CuMA: Aligning LLMs with Sparse Cultural Values via Demographic-Aware Mixture of Adapters

TL;DR

This paper introduces CuMA, a demographic-aware mixture of adapters framework that aligns large language models with diverse cultural values by disentangling conflicting gradients, thereby preserving cultural diversity and avoiding mean collapse.

Contribution

CuMA presents a novel approach using demographic-aware routing and latent cultural topology to improve cultural alignment in LLMs, outperforming dense and semantic-only models.

Findings

CuMA achieves state-of-the-art results on multiple cultural alignment benchmarks.

CuMA effectively mitigates mean collapse, maintaining cultural diversity.

Demographic-aware routing improves alignment with diverse cultural values.

Abstract

As Large Language Models (LLMs) serve a global audience, alignment must transition from enforcing universal consensus to respecting cultural pluralism. We demonstrate that dense models, when forced to fit conflicting value distributions, suffer from \textbf{Mean Collapse}, converging to a generic average that fails to represent diverse groups. We attribute this to \textbf{Cultural Sparsity}, where gradient interference prevents dense parameters from spanning distinct cultural modes. To resolve this, we propose \textbf{\textsc{CuMA}} (\textbf{Cu}ltural \textbf{M}ixture of \textbf{A}dapters), a framework that frames alignment as a \textbf{conditional capacity separation} problem. By incorporating demographic-aware routing, \textsc{CuMA} internalizes a \textit{Latent Cultural Topology} to explicitly disentangle conflicting gradients into specialized expert subspaces. Extensive evaluations on WorldValuesBench, Community Alignment, and PRISM demonstrate that \textsc{CuMA} achieves state-of-the-art performance, significantly outperforming both dense baselines and semantic-only MoEs. Crucially, our analysis confirms that \textsc{CuMA} effectively mitigates mean collapse, preserving cultural diversity. Our code is available at https://github.com/Throll/CuMA.