A Better Critic Matters for Distribution Matching Distillation
cs.CV, cs.LG
Submitted: 2026-10-30
Updated: 2026-10-30
Comments: 35 pages, 16 figures. Project page and code: https://peppaking8.github.io/critic-matters/
Code: https://github.com/black-forest-labs/flux2
Project page: https://peppaking8.github.io/critic-matters/ABSTRACT
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Optimizing Few-Step Generation with Adaptive Matching Distillation
- Context-Matched Distillation: Teacher Causality for Autoregressive Video Distillation
- Flash Diffusion: Accelerating Any Conditional Diffusion Model for Few Steps Image Generation
- Data-Forcing Distillation: Restoring Diversity and Fidelity in Few-Step Video Generation
- Diffusion Models Beat GANs on Image Synthesis
- Phased DMD: Few-step Distribution Matching Distillation via Score Matching within Subintervals
- One-Forcing: Towards Stable One-Step Autoregressive Video Generation
- One Step Diffusion via Shortcut Models
- SenseFlow: Scaling Distribution Matching for Flow-based Text-to-Image Distillation
- Mean Flows for One-step Generative Modeling
- Improved Mean Flows: On the Challenges of Fastforward Generative Models
- Matrix-game 2.0: An open-source, real-time, and streaming interactive world model
- GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
- LoRA: Low-Rank Adaptation of Large Language Models
- Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion
- VBench: Comprehensive Benchmark Suite for Video Generative Models
- Distribution Matching Distillation Meets Reinforcement Learning
- A Rank Stabilization Scaling Factor for Fine-Tuning with LoRA
- Distribution Matching Distillation without Fake Score Network
- T2V-Turbo-v2: Enhancing Video Generation Model Post-Training through Data, Reward, and Conditional Guidance Design
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs