mlsanityData to 14 Oct 2026

Optimizing Input of Denoising Score Matching is Biased Towards Higher Score Norm

arXiv 2511.11727 ↗First posted 13 Nov 2025cs.LG cs.CVcited by 1

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Many recent works utilize denoising score matching to optimize the conditional input of diffusion models. In this workshop paper, we demonstrate that such optimization breaks the equivalence between denoising score matching and exact score matching. Furthermore, we show that this bias leads to higher score norm. Additionally, we observe a similar bias when optimizing the data distribution using a pre-trained diffusion model. Finally, we discuss the wide range of works across different domains that are affected by this bias, including MAR for auto-regressive generation, PerCo for image compression, and DreamFusion for text to 3D generation.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 22 of its 25 references that are in the map. First posted on arXiv 13 Nov 2025.

Builds on

22 of 25 references.

Closest ancestors · 6

Same direction, newest first

  1. 2025DiSA: Diffusion Step Annealing in Autoregressive Image Generation26 May
  2. 2025Flow to the Mode: Mode-Seeking Diffusion Autoencoders for State-of-the-Art Image Tokenization14 Mar
  3. 2024Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion1 Jul
  4. 2024CFG++: Manifold-constrained Classifier Free Guidance for Diffusion Models12 Jun
  5. 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov
  6. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 16

  1. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,058
  2. 2017Neural Discrete Representation Learningcited by 4,712
  3. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,101
  4. 2022DreamFusion: Text-to-3D using 2D Diffusioncited by 3,025
  5. 2024Autoregressive Image Generation without Vector Quantizationcited by 651
  6. 2025Transfer between Modalities with MetaQueriescited by 236
  7. 2023A Variational Perspective on Solving Inverse Problems with Diffusion Modelscited by 222
  8. 2024Autoregressive Video Generation without Vector Quantizationcited by 186
  9. 2022Lossy Image Compression with Conditional Diffusion Modelscited by 170
  10. 2024Fluid: Scaling Autoregressive Text-to-image Generative Models with Continuous Tokenscited by 147
  11. 2024HART: Efficient Visual Generation with Hybrid Autoregressive Transformercited by 132
  12. 2023Towards image compression with perfect realism at ultra-low bitratescited by 103
4 more
  1. 2024Autoregressive Speech Synthesis without Vector Quantizationcited by 65
  2. 2025NextStep-1: Toward Autoregressive Image Generation with Continuous Tokens at Scalecited by 57
  3. 2023Taming Mode Collapse in Score Distillation for Text-to-3D Generationcited by 36
  4. 2025Frequency Autoregressive Image Generation with Continuous Tokenscited by 27

Built on by

Cited by year

Citing papers per year: 2025: 1.

1 paper.

  1. 2025CoD: A Diffusion Foundation Model for Image Compressioncited by 10

Similar, not cited

Close in wording, no citation either way

  1. 2024From optimal score matching to optimal sampling
  2. 2026Denoising Score Matching in Diffusion Models: The Effect of Rank Constraints on Generalization
  3. 2022From Denoising Diffusions to Denoising Markov Models
  4. 2025Denoising Score Matching with Random Features: Insights on Diffusion Models from Precise Learning Curves
  5. 2024Efficient Denoising using Score Embedding in Score-based Diffusion Models
  6. 2024Theory on Score-Mismatched Diffusion Models and Zero-Shot Conditional Samplers
  7. 2025Implicit score matching meets denoising score matching: improved rates of convergence and log-density Hessian estimation
  8. 2023Consistent Diffusion Models: Mitigating Sampling Drift by Learning to be Consistent