Exploring Large Language Models on Cross-Cultural Values in Connection with Training Methodology
This work addresses the problem of cultural bias in LLMs for developers and users, offering incremental insights into training methodologies to enhance cross-cultural understanding.
The paper investigates how open-source large language models (LLMs) judge cross-cultural values, finding they align with human norms on socio-cultural aspects but show biases toward Western culture, which can be mitigated by multilingual training and improved with larger model sizes or synthetic data for smaller models.
Large language models (LLMs) closely interact with humans, and thus need an intimate understanding of the cultural values of human society. In this paper, we explore how open-source LLMs make judgments on diverse categories of cultural values across countries, and its relation to training methodology such as model sizes, training corpus, alignment, etc. Our analysis shows that LLMs can judge socio-cultural norms similar to humans but less so on social systems and progress. In addition, LLMs tend to judge cultural values biased toward Western culture, which can be improved with training on the multilingual corpus. We also find that increasing model size helps a better understanding of social values, but smaller models can be enhanced by using synthetic data. Our analysis reveals valuable insights into the design methodology of LLMs in connection with their understanding of cultural values.