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

1,278 papers90 papers in the last four weeks, down 44% on the four weeks before. 0.7% of all new papers.

Nov 12, 2026

Diffusion Tree Sampling for Inference Time Adaptation of Crystal Generative Models

Diffusion-based generative models have become strong priors over plausible crystal structures, but turning that prior into promising candidates requires steering toward several conflicting properties at once. Fine-tuning is expensive and produces a new model for every new objective, while differentiable guidance is not applicable to many real-world tasks. We instead adapt these generative models at inference time using Diffusion Tree Sampling (DTS), a Monte Carlo tree search procedure over the denoising trajectory. We make two technical contributions on top of DTS that are particularly relevant for AI-for-science applications. First, we show how to use existing property-conditional generative models (e.g., models fine-tuned for classifier-free guidance) as guided proposals inside DTS, without distorting the target distribution. Second, we extend DTS to multi-objective sampling, so that a single search produces samples spanning an entire Pareto front. We instantiate both adaptations on de novo crystal generation: improving the stability of crystals generated by different models, using pretrained conditional models to generate crystals belonging to rare space groups or with high bulk modulus, and generating crystals along a Pareto front of high dielectric constant and high bandgap materials. Across these tasks, DTS-based inference-time adaptation successfully improves model outputs without modifying the models themselves.
Nov 12, 2026

DRAMA: Dissecting Attention Redundancy for Accelerating Multimodal Diffusion Large Language Models

Multimodal diffusion large language models (dLLMs) generate text through iterative masked denoising with bidirectional attention, offering a compelling alternative to autoregressive vision-language modeling. However, each denoising step requires a full forward pass over long multimodal sequences in which visual tokens dominate the sequence length, and a fixed step budget further compounds this quadratic per-step cost. Existing inference acceleration approaches for multimodal dLLMs characterize attention redundancy implicitly, motivating designs from isolated attention visualizations rather than from measurement. We close this gap with DRAMA, a training-free acceleration framework grounded in a systematic dissection of attention redundancy in multimodal dLLMs. By probing attention tensors along spatial and temporal axes, we reveal two persistent redundancy structures: (i) attention mass consistently concentrates on a sparse subset of visual tokens, and (ii) attention distributions rapidly stabilize across denoising steps after an initial formation phase. These findings motivate two complementary mechanisms: Spatial-Redundancy-Aware Visual Token Pruning (SVTP) and Sufficiency-Guided Adaptive Stopping (SGAS). Together, DRAMA achieves an average 18.67 times inference speedup across six vision-language benchmarks without training, architectural modification, or attention caching, while maintaining competitive accuracy. Our results establish a strong, measurement-grounded acceleration baseline for multimodal dLLM inference.
Nov 12, 2026

Generative AI models for exploring diffusion dynamics from multi-frame particle tracking

Particle-based volumetric flow measurement techniques have proven their efficacy in quantitatively assessing the dynamic characteristics of unsteady flows within the realm of fluid mechanics. Despite recent advances in multi-frame particle tracking, an open challenge in this area is maintaining trajectories in temporally under-sampled 3D sequences, where inter-frame displacements are large. In this work, we develop DiffTR-Flow3D, a diffusion-based model that estimates flow motion from spatiotemporal three-dimensional particle recordings via a stochastic diffusion process. DiffTR-Flow3D conditions the diffusion process with both global and local particle dynamics and enhances flow estimation by leveraging historical particle trajectories. Extensive experiments, including numerical simulations and real-world measurements, demonstrate that DiffTR-Flow3D achieves state-of-the-art displacement estimation accuracy compared with learning-based flow estimators, and, when integrated into particle tracking pipelines, substantially improves their long-term tracking performance, while exhibiting robustness and generalization across diverse flow patterns. This work is expected to provide fluid researchers with an effective tool for the consistent and precise analysis of particle trajectories in temporally under-sampled 3D sequences with large inter-frame displacements.
Nov 11, 2026

Token Commitment as a Control Surface for Steering Masked Diffusion Generation

Masked diffusion models repeatedly predict provisional tokens before a sampler decides which predictions become permanent. Existing inference-time steering methods primarily modify activations, logits, or denoising trajectories, while treating this commitment policy as a fixed decoding detail. We instead study token commitment as a diffusion-specific control surface and separate it into two decisions: which tokens to commit and how many tokens to commit at each step. For the first, we introduce counterfactual target-aware commitment, which ranks unresolved positions by a mixture of model confidence and leave-one-token-out target contribution. For the second, we introduce calibrated adaptive budgets driven by target progress and confidence. Across language and vision-language settings in the LLaDA family, multiple targets, two activation-steering methods, and an unsteered control, both mechanisms yield empirically monotonic target-quality trade-offs in every configuration we evaluate. Adaptive schedules also recover competitive operating points relative to scorer-free stochastic schedules matched in average denoising cost. Because both mechanisms require only a scalar score evaluated on a provisional completion, they apply unchanged when the target is an automatic proxy for task success rather than an output attribute: NLI entailment for dialogue summarization and CLIPScore for captioning, neither of which corresponds to a single reusable steering direction. We reproduce the behavior of both mechanisms on a second masked-diffusion family, Dream-7B, for one control target and one task-success target. These results characterize commitment--which provisional predictions become irreversible, and how rapidly--as an interpretable control surface for steering masked diffusion generation under the conditions we test.
Nov 11, 2026

Iris-Z: Scaling One Generative Model Past Specialist Geometry Models

Language moved from many task-specific models to one unified model: a single generative model with a single generative training task that handles all of them within the same backbone, and keeps improving with data, parameters and compute. We believe the bitter lesson applies just as much to computer vision, 3D and spatial understanding, in the end to visual intelligence: the effort should go into general architectures that solve all these tasks in a unified way, and into scaling them, rather than into task-specific designs. We test this on monocular geometry, a task that has belonged to specialised architectures for years. Iris-Z is a general image generator that produces appearance and geometry together. A single backbone denoises, reasons about and understands the image and its depth at the same time, and one set of weights generates images with their depth, estimates depth from photographs and paints images onto a given depth. At enough scale it surpasses state-of-the-art specialists on established depth benchmarks while keeping the generative abilities of the backbone it started from without specialized layers. We release two models for the community.
Nov 10, 2026

RB-LDC: Redundancy-Balanced Latent Coding for Noise-Robust Diffusion Latents

Latent diffusion models decode approximately correct latents, making decoder sensitivity to perturbations relevant to image quality. After whitening, forward diffusion corrupts every direction at the same signal-to-noise ratio (SNR), despite unequal decoder sensitivities. We study an invertible linear layer between the encoder and diffusion process that shifts each direction's log-SNR. Its linear minimum mean-square-error (LMMSE) recovery risk has an exact spectral expression for any whitened latent distribution. Under an energy budget and a floor ensuring invertibility, minimizing decoder-sensitivity-weighted LMMSE risk integrated over the diffusion schedule yields a unique optimal allocation. The layer aligns with the decoder's expected pullback metric and equalizes marginal risk reduction across directions, computed by a deterministic spectral solver. At large budgets, the allocation follows a power law in sensitivity whose exponent is determined by the low-SNR tail of the loss weighting: one half for noise prediction and two thirds for data prediction under the cosine schedule, with a crossover on truncated schedules. We instantiate this principle as RB-LDC using Jacobian-based sensitivity estimates. On ImageNet at 256 x 256 resolution with SD-VAE and VA-VAE, trained RB-LDC improves reconstruction from noisy latents over budget-matched isotropic coding in all four metrics and five coding configurations, averaged over 15 SNRs and three pipeline seeds. It reduces average perturbed FID by 17.1 for VA-VAE block-local coding while preserving clean reconstruction. We also report intermediate SNRs where perturbed FID worsens. With a frozen diffusion model, RB-LDC yields lower FID point estimates in all 12 tested combinations of sampling progress and channel SNR under corruption during sampling.
Nov 9, 2026

DUO: Dual Readouts of Diffusion Models for OOD Detection

Diffusion models are powerful generative methods for learning and sampling from complex data distributions. Safely deploying these models requires detecting inputs that fall outside their training distribution. Because diffusion models recover clean samples from noisy inputs, reconstruction error is often used for out-of-distribution (OOD) detection. This approach assumes that in-distribution (ID) inputs are reconstructed more accurately than OOD inputs. However, OOD inputs may be reconstructed equally well or even better, making this score unreliable. We therefore propose the leading sensitivity spectrum of intermediate denoising features as a complementary OOD signal and demonstrate that it remains informative when reconstruction error is misleading. We introduce DUal-readout of diffusion models for OOD detection (DUO), a fully unsupervised, training-free OOD detector that combines reconstruction error and feature sensitivity. DUO improves average AUROC by approximately 6% points over the strongest baselines while substantially reducing inference cost, and achieves SOTA on near-OOD and high-resolution benchmarks.
Nov 9, 2026

Stylization as Geodesic Navigation on Diffusion Manifolds

Facial stylization asks for a delicate transformation: the image should look unmistakably artistic, yet the person should remain unmistakably the same. Existing training-free and instruction-driven methods often fail to expose such control, either drifting away from identity or producing only weak stylization. We introduce Stylus, a training-free framework that treats facial style transfer as navigation on the curved latent manifold of a pretrained diffusion model. Rather than interpolating in flat Euclidean space, \method refines the interpolated content--style point by lowering a ratio-weighted discrete geodesic energy under the pullback metric, made tractable by a rank-one power-iteration approximation of the Jacobian. This principle drives two complementary controls: content--style latent interpolation for smooth semantic fusion, and self-attention query optimization for spatially adaptive style injection. The resulting dual-control mechanism provides continuous, fine-grained control over stylization strength while preserving identity-critical facial structure, all without additional training. Experiments show that Stylus achieves the strongest identity and content preservation at competitive stylization strength.
Nov 7, 2026

MGE: A Membership Guided Data Extraction Attack against Diffusion Models

Recently, diffusion models have achieved remarkable performance in image generation. However, their inherent memory properties introduce a potential safety risk of training data leakage, i.e., data extraction attack. To illustrate such risk, we propose a membership guided extraction (MGE) framework to effectively extract training data from a pre-trained diffusion model. Our core idea is to formulate the data extraction as a conditional generation, where the condition signal is from a membership inference classifier. To improve the success rate and efficiency of MGE, we propose adversarial gradient and parallel branch decoding. Our experiments on image generation benchmarks demonstrate that our MGE can extract hundreds of images, which shows that more attention should be paid to privacy-preserving of diffusion model.
Nov 7, 2026

Split and Bridge: Multimodal Generation via Diffusion Bridging

Visual data are often expressed through multiple complementary modalities (e.g., images and segmentation masks), each capturing different aspects of the same underlying structure. Most existing multimodal methods adopt a top-down paradigm, learning a shared latent variable through joint training from scratch, which limits scalability. We propose Split and Bridge (SnB), a bottom-up framework for multimodal generation that couples pretrained unimodal diffusion models at sampling time. Rather than learning a unified latent space, SnB generates coherent multimodal samples by following a shared diffusion trajectory up to a split point and then branching into modality-specific trajectories guided by a bridge module. This design enables plug-and-play reuse of powerful pretrained models without joint training, making it scalable and easily adaptable. Across standard benchmarks (PolyMNIST, CelebAMask-HQ) and more complex settings (PIE-Bench++, ROCO), SnB achieves a strong balance between sample quality and coherence. We show that SnB naturally extends to downstream zero-shot tasks such as interleaved co-generation, image editing, and style coupling/decoupling. Our results demonstrate that bottom-up generation provides a practical framework that complements existing top-down multimodal approaches. Code is available at https://github.com/AhmadArrabi/SnB
Nov 7, 2026

On the universal structure in deterministic generative diffusion

Diffusion-based generative modeling has emerged as a general computational framework for simulating high-dimensional, structured data across various domains. Understanding its generative mechanisms is paramount for advancing model training and inference in a principled way. Here, we demonstrate that deterministic sampling trajectories share a universal geometric structure invariant to data modalities and diffusion model implementations. This structure can be largely explained by an exactly solvable model based on the optimal linear denoiser.
Nov 6, 2026

Semiprime Factorization with Discrete Diffusion and Beam Search: Scaling Laws

We demonstrate that transformer architectures trained with discrete denoising diffusion and decoded via guided test-time beam search can learn to recover the prime factors of semiprime integers directly from data. Our approach establishes a new frontier for learning-based semiprime factorization, successfully factoring 68-bit integers-a substantial advance over the prior 56-bit state of the art. While computationally less efficient than classical factoring algorithms, our model requires no domain-specific number-theoretic priors and learns purely from training examples. In practice, the diffusion model acts as a generative prior, identifying promising factor candidates where exact verification or lattice-based post-processing can complete the factorization. Furthermore, we derive empirical scaling laws relating approximation quality to training data volume, model size, and test-time computation. In contrast to language, problem size is the key driver of scaling behavior. Achieving a fixed accuracy on a held-out test set requires compute that scales exponentially with bit length. Consequently, our findings pose no threat to security, as the projected compute cost required to factorize cryptographically relevant integers remains prohibitive.
Nov 6, 2026

Generative Active Learning for Label-Efficient Pan-Organ Ultrasound Diagnosis

Medical ultrasound (US) enables real-time, non-invasive diagnosis, yet building US-based AI systems is limited by data scarcity, costly expert annotation, and limited cross-center generalizability. Privacy constraints and instance-level reporting hinder the collection of diverse abnormal cases and frame-level labels, while repeated retraining for new centers increases computational cost. We introduce EvoDiff, a generative active learning framework that integrates diffusion modelling with information-theoretic objectives, evolutionary search, and curriculum-based sampling to select informative samples across heterogeneous US datasets. EvoDiff formulates sample selection as a generate-to-retrieve process, in which diffusion-based optimization in feature space guided by information gain produces synthetic probes, and only the real unlabeled samples retrieved by these probes are annotated and used for downstream training. Across six benchmarks covering fifteen datasets and 16,696 images, we evaluate EvoDiff on convolutional network and transformer-based foundation model. EvoDiff approaches fully supervised performance with substantially fewer labels, outperforms active learning baselines, and improves foundation model generalization under cross-domain external validation on unseen cohorts. A reader study with six radiologists indicates the potential to support clinical annotation workflows, with a model-based estimated 24% to 84% reduction in annotation workload at matched downstream performance.
Nov 6, 2026

Reconciling European Electricity Data with Diffusion Priors Learned from Observations Alone

The ENTSO-E Transparency Platform publishes generation, load, and cross-border power flows for European bidding zones. The values carry undocumented errors and violate energy conservation, and data reconciliation consists in correcting them to satisfy that balance. In this work, we formulate data reconciliation as Bayesian inference. Classical methods implicitly take a uniform prior, which we replace with a diffusion model estimated from the observations alone by empirical-Bayes expectation-maximization. The resulting posterior is sampled with a guided process whose projection step makes every sample satisfy the balance exactly. On synthetic systems of six to thirty-five bidding zones, built from ENTSO-E resource-adequacy data, the diffusion prior gives the most accurate reconciliation, increasingly so as the system grows. Its credible intervals are calibrated or conservative, and it remains the most accurate under a misspecified noise covariance. Finally, when the generation of entire zones goes unreported, the classical methods fail while the learned prior recovers it from cross-zone structure.
Nov 6, 2026

FANO: Anchored Forward-Noise Optimization for Diffusion Inverse Problems

Pre-trained diffusion models have become a widely adopted zero-shot prior for image inverse problems, among which annealed decoupling methods achieve leading reconstruction performance on nonlinear tasks. We find, however, that annealed methods that directly optimize the denoiser input xtx_t exhibit radial drift in later layers: within-layer optimization drives the noise-component norm of xtx_t away from its typical value, pushing the denoiser out of its high-performance regime and degrading reconstruction quality. To address this, we propose FANO (Anchored Forward-Noise Optimization). At each annealing layer, FANO fixes the previous denoiser estimate as an anchor and optimizes only the forward noise ee, whose distribution is known, instead of xtx_t, thereby preserving the forward-diffusion structure of the denoiser input. The standard Gaussian prior on ee concentrates its norm near the typical value, and deviations degrade denoiser performance. We accordingly decompose ee in polar coordinates into a direction and a radial magnitude, and apply a soft potential to regulate the radius, keeping xtx_t in the high-density region of the forward distribution. On multiple linear and nonlinear inverse problems, FANO matches the strongest current baselines on linear tasks and achieves consistent improvements on nonlinear tasks.
Nov 5, 2026

Scale as a Modality: Multi-Scale Pixel-Space Diffusion Transformers

Latent diffusion models rely on pretrained autoencoders that introduce lossy compression and constrain spatial granularity. Pixel-space diffusion avoids this bottleneck and enables flexible tokenization at multiple spatial scales, yet existing approaches largely adopt two-level coarse-to-fine architectures with limited cross-scale interaction. We introduce MS-pDiT, a multi-scale pixel-space diffusion transformer that treats spatial scale as a modality. MS-pDiT tokenizes the same noisy image at multiple patch sizes, processes each scale in a dedicated transformer stream, and enables bidirectional information exchange through compacted MM-DiT-style joint attention, with attention cost independent of the finest patch size. This design establishes the number of scales as a new architectural scaling dimension. On ImageNet 256×256256\times256, increasing the number of scales from two to five consistently improves FID from 4.39 to 2.31, 2.26, and 2.19, while adding an intermediate scale substantially outperforms increasing depth under comparable parameter budgets. Further analysis highlights the importance of bidirectional cross-scale interaction and late introduction of fine-grained streams. MS-pDiT achieves FID scores of 1.43 and 1.44 on ImageNet 256×256256\times256 and 512×512512\times512, respectively, and naturally extends to text-to-image generation, achieving 0.86 on GenEval and 83.9 on DPG-Bench.
Nov 5, 2026

eSharp: Diffusion-based Sharp Monocular Metric Event-to-Depth Estimation

Depth perception is fundamental for autonomous navigation, industrial inspection, and mixed-reality systems. It often operates under strict constraints on power and form factor. In such cases, monocular depth estimation (MDE) is often used. Even as recent advances continue to improve monocular depth quality, conventional frame-based MDE systems remain challenged by fast motion and high-dynamic-range scenes. This work introduces eSharp, an MDE framework that addresses these problems using an event camera. eSharp repurposes pretrained diffusion priors for event-only MDE and produces accurate metric depth with unparalleled sharpness on real data. Real-world event datasets provide only sparse LiDAR depth maps for ground-truth supervision, making dense prediction challenging. We address this limitation through cross-modal distillation, using LiDAR-calibrated dense pseudo-labels generated from a spatially co-registered frame camera to train eSharp. Experiments show that eSharp outperforms prior event-to-depth baselines in both metric-depth accuracy and spatial fidelity, establishing diffusion models as a promising foundation for event-driven 3D perception.
Nov 5, 2026

Flux as a Codec: Zero-Shot Lossy Compression with Flux

We propose FluxC, a zero-shot codec that uses Flux as a compression prior. Although DiffC has shown that pretrained diffusion models can serve as zero-shot codecs, we ask how the choice of coding path and reconstruction method affects rate--distortion--perception behavior. We analyze FluxC along three axes: coding cost and bit allocation, reconstruction method, and low rate distortion. At matched signal-to-noise ratio (SNR), we show that the path choice does not affect the ideal information rate, but changes how bits are allocated across time steps. For reconstruction, we compare ancestral sampling, one-step ODE reconstruction, and many-step ODE reconstruction. Our analysis shows that ODE-based reconstruction can achieve lower distortion than ancestral sampling, especially in the low rate regime. Experiments on Gaussian sources and natural images support this prediction. Moreover, FluxC performs competitively with, and in some cases outperforms, recent neural image compression methods, suggesting that Flux can serve as a principled zero-shot codec.
Nov 5, 2026

Self-Supervised Diffusion Models for Handwritten Signature Verification via Ruin Reconstruction

Handwritten signature verification (HSV) is a critical biometric technique in legal and forensic authentication, whose efficacy hinges on the discriminability of signature representations. Current HSV methods predominantly adopt supervised learning to enhance representation quality, leaving self-supervised learning largely unexplored. We propose SigDiffuser, a diffusion-based self-supervised framework for signature representation and verification. At its core, SigDiffuser introduces a novel pretext task termed Reconstruction from Ruins (RFR), which deliberately ruins'' signatures through severe crumpling, obliterating their structural integrity and writing styles. A visual encoder is employed to capture crucial handwriting patterns from intact signature regions, guiding the denoiser to reconstruct the ruined'' portions and learning meaningful representations. Additionally, we introduce a pyramidal feature aggregation module (PFAM) to bridge the visual encoder and denoiser, enhancing representation quality via hierarchical aggregation and modulation. Extensive experiments on multilingual signature datasets demonstrate that SigDiffuser significantly outperforms both state-of-the-art SSL methods and HSV models. We further show that increasing the crumpling severity on signatures effectively improves representation quality. Notably, our work demonstrates that extreme destruction, coupled with self-supervised diffusion models, can yield superior signature representation compared to traditional mild deformations, paving new avenues for HSV and other biometric domains. Code will be publicly available.
Nov 5, 2026

FisherDistill: Diffusion Distribution Distillation via Closed-Form Fisher Critics

Recent advances in diffusion distillation have enabled high-quality image generation in only one or a few network function evaluations (NFEs). Despite their efficiency, aggressively distilled generators often retain a residual distribution gap from their multi-step teachers, motivating recent methods to introduce adversarial refinement with a learned discriminator. In this work, we ask whether such refinement necessarily requires a learned adversarial discriminator. We introduce FisherDistill, a simple alternative based on Fisher linear discriminant analysis. Using intermediate representations from the frozen teacher, we estimate the first- and second-order statistics of real and generated samples and derive a closed-form Fisher direction that maximally separates their feature distributions relative to within-class variation. This direction provides a lightweight refinement signal that complements the original distillation objective without requiring an additional adversarial discriminator or its optimization loop. Across multiple distillation frameworks, our approach consistently improves generation quality over their respective distilled baselines and substantially closes the gap to adversarially refined counterparts. Code: https://bhosalems.github.io/FisherDistill/
Nov 5, 2026

OptIMerge: Optimization-Based Merging of Diffusion Models

The arithmetic merging of weights from multiple task-specific models fine-tuned from a shared base model, commonly known as model souping, has emerged as a popular strategy for multi-task learning without the computational overhead of joint retraining. However, naive scalar weight averaging frequently leads to severe performance degradation when the constituent models learn structurally divergent feature representations, leading to destructive parameter interference. We study this problem in a largely unexplored setting: merging image-to-image (I2I) diffusion models fine-tuned for dense, pixel-accurate tasks such as image restoration. Unlike the large language models (LLMs) and text-to-image (T2I) diffusion models for which most merging methods were developed, I2I experts encode conflicting spatial priors (e.g., anisotropic motion-blur vs. isotropic defocus-blur) in their convolutional filters, where both naive soup and LLM-style element-wise merging introduce visible artifacts. We introduce OptIMerge, a gradient-based optimization framework: instead of relying on heuristic scalars, OptIMerge learns highly expressive, yet parameter-efficient merger coefficients by minimizing a multi-task diffusion loss over a small set of 100-500 images. OptIMerge parameterizes the merge with architecture-aware transformations: for attention layers, a structured low-rank feature alignment with explicit channel biases; for convolutional layers, per-kernel merger coefficients, and a low-rank Cayley transform that aligns channels by orthogonal rotation without distorting the filters. By keeping the experts frozen and optimizing only these lightweight merger parameters (equivalent to <0.4% of the network footprint), OptIMerge achieves simultaneous multi-task capabilities that rival computationally expensive joint training and decisively outperform zero-shot model soups.
Nov 4, 2026

Dynamical State-Space Diffusion for Spatiotemporal Prediction

Diffusion models for temporal prediction typically corrupt each future frame with independent noise, leaving temporal dependence entirely to the denoiser. We propose Dynamical State-Space Diffusion (DSSD), which learns the temporal structure of the corruption process itself. DSSD builds a temporal kernel from the damped oscillatory modes of a stable diagonal state-space model, whose learned decay rates and frequencies let the corruption capture multiple timescales and oscillatory correlations, while the denoiser architecture remains unchanged. Running this model on a finite clip, however, requires the state before the first frame, which is unobserved; starting from a zero state leaves the early frames with less noise than the diffusion schedule prescribes. We avoid this by using the Cholesky factor of the model's stationary covariance as the temporal operator. We prove that this operator is causal, exactly invertible, keeps each frame's noise variance equal to that of standard diffusion, and reduces to the identity at the clean end of the process, and that it is exactly the Kalman innovations form of the state-space model. The denoiser is trained to predict the temporally correlated noise, and sampling removes and reapplies the temporal mixing at each step. On scientific-field and robot-video datasets, DSSD outperforms diffusion baselines that share its denoiser, and the evolution of its learned modes during training shows that the temporal operator adapts to each dataset.
Nov 4, 2026

DyPSI: Dynamic Physics Sensing via Joint Field and Sensor-Trajectory Generation

Physics sensing in evolving environments requires reconstructing dense spatiotemporal fields from sparse observations while adapting where future observations should be collected. Existing reconstruction and sensor-placement methods have made substantial progress, but they typically rely on static sensing assumptions: they either use a single layout across all frames or optimize placements independently at each frame, which limits their ability to produce temporally coherent sensing trajectories. We introduce DyPSI, a dynamic physics sensing framework that formulates this problem as the joint generation of future fields and sensor trajectories conditioned on historical observations. DyPSI lifts discrete sensor coordinates into continuous Sensor Position Fields, encodes fields and sensing configurations with a shared Functional Tucker representation, and trains a joint diffusion model to generate future fields and sensing trajectories simultaneously in the resulting latent space. A Gaussian-process temporal kernel correlates the perturbation injected during diffusion training, biasing the denoiser toward temporally smooth recoveries, and we further provide an analysis linking the EDM training objective to a sequence-aggregated A-optimality criterion and a kernel-induced bound on discrete trajectory variation. Experiments on turbulent flow, GLORYS12 sea-surface temperature, and 3D car aerodynamics show that DyPSI consistently outperforms static and frame-wise placement strategies, with substantial error reduction under tight sensor budgets. The code is available at https://github.com/mengqi132/DyPSI.
Nov 3, 2026

From Natural Images to Medical Scans: Transferability of Diffusion Models for Medical Image Generation

Transfer learning from natural images is standard in supervised and increasingly adopted in self-supervised medical image analysis, but its role in generative modelling, and in particular diffusion denoisers, remains underexplored. Medical diffusion models typically train the denoiser from scratch, with no systematic evaluation of natural-image transfer. We analyse this across latent and pixel paradigms, model sizes (131M-952M) and two radiology datasets (RadImageNet, BraTS), and ask which aspects of transfer support generation. We find that: i) Natural-image initialisation consistently improves generation, by 21-34% relative gFID in pixel space against 14% in latent space, where the tokeniser supplies much of the prior. gFID gains mirror downstream pathology classification and radiomic-feature FD, both proxies for clinical fidelity. ii) Benefits reflect spatial structure, not semantics: semantic probing does not predict generation quality, whereas a spatial-structure metric adapted to internal denoiser representations tracks gFID at ∣r∣=0.85|r|=0.85, positioning spatial organisation as a channel through which pretraining manifests. iii) Mixed-domain denoiser training shows that natural and medical images share similar layer-wise representations up to the late blocks, and mixing natural images into medical training improves fidelity. iv) Transfer benefit is tied to natural image initialisation, outperforming in-domain medical pretraining and several reinitialisation schemes. Overall, natural image denoisers transfer positively to medical ones in our setup. Code is available at https://github.com/CAIROLab-Bern/rad-dit
Nov 3, 2026

SurvAD: How Anomalies Survive Diffusion Reconstruction

Diffusion-based anomaly detectors reconstruct a test image towards normality, yet how much of an anomaly survives, its leak, remains uncharacterized. We show that a first-order deterministic sampler obeys an exact recursion along the anomaly, whose continuous limit is a survival law. The anomaly decays exponentially as the noise falls, at a rate set by the retention rr, the share the model keeps on the states it visits. Because rr depends on the noise level and not on the step count, one measured path predicts the leak of other step counts, samplers and schedules from a nearby start to within a few hundredths, on four diffusion models and real defects. The law also shows that the leak's growth over steps is exposure bias and that normal details obey the same law, so changing the path or the training trades leak for detail error. We introduce SurvAD, which fine-tunes the reconstructor on inference-like starts and fits a comparator that adds evidence to the residual map without suppressing its ranking elsewhere. For under 2%2\% more inference time, SurvAD raises the multi-class pixel average precision (P-AP) of DeCo-Diff from 76.876.8 to 80.780.7 on MVTec AD and from 57.557.5 to 60.060.0 on VisA. It matches CostFilter-AD on the same reconstructor with 9×9\times fewer trainable parameters and, under Dinomaly's protocol, exceeds its reported P-AP on Dinomaly on both benchmarks.
Nov 3, 2026

Surrogate Models for Stochastic Al-Cu Phase-Field Solidification

Phase-field models resolve microstructural evolution during rapid alloy solidification, but their computational cost limits long trajectories, stochastic ensembles, and studies across process conditions. We compare a deterministic U-Net with conditional flow matching (Flow), conditional diffusion (EDM), and a variance-preserving denoising diffusion bridge (VP-DDBM) for two-dimensional Al-Cu solidification. Given the current phase, concentration, and prescribed temperature fields, each model predicts the subsequent state and is applied autoregressively. At two thermal gradients outside the training range, we compare surrogate rollouts spanning 48,00048{,}000 phase-field Euler steps with five phase-field futures generated from a common restart at each gradient. With two network evaluations per transition, Flow achieves the lowest time-averaged fair continuous ranked probability score among the generative surrogates for mean phase, mean Cu concentration, interface length, and leading-tip position at both gradients. Flow also reproduces the cellular morphology and retains sharp solid--liquid interfaces. It generates a single future approximately 107×107\times faster than the phase-field solver on the reported hardware, showing potential for faster process-condition screening and ensemble-based microstructure studies.
Nov 2, 2026

The role of injected noise in diffusion model generalization

Diffusion models trained on distinct datasets can produce highly similar samples when prompted with the same noise input, suggesting that they learn a generalized data distribution. We investigate the role of noise injected during the sampling process, and show that it plays an important role in making or breaking this strong generalization effect. Based on this finding we investigate the robustness of sample similarity to changes in the injected noise, and find that its impact is largest in the sampling regime where broad structure already exists but the precise identity of the sample is not yet determined.
Nov 2, 2026

When Optimization Meets Human Intent: Human-Centered Diffusion Model Optimization

Diffusion models are evolving from general tools into personalized systems that adapt to user needs through fine-tuning and guidance. However, diffusion optimization mainly targets measurable objectives such as aesthetics, rarely examining whether they align with human intent. Human needs are multidimensional, contextual, and dynamic, whereas existing methods approximate them with fixed metrics. Because these metrics capture human needs only partially and may amplify deviations from true intent during optimization, higher scores need not yield results better matching human expectations. To understand how human needs enter and influence optimization, we systematically analyze how human intent is expressed, translated into optimizable objectives, applied to model behavior or generation, and evaluated through outputs. We further discuss trade-offs affecting human-intent alignment in personalized image generation and gaps between automated metrics and real user judgments. This human-centered synthesis reorganizes diffusion optimization and provides a systematic reference for personalized generation systems designed for human preferences.
Nov 2, 2026

Generative diffusion in long-range correlated landscapes: Speciation cascade and the foam structure of the score

While state of the art for image and video generation, the fundamental reasons for the effectiveness of generative diffusion remain unclear. In particular, the observed advantage of following an entire reverse process with respect to single-shot or autoregressive methods, and for what type of data it may hold, is not well understood. In this work, we aim to better understand the role of the reverse process by considering a complex yet analytically tractable data model, the Gibbs measure of a high-dimensional Gaussian random potential with long-range correlations. Focusing on the exact score, we show that the speciation threshold, in which a trajectory commits to one class of the data, becomes a speciation window, in which the reverse process progressively resolves the Parisi tree of pure states of the full replica-symmetry breaking measure, committing to ever finer branches down to a single pure state. Importantly, these commitments take place at noise levels where the noisy sample distinguishes the pure states only through an energy of order one, beyond the reach of standard replica saddle points. Constructing a Larkin argument suited to such O(1)O(1) energy gaps, we describe the score as a foam, with slab-like cells separated by shock-like boundaries. In the variance-exploding process we consider, this window is followed by a shrinkage of the samples towards the data support, and finally by reconstruction within the selected pure state. The denoising error of the exact score follows these stages, dropping during the speciation window with a profile set by the Parisi function alone, and then within the selected state. We believe that the generative foam we describe, and in particular the boundaries between cells, could play an important role in determining where the errors of trained diffusion models redistribute the mass of the generated samples when the data are long-range correlated.
Nov 2, 2026

RouteREPA: Routed Representation Alignment for Just Image Transformers

Representation alignment (REPA) is widely used to accelerate the training of latent-space diffusion transformers (DiTs). However, applying it to pixel-space models such as Just Image Transformers (JiT) is challenging. Our analysis reveals a pronounced representation conflict between discriminative alignment and pixel-space generation. Compared with its latent counterpart, JiT learns representations that are substantially less aligned with DINOv2. Applying REPA to JiT consistently reduces the reconstructability of its intermediate representations, whereas this effect is not observed in Latent JiT. We attribute this conflict to JiT directly reconstructing pixels, which requires its representations to hierarchically encode discriminative semantics and reconstruction-relevant visual information across network depths. To address these challenges, we propose Routed Representation Alignment (RouteREPA), which uses lightweight, representation-dependent routers to adaptively route task-oriented representations for alignment and generation. We further apply routing across multiple network depths to accommodate their depth-dependent representation requirements. Extensive experiments on ImageNet-1K at 256x256 resolution show that RouteREPA consistently improves JiT across model scales and training budgets, outperforming standard REPA. On JiT-H/16, RouteREPA further surpasses PixelREPA using only one-third of its training epochs without classifier-free guidance.