Unifying Distributional Training for One-Step Visual Generation
cs.LG
Submitted: 2026-09-28
Updated: 2026-10-02
Code: https://github.com/Jiawei-Yang/FD-Loss
Project page: https://shihaoyang0423.github.io/MGFlow-website
Sources
- Gradient Flow Drifting: Generative Modeling via Wasserstein Gradient Flows of KDE-Approximated Divergences
- A Wasserstein-type distance in the space of Gaussian Mixture Models
- Generative Modeling via Drifting
- Representation Distribution Matching for One-Step Visual Generation
- Drifting Fields are not Conservative
- AdvFD: Boosting Visual Generation via Adversarial Fr'echet Distance Loss
- Consistency Models Made Easy
- Improved Mean Flows: On the Challenges of Fastforward Generative Models
- GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment
- One-Step Generative Modeling via Wasserstein Gradient Flows
- Masked Autoencoders Are Scalable Vision Learners
- GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
- Classifier-Free Diffusion Guidance
- Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion
- Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation
- A Unified View of Score-Based and Drifting Models
- REPA-E: Unlocking VAE for End-to-End Tuning with Latent Diffusion Transformers
- Back to Basics: Let Denoising Generative Models Denoise
- Autoregressive Image Generation without Vector Quantization
- Amortized Moment Matching for Visual Generation
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks