Why are all LLMs Obsessed with Japanese Culture? On the Hidden Cultural and Regional Biases of LLMs

TL;DR

This paper investigates regional and cultural biases in LLMs, revealing a tendency towards Japanese culture and showing that biases emerge after supervised fine-tuning rather than during pre-training.

Contribution

It introduces a new dataset based on Culture-Related Open Questions to analyze LLM regional preferences and biases.

Findings

LLMs tend to favor Japanese culture in responses.

Prompting in high-resource languages reduces cultural bias.

Biases emerge after supervised fine-tuning, not pre-training.

Abstract

LLMs have been showing limitations when it comes to cultural coverage and competence, and in some cases show regional biases such as amplifying Western and Anglocentric viewpoints. While there have been works analysing the cultural capabilities of LLMs, there has not been specific work on highlighting LLM regional preferences when it comes to cultural-related questions. In this work, we propose a new dataset based on a comprehensive taxonomy of Culture-Related Open Questions (CROQ). The results show that, contrary to previous cultural bias work, LLMs show a clear tendency towards countries such as Japan. Moveover, our results show that when prompting in languages such as English or other high-resource ones, LLMs tend to provide more diverse outputs and show less inclinations towards answering questions highlighting countries for which the input language is an official language. Finally, we also investigate at which point of LLM training this cultural bias emerges, with our results suggesting that the first clear signs appear after supervised fine-tuning, and not during pre-training.