Fact · Speech recognition
1.8percent word error rate
With all of LibriSpeech's labeled training data, wav2vec 2.0 reached word error rates of 1.8 percent on the clean test set and 3.3 percent on the harder other test set.
Checked against its source on October 16, 2026
In the source's own words
“Experiments using all labeled data of Librispeech achieve 1.8/3.3 WER on the clean/other test sets.”
wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations, Alexei Baevski et al., arXiv (NeurIPS 2020), 2020, Abstract.
Open the original- What it measures
- LibriSpeech is read audiobook English. The model was first pre-trained on unlabeled audio, then fine-tuned on the labeled data.
- What it does not show
- Clean read speech only. Error rates on conversation, noise or accents are much higher.