cs.AI, cs.CV, cs.LG

ArcVQ-VAE: A Spherical Vector Quantization Framework with ArcCosine Additive Margin

arXiv:2605.13517v1 Announce Type: cross
Abstract: Vector Quantized Variational Autoencoder (VQ-VAE) has become a fundamental framework for learning discrete representations in image modeling. However, VQ-VAE models must tokenize entire images using a …