AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Spherical Leech Quantization for Visual Tokenization and Generation

Yue Zhao, Hanwen Jiang, Zhenlin Xu, Chutong Yang, Ehsan Adeli, Philipp Kraehenbuehl

CVPR 2026 · Architectures and efficiency › Quantization · approach: Post-training weight quantization · 5 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
replace BSQ (d=14) with Lambda24-SQ at 2^14 codes (ablation)ViT-S tokenizer · ImageNet-128 reconstructionrFID: 12.98 -> 11.16Improved
replace BSQ (d=18) with Lambda24-SQ at large codebook (ablation)ViT-S tokenizer · ImageNet-128 reconstructionrFID: 10.96 -> 8.98Improved
replace random-projection VQ with Lambda24-SQ at 196,560 codes (ablation)ViT-S tokenizer · ImageNet-128 reconstructionrFID: 9.10 -> 8.98Improved
make the codebook learnable (ablation)Lambda24-SQ · ImageNet-128 reconstructionrFID: 8.98 -> 8.78No clear effect
increase codebook size from 2^14 to 196,560 (ablation)Lambda24-SQ · ImageNet-128 reconstructionrFID: 11.16 -> 8.98Improved
add DINOv2 VF alignment to the tokenizer (ablation)VAR + Lambda24-SQ · ImageNet-1k tokenizerrFID: 0.84 -> 0.92Hurt
add DINOv2 VF alignment to the tokenizer (ablation)VAR generation with Lambda24-SQ · ImageNet-1k generationgFID: 6.81 -> 5.79Improved
use full 196,560-way CE head instead of factorized 9-way CE (ablation)Infinity-CC + Lambda24-SQ · ImageNet-1k generationgFID: 11.7 -> 8.7Improved
use full CE head instead of BCE head (ablation)Infinity-CC + BSQ · ImageNet-1k generationgFID: 10.7 -> 10.3Improved
add layerwise linear CFG scaling (ablation)Infinity-CC + Lambda24-SQ · ImageNet-1k generationgFID: 7.46 -> 6.81Improved
add linearly scaling top-k (ablation)Infinity-CC + Lambda24-SQ · ImageNet-1k generationgFID: 6.81 -> 6.68Improved
train tokenizer for 100 epochs instead of 25 (ablation)Lambda24-SQ · VAR tokenizer rFIDrFID: 1.08 -> 0.84Improved

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