Initial dataset: 1,931 transcripts, 59K SFT pairs, 85K RAG chunks, 765 bilingual terms bf38a2e verified audreyt commited on 14 days ago