cs.AI, cs.CL, cs.CR, cs.LG

Adaptive Instruction Composition for Automated LLM Red-Teaming

arXiv:2604.21159v1 Announce Type: cross
Abstract: Many approaches to LLM red-teaming leverage an attacker LLM to discover jailbreaks against a target. Several of them task the attacker with identifying effective strategies through trial and error, res…