VideoREPA: Learning Physics for Video Generation through Relational Alignment with Foundation Models

arXiv 2025 34.1 method

TLDR

VideoREPA distills physics understanding from video foundation models into text-to-video diffusion models via token relation alignment, improving physical plausibility of generated videos.

Reasoning

The paper proposes a novel token relation distillation loss to inject physics knowledge into T2V models and reports benchmark improvements over CogVideoX. Strengths include a clear problem framing and a new alignment method, but the abstract lacks quantitative details and broader comparisons, and the connection to world models is not explicitly established.

Read-first score

Read-first score 34.1, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 4.

Recency 8%

86.7

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

Methodology quality 25%

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

Reproducibility 25%

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

Topical relevance 42%

5.7

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: 203.

Keyword Scores

world dynamics prediction

world model

video world model

world simulator

generative world model

interactive world model

model-based reinforcement learning world model

Resources

Paper