AIBearisharXiv – CS AI · 10h ago6/10
🧠
Lessons Without Borders? Evaluating Cultural Alignment of LLMs Using Multilingual Story Moral Generation
Researchers evaluated how well frontier LLMs like GPT-4o and Gemini interpret story morals across 14 language-culture pairs, finding that while models generate semantically similar outputs to humans, they lack cultural diversity and concentrate on universally shared values rather than culturally-specific moral interpretations.
🧠 GPT-4🧠 Gemini