Cultural Binding Heads in Language Models 事件

Cultural Binding Heads in Language Models arXiv:2605.28543v1 Announce Type: cross Abstract: LLMs often default to equal treatment across cultural groups, even though context warrants differentiation: this is a lack of difference awareness. Using mechanistic interpretability and a factorial design on the N4 cultural appropriation benchmark from Wang et al. (2025), we identify 2-3 mid-layer attention heads per model that contribute causally to cultural binding across eight models (four architectu