Fetching the paper…

Exploiting the Vulnerability of Large Language Models via Defense-Aware Architectural Backdoor · Around