Backdoor Unlearning Generalization: A Path Toward the Removal of Unknown Triggers in LLMs 文章

ArXiv CS.CL2026-06-03NEWSen作者: Lisa Bouger, Th\'eo Lasnier, Philippe Looubet Moundi, Yannick Teglia, Djam\'e Seddah
查看原文 →
关系图谱

Backdoor Unlearning Generalization: A Path Toward the Removal of Unknown Triggers in LLMs · 相关技术

相关技术

Cross Activation Shift Distancelarge language modelsBackdoor attacks