AI RESEARCH

Tatemae: Detecting Alignment Faking via Tool Selection in LLMs

arXiv CS.AI

ArXi:2604.26511v1 Announce Type: cross Alignment faking (AF) occurs when an LLM strategically complies with