TY  - RPRT
TI  - Predicting Upcoming Stuttering Events from Three-Second Audio: Stratified Evaluation Reveals Severity-Selective Precursors, and the Model Deploys Fully On-Device
AU  - Nazar Kozak
PY  - 2026
UR  - https://arxiv.org/abs/2604.27279
ID  - 2604.27279
ER  -