⇪
Drop predictions.jsonl here
one record per utterance_id · UTF-8 · max 20 MB
SUBMISSION NAME
whisper-ft-v3 + lm rescore
TRAINING DATA USED
open subset + own data▾
✓
If we win, we agree our solution is published under an open licence and our scores join the benchmark corpus of the national data commons.
scoring takes ~3 min · you'll be notified in-app
Submit for scoring →