Do We Really Need Curated Malicious Data for Safety Alignment in Multi-modal Large Language Models?

Devs

Do We Really Need Curated Malicious Data for Safety Alignment in Multi-modal Large Language Models? | Read Paper on Bytez