mlsanityData to 7 Oct 2026

Latent Structure Emergence in Diffusion Models via Confidence-Based Filtering

arXiv 2602.06155 ↗First posted 5 Feb 2026cs.LG stat.MLcited by 0

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Diffusion models generate samples through a sequence of learned denoising steps, and recent work has studied how semantic structure appears along this sampling process. We study this question in deterministic samplers by measuring semantic accessibility: how much information about a final semantic property, such as an image class label or attribute, can be extracted from the seed and intermediate states along the trajectory that produces the sample. Using DDIM sampling, for which each initial noise seed determines a unique trajectory and final image, we train separate classifiers (probes) at several points along the trajectory to predict a semantic property of the final image. We measure how well such a property can be predicted from the state at a given point using top-1 accuracy and normalized mutual information. Across MNIST, Fashion-MNIST, CIFAR-10, and CelebA, class labels and image attributes can be predicted above chance from the initial noise seed, and along DDIM trajectories, this accessibility exceeds matched-noise forward baselines. We find that semantic accessibility is substantially higher along the trajectories whose final images are classified with high confidence than along those classified with low confidence. These measurements provide a quantitative view of when semantic properties can be recovered during deterministic generation.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 30 of its 35 references that are in the map. First posted on arXiv 5 Feb 2026.

Builds on

30 of 35 references.

Closest ancestors · 14

Same direction, newest first

  1. 2025Measuring Semantic Information Production in Generative Diffusion Models12 Jun
  2. 2024CleanDIFT: Diffusion Features without Noise4 Dec
  3. 2024textitRevelio: Interpreting and leveraging semantic information in diffusion models23 Nov
  4. 2024Latent Abstractions in Generative Diffusion Models4 Oct
  5. 2024Dynamical Regimes of Diffusion Models28 Feb
  6. 2024A Phase Transition in Diffusion Models Reveals the Hierarchical Nature of Data26 Feb
  7. 2023Diffusion Model as Representation Learner21 Aug
  8. 2023Spontaneous Symmetry Breaking in Generative Diffusion Models31 May
  9. 2022Diffusion Models already have a Semantic Latent Space20 Oct
  10. 2022Classifier-Free Diffusion Guidance26 Jul
  11. 2021Diffusion Models Beat GANs on Image Synthesis11 May
  12. 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov
2 more
  1. 2020Denoising Diffusion Implicit Models6 Oct
  2. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 16

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  3. 2017On Calibration of Modern Neural Networkscited by 4,860
  4. 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,283
  5. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  6. 2016Understanding intermediate layers using linear classifier probescited by 1,279
  7. 2021Probing Classifiers: Promises, Shortcomings, and Advancescited by 831
  8. 2017Selective Classification for Deep Neural Networkscited by 799
  9. 2019Designing and Interpreting Probes with Control Taskscited by 608
  10. 2023Emergent Correspondence from Image Diffusioncited by 437
  11. 2023Diffusion Hyperfeatures: Searching Through Time and Space for Semantic Correspondencecited by 211
  12. 2020A Theory of Usable Information Under Computational Constraintscited by 153
4 more
  1. 2024InitNO: Boosting Text-to-Image Diffusion Models via Initial Noise Optimizationcited by 123
  2. 2023Generating images of rare concepts using pre-trained diffusion modelscited by 78
  3. 2023Readout Guidance: Learning Control from Diffusion Featurescited by 52
  4. 2024Good Seed Makes a Good Crop: Discovering Secret Seeds in Text-to-Image Diffusion Modelscited by 47

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Probing the Latent Hierarchical Structure of Data via Diffusion Models
  2. 2025Optimal Stopping in Latent Diffusion Models
  3. 2026Noise in Diffusion Models Is a Learnable Input
  4. 2026Provably Learning Diffusion Models under the Manifold Hypothesis: Collapse and Refine
  5. 2025Diffusion models under low-noise regime
  6. 2026Emergence of Distortions in High-Dimensional Guided Diffusion Models
  7. 2026Diffusion Models Are Statistically Optimal for Learning Low-Dimensional Multi-Modal Distributions
  8. 2026The Entropic Signature of Class Speciation in Diffusion Models