Mitigating Cross-Lingual Cultural Inconsistencies in LLMs via Consensus-Driven Preference Optimisation
Authors: Lucas Resck, Isabelle Augenstein, Anna Korhonen
Summary
arXiv:2605. 12515v1 Announce Type: new Abstract: Despite their impressive capabilities, multilingual large language models (MLLMs) frequently exhibit inconsistent behaviour when the prompt's language changes.
Relevance
Read next because Mitigating Cross-Lingual Cultural Inconsistencies in LLMs via Consensus-Driven Preference Optimisation overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: strong, under, alignment, eval, source, persona, line, prompt. Source: arxiv cs.CL (NLP).
Threat model
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses failure, evaluation.
Abstract
arXiv:2605.12515v1 Announce Type: new Abstract: Despite their impressive capabilities, multilingual large language models (MLLMs) frequently exhibit inconsistent behaviour when the prompt's language changes. While such adaptation is generally desirable, it becomes a critical failure when a user's identity is explicitly defined. For instance, given a fixed British persona and an ambiguous everyday knowledge query about literature, the prompt's language frequently overwrites the system persona -- yielding Shakespeare in English but Cervantes in Spanish. To robustly quantify this Cross-lingual Cultural Inconsistency, we introduce Singleton Fleiss's $\kappa_S$, a metric mathematically resilient to hallucinations. For mitigation, we propose Cross-lingual Cultural Consistent Preference Optimisation (C-3PO), a consensus-driven alignment framework. C-3PO achieves up to a 0.10-point absolute increase in $\kappa_S$ over unaligned models, outperforming strong prompting and representation steering baselines. Empirical evaluations show this inconsistency disproportionately affects lower-resource languages like Indonesian and Persian. A layer-wise interpretability analysis reveals the underlying mechanism: by early-decoding intermediate layer representations, we find that MLLMs implicitly personalise outputs towards the prompt language's stereotypical culture as forward-pass representations stabilise.