arXiv cs.CL· Lucas Resck, Isabelle Augenstein, Anna Korhonen·· 12 小时前AI 评分29
用共识驱动的偏好优化缓解 LLM 跨语言文化不一致
Mitigating Cross-Lingual Cultural Inconsistencies in LLMs via Consensus-Driven Preference Optimisation
AI 导读
研究者提出 C-3PO 共识驱动对齐框架，将跨语言文化一致性指标 κ_S 在未对齐模型基础上提升 0.07–0.10，超过所有提示词与表征引导基线。该框架仅用多选题训练，也能改善开放式生成的一致性，并基本保留用户身份、文化中立性与内在文化知识。评估显示，印尼语、波斯语等低资源语言的不一致更明显。
正文 · 原文
Abstract:Multilingual large language models (MLLMs) often answer culturally ambiguous questions inconsistently across languages, even for a fixed user persona -- e.g., Shakespeare in English but Cervantes in Spanish for the same British user. 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 raises $\kappa_S$ by 0.07--0.10 over unaligned models, above every prompting and representation steering baseline averaged across personas, whilst largely preserving explicit user identities, cultural neutrality and intrinsic cultural knowledge; where it does not, a soft-consensus variant recovers most of the cost. Although trained only on multiple-choice questions, C-3PO also improves consistency in open-ended generation. Empirical evaluations demonstrate this inconsistency is more pronounced for lower-resource languages like Indonesian and Persian. Finally, early decoding of intermediate layers reveals that MLLMs implicitly personalise outputs towards the prompt language's stereotypical culture as forward-pass representations stabilise.
| Comments: | 25 pages, 14 figures, 17 tables |
| Subjects: | Computation and Language (cs.CL) |
| ACM classes: | I.2.7; I.2.6 |
| Cite as: | arXiv:2605.12515 [cs.CL] |
| (or arXiv:2605.12515v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.12515 arXiv-issued DOI via DataCite |
Submission history
From: Lucas Resck [view email]
[v1]
Thu, 2 Apr 2026 14:04:06 UTC (207 KB)
[v2]
Wed, 27 May 2026 15:20:41 UTC (239 KB)
[v3]
Fri, 16 Oct 2026 15:46:19 UTC (193 KB)
来源：arXiv cs.CL · arxiv.org