#diffusion models
115 resultsFeatFix: Reuse What You Verify through Local Exact-Feature Correction for Faster Cached Diffusion Inference
Hanshuai Cui, Zhiqing Tang, Zhi Yao +3
FeatFix reuses exact intermediate features computed for verification to locally correct draft outputs in cached diffusion inference, speeding up image and video generation while pr…
S-Avatar: Diffusion-Guided Gaussian Head Avatars from a Single Image
Hail Song, Seokhwan Yang, Jiwon Yang +2
S-Avatar is a method that creates photorealistic 3D head avatars from a single image by using diffusion-guided Gaussian splatting and aligning the result with the FLAME parametric…
AnchorMark: Robust Diffusion Watermarking via Latent-Space Rotation Synchrony
Yuqi Qian, Yun Cao, Haocheng Fu +3
The paper proposes AnchorMark, a training‑free inversion‑based watermarking method for diffusion models that uses a latent‑space rotation synchrony property to embed a central anch…
Training Skills Like Parameters via Self-Supervised Semantic Diffusion
Mo Li, Zixin Yin, Ting Cao +1
The paper introduces a self‑supervised framework that lets a language model acquire and store textual skills in an external library using diffusion‑style reconstruction loss, witho…
ROAD: Reciprocal-Objective Alignment of Discriminative Semantics for 3D Shape Generation
Xiao Luo, Mingyang Du, Xin Zhou +5
The paper introduces ROAD, a framework that transfers semantic and structural knowledge from discriminative 3D foundation models into diffusion transformers for 3D shape generation…
FDDWAN: A Frequency-Decoupled Diffusion Network for Watermarking Attack
Chunpeng Wang, Yuxin Li, Xiaoyu Wang +3
The paper introduces FDDWAN, a two-stage neural network that removes invisible watermarks by first decomposing images with wavelets and then refining residuals with a diffusion mod…
VocalRender: Score-Native Singing Voice Synthesis for Real-World Composition
Yukun Chen, Tianrui Wang, Zhaoxi Mu +2
The paper introduces VocalRender, a system that can directly synthesize singing voices from musical scores—including lyrics, pitches, note values, and tempo—without needing separat…
TongueReenact: Geometry-Anchored Tongue Synthesis for Face Reenactment
MD Wahiduzzaman Khan, Mingshan Jia, Xiaolin Zhang +2
The paper presents a framework that adds realistic, cross‑identity tongue motion to face reenactment by automatically training a tongue segmentation model and using a spatially con…
4DHumanDiff: Direct Text-to-4DGS Generation for Consistent 360-Degree Dynamic Humans
Renlong Wu, Haoran Chen, Yuxiang Wei +3
The paper introduces 4DHumanDiff, a diffusion-based framework that directly generates 360-degree dynamic human models as 4D Gaussian Splatting representations from text prompts, el…
VETO: Towards Protecting Images From Frontier AI Editing
Jonas Grebe, Hossein Shakibania, Tobias Braun +2
The paper presents VETO, a subtle anti-edit cloak that disrupts how modern diffusion-based image editors read source images, and introduces VetoBench, a benchmark for evaluating pr…
Bunraku: Turning a Single Illustration into an Editable Live2D Character
Junhao Chen, Jingjia Mao, Dayong Li +6
The paper introduces Bunraku, a system that automatically creates a complete Live2D character—including layered RGBA images, deformation meshes, and animation keyposes—from a singl…
Mitigating Compounding Error via Video Representation Regularization
Taiye Chen, Qi Zhang, Yisen Wang
The paper studies why autoregressive video generation models accumulate errors over time and introduces a lightweight regularization that stabilizes hidden representations, reducin…
Existence-Field Diffusion Model for Spatial Point Processes with Variable Cardinality
Xiaoyin Pan, Christian R. Shelton, Rakshith Mahishi +1
The paper introduces the existence-field diffusion model, which uses an existence variable for each potential point to jointly model spatial locations and the number of points in s…
Anchoring and Steering Diffusion: Enhancing the Faithfulness of Text-to-Image Generation at Inference Time
Xinyi Wang, Yuyang Huang, Yalin Su +4
The paper introduces AnchorSteer, a training‑free method that improves text‑to‑image diffusion models by initializing with CLIP‑aligned latent noise and actively correcting semanti…
FARI: Robust One-Step Inversion for Watermarking in Diffusion Models
Jindong Yang, Han Fang, Weiming Zhang +2
The paper introduces FARI, a fast one-step inversion method combined with lightweight adversarial LoRA fine-tuning to robustly extract watermarks from diffusion-generated images, a…
Amortized Moment Matching for Visual Generation
Wenze Liu, Xintao Wang, Pengfei Wan +1
The paper introduces amortized moment matching, using neural networks to learn data moments as training signals, and proposes the Amortized Fréchet Distance loss to improve one-ste…
Flow Map Learning via Nongradient Vector Flow
Mark Goldstein, Anshuk Uppal, Raghav Singhal +2
The paper proposes SGFlow, a method that learns flow maps for diffusion and flow‑based generative models without requiring model invertibility or backpropagation through repeated m…
TPD: Temporal Prior Decoupling for Text-to-Video Diffusion Models
Taewon Kang, Matthias Zwicker
The paper introduces Temporal Prior Decoupling (TPD), a training‑free method that restores suppressed late‑segment information during diffusion sampling for text‑to‑video models, i…
When Do Learned Diffusion Proposals Help Constraint Solving? A Controlled Study on Continuous Algebraic Systems
Quang Bui, Sparsh Roy, Akash Gundimeda +1
The paper studies when graph‑neural diffusion proposals improve solving continuous algebraic constraint systems, comparing learned proposals to random multi‑start baselines and ide…
Genie Sim PanoWorld: An Infinite Indoor 3D World Generation Pipeline via Panoramic Scene Modeling and Simulation
Yongxin Su, Linjie Hou, Feng Wang +4
The paper introduces Genie Sim PanoWorld, a feed‑forward system that creates a controllable panoramic video from a single 360° image and then reconstructs it into a high‑fidelity,…
Dual Inversion for Text-to-Image Diffusion Models: From Both Prompt and Noise Perspectives
Xiaolong Liu, Junjian Li, Yuan Xiao +4
The paper introduces Dualin, a two‑stage method that simultaneously recovers a human‑readable text prompt and the latent noise of a target image to improve prompt inversion for tex…
FreeShadow: Training-Free Shadow Removal via Illumination Transfer and Selective Content Preservation in Diffusion Models
Yinan Wang, Yan Huang, Yong Xu +1
FreeShadow removes shadows from images without any training by leveraging pretrained diffusion models, using illumination transfer attention to bring lighting cues from non‑shadow…
ZUNA1.1: A more flexible EEG foundation model for Denoising and Super-resolution
Christopher Warner, Jonas Mago, JR Huml +1
The paper presents ZUNA1.1, a 380‑million‑parameter diffusion autoencoder that can denoise and super‑resolve EEG recordings of variable length and channel configurations, outperfor…
Qwen-Audio-3.0-Gen-Preview Technical Report
Junyu Dai, Xiaoyue Duan, Xinyue Fan +14
The paper introduces Qwen-Audio-3.0-Gen-Preview, a unified non‑autoregressive model that uses a diffusion transformer and a shared VAE to generate complete mixed‑waveform audio fro…