Trojan Detection Through Pattern Recognition for Large Language Models