Bias Beyond Borders: Political Ideology Evaluation and Steering in Multilingual LLMs

TL;DR

This paper evaluates political bias in multilingual LLMs across 50 countries and 33 languages, introducing a novel cross-lingual alignment steering method to reduce bias while maintaining response quality.

Contribution

It presents a large-scale multilingual bias evaluation framework and a new post-hoc mitigation technique, Cross-Lingual Alignment Steering (CLAS), for consistent ideological alignment across languages.

Findings

Significant bias reduction in economic and social axes.

Maintains high response quality after mitigation.

Ensures cross-lingual ideological consistency.

Abstract

Large Language Models (LLMs) increasingly shape global discourse, making fairness and ideological neutrality essential for responsible AI deployment. Despite growing attention to political bias in LLMs, prior work largely focuses on high-resource, Western languages or narrow multilingual settings, leaving cross-lingual consistency and safe post-hoc mitigation underexplored. To address this gap, we present a large-scale multilingual evaluation of political bias spanning 50 countries and 33 languages. We introduce a complementary post-hoc mitigation framework, Cross-Lingual Alignment Steering (CLAS), designed to augment existing steering methods by aligning ideological representations across languages and dynamically regulating intervention strength. This method aligns latent ideological representations induced by political prompts into a shared ideological subspace, ensuring cross lingual consistency, with the adaptive mechanism prevents over correction and preserves coherence. Experiments demonstrate substantial bias reduction along both economic and social axes with minimal degradation in response quality. The proposed framework establishes a scalable and interpretable paradigm for fairness-aware multilingual LLM governance, balancing ideological neutrality with linguistic and cultural diversity.