Fetching the paper…

Stealthy and Persistent Unalignment on Large Language Models via Backdoor Injections · Around