AIInterviewTraining logoAIInterview/Training

cross lingual

AI, ML & GenAI interview questions tagged cross lingual, across every topic.

2 questions · 0 unlocked for you
Open full answer →
Open full answer →

Concepts behind "cross lingual"

The curriculum that explains the ideas these questions test.

Core
🧠 Foundations of LLMs & GenAISign in
Multilingual Models and the Tokenization TaxMultilingual LLMs perform unevenly: best on high-resource languages (English), worse on low-resource ones, because training data is English-heavy. A subtler issue is tokenization: tokenizers trained mostly on English split other languages and non-Latin scripts into far more tokens, so the same meaning costs more tokens, more money, more latency, and less context, a real fairness and cost penalty. AI, ML, and GenAI interviews probe it because global products hit both the quality gap and the token tax, and per-language evaluation exposes what aggregates hide.
Explore the full curriculum →