Speech and audio·EASYHUB JOURNAL
NVIDIA fine-tunes Nemotron 3.5 ASR on 133.7 hours of Saudi dialect speech, cutting target WER from 55.05% to 29.96%
What changed
NVIDIA published on October 1 a fine-tuning recipe for adapting Nemotron 3.5 ASR to Saudi Najdi and Hijazi dialects. Using 133.7 hours of dialect speech plus minimal curation, FLEURS replay mixing, duration bucketing and partial encoder unfreezing, the team reduced target-set WER from 55.05% to 29.96% while improving English WER from 11.04% to 10.42%. This is a deployment adaptation workflow, not a new base ASR model release.
- Original title
- Fine-Tuning NVIDIA Nemotron for Saudi Arabic Dialects, with a Path to Other Languages
- Source
- NVIDIA · developer.nvidia.com
- Topic
- Speech and audio
- Source month
- 2026-10
This is a concise EasyHub summary of the linked source, not the full report or original reporting. Availability and preview conditions are described in the summary and original.
Summary page published · Editorial information