Vector Quantization

VQ (STE)

[NeurIPS 2017] [arXiv:1711.00937] Neural Discrete Representation Learning

[CVPR 2021] [arXiv:2012.09841] Taming Transformers for High-Resolution Image Synthesis

VQ (Gumbel Softmax)

[ICLR 2020] [arXiv:1910.05453] vq-wav2vec: Self-Supervised Learning of Discrete Speech Representations

Improved VQ

[ICLR 2022] [arXiv:2110.04627] Vector-quantized Image Modeling with Improved VQGAN

FSQ

[ICLR 2024] [arXiv:2309.15505] Finite Scalar Quantization: VQ-VAE Made Simple

LFQ

[ICLR 2024] [arXiv:2310.05737] Language Model Beats Diffusion: Tokenizer is Key to Visual Generation

BSQ

[ICLR 2025] [arXiv:2406.07548] Image and Video Tokenization with Binary Spherical Quantization

Rotation Trick

[ICLR 2025] [ arXiv:2410.06424] Restructuring Vector Quantization with the Rotation Trick

SimVQ

[ICCV 2025] [arXiv:2411.02038] Addressing Representation Collapse in Vector Quantized Models with One Linear Layer

GSQ

[arXiv:2412.02632] Scaling Image Tokenizers with Grouped Spherical Quantization

IBQ

[ICCV 2025] [arXiv:2412.02692] Taming Scalable Visual Tokenizer for Autoregressive Image Generation

SoftVQ

[CVPR 2025] [arXiv:2412.10958] SoftVQ-VAE: Efficient 1-Dimensional Continuous Tokenizer

OptVQ

[arXiv:2412.15195] Preventing Local Pitfalls in Vector Quantization via Optimal Transport