Cultural Binding Heads in Language Models

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 (base and instruct versions of four architectures). Cultural binding is the process of associating a cultural item with its related identity. Knockout of the identity-to-item edges on these heads lowers the binding strength by 9-23%. The identified heads transfer from instruct to base models, suggesting that cultural binding is created during pre-training. An $\alpha$-scaling shows a graded dose-response. Moderate amplification steering at generation ($\alpha = 2-3$) increases cultural differentiation accuracy by 1-3 pp while leaving reasoning on culturally neutral questions mostly intact. A knowledge probing task shows that models know 3-6 times more than they act upon, indicating that the bottleneck lies in routing and not knowledge.

Authors 2

Avrile Floro, Luca Benedetto

Specification

Official page

arxiv.org/abs/2605.28543

Source:arXiv (Atom API + RSS)·T1·observed 2 h ago·high

Arxiv announce type

replace

Source:arXiv (Atom API + RSS)·T1·observed 2 h ago·high

arXiv id

2605.28543

Source:arXiv (Atom API + RSS)·T1·observed 2 h ago·high

Categories

cs.AI, cs.CL, cs.LG

Source:arXiv (Atom API + RSS)·T1·observed 2 h ago·high

PDF

arxiv.org/pdf/2605.28543

Source:arXiv (Atom API + RSS)·T1·observed 2 h ago·high

Primary category

cs.AI

Source:arXiv (Atom API + RSS)·T1·observed 2 h ago·high

Published

11 Sept 2026

Source:arXiv (Atom API + RSS)·T1·observed 2 h ago·high

Each value shows its source, tier and observation time. Conflicting claims are kept side by side and flagged — never averaged. How AI Atlas records facts →

Provenance