mlsanityData to 13 Oct 2026

DiVeQ: Differentiable Vector Quantization Using the Reparameterization Trick

arXiv 2509.26469 ↗First posted 30 Sep 2025cs.LGcited by 5

Not on the map. Why?

Abstract (arXiv)

Vector quantization is common in deep models, yet its hard assignments block gradients and hinder end-to-end training. We propose DiVeQ, which treats quantization as adding an error vector that mimics the quantization distortion, keeping the forward pass hard while letting gradients flow. We also present a space-filling variant (SF-DiVeQ) that assigns input to a curve constructed by the lines connecting codewords, resulting in less quantization error and full codebook usage. Both methods train end-to-end without requiring auxiliary losses or temperature schedules. In VQ-VAE image compression, VQGAN image generation, and DAC speech coding tasks across various data sets, our proposed methods improve reconstruction and sample quality over alternative quantization approaches.

Key facts. Cited by 5 papers in the map (data through 13 Oct 2026). Builds on 42 of its 57 references that are in the map. First posted on arXiv 30 Sep 2025.

Builds on

42 of 57 references.

References · 42

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,777
  2. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,082
  3. 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 6,000
  4. 2017Neural Discrete Representation Learningcited by 4,680
  5. 2020Taming Transformers for High-Resolution Image Synthesiscited by 3,237
  6. 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,975
  7. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,726
  8. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,056
  9. 2015LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loopcited by 1,350
  10. 2016The Concrete Distribution: A Continuous Relaxation of Discrete Random Variablescited by 1,142
  11. 2019StarGAN v2: Diverse Image Synthesis for Multiple Domainscited by 1,095
  12. 2022MaskGIT: Masked Generative Image Transformercited by 1,090
30 more
  1. 2021SoundStream: An End-to-End Neural Audio Codeccited by 762
  2. 2022Autoregressive Image Generation using Residual Quantizationcited by 709
  3. 2021Vector-quantized Image Modeling with Improved VQGANcited by 657
  4. 2021VideoGPT: Video Generation using VQ-VAE and Transformerscited by 653
  5. 2023Language Model Beats Diffusion -- Tokenizer is Key to Visual Generationcited by 636
  6. 2020Jukebox: A Generative Model for Musiccited by 537
  7. 2023Finite Scalar Quantization: VQ-VAE Made Simplecited by 433
  8. 2023High-Fidelity Audio Compression with Improved RVQGANcited by 427
  9. 2024Latent Action Pretraining from Videoscited by 308
  10. 2021PeCo: Perceptual Codebook for BERT Pre-training of Vision Transformerscited by 204
  11. 2022Self-supervised Learning with Random-projection Quantizer for Speech Recognitioncited by 159
  12. 2017Soft-to-Hard Vector Quantization for End-to-End Learning Compressible Representationscited by 152
  13. 2024Image and Video Tokenization with Binary Spherical Quantizationcited by 110
  14. 2023Straightening Out the Straight-Through Estimator: Overcoming Optimization Challenges in Vector Quantized Networkscited by 81
  15. 2022UViM: A Unified Modeling Approach for Vision with Learned Guiding Codescited by 77
  16. 2022SQ-VAE: Variational Bayes on Discrete Representation with Self-annealed Stochastic Quantizationcited by 68
  17. 2023Towards Accurate Image Coding: Improved Autoregressive Image Generation with Dynamic Vector Quantizationcited by 59
  18. 2023Online Clustered Codebookcited by 55
  19. 2020Robust Training of Vector Quantized Bottleneck Modelscited by 54
  20. 2024Restructuring Vector Quantization with the Rotation Trickcited by 50
  21. 2023Regularized Vector Quantization for Tokenized Image Synthesiscited by 46
  22. 2023Designing a Better Asymmetric VQGAN for StableDiffusioncited by 45
  23. 2024A Data-Driven Representation for Sign Language Productioncited by 17
  24. 2024HyperVQ: MLR-based Vector Quantization in Hyperbolic Spacecited by 13
  25. 2023Soft Convex Quantization: Revisiting Vector Quantization with Convex Optimizationcited by 10
  26. 2025Compressing 3D Gaussian Splatting by Noise-Substituted Vector Quantizationcited by 7
  27. 2024Balance of Number of Embedding and their Dimensions in Vector Quantizationcited by 4
  28. 2025HH-Codec: High Compression High-fidelity Discrete Neural Codec for Spoken Language Modelingcited by 2
  29. 2025Scalable Transceiver Design for Multi-User Communication in FDD Massive MIMO Systems via Deep Learningcited by 1
  30. 2025Spotlight-TTS: Spotlighting the Style via Voiced-Aware Style Extraction and Style Direction Adjustment for Expressive Text-to-Speechcited by 1

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Online Vector Quantized Attentioncited by 1
  2. 2026AVQ-Attention: Adaptive Vector-Quantized Attentioncited by 1
  3. 2026Generalized Radius and Integrated Codebook Transforms for Differentiable Vector Quantizationcited by 0
  4. 2026Divide et Calibra: Multiclass Local Calibration via Vector Quantizationcited by 0
  5. 2026What Makes a Good Medical Image Tokenizer? Rethinking Reconstruction and Generation in Medical Image Tokenizationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Addressing Representation Collapse in Vector Quantized Models with One Linear Layer
  2. 2026HAS-VQ: Hessian-Adaptive Sparse Vector Quantization for High-Fidelity LLM Compression
  3. 2021Differentiable Model Compression via Pseudo Quantization Noise
  4. 2026Beyond Stationarity: Rethinking Codebook Collapse in Vector Quantization
  5. 2025Training-Free Vector Quantization via Gaussian VAEs
  6. 2026Efficient VQ-QAT and Mixed Vector/Linear quantized Neural Networks
  7. 2026MGVQ: Synergizing Multi-dimensional Sensitivity-Aware and Gradient-Hessian Fusion for Vector Quantization
  8. 2024MVQ:Towards Efficient DNN Compression and Acceleration with Masked Vector Quantization