1
Framing Instability in LLM Ethical Stance: Auditing Negation Sensitivity in Moral Dilemmas
道德困境下,模型对“应该”和“不应该”的立场可能瞬间翻转,16个模型全员中招,值得深思。
arXiv:2601.21433v2 Announce Type: replace Abstract: Language models are increasingly consulted on ethically consequential questions, yet the stance a …