Spaces:
Runtime error
is the leaderboard being maintained
@sanchit-gandhi - This is a very useful resource (ASR acc leaderboard). Is it actively maintained ? There are some newer models from Meta/Google since the whisper release that I dont see here yet.
Hey @deepaknets - we'll shortly be releasing a follow-up to this leaderboard with the latest models! cc @reach-vb
is the current leaderboard showing those new updates or are they still coming? are new submissions still being accepted/added?
The leaderboard shows any new (complete) submissions to the ESB benchmark. The ESB paper proposes a 'train-all test-all' approach to evaluation, where each model is both trained and evaluated on each dataset. For a 'train-one test-all' approach, where each model is trained on one dataset and then evaluated on all, see the Open ASR leaderboard: https://huggingface.co/spaces/hf-audio/open_asr_leaderboard