An End-to-End Latent-Rollout Approach for Pushing Few-Step ImageNet- 256 Generation to FID 1.11 without Fr'echet Losses
cs.CV
Submitted: 2026-09-26
Updated: 2026-09-26
Terminology
Sources
- Directly Fine-Tuning Diffusion Models on Differentiable Rewards
- Generative Modeling via Drifting
- Diffusion Models Beat GANs on Image Synthesis
- Mean Flows for One-step Generative Modeling
- One-Step Generative Modeling via Wasserstein Gradient Flows
- Denoising Diffusion Probabilistic Models
- Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion
- REPA-E: Unlocking VAE for End-to-End Tuning with Latent Diffusion Transformers
- Diffusion Adversarial Post-Training for One-Step Video Generation
- Adversarial Flow Models
- Flow Matching for Generative Modeling
- Spectral Normalization for Generative Adversarial Networks
- Semantics Lead the Way: Harmonizing Semantic and Texture Modeling with Asynchronous Latent Diffusion
- FACM: Flow-Anchored Consistency Models
- High-Resolution Image Synthesis with Latent Diffusion Models
- Fast High-Resolution Image Synthesis with Latent Adversarial Diffusion Distillation
- Adversarial Diffusion Distillation
- Unified Continuous Generative Models
- Flow Map Distillation Without Data
- Representation Fr'echet Loss for Visual Generation
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models