Representation Distribution Matching for One-Step Visual Generation
We elucidate the design space of Representation Distribution Matching (RDM), our name for the paradigm that trains a one-step image generator by matching generated and reference feature distributions under frozen pretrained encoders. We identify two design axes, how the distributions are compared and the representations they are compared in, and controlled studies along them yield three findings. First, the classical MMD, which could not train convincing generators a decade ago, becomes a strong and scalable objective once estimated right. Second, the generated batch is then the operative vari
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 46%abnormal-codex/Ai-Pixel-Design-Archive →
- LinkedLinked via arxiv author · 85%Lan Feng →
“Representation Distribution Matching for One-Step Visual Generation”
- LinkedLinked via arxiv author · 85%Wuyang Li →
“Representation Distribution Matching for One-Step Visual Generation”
- LinkedLinked via arxiv author · 85%Eloi Zablocki →
“Representation Distribution Matching for One-Step Visual Generation”
- LinkedLinked via arxiv author · 85%Matthieu Cord →
“Representation Distribution Matching for One-Step Visual Generation”
- LinkedLinked via arxiv author · 85%Alexandre Alahi →
“Representation Distribution Matching for One-Step Visual Generation”
- PossiblePossibly related (embedding) · 51%I built my 'first' flow matching image generator, here's what I learned [P] →