Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond

arXiv 2026 65.8 theory

TLDR

Introduces a levels x laws taxonomy for world models, defining three capability levels and four regimes, synthesizing over 400 works.

Reasoning

Strengths: comprehensive taxonomy covering multiple domains and capability levels, with broad literature synthesis. Weaknesses: abstract lacks specific empirical validation or real-world experiments; it is a survey/roadmap rather than presenting new experimental results.

Read-first score

Read-first score 65.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 54.

Recency 6%

100

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

Reproducibility 18%

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

Topical relevance 29%

77.1

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=evaluation,experiment,result

Citation impact 18%

59.8

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.59830478

Citation velocity 12%

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

Field roles

FrontierBridgeMethodology anchorReproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 417.

Keyword Scores

world model

world simulator

world dynamics prediction

interactive world model

model-based reinforcement learning world model

video world model

generative world model

Deep Analysis

Innovations

Introduction of a 'levels x laws' taxonomy for agentic world modeling, defining three capability levels (L1 Predictor, L2 Simulator, L3 Evolver) and four governing-law regimes (physical, digital, social, scientific).
Synthesis of over 400 works and summary of more than 100 representative systems across multiple research communities (model-based RL, video generation, web/GUI agents, multi-agent social simulation, AI-driven scientific discovery).
Proposal of decision-centric evaluation principles and a minimal reproducible evaluation package for world models.
Roadmap connecting previously isolated communities and outlining architectural guidance, open problems, and governance challenges.

Methodology

The paper constructs a 'levels x laws' taxonomy to categorize world models along two axes: capability levels (predictor, simulator, evolver) and governing-law regimes (physical, digital, social, scientific). Using this framework, the authors conduct a literature survey, synthesizing over 400 works and analyzing methods, failure modes, and evaluation practices across each level-regime pair. They then propose evaluation principles and a reproducible evaluation package, and outline architectural guidance and open problems.

Key Results

No experimental results are reported; the paper is a survey and taxonomy that provides a structured roadmap for agentic world modeling, synthesizing existing work and proposing evaluation principles.

Resources

Paper Code