Confiance d'une IA : ce que la moyenne peut cacherYethikrishna R examine une étude d'Umair Mehmood Imam sur Zuse, une famille de modèles qui choisissent parmi des réponses prédéfinies et attribuent une probabilité à chaque option. Le résultat le plus3 min read·yethikrishnar.hashnode.dev2h ago
How much speech data helps? Read the learning curveYethikrishna R examines research by Alvin Nahabwe and colleagues on a question that matters before a speech project spends more on transcription: how does recognition improve as the amount of training3 min read·yethikrishnar.hashnode.dev8h ago
Before the speech score: check what the benchmark repeatsYethikrishna R examines a study by Christ Kekeli Keli, Steve Sanogo, Patrice Sebastiano, Hamad Aldenawi and Ryan Whetten on building a more comparable speech-recognition benchmark for Ewe. Their Nufo3 min read·yethikrishnar.hashnode.dev14h ago
Kupima AI kwa Kiswahili: orodha ya lugha haitoshiYethikrishna R anachambua utafiti wa Manan Uppadhyay, Himanshu Beniwal, Prashant Kodali na Sunayana Sitaram kuhusu swali la msingi katika kupima AI: lugha ikitajwa katika jaribio, je, matumizi halisi3 min read·yethikrishnar.hashnode.dev20h ago
Why an English confidence threshold needs a Swahili testYethikrishna R examines research by Marek Šuppa and Jaroslav Kopčan on a practical problem for multilingual AI: a confidence rule tuned in English can behave differently when the application changes l3 min read·yethikrishnar.hashnode.dev1d ago