Mitigating Cross-Lingual Cultural Inconsistencies in LLMs via Consensus-Driven Preference Optimisation

Multilingual large language models (MLLMs) often exhibit inconsistent behavior regarding cultural identity when the prompt's language changes. Researchers introduce a new metric, Singleton Fleiss's "k_S", and a consensus-driven alignment framework, C-3PO, to mitigate these cross-lingual cultural inconsistencies, achieving significant improvements.