Interesting thought. Maybe an LLM would build deeper insight with only one training language. On the other hand, the model might overfit with just one language

Interesting thought. Maybe an LLM would build deeper insight with only one training language. On the other hand, the model might overfit with just one language -- maybe multilingual models generalize better?