How Many Directions Must a Truncated Diffusion Sampler Retain? Matching Bounds Under Power-Law Spectra
On the map: Diffusion models → Diffusion foundations and sampling also fits: Flow matching, time series and science
Abstract (arXiv)
Diffusion samplers can reduce computation by generating selected spectral coordinates and filling the remaining directions with noise. How many directions must they retain? We study this question for data with power-law covariance spectra. For Gaussian data compared to a smoothed target, we prove matching bounds on the required number of retained directions, provided that the ambient dimension is sufficiently large. The truncation error depends on the combined Wiener gains of the omitted directions, regardless of the accuracy of the sampler on the retained coordinates. Keeping only directions whose signal exceeds the output noise level can therefore leave a non-vanishing error: many individually weak directions remain significant in aggregate. Combining this characterization with a diffusion convergence bound yields sufficient sampling-step complexity under exact scores. The upper bounds also extend to estimated principal components and, componentwise, to Gaussian mixtures. The practical prescription is to select the retained subspace using an aggregate spectral-tail error budget, then to choose the diffusion noise level accordingly.
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 9 of its 32 references that are in the map. First posted on arXiv 7 Oct 2026.
Builds on
9 of 32 references.
Closest ancestors · 8
Same direction, newest first
- 2025Low-dimensional adaptation of diffusion models: Convergence in total variation22 Jan
- 2024Linear Convergence of Diffusion Models Under the Manifold Hypothesis11 Oct
- 2024Convergence of Diffusion Models Under the Manifold Hypothesis in High-Dimensions27 Sep
- 2023Nearly d-Linear Convergence Bounds for Diffusion Models via Stochastic Localization7 Aug
- 2022Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions22 Sep
- 2022Convergence of denoising diffusion models under the manifold hypothesis10 Aug
- 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
- 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov
Other references · 1
- 2016FMA: A Dataset For Music Analysiscited by 205
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Asymptotic Preservation and Uniform Accuracy of Diffusion and Flow-Matching Samplers
- 2026Lower Bounds for Parallel Diffusion Sampling
- 2026When Diffusion Model Can Ignore Dimension: An Entropy-Based Theory
- 2024Faster Diffusion Sampling with Randomized Midpoints: Sequential and Parallel
- 2026A Random Matrix Theory Perspective on the Consistency of Diffusion Models
- 2026Covariance-aware sampling for Diffusion Models
- 2023Restoration-Degradation Beyond Linear Diffusions: A Non-Asymptotic Analysis For DDIM-Type Samplers
- 2025From Score Matching to Diffusion: A Fine-Grained Error Analysis in the Gaussian Setting