JEPA as a Neural Tokenizer: Learning Robust Speech Representations with Density Adaptive Attention

AI Papers Podcast Daily - JEPA as a Neural Tokenizer: Learning Robust Speech Representations with Density Adaptive Attention