1
Dummy Backdoor as a Defense: Removing Unknown Backdoors via Shared Internal Mechanisms for Generative LLMs
用假后门骗过真后门：无需已知信息，通过共享内部机制清除生成式大模型里的未知后门。
arXiv:2606.11648v1 Announce Type: cross Abstract: Backdoor attacks pose a serious threat to the safety and reliability of Large Language Models (LLMs)…