π¨ ALERT: A Comprehensive Benchmark for Assessing Large Language Models' Safety through Red Teaming Jun 25 β’ 3