AI RESEARCH
Why Do Aligned LLMs Remain Jailbreakable: Refusal-Escape Directions, Operator-Level Sources, and Safety-Utility Trade-off
arXiv CS.AI
•
ArXi:2605.08878v1 Announce Type: cross Aligned large language models (LLMs) remain vulnerable to jailbreak attacks. Recent mechanistic studies have identified latent features and representation shifts associated with jailbreak success, but they leave a fundamental question open: why do aligned LLMs remain jailbreakable, and what structural vulnerabilities in the model make this possible? We study this question through a continuous input-transformation view.