Zero-Shot Knowledge Base Resizing for Rate-Adaptive Digital Semantic Communication

Shumin Yao 1 , Hui Du 1 , Lifeng Xie 1,∗ , Yaping Sun 1 , Hao Chen 1 , Nan Ma 2,1 , , and Xiaodong Xu 2,1 · 2026

Read Zero-Shot Knowledge Base Resizing for Rate-Adaptive Digital Semantic Communication by Shumin Yao 1 , Hui Du 1 , Lifeng Xie 1,∗ , Yaping Sun 1 , Hao Chen 1 , Nan Ma 2,1 , , and Xiaodong Xu 2,1 (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-2602-01829.txt.

Description

Shumin Yao and colleagues resize the knowledge base of a VQ-VAE system on the fly, instead of one model per rate. Rate is governed by codebook size, a fixed hyperparameter, so control needs one model per size at prohibitive cost. Their zero shot method ranks all vectors in one parent base by semantic hierarchy. It embeds vectors in hyperbolic space, builds a semantic tree by minimum spanning tree, then prunes least important leaves for resizing. Tests show reconstruction near dedicated bases, robust at very low rates where bases fail.

Catalog record

CREATORShumin Yao 1 , Hui Du 1 , Lifeng Xie 1,∗ , Yaping Sun 1 , Hao Chen 1 , Nan Ma 2,1 , , and Xiaodong Xu 2,1
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE23803
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.01829

Browse the full library · About the Tower of Babel