REVIEW 24 cited by
Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptions
read the original abstract
We provide theoretical convergence guarantees for score-based generative models (SGMs) such as denoising diffusion probabilistic models (DDPMs), which constitute the backbone of large-scale real-world generative models such as DALL$\cdot$E 2. Our main result is that, assuming accurate score estimates, such SGMs can efficiently sample from essentially any realistic data distribution. In contrast to prior works, our results (1) hold for an $L^2$-accurate score estimate (rather than $L^\infty$-accurate); (2) do not require restrictive functional inequality conditions that preclude substantial non-log-concavity; (3) scale polynomially in all relevant problem parameters; and (4) match state-of-the-art complexity guarantees for discretization of the Langevin diffusion, provided that the score error is sufficiently small. We view this as strong theoretical justification for the empirical success of SGMs. We also examine SGMs based on the critically damped Langevin diffusion (CLD). Contrary to conventional wisdom, we provide evidence that the use of the CLD does not reduce the complexity of SGMs.
Forward citations
Cited by 24 Pith papers
-
Smoothed Score Queries and the Complexity of Sampling
Smoothed-score queries to the resolvent of a Gaussian precision matrix yield an O((log κ) log(1/δ)) query sampler for TV error δ and an Ω(log κ) bit lower bound, improving the condition-number dependence from √κ.
-
Asymptotic-Preserving A Posteriori Analysis of Diffusion and Flow-Matching Samplers
DDIM (σ-clock Euler) is the unique layer-exact fixed-step sampler; deterministic residual budgets stay O(1) with no log(1/σ_min), while stochastic path-KL scales as Λ²/N from the Itô term alone.
-
Weak Poincar\'e Inequalities via Approximate Stochastic Localization: Application to Sampling the Sherrington-Kirkpatrick Model
Approximate stochastic localization plus conductance transfers yield a weak Poincaré inequality for the SK model at β < 1/2, enabling efficient Glauber sampling from a warm start.
-
When Does Synthetic Data Augmentation Improve Score-Based Imbalanced Classification?
Synthetic minority augmentation improves threshold-integrated and optimized classification metrics only under model misspecification by correcting ranking errors, while providing no fundamental benefit beyond possible...
-
Dimension-Uniform Discretization Analysis of Preconditioned Annealed Langevin Dynamics for Multimodal Gaussian Mixtures
Proves dimension-uniform KL bounds for exponential-integrator discretization of preconditioned ALD on Gaussian mixtures under spectral summability, showing EM stability restrictions are scheme-dependent rather than intrinsic.
-
Training-Free Generative Sampling via Moment-Matched Score Smoothing
MM-SOLD is a training-free particle sampler whose large-particle limit converges to a moment-matched Gibbs distribution obtained by exponentially tilting a score-smoothed target.
-
Proximal-Based Generative Modeling for Bayesian Inverse Problems
PGM replaces the intractable likelihood score in diffusion models with a closed-form Moreau score computed via proximal operators, enabling non-asymptotic sampling for inverse problems trained only on prior data.
-
Proximal-Based Generative Modeling for Bayesian Inverse Problems
PGM framework links diffusion to proximal regularization for closed-form Moreau-score sampling in Bayesian inverse problems, learned only from prior samples.
-
Potential Hessian Ascent III: Sampling the Sherrington--Kirkpatrick Model at Beta < 1/2
Polynomial-time algorithm samples the Sherrington-Kirkpatrick Gibbs measure at beta < 1/2 with o(1) TVD error by combining potential Hessian ascent, stochastic localization, covariance estimates, and Jarzynski equalit...
-
Potential Hessian Ascent III: Sampling the Sherrington--Kirkpatrick Model at Beta < 1/2
A polynomial-time algorithm samples the SK model Gibbs measure with o(1) TVD error for β < 1/2 by combining potential Hessian ascent, stochastic localization, Jarzynski equality, and Glauber dynamics.
-
Generative Modeling from Black-box Corruptions via Self-Consistent Stochastic Interpolants
SCSI iteratively refines a self-consistent transport map to invert black-box corruptions and enable generative modeling of clean data.
-
Minimum Norm Interpolation via The Local Theory of Banach Spaces: The Role of Gaussianity
The sharp MSE bound for the ℓ1-minimum-norm interpolator under isotropic Gaussian covariates is recovered via the geometry of symmetric Gaussian polytopes, without the convex Gaussian min-max theorem.
-
A Variational-Flow Analysis of Diffusion-Based Speech Enhancement under Noise-Power Mismatch
A pathwise variational analysis shows the kink in SI-SDR degradation for StoRM is localized to the predictor via an exact factorization of parametric sensitivity under hypotheses on reverse-process flow.
-
A Variational-Flow Analysis of Diffusion-Based Speech Enhancement under Noise-Power Mismatch
Under three flow hypotheses, non-smoothness of StoRM output at training noise amplitude is equivalent to non-smoothness of the predictor map, via an exact sensitivity factorization that survives Euler–Maruyama discretization.
-
Diffusion Models Adapt to Low-Dimensional Structure Under Flexible Coefficient Choices
For a broad class of coefficients, diffusion models achieve Õ(k/ε) iteration complexity for ε-accurate TV sampling under low-dimensional structure, independent of ambient dimension.
-
Global Convergence of Gradient Descent for Score Matching in Gaussian Mixtures via Reverse Fisher Divergence
Proves global GD convergence on reverse Fisher divergence for GMM score matching to single-Gaussian targets from arbitrary init and to separated GMM targets under random init.
-
Guided Flow Matching for Forward and Inverse PDE Problems with Sparse Observations: Algorithm and Theory
FM4PDE applies flow matching to learn joint PDE coefficient-solution distributions, using guided sampling with composite losses for forward and inverse problems and providing error guarantees under stated assumptions.
-
The Value of Covariance Matching in Gaussian DDPMs and the Lanczos Sampler
Full covariance matching in Gaussian DDPMs yields O(1/T^2) path KL error and is enabled by the training-free Lanczos Gaussian sampler using Jacobian-vector products.
-
Decentralized Diffusion Policy Learning for Enhanced Exploration in Cooperative Multi-agent Reinforcement Learning
Decentralized diffusion policies trained with importance sampling score matching enhance exploration and performance in cooperative MARL over Gaussian policy baselines.
-
Energy Generative Modeling: A Lyapunov-based Energy Matching Perspective
Training and sampling in static scalar energy generative models are two instances of the same Lyapunov-driven density transport dynamics on Wasserstein space, differing only by initial condition, which yields a finite...
-
Diffusion Model for Manifold Data: Score Decomposition, Curvature, and Statistical Complexity
Diffusion models on manifold-supported data admit score decompositions whose statistical rates are controlled by intrinsic dimension and curvature.
-
Scaling World-Model Reinforcement Learning Through Diffusion Policy Optimization
MBDPO reformulates policy optimization as a diffusion process over searched trajectories in latent world models to reduce misalignment between search and value learning.
-
Multi-Objective Learning for Diffusion Models: A Statistical Theory under Semi-Supervised Learning
A semi-supervised MOL framework for diffusion models with generalization bounds depending only on specialist model complexity, extended to diffusion policies for sequential decisions.
-
Stabilizing black-box algorithms through task-oriented randomization
Introduces task-oriented randomization for black-box stability with theoretical guarantees, stability-exploration trade-off analysis, and extension to top-k ranking motivated by LLMs.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.