Streaming Multi-Agent Autoregressive Diffusion Model with World State Registers

arXiv 2026 39.7 method

TLDR

WorldWeaver: a streaming multi-agent video diffusion model with world state registers for cross-agent consistency.

Reasoning

The paper introduces world state registers to maintain shared world information across agents, addressing a key limitation in multi-agent video generation. Its strengths include the novel architectural design and explicit state supervision, but weaknesses are the limited evaluation to two-agent Minecraft and lack of comparison to interactive or RL-based world models.

Read-first score

Read-first score 39.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 47.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Topical relevance 29%

67.1

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 18%

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

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Citation impact 18%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 292.

Keyword Scores

generative world model

video world model

world model

world dynamics prediction

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Cross-agent world state registers: learnable tokens that store shared world information, track individual agent status, and are dynamically updated after each generated chunk.
Streaming multi-agent autoregressive video diffusion model (WorldWeaver) that incorporates world state registers into the rollout.
Supervision signals for world state registers including individual agent status, global state views (bird's-eye views), and scene text.
Mixture-of-Transformers architecture with separate weights for world state modeling and visual frame modeling.

Methodology

The paper presents WorldWeaver, a streaming multi-agent video diffusion model that augments autoregressive chunk generation with learnable cross-agent world state registers. These registers are updated dynamically and supervised with agent status, bird's-eye views, and scene text. A Mixture-of-Transformers design separates world state and visual frame processing. The model is evaluated on two-agent Minecraft video generation.

Key Results

Explicit world-state modeling via registers improves logical consistency and generation quality in two-agent Minecraft video generation.

Resources

Paper