Transformers converge globally to the optimal DDPM denoiser for multi-token GMMs via self-attention mean denoising, with explicit token and iteration requirements.
hub
arXiv preprint arXiv:2208.11970 , year=
27 Pith papers cite this work. Polarity classification is still indexing.
27
Pith papers citing it
hub tools
citation-role summary
background 3
other 1
citation-polarity summary
representative citing papers
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.
Spatio-Temporal MeanFlow adapts MeanFlow to PDEs by replacing the generative velocity field with the physical operator and extending the integral constraint to the spatio-temporal domain, yielding a unified solver for time-dependent and stationary equations with improved accuracy and generalization.
Random team assignments in a professional firm reveal that indirect ties strongly increase new direct tie formation, while effects of degree and local density are smaller and less robust.
Presents SimA, a score-based single-query membership inference attack for diffusion models and LDMs that uses denoiser output norm to reveal training set proximity and outperforms multi-query baselines on eight datasets.
Doloris introduces dual conditional diffusion implicit bridges plus a sparsity masking strategy to model unpaired single-cell perturbation responses and reports state-of-the-art results on public datasets.
EnSF-LR combines nonlinear score-based analysis on observed components with EnKF-style linear regression on unobserved components via ensemble covariance, achieving lower full-state RMSE than EnSF and EnKF in nonlinear-observation tests on Lorenz-63 and Lorenz-96.
LBDTPP generates high-quality variable-length event sequences by autoregressing over latent blocks and diffusing within blocks, with Wasserstein bounds claiming reduced error accumulation under local approximation and prefix-stability assumptions.
T-CLIP introduces a physics-aware thermal captioning dataset (IR-Cap) and a decoupled dual-LoRA adaptation of CLIP that improves cross-modal retrieval on thermal benchmarks by separating scene-level and object-level thermal understanding.
A mixture-of-experts diffusion model with variational Bayesian inference jointly infers the channel and expert indicator to adapt to different propagation environments in massive MIMO channel estimation.
GenTS is a modular benchmark library providing unified data pipelines, generative models, and evaluation metrics for time series synthesis, forecasting, and imputation, with open-source code and initial benchmarking experiments.
An optimal control formulation adds time-dependent perturbations to the reverse diffusion process to match target attribute distributions while preserving sample fidelity.
BBDRec applies Brownian bridge diffusion to enable direct item-to-history transitions in sequential recommendation, outperforming prior diffusion and sequential baselines on public datasets.
DMin uses gradient compression to scalably estimate training data influence in billion-parameter diffusion models.
Moment-matched GMM kernels in DDIM yield lower FID and higher IS than Gaussian kernels at small sampling steps on CelebA-HQ, FFHQ, ImageNet, and Stable Diffusion tasks.
Semantic watermarks in LDMs have an irreducible geometric distortion floor from proxy-target model mismatches that limits forgery fidelity and supports scheme-agnostic detection via global drift and local deformation.
Proposes SDE sampler with state-dependent diffusion for HTDMs that induces self-regulating annealing, claimed necessary for heavy-tailed sampling.
Physical Foundation Models are fixed physical hardware realizations of foundation-scale neural networks that compute via inherent material dynamics, potentially delivering orders-of-magnitude gains in energy efficiency, speed, and density over digital systems.
Diffusion models are reorganized under a Langevin perspective that unifies ODE and SDE formulations and shows flow matching is equivalent to denoising under maximum likelihood.
A conditional diffusion model downscales global atmospheric forecasts from 100 km to 30 km resolution while improving probabilistic skill, matching power spectra, and preserving physical relationships.
Proposes Lipschitz regularization during fine-tuning to prevent distributional drift in personalized diffusion models, improving subject fidelity and prompt adherence.
A diffusion model variant that adds structured non-zero-mean noise via modified forward/reverse processes, yielding an ELBO loss analogous to offset noise but with time-dependent coefficients, and showing gains on synthetic high-dimensional data.
CGSoRec denoises social relations and reweights user social preferences to serve as conditions that steer a diffusion recommender away from popularity bias.
Timestep embeddings are redundant in diffusion models under certain conditions, with time-agnostic variants matching or exceeding conditioned models on FID, precision, and recall for CelebA and CIFAR-10.
citing papers explorer
-
Transformers Learn the Optimal DDPM Denoiser for Multi-Token GMMs
Transformers converge globally to the optimal DDPM denoiser for multi-token GMMs via self-attention mean denoising, with explicit token and iteration requirements.
-
Pathway variability, coat stiffening and mechanical adaptation during clathrin-mediated endocytosis
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.
-
Physics-Informed Neural PDE Solvers via Spatio-Temporal MeanFlow
Spatio-Temporal MeanFlow adapts MeanFlow to PDEs by replacing the generative velocity field with the physical operator and extending the integral constraint to the spatio-temporal domain, yielding a unified solver for time-dependent and stationary equations with improved accuracy and generalization.
-
Causal inference for social network formation
Random team assignments in a professional firm reveal that indirect ties strongly increase new direct tie formation, while effects of degree and local density are smaller and less robust.
-
Score-based Membership Inference on Diffusion Models
Presents SimA, a score-based single-query membership inference attack for diffusion models and LDMs that uses denoiser output norm to reveal training set proximity and outperforms multi-query baselines on eight datasets.
-
Doloris: Dual Conditional Diffusion Implicit Bridges with Sparsity Masking Strategy for Unpaired Single-Cell Perturbation Estimation
Doloris introduces dual conditional diffusion implicit bridges plus a sparsity masking strategy to model unpaired single-cell perturbation responses and reports state-of-the-art results on public datasets.
-
A Two-Step Ensemble Score Filter for Data Assimilation in Partially Observed Systems
EnSF-LR combines nonlinear score-based analysis on observed components with EnKF-style linear regression on unobserved components via ensemble covariance, achieving lower full-state RMSE than EnSF and EnKF in nonlinear-observation tests on Lorenz-63 and Lorenz-96.
-
Latent Block-Diffusion Temporal Point Processes: A Semi-Autoregressive Framework for Asynchronous Event Sequence Generation
LBDTPP generates high-quality variable-length event sequences by autoregressing over latent blocks and diffusing within blocks, with Wasserstein bounds claiming reduced error accumulation under local approximation and prefix-stability assumptions.
-
T-CLIP: Enabling Thermal Perception for Contrastive Language-Image Pretraining
T-CLIP introduces a physics-aware thermal captioning dataset (IR-Cap) and a decoupled dual-LoRA adaptation of CLIP that improves cross-modal retrieval on thermal benchmarks by separating scene-level and object-level thermal understanding.
-
Mixture-of-Experts Diffusion Models for Adaptive Massive MIMO Channel Estimation via Variational Bayesian Inference
A mixture-of-experts diffusion model with variational Bayesian inference jointly infers the channel and expert indicator to adapt to different propagation environments in massive MIMO channel estimation.
-
GenTS: A Comprehensive Benchmark Library for Generative Time Series Models
GenTS is a modular benchmark library providing unified data pipelines, generative models, and evaluation metrics for time series synthesis, forecasting, and imputation, with open-source code and initial benchmarking experiments.
-
Inference-Time Attribute Distribution Alignment for Unconditional Diffusion
An optimal control formulation adds time-dependent perturbations to the reverse diffusion process to match target attribute distributions while preserving sample fidelity.
-
Brownian Bridge Diffusion for Sequential Recommendation
BBDRec applies Brownian bridge diffusion to enable direct item-to-history transitions in sequential recommendation, outperforming prior diffusion and sequential baselines on public datasets.
-
DMin: Scalable Training Data Influence Estimation for Diffusion Models
DMin uses gradient compression to scalably estimate training data influence in billion-parameter diffusion models.
-
Improved DDIM Sampling with Moment Matching Gaussian Mixtures
Moment-matched GMM kernels in DDIM yield lower FID and higher IS than Gaussian kernels at small sampling steps on CelebA-HQ, FFHQ, ImageNet, and Stable Diffusion tasks.
-
Rethinking Forgery Attacks on Semantic Watermarks in Black-Box Settings: A Geometric Distortion Perspective
Semantic watermarks in LDMs have an irreducible geometric distortion floor from proxy-target model mismatches that limits forgery fidelity and supports scheme-agnostic detection via global drift and local deformation.
-
Self-Regulating Annealing in Heavy-Tailed Diffusion Models
Proposes SDE sampler with state-dependent diffusion for HTDMs that induces self-regulating annealing, claimed necessary for heavy-tailed sampling.
-
Physical Foundation Models: Fixed hardware implementations of large-scale neural networks
Physical Foundation Models are fixed physical hardware realizations of foundation-scale neural networks that compute via inherent material dynamics, potentially delivering orders-of-magnitude gains in energy efficiency, speed, and density over digital systems.
-
Rethinking the Diffusion Model from a Langevin Perspective
Diffusion models are reorganized under a Langevin perspective that unifies ODE and SDE formulations and shows flow matching is equivalent to denoising under maximum likelihood.
-
Downscaling weather forecasts from Low- to High-Resolution with Diffusion Models
A conditional diffusion model downscales global atmospheric forecasts from 100 km to 30 km resolution while improving probabilistic skill, matching power spectra, and preserving physical relationships.
-
Preserve and Personalize: Personalized Text-to-Image Diffusion Models without Distributional Drift
Proposes Lipschitz regularization during fine-tuning to prevent distributional drift in personalized diffusion models, improving subject fidelity and prompt adherence.
-
A Probabilistic Formulation of Offset Noise in Diffusion Models
A diffusion model variant that adds structured non-zero-mean noise via modified forward/reverse processes, yielding an ELBO loss analogous to offset noise but with time-dependent coefficients, and showing gains on synthetic high-dimensional data.
-
Balancing User Preferences by Social Networks: A Condition-Guided Social Recommendation Model for Mitigating Popularity Bias
CGSoRec denoises social relations and reweights user social preferences to serve as conditions that steer a diffusion recommender away from popularity bias.
-
On the Redundancy of Timestep Embeddings in Diffusion Models
Timestep embeddings are redundant in diffusion models under certain conditions, with time-agnostic variants matching or exceeding conditioned models on FID, precision, and recall for CelebA and CIFAR-10.
-
Elucidating Representation Degradation Problem in Diffusion Model Training
Diffusion models suffer representation degradation at high noise due to recoverability mismatch; ERD mitigates this by dynamic optimization reallocation, accelerating convergence across backbones.
-
On the Robustness of Distribution Support under Diffusion Guidance
Establishes robustness of distribution support for guided diffusion processes under exact score access across DDIM, DDPM, and exponential integrator discretizations.
-
AI-Generated Image Recognition via Fusion of CNNs and Vision Transformers
A fused CNN-ViT model achieves 97.32% accuracy distinguishing AI-generated from real images on the CIFAKE dataset.