AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Addressing Representation Collapse in Vector Quantized Models with One Linear Layer

Yongxin Zhu, Bocheng Li, Yifei Xin, Zhihua Xia, Linli Xu

ICCV 2025 · Architectures and efficiency › RNNs and normalization · approach: Sparse dictionary representations · 61 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
reparameterize the codebook with a learnable linear layer (SimVQ) instead of vanilla VQGANVQGAN · ImageNet-1k 128x128rFID: 3.74 -> 2.24Improved
reparameterize the codebook with a linear layerVQGAN · ImageNet-1k 128x128codebook utilization: 1.4% -> 100.0%Improved
freeze the coefficient codebook and train only the linear basis (ablation)SimVQ · ImageNet-1k 128x128rFID: 2.31 -> 2.24Improved
use Gaussian instead of uniform initialization of the frozen codebook (ablation)SimVQ · ImageNet-1k 128x128rFID: 2.31 -> 2.24No clear effect
scale codebook size from 1,024 to 262,144 (ablation)SimVQ · ImageNet-1k 128x128rFID: 3.67 -> 1.99Improved
scale codebook size from 8,192 to 65,536 (ablation)SimVQ · ImageNet-1k 128x128rFID: 2.98 -> 2.24Improved
scale codebook size from 65,536 to 262,144 (ablation)SimVQ · ImageNet-1k 128x128SSIM: 78.4 -> 80.3Improved
scale codebook size from 100,000 to 200,000VQGAN-LC-CLIP · ImageNet-1k 256x256rFID: 2.62 -> 2.66Hurt
replace CLIP-initialized codebook with SimVQVQGAN-LC-CLIP · ImageNet-1k 128x128rFID: 2.40 -> 2.24Improved
lower latent dimension from 128 to 8VQGAN-FC · ImageNet-1k 128x128rFID: 5.33 -> 2.63Improved
add EMA codebook updateVQGAN · ImageNet-1k 128x128rFID: 3.74 -> 3.23Improved
replace EMA codebook of WavTokenizer with SimVQWavTokenizer · LibriTTS test-cleanUTMOS: 3.74 -> 4.00Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy