Earnings-22 is a 119-hour corpus of real-world English earnings calls covering 22 publicly traded companies. Heavy domain vocabulary (financial terminology, proper names) and accented speech make it a tough benchmark for production ASR systems.
Wer is the reported evaluation metric for Earnings-22. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.
Lower is better
Muted rows were not state of the art when published — an earlier or same-year result already scored better.