ai-evaluation

Twi-English maternal-health dataset: What Mixed Questions Reveal About AI Understanding

Why 20,000 Twi-English maternal-health Q&A pairs can reveal what benchmark scores miss about real code-switching.