HF guides Nemotron ASR fine-tuning

Hugging Face published a per-language fine-tuning guide for the 'Nemotron 3.5 ASR' speech model. The composition is strongly academic—one official HF source plus four arXiv papers—research accumulation bundled into implementation steps rather than an announcement, taking the form of a how-to anyone can follow. The focus is fine-tuning a general ASR model to each language's characteristics to raise accuracy; the through-line is a move from 'one big model' toward lightly adapting to each language in practice. The notable part is organizing multilingual ASR so practitioners can optimize it themselves. But it's right after release—the degree of per-language accuracy gains and adoption in non-English regions are to confirm.