Investigating Sample Bias Towards Languages in Audio Super-resolution

This award-winning research demonstrated that AI models trained only on English perform unpredictably on other languages. Using a massive crowdsourced multilingual dataset, we developed a framework for de-risking global AI product launches through market-specific testing.