Continuous First, Discrete Later: VQ-VAEs Without Dimensional Collapse
cs.LG
Submitted: 2026-05-07
Updated: 2026-09-27
Comments: First two authors contributed equally. Revised to include larger-scale ImageNet-1K experiments
Code: https://github.com/xyz-zy/vqvae_latent_span_collapse
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Moshi: a speech-text foundation model for real-time dialogue
- Jukebox: A Generative Model for Music
- WavTokenizer: an Efficient Acoustic Discrete Codec Tokenizer for Audio Language Modeling
- Bridging Continuous and Discrete Tokens for Autoregressive Visual Generation
- Early Quantization Shrinks Codebook: A Simple Fix for Diversity-Preserving Tokenization
- Representation Collapsing Problems in Vector Quantization
- Addressing Representation Collapse in Vector Quantized Models with One Linear Layer
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks