AutoDetect: Towards a Unified Framework for Automated Weakness Detection in Large Language Models