Generalized Radius and Integrated Codebook Transforms for Differentiable Vector Quantization

Haochen You, Heng Zhang, Hongyang He, Yuqi Li, Baojing Liu · 2026

Read Generalized Radius and Integrated Codebook Transforms for Differentiable Vector Quantization by Haochen You, Heng Zhang, Hongyang He, Yuqi Li, Baojing Liu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-01140.txt.

Description

Haochen You and co-authors make vector quantization fully trainable while keeping hard picks in the forward run. Old straight-through tricks tie step size to the pick gap and train each code alone, giving wobbly slopes and dead codes at scale. Their GRIT-VQ swaps that trick for a radius-based step that moves latents with a set, shape-wise stride, and bends the whole code set through one shared map. Tests on picture rebuild, picture make, and item token jobs lift scores and wake up far more codes.

Catalog record

CREATORHaochen You, Heng Zhang, Hongyang He, Yuqi Li, Baojing Liu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE110388
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.01140

Browse the full library · About the Tower of Babel