoor Detection in MLLM Fine-Tuning

📅 日期