Audio validation set organized as validation/audio/* plus validation/metadata.jsonl. The metadata contains only file_name and transcription; transcriptions include timestamp and speaker markers.
validation/audio/*
validation/metadata.jsonl
file_name
transcription
