§ feed · storyline
How speech models fail where it matters the most and what to do about it
Together AI publishes research showing state-of-the-art speech models fail on street names 39% of the time and proposes a fix for real-world transcription accuracy.
State-of-the-art speech models like Whisper and Deepgram score near-human on benchmarks — then fail 39% of the time on street names.
New research from Together AI exposes the gap and a fix.
§ sources1 publication · timeline below