Cultural Binding Heads in Language Models

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). It comes from a paper posted to arXiv on 11 September 2026. LLMs often default to equal treatment across cultural groups, even though context warrants differentiation: this is a lack of difference awareness. 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.

Why it counts

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).

Sources

Primary source: primary source

What is not known

This brief does not claim independent replication. Claims that appear only on X and not in the primary source stay unknown.