Skip to content

Dataset · hf-audio

Open ASR Leaderboard test sets

hf-audio/open-asr-leaderboard

ASR tests with audio and text, sorted by length and saved in parquet format.

What's inside

Dataset compiled from open-asr-leaderboard/datasets-test-only. Each split is sorted by audio length, and the format was converted from a loading script to parquet without remote code. Contains audio and transcriptions; size ranges from 100k to 1M examples. Languages are not listed in the card.

What you can do with it

  • Evaluate ASR models on leaderboard test sets.
  • Assess speech recognition quality across audio lengths.
  • Load secure parquet data for benchmarking.
Size
100K–1M rows
Downloads this month
27.1K
Likes
105

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("hf-audio/open-asr-leaderboard")