StableVQ: Practical Guidelines for Stable Vector-Quantized Tokenizer Training
cs.CV
Submitted: 2026-09-22
Updated: 2026-09-22
Comments: Project page: https://tt-day.github.io/StableVQ/
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Scalable Training for Vector-Quantized Networks with 100% Codebook Utilization
- VQRAE: Representation Quantization Autoencoders for Multimodal Understanding, Generation and Reconstruction
- Enhancing Vector Quantization with Distributional Matching: A Theoretical and Empirical Study
- Categorical Reparameterization with Gumbel-Softmax
- Flow Matching for Generative Modeling
- Open-MAGVIT2: An Open-Source Project Toward Democratizing Auto-regressive Visual Generation
- Finite Scalar Quantization: VQ-VAE Made Simple
- Score-Based Generative Modeling through Stochastic Differential Equations
- Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation
- Vector-quantized Image Modeling with Improved VQGAN
- Language Model Beats Diffusion -- Tokenizer is Key to Visual Generation
- Diffusion Transformers with Representation Autoencoders
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models