Awesome World Model Hub Papers · Datasets · Projects
← Back to papers

Matrix-Game 2.0: An Open-Source, Real-Time, and Streaming Interactive World Model

arXiv 25.8 2025 72.1 method, system

TLDR

Matrix-Game 2.0 is an open-source interactive world model for real-time streaming video generation using few-step auto-regressive diffusion at 25 FPS.

Reasoning

The paper presents a novel framework that overcomes the real-time limitations of existing interactive world models through few-step distillation and causal architecture. Its strengths include open-source release and a scalable data pipeline, but it lacks real-world evaluation, relying solely on simulated environments (Unreal Engine, GTA5).

Read-first score

Read-first score 72.1, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 48.

Recency 8%
86.7

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

Reproducibility 25%
85

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

Topical relevance 42%
68.6

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%
60

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

Field roles

FrontierReproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 95.

Keyword Scores

interactive world model
10
world model
9
video world model
9
generative world model
8
world simulator
6
world dynamics prediction
5
model-based reinforcement learning world model
1

Deep Analysis

Innovations

  • Scalable data production pipeline for Unreal Engine and GTA5 environments generating about 1200 hours of video data with diverse interaction annotations
  • Action injection module enabling frame-level mouse and keyboard inputs as interactive conditions
  • Few-step distillation based on causal architecture for real-time and streaming video generation at 25 FPS

Methodology

Matrix-Game 2.0 uses a few-step auto-regressive diffusion model with a causal architecture to generate long videos on-the-fly. It incorporates an action injection module for frame-level mouse and keyboard inputs and is trained on a large-scale dataset produced via a pipeline from Unreal Engine and GTA5 environments. The model employs distillation to reduce inference steps, enabling real-time performance.

Key Results

Matrix-Game 2.0 generates high-quality minute-level videos across diverse scenes at an ultra-fast speed of 25 FPS.

Limitations

  • Dependence on synthetic data from Unreal Engine and GTA5 may limit generalization to real-world scenes
  • Few-step distillation may trade off generation quality for speed compared to full diffusion models

Tags