arXiv Artificial Intelligence· Harsh Kumar, Rahul Maity, Tanmay Joshi, Aman Chadha, Vinija Jain, Suranjana Trivedy, Amitava Das·· 19 小时前AI 评分44
PermaFrost-Attack：通过 Stealth Pretraining Seeding 在 LLM 训练中埋下逻辑地雷
PermaFrost-Attack: Stealth Pretraining Seeding(SPS) for planting Logic Landmines During LLM Training
AI 导读
研究提出 Stealth Pretraining Seeding（SPS）威胁模型，攻击者将少量投毒内容散布在隐蔽网站上，使其被 Common Crawl 等来源吸收进未来训练语料。
来源：arXiv Artificial Intelligence · arxiv.org