#machine learning
Mar 2026
Probing Cultural Signals in Large Language Models through Author Profiling
Across several open-source models evaluated on more than 10,000 lyrics, it is found that LLMs achieve non-trivial profiling performance but demonstrate systematic cultural alignment: most models default toward North American ethnicity, while DeepSeek-1.5B aligns more strongly with Asian ethnicity.
Valentin Lafargue, A. Guerra-Adames, E. Claeys et al.
· arXiv.org · 0 citations