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

2026-01-08

arXiv.org

AI Summary

This work addresses the challenge of aligning large language models with diverse global user values, where cultural sparsity often leads to “mean collapse,” undermining the representation of cultural plurality. To mitigate this, the authors formulate cultural alignment as a conditional capacity disentanglement problem and propose the Cultural Mixture-of-Adapters (CuMA) framework. CuMA explicitly handles cultural sparsity and gradient interference through a demographic-aware mixture-of-experts architecture, latent cultural topology modeling, and gradient-decoupled training, enabling disentangled value alignment. Evaluated on the WorldValuesBench, Community Alignment, and PRISM benchmarks, CuMA achieves state-of-the-art performance, significantly outperforming dense models and semantics-driven MoE approaches while effectively preserving cultural diversity.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Humans and AI: Learning Human Values and Preferences
• Natural Language Processing: (Large) Language Models

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

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.

Problem

• Mean Collapse
• Cultural Sparsity
• Cultural Pluralism
• Value Alignment
• Large Language Models

Innovation

• Cultural Sparsity
• Mean Collapse
• Mixture of Adapters
• Demographic-Aware Routing
• Latent Cultural Topology