DIPHINE is the first diffusion-based neural estimator for the 16 ΦID atoms in continuous non-Gaussian dynamical systems, obtained by joint MI estimation followed by Möbius inversion.
super hub Canonical reference
Score-Based Generative Modeling through Stochastic Differential Equations
Canonical reference. 76% of citing Pith papers cite this work as background.
533
Pith papers citing it
Background
76% of classified citations
abstract
Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model.
hub tools
citation-role summary
background 72
method 17
baseline 3
other 1
citation-polarity summary
claims ledger
- abstract Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate
authors
co-cited works
representative citing papers
The paper establishes a reverse-time quantum diffusion framework that generates complex quantum ensembles from simple distributions by deriving and learning a feedback Hamiltonian from forward trajectory data.
WavTTS is the first raw-waveform diffusion TTS model using DiT flow matching and multi-scale mel supervision that approaches SOTA latent zero-shot performance while beating prior end-to-end models.
Flux Matching generalizes score-based generative modeling by using a weaker objective that admits infinitely many non-conservative vector fields with the data as stationary distribution, enabling new design choices beyond traditional score matching.
A-CODE presents a fully atomic one-stage multimodal diffusion model for protein co-design that claims superior unconditional generation performance over prior one- and two-stage models plus a tenfold success-rate gain on hard binder-design tasks.
FMRG reformulates guidance as deterministic optimal control, deriving a single-trajectory method using the flow map that matches or exceeds baselines on reward-guided generation and inverse problems with 3 NFEs at text-to-image scale.
Quotient-space diffusion models generate correct symmetric distributions by removing redundancy on the quotient space, simplifying learning and improving results on small molecules and proteins under SE(3) symmetry.
The García-Pintos feedback Hamiltonian equals the score function of the quantum trajectory distribution, linking quantum feedback to diffusion-model reversal.
Diffusion sampling from d-dimensional distributions requires at least ~sqrt(d) adaptive score queries when score estimates have polynomial accuracy.
OP-GRPO is the first off-policy GRPO method for flow-matching models that reuses trajectories via replay buffer and importance sampling corrections, matching on-policy performance with 34.2% of the training steps.
Generative diffusion and flow models are constructed to remain exactly on the Lorentz-invariant massless N-particle phase space manifold during sampling for particle physics applications.
ASTRA reframes transition-state search as guided diffusion inference that samples the isodensity surface between metastable basins and converges to first-order saddles via score differences and physical forces.
MF-PID turns independent diffusion samples into mean-field interacting agents, proving that quadratic interactions yield exact linear mean interpolation and delivering 19-24% energy savings in demand-response control.
Föllmer processes are variationally optimal among generative diffusions because they minimize the impact of drift estimation error on path-space KL divergence, rendering different interpolation schedules statistically equivalent.
Flow-GRPO is the first online RL method for flow matching models, raising GenEval accuracy from 63% to 95% and text-rendering accuracy from 59% to 92% with little reward hacking.
LLaDA is a scalable diffusion-based language model that matches autoregressive LLMs like LLaMA3 8B on tasks and surpasses GPT-4o on reversal poem completion.
DDIMs construct non-Markovian diffusion processes that share DDPM training objectives but allow much faster reverse sampling, demonstrated empirically at 10-50x wall-clock speedup.
Set diffusion factorizes likelihood over arbitrary token sets and uses a set-causal diffusion architecture to support KV caching and any-order decoding, yielding improved speed-quality tradeoffs versus prior diffusion LMs.
Proposes diffeomorphic optimization for manifold-constrained problems in generative models via flow maps, with Lie-group extensions for protein design showing metric improvements.
Flow-Map GRPO uses anchored stochastic flow map composition to enable GRPO-based RL alignment of deterministic few-step flow-map generators while preserving their marginal paths.
Introduces a Bridge latent interface that maps mismatched student latents into teacher space, enabling distillation from modern diffusion teachers to compact one-step students and raising SD 1.5 HPSv3 from 5.4 to 9.4 while keeping one-step speed.
FlexiSLM is the first spoken language model supporting dynamic and controllable frame rates on speech input and output, outperforming fixed-rate 7B models at high quality and enabling faster inference at lower rates like 6.25 Hz.
ReMatch corrects train-test residual distribution mismatch in probabilistic downscaling via optimal transport in low-dimensional PCA space, reducing under-dispersion and improving SSR and CRPS on HRRR-ERA5 wind data.
Hybrid simulation and non-Euclidean elasticity theory demonstrate that clathrin coats develop adaptive rigidity and memory during growth, producing flat, stalled, or closed outcomes through two energy-landscape gates and matching experiments without fitted parameters.
citing papers explorer
-
Wasserstein Mirror Gradient Flow as the limit of the Sinkhorn Algorithm
Sinkhorn iterations converge to a Wasserstein mirror gradient flow (the Sinkhorn flow) as regularization epsilon goes to zero with iterations scaled as 1/epsilon.
-
Human Motion Diffusion Model
MDM is a classifier-free diffusion model that generates expressive human motions by predicting clean samples rather than noise, supporting text and action conditioning and outperforming prior methods on standard benchmarks.
-
High-Resolution Image Synthesis with Latent Diffusion Models
Latent diffusion models achieve state-of-the-art inpainting and competitive results on unconditional generation, scene synthesis, and super-resolution by performing the diffusion process in the latent space of pretrained autoencoders with cross-attention conditioning, while cutting computational and
-
Diffusion Models Beat GANs on Image Synthesis
Diffusion models with architecture improvements and classifier guidance achieve superior FID scores to GANs on unconditional and conditional ImageNet image synthesis.
-
ExpoCM: Exposure-Aware One-Step Generative Single-Image HDR Reconstruction
ExpoCM enables fast one-step single-image HDR reconstruction via exposure-dependent perturbations and region-conditioned consistency trajectories derived from a probability flow ODE.
-
Cast3: Translating numerical weather prediction principles into data-driven forecasting
Cast3 translates NWP principles into a data-driven model using cubed-sphere grids, super-ensembles, and generative nudging to achieve state-of-the-art ensemble predictions that outperform baselines.
-
A PDE-Based Framework for Generative Modeling Beyond Classical Score-Based Diffusion
Nonlinear modification of Ornstein-Uhlenbeck dynamics produces condensation in a Fokker-Planck equation and yields a stabilized reverse PDE that reconstructs the initial distribution for generative modeling.
-
Optimizing Visual Generative Models via Distribution-wise Rewards
Distribution-wise rewards with subset-replace strategy and post-hoc merging improve FID-50K on SiT (8.30 to 5.77) and EDM2 (3.74 to 3.52) while preserving diversity.
-
Decision-Aware Training for Sample-Based Generative Models
Augments the energy score objective for sample-based generative models with a differentiable decision loss that is itself a proper scoring rule, yielding targeted improvements on cost-sensitive regions in synthetic and real tasks.
-
DriftScope: Measuring The Hidden Effects of Diffusion Model Adaptation
Adapting diffusion models causes hidden damage to unrelated concepts detectable via sparse autoencoders and zero-shot classification, and DriftScope provides a prompt-level token-drift diagnostic.
-
Histogram-constrained Image Generation
HIG enforces exact histogram constraints on diffusion-generated images by modeling the control task as an optimal transport problem and applying guidance transformations during sampling.
-
PhotoQuilt: Training-Free Arbitrary-Resolution Photomosaics via Bootstrapped Tiled Denoising
A training-free bootstrapped tiled denoising procedure enables arbitrary-resolution photomosaics in diffusion models by fixing global layout at low resolution then denoising independent tiles after latent upscaling and noise re-injection.
-
Few-Step Boltzmann Generators via Scalable Likelihood Flow Maps
SCALLOP replaces Hutchinson's trace estimator with a scalable, vectorized likelihood distillation objective for F2D2 flow maps, cutting training variance and time while improving performance on molecular Boltzmann generators and image data.
-
EMOSH: Expressive Motion and Shape Disentanglement for Human Animation
EMOSH proposes an Expressive Human Model with disentangled parameters, coarse-to-fine motion injection, and spatially-aligned conditioning to generate high-fidelity expressive human videos without driving-subject shape leakage.
-
Class-frequency Guided Noise Schedule for Diffusion Models
Proposes CFRG noise schedule for diffusion models that assigns larger noises to low-frequency classes to improve generation on imbalanced datasets.
-
Don't Settle at the Mode! Mitigating Diversity Collapse in Pretrained Flow Models via Feature Self-Guidance
Feature self-guidance disperses internal features of flow models during batch generation and applies manifold regularization to increase output diversity while preserving condition alignment.
-
Error-Conditioned Neural Solvers
Error-Conditioned Neural Solvers improve PDE prediction accuracy by using the residual field as network input for learned corrections, outperforming residual-minimization methods by up to 10x on turbulent flows and generalizing better under distribution shifts.
-
Generative Models on Analog Hardware with Dynamics
AIS framework with time-varying piecewise parameters and hidden states enables analog hardware to perform generative modeling, achieving FID 27.6 on MNIST and 80.8 on Fashion-MNIST with 23uJ per image using 4-bit sparse architectures.
-
SpatialFlow-GRPO: Where Spatial Credit Drives Image Editing
SpatialFlow-GRPO adds region-level reward feedback and spatial alignment to Flow-GRPO-style RL for image editing, reporting gains on GEdit-Bench, ImgEdit-Bench, and a new MultiEditBench.
-
Adversarial Diffusion Across Modalities: A Fusion Survey of Attacks, Defenses, and Evaluation for Text, Vision, and Vision-Language Models
A narrative survey that catalogs fifty papers on diffusion-based adversarial techniques across text, vision, and vision-language models, proposes a six-class taxonomy of diffusion roles plus a unified five-dimension evaluation framework, and releases a companion catalog.
-
Forget, Anticipate and Adapt: Test Time Training for Long Videos
FFN performs TTT on multi-hour videos by restricting updates to three frames and using a surprise metric for adaptive window sizing, plus a new EpicTours dataset.
-
In-context Region-based Drag: Drag Any Region to Any Shape
ICRDrag performs region-based drag editing in diffusion models by feeding source image, source mask, and target mask into an in-context framework with image-mask attention consistency and source-target attention correspondence regularizations, plus a new Paired Region Dataset.
-
Causal-rCM: A Unified Teacher-Forcing and Self-Forcing Open Recipe for Autoregressive Diffusion Distillation in Streaming Video Generation and Interactive World Models
Causal-rCM unifies teacher-forcing and self-forcing distillation for autoregressive video diffusion, delivering a 2-step model with VBench-T2V score 84.63 and enabling interactive world models on Cosmos 3 using only synthetic data.
-
FLUX3D: High-Fidelity 3D Gaussian Generation with Diffusion-Aligned Sparse Representation
FLUX3D introduces Diffusion-Aligned Structured Latents (DA-SLAT) and Sparse-structure Multimodal Diffusion Transformer (SMDiT) with MARoPE to address representation and alignment bottlenecks in sparse-voxel 3DGS generation.
-
Bridging the Manifold Gap: Riemannian Residual Line Search for One-Step Image Editing
Riemannian Residual Line Search improves one-step diffusion image editing by curvature-based residual path construction and CLIP-based candidate selection, reporting SOTA on 700-sample PIE-Bench++ across 10 edit types.
-
Posterior Refinement: Fast Language Generation via Any-Order Flow Maps
FMLM+ with Posterior Refinement bridges masked diffusion and flow map models to match discrete baseline quality in language generation using 32x fewer neural function evaluations via posterior scoring and refinement.
-
Catastrophic Compositional Generation: Why Vanilla Diffusion Models Fail to Extrapolate
Vanilla diffusion models fail at compositional generation because score estimation error dominates and cannot be fixed by inference-time techniques when targets are out-of-distribution.
-
Diffusion Models Adapt to Low-Dimensional Structure Under Flexible Coefficient Choices
For a broad class of coefficients, diffusion models achieve Õ(k/ε) iteration complexity for ε-accurate TV sampling under low-dimensional structure, independent of ambient dimension.
-
Online forecast reconciliation using linear models
A framework for online forecast reconciliation is developed via multivariate linear models on graph hierarchies, ridge regression, and recursive least squares, with a demonstration on district heating load data.
-
Scalable Maximum Entropy Reinforcement Learning for Diffusion Policies via Adjoint Matching
Presents adjoint matching for scalable max-ent RL training of diffusion policies, enabling simulation-free optimization.
-
Lighting-Consistent Object Transfer Across Radiance Fields
Diffusion-based per-view harmonization for lighting-consistent object transfer between 3DGS scenes, using heterogeneous training data and final 3D consolidation.
-
ZeroGVC: Zero-Shot Generative Video Compression with Autoregressive Diffusion Priors
ZeroGVC performs zero-shot generative video compression by guiding pretrained autoregressive diffusion priors with codebook noise vectors for P-frames after encoding the initial I-frame.
-
CoRDE: Concept-Prior Routed Diffusion Experts for Structural Generalization in Robot Manipulation
CoRDE uses concept-prior variational distillation and LoRA-based expert pools to route diffusion models for structurally generalizable robot manipulation policies.
-
VT-DUDA: Visual Token Conditioning for Diffusion-guided Unsupervised Domain Adaptation
VT-DUDA proposes visual token conditioning from source images concatenated with text embeddings in latent diffusion models to improve synthetic target-style data generation for unsupervised domain adaptation.
-
Do as I Do: Dexterous Manipulation Data from Everyday Human Videos
DO AS I DO reconstructs and retargets hand-object interactions from in-the-wild monocular RGB videos to produce dexterous robot manipulation trajectories, outperforming prior methods on ground-truth and online video datasets.
-
VideoMDM: Towards 3D Human Motion Generation From 2D Supervision
VideoMDM learns coherent 3D motion manifolds from 2D supervision alone by using a pretrained lifter as noisy teacher, depth-weighted 2D reprojection loss, and adapted regularizers, nearly matching fully 3D-supervised performance on HumanML3D.
-
HYDRA-X: Native Unified Multimodal Models with Holistic Visual Tokenizers
HYDRA-X presents the first unified multimodal model using a single ViT for holistic image-video tokenization, with ablations on attention and compression plus a latent-level editing improvement.
-
High-Fidelity Two-Step Image Generation via Teacher-Aligned End-to-End Distillation
Z-Image Turbo++ narrows the quality gap to 8-step generation via three distillation techniques tailored for the 2-step regime.
-
Least-Action-Guided Diffusion for Physical Extrapolation
LAPG combines conditional score-based diffusion with an action-derived guidance score to reduce phase drift and preserve physical invariants during temporal, parameter, and geometric extrapolation on free-fall, spring-mass, vortex, and airfoil systems.
-
Fast and Highly Expressive Policy Learning for Offline Reinforcement Learning via Bootstrapped Flow Q-Learning
BFQ enables single-step noise-to-action mapping in offline RL by dividing flow-path displacements into bootstrappable short-range components learned from marginal velocity.
-
Few-step Generative Models as Lossy Compression
Few-step generative models can be reformulated as lossy codecs in the reverse channel coding framework without retraining, yielding faster encoding/decoding on low-resolution image benchmarks.
-
Learning the Universe: Posterior Reliability of Neural Generative Models in High-Dimensional Field-Level Inference of Cosmic Initial Conditions
Generative models for cosmological field-level inference can reproduce posterior means and cross-correlations yet fail to capture correct uncertainty geometry when validated against HMC reference samples.
-
KPGrasp: Scalable Keypoint Flow Matching for Dexterous Grasp Generation
KPGrasp is a scalable Transformer flow-matching model using 3D hand keypoints that achieves 76.3% success on Dexonomy (47.4% improvement) and best average on DexGrasp Anything without contact losses or test-time refinement.
-
LiteVSR: Lightweight Adaptation of Frozen Diffusion Transformers for Video Super-Resolution
LiteVSR performs video super-resolution on a completely frozen Diffusion Transformer via a lightweight State-Aware Adapter that uses dual-stream extraction and time-dependent cross-attention, reaching competitive quality with 11.25% trainable parameters after 12 GPU-hours.
-
Back on Track: Aligning Rewards and States for Reasoning in Diffusion Large Language Models
PAPO improves reasoning performance in diffusion LLMs by converting sparse terminal rewards into dense step-wise credit and replaying real high-uncertainty trajectories, reporting gains up to 42.2% on Countdown.
-
EditSR: Enhancing Neural Symbolic Regression via Edit-based Rectification
EditSR improves neural symbolic regression accuracy on complex expressions by pretraining an edit-based rectifier on state-transition correction chains that enforce syntactic validity and condition edits only on the current expression state.
-
Second-Order Path Kernel Interpolation Formulas in Machine Learning
Derives second-order path-kernel interpolation formulas for gradient descent, SGD, and momentum training, adding curvature terms and a concentration estimate around the expected prediction.
-
GenPO++: Generative Policy Optimization with Jacobian-free Likelihood Ratios
GenPO++ achieves exact Jacobian-free likelihood ratio computation for generative flow policies by embedding history states as auxiliary memory in a high-order reversible ODE solver.
-
Style-CCL: Content-Preserving Style Transfer via Curriculum Continual Learning
Style-CCL uses curriculum continual learning on a million-scale synthetic dataset with a dual-branch SC-DiT to achieve state-of-the-art content-preserving style transfer.
-
ReCache: Learning Budget-Aware Caching Schedules for Diffusion Models via REINFORCE
ReCache learns recomputation schedules via policy gradients to maximize quality under a target compute budget for any caching mechanism in diffusion models.