chunked/content_aware_chunking/_glossary/chunk_115.txt

For example, Wav2Vec2 is a speech model pretrained on a contrastive task which requires the model to identify the "true" speech representation from a set of "false" speech representations.