Author name: Xinyu Zhao, Nikita Karagodin, Hamed Hassani, Sinan Hersek, Paul Pu Liang, Yury Polyanskiy

arXiv:2605.06870v2 Announce Type: replace
Abstract: While many approaches to improve VQ-VAE performance focus on codebook size and utilization, the effect of dimensional collapse, where trained VQ-VAE representations live in an extremely low-dimension…

arXiv:2605.06870v1 Announce Type: new
Abstract: While many approaches to improve VQ-VAE performance focus on codebook size and utilization, the effect of dimensional collapse, where trained VQ-VAE representations live in an extremely low-dimensional s…