让多语言大模型在不同语言下保持文化一致性，避免身份错乱。
Mitigating Cross-Lingual Cultural Inconsistencies in LLMs via Consensus-Driven Preference Optimisation
- 基于共识的偏好优化框架，稳定跨语言文化表现
- 在低资源语言上提升0.13点文化一致性评分
- 适合关注多语言模型公平性与文化安全的研究者
尽管多语言大语言模型能力强大，但当提示语种变化时，其行为常出现不一致。例如，设定英国人设后，英语提问输出莎士比亚，西班牙语提问却输出塞万提斯。为量化这种跨语言文化不一致，我们提出单一弗莱斯κ_S，该指标对幻觉具有数学鲁棒性。为此，我们提出跨语言文化一致偏好优化（C-3PO）框架。实验证明，该方法相较未对齐模型在κ_S上提升最高达0.13点，优于强提示与表征引导基线，且能保持用户身份、文化中立性与内在文化知识。评估显示，该不一致现象在印尼语、波斯语等低资源语言中尤为严重。早期解码表明，模型在前向传播中逐渐将输出个性化为提示语种的刻板文化印象。
原文摘要 · Abstract (English)
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 $κ_S$, 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 0.13-point absolute increase in $κ_S$ over unaligned models, consistently outperforming strong prompting and representation steering baselines whilst preserving explicit user identities, cultural neutrality and intrinsic cultural knowledge. Empirical evaluations demonstrate this inconsistency disproportionately affects 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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。