DiSI disentangles stochastic interpolants into separate generation and regression paths, allowing controllable transitions between regression and generative image restoration with a unified few-step sampler.
hub
Advances in neural information processing systems34, 8780–8794 (2021)
17 Pith papers cite this work. Polarity classification is still indexing.
17
Pith papers citing it
hub tools
citation-role summary
background 1
citation-polarity summary
years
2026 17roles
background 1polarities
background 1representative citing papers
GSS unifies diffusion models and random structure search as limiting regimes of one sampling process to recover diverse metastable structures at over tenfold lower cost than RSS, including for compositions outside the training data.
GeoEdit constructs local tangent frames from small perturbations to initial noise, enabling Jacobian-free on-manifold edits in diffusion models via alternating tangent steps and diffusion projections.
WaterGen decouples scene generation from medium degradation in a two-stage latent diffusion process to produce controllable realistic underwater images that improve downstream restoration and segmentation.
DPDiff-AD conditions a diffusion model on local prototypes (via nearest aggregation) and global prototypes (via optimal transport) to model normality scalably in multi-class anomaly detection, reporting AUROC gains on 160-category data.
Pretrained autoencoders in medical latent diffusion encode discriminative features well for reconstruction but structure their latent spaces in ways that hinder classifier learning, a gap that persists across architectures and is not closed by domain fine-tuning.
LPH-VTON uses a single denoising process with staged handover from structure-biased to texture-biased diffusion models to improve both geometric alignment and textural fidelity in virtual try-on.
A hybrid-conditioned diffusion transformer generates 2D topologies matching SIMP solutions within 1% compliance error using only five denoising steps.
Geometry-preserving losses based on tangent-space distances improve blackbox GAN adaptation to shifted distributions compared with standard losses.
VS-DDPM accelerates 3D diffusion models for medical modality translation, reaching SOTA Dice scores of 0.80-0.88 and SSIM 0.95 on missing MRI synthesis in BraTS2025 while remaining competitive on tumor removal and sCT tasks.
TabSCM produces causally consistent tabular data by orienting a CPDAG into a DAG, fitting root marginals with KDE, and using conditional diffusion plus trees for child nodes, outperforming GANs and diffusion baselines on fidelity, utility, and privacy across seven datasets.
DiffHDR converts LDR videos to HDR by formulating the task as generative radiance inpainting in a video diffusion model's latent space, using Log-Gamma encoding and synthesized training data to achieve better fidelity and stability than prior methods.
x-prediction maintains manifold adherence during training-free diffusion guidance better than ε- or v-prediction, per theoretical analysis and experiments on bird classification and style transfer.
SOCS derives per-step closed-form control signals from stochastic optimal control to steer diffusion sampling trajectories toward measurements while preserving the generative prior.
SEGS constructs structural energy in the PCA subspace of U-Net features and injects its gradient into the denoising process to improve multi-view consistency in text-to-3D generation.
SHIFT learns and applies steering vectors to selected layers and timesteps in DiT models to suppress concepts, shift styles, or bias objects while keeping image quality and prompt adherence intact.
DE-CM trains a flow-map consistency model on three sub-trajectories (coupling, instantaneous, noise-to-noisy) and reports 1.70 FID one-step on ImageNet 256.
citing papers explorer
-
Generative structure search for efficient and diverse discovery of molecular and crystal structures
GSS unifies diffusion models and random structure search as limiting regimes of one sampling process to recover diverse metastable structures at over tenfold lower cost than RSS, including for compositions outside the training data.