Diffusion Models

Ask about this topic

Overview

Diffusion models learn a data distribution by gradually corrupting examples with noise and training a denoiser to reverse that process. Major approaches include DDPMs and score-based SDEs, latent diffusion and diffusion Transformers, flow and consistency methods, and discrete masked or continuous-time Markov diffusion; conditioning and guidance adapt them to specific tasks. They are used for images, video, audio, language, molecules, and scientific forecasting, with evaluations ranging from ImageNet FID to language perplexity and domain-specific measures of accuracy and uncertainty. A central tension is how to make sampling faster and generation more controllable without sacrificing quality, calibration, or diversity, while theory and ongoing audits address generalization, memorization, and privacy.

Latest papers

CardsList

1,192 papers152 papers in the last four weeks, up 77% on the four weeks before. 0.9% of all new papers.

May 21, 2026

Rethinking Token Reduction for Diffusion Models via Output-Similarity-Awareness

Diffusion Transformers (DiTs) achieve superior image generation quality but suffer from quadratic computational complexity relative to token count. While various token reduction (TR) methods have been proposed to mitigate this cost, they overlook the primary objective of generative models: minimizing recovery error, which requires reflecting output token similarity. They rely solely on input token similarity inherited from reduction-only ViT paradigms, leading to a fundamental misalignment with this objective. To bridge this gap, we propose DiTo, a novel TR paradigm that shifts the focus toward output-centric token reduction. Based on the observation that output token similarity is consistently preserved across adjacent timesteps, DiTo utilizes prior-step similarities as an effective proxy to establish token correspondences at a Matching timestep, which are then reused across multiple subsequent Reduction timesteps. To optimize this interleaved scheduling, we propose Pair Match Ratio (PMR)-guided Interval Scheduling to determine the optimal matching frequency. Furthermore, to mitigate localized approximation errors and resulting blocking artifacts caused by repeated reuse, we propose Frequency-aware Token Matching by incorporating a selection-frequency penalty. Extensive experiments demonstrate that DiTo consistently outperforms existing TR methods with 1.6-3.9 dB higher PSNR at comparable speedups, achieving a superior Pareto frontier.

Hangyeol Lee, Hyojeong Lee, Joo-Young Kim

KAIST, Daejeon, Republic of Korea

Diffusion TransformerDiffusion-Based Image GenerationVisual Token PruningDiffusion Model Inference AccelerationDiffusion Models
AskPDF
May 21, 2026

RiT: Vanilla Diffusion Transformers Suffice in Representation Space

Flow matching with xx-prediction -- regressing the clean data point rather than the ambient velocity -- is known to exploit low-dimensional manifold structure effectively in pixel space \cite{li2025back}. We ask whether a pretrained representation space, while containing a low-dimensional data manifold of comparable intrinsic dimensionality, offers a distribution more favorable for flow-matching learning. Comparing pixel, SD-VAE, and DINOv2 features along four geometric axes, we find that pixel and DINOv2 share nearly identical intrinsic dimensionalities (both d^ ⁣≈ ⁣33\hat{d}\!\approx\!33) yet DINOv2 exhibits 7.3×7.3\times higher effective rank, 35×35\times better covariance conditioning, 11.5×11.5\times lower excess kurtosis, and 1.7×1.7\times lower on-manifold interpolation error; SD-VAE latents are consistently intermediate, indicating that the advantage stems from representation-learning objectives rather than mere compression. These statistical properties render the flow-matching regression well-conditioned and remove the need for the specialized prediction heads or Riemannian transport used by prior DINOv2 diffusion methods. We propose the Representation Image Transformer (RiT): a vanilla Diffusion Transformer trained by xx-prediction on frozen DINOv2 features, augmented only by a dimension-aware noise schedule and joint [CLS]-patch modeling. On ImageNet 256×256256{\times}256, RiT attains FID 1.45 without guidance and 1.14 with classifier-free guidance, outperforming DiTDH^\text{DH}-XL with 19%19\% fewer parameters (676M vs.\ 839M). The resulting ODE is efficiently solvable at coarse discretizations: with classifier-free guidance, 55 Heun steps already reach FID 2.0 and 1010 steps reach 1.25, without distillation or consistency training. Code at https://github.com/lezhang7/RiT.

Le Zhang, Ning Mang, Aishwarya Agrawal

Mila – Québec AI Institute, UdeM · Utrecht University · Canada CIFAR AI Chair

Flow MatchingVisual Representation LearningDiffusion TransformerDiffusion-Based Image GenerationDiffusion Models
AskPDF
May 21, 2026

Noise Schedule Design for Diffusion Models: An Optimal Control Perspective

We develop a principled framework for analyzing and designing noise schedules in diffusion models. We show that one can recast this design problem as an optimal control problem, whose state is the Fisher information of the diffusion process, which evolves according to an ODE, and whose control input is the noise schedule. The objective of the optimal control problem is a functional involving the Fisher information, which is shown to be an upper bound on the Kullback-Leibler sampling error. By solving this optimal control problem, we obtain sufficient conditions on noise schedules under which O~(d/n)\tilde{\mathcal{O}}(d/n) sampling error is achievable, where dd is the data dimension and nn is the number of discretization steps. While existing theoretical works also prove that O~(d/n)\tilde{\mathcal{O}}(d/n) sampling error bounds are achievable, these results hold for specific noise schedules that differ from those used in practice. Under additional assumptions on the data distribution, we derive a family of noise schedules that recovers the functional forms of schedules widely used in practice. We show that these schedules achieve the O~(d/n)\tilde{\mathcal{O}}(d/n) error bound when their parameters are chosen according to our analysis. Our theory offers a principled approach to noise schedule design, yielding explicit functional forms and guidance for selecting their parameters.

Seo Taek Kong, Weina Wang, R. Srikant

ECE & CSL, University of Illinois Urbana-Champaign · Computer Science Department, Carnegie Mellon University · ECE, CSL & NCSA, University of Illinois Urbana-Champaign

Diffusion SamplingDiffusion-Based Image GenerationDiffusion Timestep SchedulingOptimal ControlDiffusion Models
AskPDF
May 20, 2026

Hierarchical Variational Policies for Reward-Guided Diffusion

Adapting pretrained diffusion models to downstream objectives such as inverse problems often requires expensive test-time guidance or optimization. We propose a principled framework for generating high-quality reward-aligned samples at substantially reduced inference cost. Our approach formulates test-time adaptation as a hierarchical variational model, where control is amortized into a lightweight yet expressive stochastic policy. This formulation naturally supports few-step diffusion sampling: large step sizes enable fast inference, while the learned policy maintains sample quality by providing structured per-step control. The resulting fully amortized sampler achieves a strong quality--speed tradeoff, matching or exceeding recent test-time scaling baselines while requiring significantly less compute. For example, on 4x super-resolution, our method achieves better perceptual quality with more than 5x faster inference compared to the best-performing baseline. We further extend our approach to a semi-amortized regime that combines cheap amortized proposals with limited test-time optimization, achieving state-of-the-art perceptual quality across several challenging inverse problems.

Kushagra Pandey, Farrin Marouf Sofian, Jan Niklas Groeneveld +2

Department of Computer Science University of California Irvine

Few-Step Diffusion SamplingDiffusion Model GuidanceDiffusion ModelsDiffusion Models for Image RestorationDiffusion-Based Inverse ProblemsDiffusion Model Acceleration+1
AskPDF
May 20, 2026

One-Step Distillation of Discrete Diffusion Image Generators via Fixed-Point Iteration

Discrete diffusion models excel at visual synthesis but rely on slow, iterative decoding. Existing single-step distillation methods attempt to bypass this bottleneck, either by training auxiliary score networks that effectively double compute, or by introducing specialized parameterizations and multi-stage pipelines that fragment optimization. In this paper, we introduce Fixed-Point Distillation (FPD), an end-to-end framework that constructs local correction targets by partially corrupting the student's one-step draft and refining it with a single teacher step. To compute the training objective in a semantically meaningful space, we lift discrete tokens into continuous features and apply a multi-bandwidth drift loss that iteratively accumulates these corrections. To backpropagate through the discrete bottleneck, we employ a straight-through estimator that feeds exact hard-sampled tokens to the teacher and decoder during the forward pass, ensuring that training and inference operate on the same codebook manifold, while routing continuous gradients back to the student logits. This fully differentiable pathway additionally accommodates an optional unconditional adversarial objective to enhance perceptual realism. Evaluations on both class- and text-conditional generation validate the effectiveness of our framework. FPD achieves competitive visual fidelity and structural alignment within a single inference step, narrowing the gap to multi-step teachers while outperforming existing discrete distillation baselines.

Chaoyang Wang, Yunhai Tong

Peking University

Diffusion Model DistillationFew-Step Diffusion SamplingDiffusion-Based Image GenerationDiffusion Models
AskPDF
May 20, 2026

Memorisation, convergence and generalisation in generative models

Generative neural networks learn how to produce highly realistic images from a large, but finite number of examples - or do they simply memorise their training set? To settle this question, Kadkhodaie, Guth, Simoncelli and Mallat (ICLR '24) trained diffusion models independently on disjoint subsets of a dataset and showed that they converge to nearly the same density when the number of training images is large enough. This result raises two basic questions: how much data do you need for convergence, and what does convergence capture about learning the data distribution? Here, we address these questions by providing an exact analytical characterisation of the transition from memorisation to generalisation in linear generative models. We find that these models memorise at small load, while convergence emerges continuously when the number of samples is linear in the input dimension. Strikingly, we find that convergence is insensitive to recovery of the principal latent factors of the data, which are recovered in a sharp transition. After extending our approach to data with power-law spectra, we find the same distinction between convergence and latent recovery in our experiments with convolutional denoisers and in the data of Kadkhodaie et al. We thus show that generalisation in generative models decomposes into at least two distinct objectives: matching the bulk of the data distribution and recovering the principal latent factors. These objectives correspond to two different distances between true and learnt data distribution, and only the first one is captured by convergence.

Antoine Maillard, Sebastian Goldt

1INRIA Paris & DI ENS, PSL University, Paris, France · 2International School of Advanced Studies (SISSA), Trieste, Italy

Neural Network MemorizationMemorization in Generative ModelsGenerative ModelingDiffusion Models
AskPDF
May 20, 2026

Semantic Granularity Navigation in Image Editing

Despite the generative capabilities of diffusion and flow models, real-image editing remains constrained by a persistent trade-off between semantic editability and structural fidelity. We trace a primary cause of this limitation to the implicit coupling of edit progress with model scale in existing paradigms. Under this coupling, stronger edits typically require visiting noisier states, which spends computation on destabilizing layout before the semantic change is well localized. We introduce NaviEdit, a training-free inference-time controller that decouples edit progress from model scale traversal through a strict self-consistency contract. NaviEdit operates at the rollout level and leaves the underlying pretrained model unchanged. It treats scale as a control input and reallocates a fixed step budget toward semantically responsive intermediate scales instead of destructive high-noise regimes. Experiments show positive average gains across compatible editors and flow backbones, supporting decoupling as a portable inference-time control principle.

Liangsi Lu, Minzhe Guo, Xuhang Chen +1

Guangdong University of Technology, Guangzhou, China · Huizhou University, Huizhou, China.

Diffusion Model SamplingDiffusion Timestep SchedulingDiffusion-Based Image EditingDiffusion Models
AskPDF
May 20, 2026

DrawMotion: Generating 3D Human Motions by Freehand Drawing

Text-to-motion generation, which translates textual descriptions into human motions, faces the challenge that users often struggle to precisely convey their intended motions through text alone. To address this issue, this paper introduces DrawMotion, an efficient diffusion-based framework designed for multi-condition scenarios. DrawMotion generates motions based on both a conventional text condition and a novel hand-drawing condition, which provide semantic and spatial control over the generated motions, respectively. Specifically, we tackle the fine-grained motion generation task from three perspectives: 1) freehand drawing condition. To accurately capture users' intended motions without requiring tedious textual input, we develop an algorithm to automatically generate hand-drawn stickman sketches across different dataset formats; 2) multi-condition fusion. We propose a Multi-Condition Module (MCM) that is integrated into the diffusion process, enabling the model to exploit all possible condition combinations while reducing computational complexity compared to conventional approaches; and 3) training-free guidance. Notably, the MCM in DrawMotion ensures that its intermediate features lie in a continuous space, allowing classifier-guidance gradients to update the features and thereby aligning the generated motions with user intentions while preserving fidelity. Quantitative experiments and user studies demonstrate that the freehand drawing approach reduces user time by approximately 46.7% when generating motions aligned with their imagination. The code, demos, and relevant data are publicly available at https://github.com/InvertedForest/DrawMotion.

Tao Wang, Lei Jin, Zhihua Wu +7

Beijing University of Posts and Telecommunications · University of Science and Technology of China · 3NLP Lab, School of Computer Science and Engineering, Northeastern University, Shenyang, China +4

Text-to-Motion GenerationHuman Motion GenerationDiffusion Models for Motion GenerationDiffusion Models
AskPDF
May 20, 2026

AirCast-SR: A Foundation Model for Kilometer-Scale Atmospheric Super-Resolution via Latent Consistency Diffusion

Operational weather prediction at kilometer scales remains computationally prohibitive for traditional numerical weather prediction (NWP) models, limiting forecast access for applications in energy, agriculture, and disaster management that require fine-grained spatiotemporal detail. Here we introduce AirCast-SR, a foundation model for atmospheric super-resolution that downscales global AI weather forecasts from 0.25 degree (~28 km) to 1 km horizontal resolution at hourly temporal resolution, producing 67-hour forecasts of eight coupled surface variables simultaneously. EarthMind-SR employs a three-dimensional U-Net conditioned within a Latent Consistency Model (LCM) diffusion framework, trained on patch-based samples over the contiguous United States (CONUS) using GraphCast forecasts as input and NOAA's Analysis of Record for Calibration (AORC) as the target. The model achieves near-zero bias across all variables and lead times, and its radial power spectral density analysis demonstrates preservation of fine-scale atmospheric structure at wavelengths of 10 km to 100 km where coarser models lose spectral power. We validate EarthMind-SR across three CONUS case studies spanning winter, summer, and spring seasons, and demonstrate zero-shot global transferability over India and Germany using independent surface station observations without any retraining or fine-tuning. As an open-weights foundation model, EarthMind-SR establishes a new paradigm for kilometer-scale AI weather prediction and provides a platform for regional fine-tuning, distillation, and downstream applications in climate services and hazard forecasting.

Somnath Luitel, Manmeet Singh, Joshua Durkee +11

Department of Earth, Environmental, and Atmospheric Sciences, Western Kentucky University, Bowling Green, KY, USA · NASA Goddard Space Flight Center, Greenbelt, MD, USA · The University of Texas at Austin, Austin, TX, USA +3

Geospatial Foundation ModelsSpatiotemporal ForecastingDiffusion ModelsWeather Forecasting
AskPDF
May 20, 2026

Learning to Think in Physics: Breaking Shortcut Learning in Scientific Diffusion via Representation Alignment

Physics-informed diffusion models typically enforce PDE constraints only on final outputs, leaving intermediate representations unconstrained and prone to shortcut learning under shifted boundary conditions. We introduce REPA-P, a teacher-free, architecture-agnostic framework that aligns intermediate features with physical states using first-principles residuals. REPA-P attaches lightweight 1×11{\times}1 projection heads to selected layers, decodes hidden activations into physical quantities, and applies PDE residual losses during training. These heads are discarded at inference, introducing zero overhead. Across four PDE tasks, including Darcy flow, topology optimization, electrostatic potential, and turbulent channel flow, REPA-P accelerates convergence by up to 2×2{\times}, reduces physics residuals by up to 66.4%66.4\%, and improves out-of-distribution robustness by up to 49.3%49.3\%, with consistent gains on both U-Net and Diffusion Transformer backbones. Ablations show that supervising a small set of intermediate layers captures most benefits and complements output-level physics losses. Code is available at https://github.com/Hxxxz0/REPA-P.

Haozhe Jia, Pengyu Yin, Wenshuo Chen +6

The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China · Shandong University, Jinan, China · Data61/CSIRO, Australia +5

Physics-Informed Diffusion ModelsPhysics-Informed Generative ModelingPDE Surrogate ModelingPhysics-Informed MLShortcut LearningDiffusion Models
AskPDF
May 19, 2026

Tippett-minimum Fusion of Representation-space Diffusion Models for Multi-Encoder Out-of-Distribution Detection

We address out-of-distribution (OOD) detection across the full spectrum of distribution shifts -- global domain changes, semantic divergence, texture differences, and covariate corruptions -- through a multi-encoder fusion of per-encoder representation-space diffusion models (RDMs). We statistically identify each encoder's sensitivity to specific shift types from ID data alone and introduce EncMin2L -- an encoder-agnostic two-level min⁡(⋅)\min(\cdot)-gate that combines and calibrates per-encoder diffusion-based likelihood detectors without OOD labels, outperforming monolithic multi-encoder baselines at 2.3×2.3\times lower parameter cost. Two ID-data diagnostics: η2η^2 (class-conditional F-test) and ΔμΔμ (log-likelihood shift under synthetic corruptions) -- quantify encoder specialization, while a Tippett minimum pp-value combination aggregates per-encoder scores into a single, calibration-stable OOD signal. EncMin2L achieves ≥0.94\geq 0.94 AUROC across all four shift types simultaneously, outperforming the state-of-the-art representation-space diffusion OOD detectors across overlapping benchmarks.

Neelkamal Bhuyan

Georgia Institute of Technology

Diffusion-Based OOD DetectionDistribution Shift DetectionOOD DetectionDiffusion Models
AskPDF
May 19, 2026

Improving Ensemble CAPE Forecasts with a Diffusion Model Incorporating Aerosol Information

Convective available potential energy (CAPE) is an important variable for forecasting severe weather and understanding deep convection and precipitation. The latest versions of the Global Forecast System (GFS) and related Global Ensemble Forecast System (GEFS) have exhibited a bias towards underestimating CAPE values during the summertime. We train an artificial intelligence (AI) diffusion model to improve the skill and uncertainty quantification of afternoon 6-hour lead time ensemble forecasts over the United States. Our model takes a GFS CAPE forecast as input and outputs an ensemble that significantly outperforms both GFS and GEFS 6-hour forecasts on root mean square error, continuous ranked probability score, and Brier score. We propose a two-stage training pipeline to leverage both a larger historical GFS forecast dataset and a smaller historical GEFS dataset, despite the two using initialization and parameterization schemes that vary over time. We also show that classifier-free guidance can be used to control the skill and spread of the forecasts. We then demonstrate the versatility of our framework by adding aerosol optical depths (AODs) of black carbon, organic carbon, dust, sea salt, and sulfates as additional input features. Aerosols can invigorate or suppress convection depending on atmospheric conditions. Our AI models effectively incorporate aerosols to produce improved CAPE forecasts. We interpret the model components by using permutation feature importance to rank the influence of the different AODs and find that black carbon, organic carbon, and sulfate aerosols have a greater impact on the model's CAPE predictions than sea salt and dust aerosols.

Zachary James, Joseph Guinness, Arthur DeGaetano

Cornell University, Department of Statistics and Data Science · Washington University in St. Louis, Department of Statistics and Data Science · Cornell University, Department of Earth and Atmospheric Sciences

Probabilistic ForecastingDiffusion ModelsWeather Forecasting
AskPDF
May 19, 2026

Landscape-Awareness for Geometric View Diffusion Model

Accurate camera viewpoint estimation under sparse-view conditions remains challenging, particularly in two-view scenarios. Recent approaches leverage diffusion models such as Zero123 to synthesize novel views conditioned on relative viewpoint, showing promising results when repurposed for viewpoint estimation via optimization with MSE loss. However, existing methods often suffer from nonconvex loss landscape with numerous local minima, making them sensitive to initialization and reliant on naive multistart strategies. We analyze these optimization challenges and visualize failure cases, showing that geometric ambiguities, such as symmetry and self-similarity, can mislead gradient-based updates toward incorrect viewpoints. To address these limitations, we propose a score-based method that reshapes the optimization landscape to guide updates toward the ground-truth viewpoint, followed by a refinement stage using a viewpoint-conditioned diffusion model. Experiments show that our method improves convergence, reduces reliance on brute-force sampling, and achieves competitive accuracy with higher sample-efficiency.

Yan-Ting Chen, Hao-Wei Chen, Tsu-Ching Hsiao +1

Elsa Lab, National Taiwan University

Diffusion Model GuidanceCamera Pose EstimationDiffusion Models
AskPDF
May 19, 2026

When Preference Labels Fall Short: Aligning Diffusion Models from Real Data

Preference alignment aims to guide generative models by learning from comparisons between preferred and non-preferred samples. In practice, most existing approaches rely on preference pairs constructed from model-generated images. Such supervision is inherently relative and can be ambiguous when both samples exhibit artifacts or limited visual quality, making it difficult to infer what constitutes a truly desirable output. In this work, we investigate whether real data can serve as an alternative source of supervision for preference alignment. We adopt a data-centric perspective and study a curation strategy that treats real images as reference points and constructs preference signals by contrasting them with generated or perturbed samples, without requiring manually annotated preference pairs. Through empirical analysis, we show that real-data-based supervision provides effective guidance for aligning diffusion models and achieves performance comparable to existing preference-based methods. Our results suggest that real data offers a practical and complementary source of supervision for preference alignment and highlight directions of label-efficient alignment strategies. Code and models are available at https://cwyxx.github.io/RealAlign.

Weiyan Chen, Weijian Deng, Yao Xiao +5

School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China · Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, China · Australian National University, Canberra, Australia +2

Diffusion Model AlignmentPreference AlignmentDiffusion Models
AskPDF
May 19, 2026

Latent Laplace Diffusion for Irregular Multivariate Time Series

Irregular multivariate time series impose a trade-off for long-horizon forecasting: discrete methods can distort temporal structure via re-gridding, while continuous-time models often require sequential solvers prone to drift. To bridge this gap, we present Latent Laplace Diffusion (LLapDiff), a generative framework that models the target as a low-dimensional latent trajectory, enabling horizon-wide generation without step-by-step integration over physical time. We guide the reverse process utilizing a stable modal parameterization motivated by stochastic port-Hamiltonian dynamics, and parameterize its mean evolution in the Laplace domain via learnable complex-conjugate poles, enabling direct evaluation over irregular timestamps. We also link continuous dynamics to irregular observations through renewal-averaging analysis, which maps sampling gaps to effective event-domain poles and motivates a gap-aware history summarizer. Extensive experiments show that LLapDiff improves over baselines in long-horizon forecasting, and its continuous-time generative nature supports missing-value imputation by querying the same model at historical timestamps. Code is available at https://github.com/pixelhero98/LLapDiffusion.

Zinuo You, Jin Zheng, John Cartlidge

School of Computer Science, University of Bristol · School of Engineering Mathematics and Technology, University of Bristol

Irregular Time-Series ModelingMultivariate Time Series ForecastingLong-Term Time Series ForecastingDiffusion Models
AskPDF
May 19, 2026

Probabilistic Multivariate Time Series Forecasting with Diffusion Copulas

Accurately assessing financial risk requires capturing both individual asset volatility and the complex, asymmetric dependence structures that emerge during extreme market events. While modern diffusion-based models have advanced multivariate forecasting, they often suffer from a "normality bias" when trained end-to-end, sacrificing marginal calibration for joint coherence and consistently underestimating tail risk. To address this, we propose a Diffusion-Copula framework that explicitly decouples the learning of marginal distributions from their dependence structure. We employ deep Mixture Density Networks to capture heavy-tailed asset dynamics, followed by a Classification-Diffusion Copula to model the joint dependence. Applied to cryptocurrency markets, our approach demonstrates superior performance over state-of-the-art baselines in forecasting systemic extremes of both marginal and joint events. Crucially, we demonstrate that while baseline models classify simultaneous market crashes as statistically impossible "Black Swans" (high surprise), our framework identifies them as "Expected Crashes" (low surprise), successfully preserving the correlation structure necessary for robust risk management during contagion events.

David Huk, Dongshan Wang, Miha Bresar

Department of Statistics The University of Warwick · School of Data Science The Chinese University of Hong Kong, Shenzhen

Multivariate Time Series ForecastingCopula ModelsQuantitative FinanceProbabilistic ForecastingProbabilistic Time Series ForecastingDiffusion Models
AskPDF
May 19, 2026

EpiDiffVO: Geometry-Aware Epipolar Diffusion for Robust Visual Odometry

Estimating relative pose from image pairs fundamentally requires only a minimal subset of geometrically consistent correspondences. However, most learning-based approaches rely on dense matching or direct regression, leading to redundancy and reduced geometric interpretability. In this work, we propose a sparse epipolar matching framework that predicts a compact set of correspondences optimized for geometric consistency across varying temporal baselines. To address residual noise and misalignment, we introduce an epipolar diffusion process that models correspondence uncertainty and refines keypoints toward epipolar consistency. The refined correspondences, along with depth cues, are lifted into a graph representation forming a Steiner graph that encodes relational structure between points. A graph neural network learns a compact subset of informative correspondences, which are passed to a differentiable singular value decomposition solver for end-to-end geometric estimation. Relative pose is recovered from the resulting essential matrix and evaluated in a visual odometry setting on the TartanAir and KITTI SLAM datasets. Experimental results demonstrate that combining sparse matching, diffusion-based refinement, and graph-based subset selection reduces correspondence redundancy while maintaining robust pose estimation across challenging baselines.

Prateeth Rao

International Institute of Information Technology Bangalore, Bengaluru, India

Multi-View GeometryRelative Pose EstimationCamera Pose EstimationVisual OdometryDiffusion Models
AskPDF
May 19, 2026

Boosting Text-to-Image Diffusion Models via Core Token Attention-Based Seed Selection

Text-to-image diffusion models can synthesize high-quality images, yet the outcome is notoriously sensitive to the random seed: different initial seeds often yield large variations in image quality and prompt-image alignment. We revisit this "seed effect" and show that attention dynamics over prompt core tokens, the content-bearing words, measured during the first few denoising steps, strongly predict final generation quality. Building on this observation, we introduce Attention-Based Seed Selection (ABSS), a training-free, plug-and-play method that ranks seeds for a given prompt by leveraging cross-attention to core tokens during the denoising process. ABSS requires no finetuning and does not alter the initial noise; it scores and ranks all candidate seeds, keeps only the top-k for full generation, and discards the rest, without relying on a fixed accept/reject threshold. Operating purely at inference time, ABSS can serve as a lightweight pre-selection add-on for existing seed-optimization pipelines, enabling additional gains. Across three benchmarks, extensive experiments show that ABSS enables consistent improvements in text-image alignment and visual quality for Stable Diffusion variants, as corroborated by human preference and alignment metrics.

Yunzhe Zhang, Hongfu Liu, Pengyu Hong

Brandeis University

Diffusion Model AlignmentT2I GenerationT2I Diffusion ModelsDiffusion Models
AskPDF
May 19, 2026

Drifting Objectives for Refining Discrete Diffusion Language Models

Discrete diffusion language models (DDLMs) generate text by iteratively denoising categorical token sequences, while recent drifting methods for continuous generators suggest that part of this sampling-time correction can instead be absorbed into training through an anti-symmetric fixed-point objective. We study how to transfer this principle to DDLMs, where the main challenge is the interface with discrete text: hard token samples are non-differentiable, and categorical predictions do not directly provide continuous samples to drift. We formulate TokenDrift, a drifting objective that lifts categorical predictions to soft-token features, applies anti-symmetric drifting in a frozen semantic space, and backpropagates the resulting stop-gradient feature target to DDLM logits. In controlled continual-training experiments with masked and uniform-state diffusion backbones, TokenDrift improves fixed-NFE generation quality over matched continuation baselines, reducing Gen.-PPL at 4 NFEs by 89% on MDLM and 86% on DUO. These results suggest that drifting can provide a practical refinement objective for DDLMs.

Daisuke Oba, Hiroki Furuta, Naoaki Okazaki

Institute of Science Tokyo · AIST · NII LLMC

Diffusion Model Fine-TuningDiffusion Language ModelsDrifting ModelsDiffusion ModelsDiffusion Language Model Training
AskPDF
May 19, 2026

Tweedie's Formulae and Diffusion Generative Models Beyond Gaussian

Diffusion models have achieved remarkable success in generating samples from unknown data distributions. Most popular stochastic differential equation-based diffusion models perturb the target distribution by adding Gaussian noise, transforming it into a simple prior, and then use denoising score matching, a consequence of Tweedie's formula, to learn the score function and generate clean samples from noise. However, non-Gaussian diffusion models with state-dependent diffusion coefficient have been largely underexplored, as have the corresponding Tweedie's formulae. In this work, we extend Tweedie's formula to important non-Gaussian processes, including geometric Brownian motion (GBM), squared Bessel (BESQ) processes, and Cox-Ingersoll-Ross (CIR) processes, thereby yielding the corresponding denoising score-matching objectives. We then apply the derived formulae to image and financial time series generation using GBM- and CIR-based diffusion models, and to empirical Bayes estimation under the BESQ setting. The reported experimental results demonstrate the potential of non-Gaussian models.

Wenpin Tang, Nizar Touzi, Zikun Zhang +1

Stochastic Differential EquationsTime Series GenerationScore-Based Generative ModelingDiffusion Models
AskPDF
May 19, 2026

Multi-Scale Generative Modeling with Heat Dissipation Flow Matching

Diffusion models are widely used in image generation, with most relying on noise-based corruption and denoising. A distinct branch instead uses blur as the main corruption, preserving better color budgets and multi-scale detail by providing multi-scale priors. However, blur-based models remain in SDE-based frameworks and are not integrated into ODE-based frameworks, such as Flow Matching (FM). Meanwhile, in the blur-based formulation, the classical inverse heat-dissipation (IHD) process faces an ill-posed challenge. Moreover, under the data-manifold assumption, regressing blurred images from high-dimensional noise (or velocity) space is also difficult. We propose Heat Dissipation Flow Matching (HDFM), which introduces a continuous blurred (heat-dissipation) process into FM to inject multi-scale priors. HDFM aligns an interpolated heat-dissipation path to address ill-posedness and adopts xx-prediction to mitigate high-dimensional regression difficulty. Toy experiments and ablation studies show that HDFM consistently benefits from both blur and xx-prediction. The performance of HDFM outperforms most baseline methods on all datasets.

Jun Ma, Hanquan Zhang, Yanjun Qin +2

Department of Systems Science, Faculty of Arts and Sciences, Beijing Normal University, Zhuhai 519087, China · School of Computer Science and Technology, Xinjiang University, Urumqi 830049, China · International Academic Center of Complex Systems, Beijing Normal University, Zhuhai 519087, China +1

Flow MatchingImage GenerationGenerative ModelingFlow-Based Generative ModelingDiffusion Models
AskPDF
May 19, 2026

Inference-Time Scaling in Diffusion Models through Iterative Partial Refinement

Inference-time scaling has emerged as a major approach for improving reasoning capabilities, and has been increasingly applied to diffusion models. However, existing inference-time scaling methods for diffusion models typically rely on external verifiers or reward models to rank and select samples, limiting their scalability to settings where such evaluators are available and reliable. Moreover, while recent diffusion models perform sequential inference with region-wise, mixed-noise conditioning, inference-time scaling tailored to this setting remains relatively underexplored. We propose Iterative Partial Refinement (IPR), an inference-time scaling method for sequential diffusion that requires no external verifier. Starting from an already-generated sample, IPR re-noises a subset of regions and regenerates them conditioned on the remaining regions, enabling the model to revise earlier decisions under a richer context than was available during the initial generation. This iterative partial refinement produces more globally consistent samples without external verification. On reasoning tasks requiring global constraint satisfaction, IPR consistently improves performance: on MNIST Sudoku, the valid solution rate increases from 55.8% to 75.0%. These results show that iterative partial refinement alone can serve as an effective inference-time scaling strategy for diffusion models in sequential, mixed-noise settings. Code is available at: https://github.com/ahn-ml/IPR

Taegu Kang, Jaesik Yoon, Sungjin Ahn

KAIST · KAIST & SAP · KAIST & NYU

Inference-Time ScalingDiffusion Models
AskPDF
May 19, 2026

PolycubeNet: A Dual-latent Diffusion Model for Polycube-Based Hexahedral Mesh Generation

Hexahedral meshes are widely used in simulation pipelines, yet automatic generation remains challenging for complex CAD geometries. Polycube-based hexahedral meshing is a representative approach due to its regular, parameterization-friendly structure, but existing polycube construction methods often rely on intricate surface segmentation and local heuristics, which can produce artifacts or fail on difficult shapes. In this paper, we propose an end-to-end framework for polycube generation based on conditional diffusion models. Given an input geometry represented as a point cloud, our method directly produces a corresponding polycube point cloud, eliminating the need for explicit surface segmentation or predefined polycube templates. At the core of our approach is a dual-latent conditional diffusion architecture that confines computationally expensive self-attention operations to a fixed-capacity, low-dimensional latent space. This design effectively decouples computational complexity from the resolution of both the input geometry and the output polycube, thereby avoiding the quadratic cost typical of point cloud self-attention mechanisms while supporting flexible input and output resolutions. To obtain a hexahedral mesh, the generated polycube is aligned to the input shape via rigid and non-rigid point cloud registration to establish surface correspondence, followed by a polycube-to-hex pipeline. We additionally create and release a paired dataset of CAD meshes and their corresponding polycube meshes, together with the core implementation of our model. Experiments show that PolycubeNet generalizes to complex CAD models with arbitrary genus and produces high-quality polycube structures within seconds, improving robustness and efficiency over prior learning-based approaches.

Lu He, Qitao Deng, Junjiang Deng +5

Dalian University of Technology · Jiangxi University of Science and Technology · School of Mathematical Sciences, Guangxi Minzu University +2

3D Mesh GenerationDiffusion Models
AskPDF
May 19, 2026

What Makes Synthetic Data Effective in Image Segmentation

Driven by rapid advances in large-scale generative models, synthetic data has emerged as a promising solution for visual understanding. While modern diffusion models achieve remarkable photorealistic image synthesis, their potential in complex visual segmentation tasks remains underexplored. In this work, we conduct a systematic analysis of synthetic images from state-of-the-art diffusion models to uncover the factors governing their utility. In particular, synthetic images characterized by dense scene composition and fine instance fidelity demonstrate distinctive benefits, yielding significantly more discriminative spatial representations. Building on these insights, we propose SENSE, a unified framework that leverages flexible and scalable synthetic data to substantially enhance segmentation performance. Notably, SENSE is model-agnostic, compatible with diverse architectures (e.g., DPT and Mask2Former), and scales effectively across models with varying parameter capacities. Extensive experiments on Cityscapes, COCO, and ADE20K validate the effectiveness and generalization capability of our approach. Code is available at https://github.com/zhang0jhon/SENSE.

Jinjin Zhang, Xiefan Guo, Yizhou Jin +2

State Key Laboratory of Complex and Critical Software Environment, Beihang University, Beijing 100191, China · School of Computer Science and Engineering, Beihang University, Beijing 100191, China

Synthetic Data AugmentationImage SegmentationSemantic SegmentationSynthetic Data EvaluationDiffusion Models
AskPDF
May 18, 2026

Reducing Diffusion Model Memorization with Higher Order Langevin Dynamics

Diffusion/score-based models have emerged as powerful generative models, capable of generating high-quality samples that mimic the training data distribution. However, it has been observed that they are prone to reproducing training samples-known as "memorization"-potentially violating copyright and privacy. In this paper, we study the effect of Higher-Order Langevin Dynamics (HOLD) on this phenomenon. HOLD diffusion processes introduce auxiliary variables; if the data variable is interpreted as "position," then the auxiliary variables can be interpreted as "velocity" and "acceleration," depending on the chosen order of the model. They were originally proposed based on the intuition that they regularize the trajectories of the data variable by implicitly imposing additional dynamical constraints. Our work provides, to our knowledge, the first theoretical characterization of the regularization effect of HOLD. Specifically, we show that in HOLD, the dynamics of the data variable are governed by a low-pass-filtered version of the learned score function, with smoothness increasing with the order of HOLD. We then analyze the optimal empirical score and the possibility of distribution collapse. Together, our results explain the mitigation of memorization as the model order increases. Finally, we present an empirical study on real-world data that supports our theory and highlights this distinct advantage of HOLD over standard diffusion in practice.

Benjamin Sterling, Mónica F. Bugallo, Tom Tirer

Department of Applied Math & Statistics Stony Brook University Stony Brook, NY 11790 · Department of Electrical and Computer Engineering Stony Brook University Stony Brook, NY 11790 · Faculty of Engineering Bar-Ilan University Ramat-Gan 5290002, Israel

Langevin DynamicsMemorization in Generative ModelsScore-Based Generative ModelingDiffusion Models
AskPDF
May 18, 2026

Generation of Heterogeneous PET Images from Uniform Organ Activity Maps Using a Pretrained Domain-Adapted Diffusion Model

Synthetic PET images are valuable for quantitative imaging workflow development, scalable virtual imaging trials, and deep learning model training, but conventional physics-based simulation approaches are computationally intensive, limited in anatomical variability, and often fail to capture heterogeneous PET uptake. This study developed a pretrained domain-adapted diffusion (PAD) model for anatomy-conditioned PET synthesis from uniform organ activity maps. PAD adopts a natural-image pretrained text-to-image decoder with an upstream conditioning encoder and a downstream PET-domain adapter. A two-phase training strategy was used, with the first phase learning coarse uptake distributions and the second refining local image details. Uniform organ activity maps were generated from CT-based segmentations by assigning each organ its mean uptake from the paired PET image. Evaluation included quantitative accuracy, noise assessment, radiomic analysis, tumor segmentation performance, and a human observer study. PAD-generated images achieved high quantitative accuracy, with concordance correlation coefficients above 0.92 between organ mean SUVs and assigned activity values. The synthesized images showed noise levels and texture characteristics similar to target PET images and produced comparable tumor segmentation performance. In a two-alternative forced-choice observer study, four readers achieved approximately 50% accuracy, indicating visual indistinguishability between synthesized and target images. PAD also generated realistic PET images from XCAT-derived activity maps, demonstrating compatibility with phantom-based anatomical priors. Overall, PAD provides a diffusion-based framework for generating clinically relevant heterogeneous PET images from uniform organ activity maps derived from clinical segmentations or digital phantoms, supporting data augmentation and downstream imaging studies.

Suya Li, Kaushik Dutta, Debojyoti Pal +2

Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, USA · Imaging Science Program, McKelvey School of Engineering, Washington University in St Louis, St. Louis, USA · Department of Surgery, Washington University School of Medicine, St. Louis, USA +1

Diffusion Model AdaptationPET/CT ImagingDiffusion-Based Image GenerationMedical Image SynthesisDiffusion Models
AskPDF
May 18, 2026

SURGE: Approximation and Training Free Particle Filter for Diffusion Surrogate

Data assimilation (DA) addresses the problem of sequentially estimating the state of a dynamical system from noisy and incomplete observations. In this work, we employ a diffusion model as a world model to simulate and predict the system's dynamics. Recently, score-based diffusion models have learned global diffusion priors that effectively model (stochastic) dynamics, revealing strong potential for data assimilation. In this paper, we investigate how information from noisy observations can be incorporated to enable continuous correction and refinement of the predicted system state when using a diffusion prior. Motivated by particle filtering methods, we represent the posterior distribution using a set of particles. After receiving noisy observations, the diffusion model is guided using the observation likelihood to steer the generation process toward observation-consistent states. Nevertheless, such guidance does not guarantee sampling from the true posterior. We therefore employ a Sequential Monte Carlo approach over the diffusion trajectory, viewed as a path measure, to reweight and resample particles, thereby correcting the generation process and ensuring convergence toward the desired posterior distribution. This leads to an unbiased particle filtering method that rigorously fuses observational data with diffusion model simulations.

Lifu Wei, Yinuo Ren, Naichen Shi +1

Department of Mechanical Engineering, Northwestern University, Evanston, IL, United States · Institute for Computational & Mathematical Engineering, Stanford University, Stanford, CA, United States · Department of Industrial Engineering & Management Sciences, Northwestern University, Evanston, IL, United States

Particle FilteringBayesian FilteringData AssimilationDiffusion Models
AskPDF
May 18, 2026

Spectral Progressive Diffusion for Efficient Image and Video Generation

Diffusion models have been shown to implicitly generate visual content autoregressively in the frequency domain, where low-frequency components are generated earlier in the denoising process while high-frequency details emerge only in later timesteps. This structure offers a natural opportunity for efficient generation, as high-resolution computation on noise-dominated frequencies is largely redundant. We propose Spectral Progressive Diffusion, a general framework that progressively grows resolution along the denoising trajectory of pretrained diffusion models. To this end, we develop a spectral noise expansion mechanism and derive an optimal resolution schedule from the model's power spectrum. Our framework supports training-free acceleration and a novel fine-tuning recipe that further improves efficiency and quality. We demonstrate significant speedups on state-of-the-art pretrained image and video generation models while preserving visual quality.

Howard Xiao, Brian Chao, Lior Yariv +1

Stanford University

Video Diffusion ModelsDiffusion SamplingDiffusion-Based Image GenerationDiffusion Model Inference AccelerationDiffusion Models
AskPDF
May 18, 2026

SafeDiffusion-R1: Online Reward Steering for Safe Diffusion Post-Training

Diffusion models have been widely studied for removing unsafe content learned during pre-training. Existing methods require expensive supervised data, either unsafe-text paired with safe-image groundtruth or negative/positive image pairs, making them impractical to scale. Furthermore, offline reinforcement learning and supervised fine-tuning approaches that generate synthetic data offline suffer from catastrophic forgetting, degrading generation quality. We propose a novel online reinforcement learning framework that addresses both data scarcity and model degradation through post-training with Group Relative Policy Optimization (GRPO) on both negative and positive text prompts. To eliminate the need for fine-tuning specialized safe/unsafe reward models, we introduce a steering reward mechanism that exploits an inherent property of CLIP embeddings: steering text representations toward positive safety directions and away from negative ones in the embedding space. Our online-policy approach enables the model to learn from diverse prompts, including explicit unsafe content, without catastrophic forgetting. Extensive experiments demonstrate that our method reduces inappropriate content to 18.07% (vs. 48.9% for SD v1.4) and nudity detections to 15 (vs. 646 baseline) while improving compositional generation quality from 42.08% to 47.83% on GenEval. Remarkably, these safety gains generalize to out-of-domain unsafe prompts across seven harm categories, achieving state-of-the-art performance without supervised paired data or reward tuning. Github: https://github.com/MAXNORM8650/SafeDiffusion-R1.

Komal Kumar, Ankan Deria, Abhishek Basu +3

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), UAE · Michigan State University (MSU), USA

Diffusion Model AlignmentDiffusion Model Fine-TuningRL for Diffusion ModelsT2I Diffusion ModelsDiffusion Models
AskPDF
May 18, 2026

Continuous Diffusion Scales Competitively with Discrete Diffusion for Language

While diffusion has drawn considerable recent attention from the language modeling community, continuous diffusion has appeared less scalable than discrete approaches. To challenge this belief we revisit Plaid, a likelihood-based continuous diffusion language model (DLM), and construct RePlaid by aligning the architecture of Plaid with modern discrete DLMs. In this unified setting, we establish the first scaling law for continuous DLMs that rivals discrete DLMs: RePlaid exhibits a compute gap of only 20×20\times compared to autoregressive models, outperforms Duo while using fewer parameters, and outperforms MDLM in the over-trained regime. We benchmark RePlaid against recent continuous DLMs: on OpenWebText, RePlaid achieves a new state-of-the-art PPL bound of 22.122.1 among continuous DLMs and superior generation quality. These results suggest that continuous diffusion, when trained via likelihood, is a highly competitive and scalable alternative to discrete DLMs. Moreover, we offer theoretical insights to understand the advantage of likelihood-based training. We show that optimizing the noise schedule to minimize the ELBO's variance naturally yields linear cross-entropy (information loss) over time. This evenly distributes denoising difficulty without any case-specific time reparameterization. In addition, we find that optimizing embeddings via likelihood creates structured geometries and drives the most significant likelihood gain.

Zhihan Yang, Wei Guo, Shuibai Zhang +5

NVIDIA & Cornell · NVIDIA & Georgia Tech · UW-Madison +3

Language Model Scaling LawsDiffusion Language ModelsContinuous Diffusion Language ModelsDiffusion ModelsDiffusion Language Model Training
AskPDF