Vid2World: Crafting Video Diffusion Models to Interactive World Models

arXiv 25.5 2025 66.3 method

TLDR

Vid2World repurposes pre-trained video diffusion models into interactive world models via causalization and action guidance, tested in robotics, games, and navigation.

Reasoning

The paper presents a novel method for converting video diffusion models into interactive world models, with strong empirical validation across diverse domains. However, the abstract lacks detailed comparisons to existing methods and does not discuss limitations or failure cases.

Read-first score

Read-first score 66.3, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 58.

Recency 8%

86.7

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

Topical relevance 42%

82.9

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

Methodology quality 25%

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

Reproducibility 25%

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

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 291.

Keyword Scores

world model

interactive world model

video world model

world simulator

world dynamics prediction

generative world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Leveraging pre-trained video diffusion models as interactive world models via video diffusion causalization
Causal action guidance mechanism to enhance action controllability in world models
General approach for transferring video diffusion models to interactive world models across multiple domains

Methodology

Vid2World systematically explores video diffusion causalization, reshaping both the architecture and training objective of pre-trained video diffusion models to enable autoregressive generation. It also incorporates a causal action guidance mechanism to enhance action controllability. The method is evaluated across multiple domains including robot manipulation, 3D game simulation, and open-world navigation.

Key Results

Extensive experiments across robot manipulation, 3D game simulation, and open-world navigation demonstrate that Vid2World offers a scalable and effective pathway for repurposing highly capable video diffusion models into interactive world models.

Resources

Paper Project