Phased DMD: Few-step Distribution Matching Distillation via Score Matching within Subintervals

arXiv 2025 26.8 method

TLDR

Phased DMD proposes a multi-step distillation framework using progressive distribution matching and score matching within subintervals to improve DMD for complex generation like text-to-video.

Reasoning

The paper addresses real limitations of DMD, such as reduced diversity and slow motion dynamics, with a phase-wise distillation and MoE-based approach. It provides mathematical formulations for subinterval training, but the abstract stops short of reporting detailed quantitative results and has no connection to world models.

Read-first score

Read-first score 26.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 1.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code,github

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=experiment

Topical relevance 42%

1.4

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 251.

Keyword Scores

video world model

world model

world simulator

generative world model

interactive world model

world dynamics prediction

model-based reinforcement learning world model

Resources

Paper