GE-Sim 2.0: A Roadmap Towards Comprehensive Closed-loop Video World Simulators for Robotic Manipulation

arXiv 2026 62.9 method, system

TLDR

GE-Sim 2.0 is a closed-loop video world simulator for robotic manipulation, trained on real-world data, with modules for state decoding, rollout scoring, and fast inference, achieving top leaderboard performance and real-world policy gains.

Reasoning

The paper presents a comprehensive system with strong empirical results on a public leaderboard and real-world transfer, but the abstract lacks detailed methodology and discussion of limitations, making it hard to fully assess reproducibility and failure modes.

Read-first score

Read-first score 62.9, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 63.

Recency 6%

100

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

Topical relevance 29%

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

Citation impact 18%

85.2

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.85196052

Methodology quality 18%

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

Reproducibility 18%

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

Citation velocity 12%

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

Field roles

FoundationFrontierBridge

Rank sensitivity

Stability: volatile; rank range: 523.

Keyword Scores

world simulator

video world model

world model

generative world model

interactive world model

world dynamics prediction

model-based reinforcement learning world model

Deep Analysis

Innovations

Re-training on thousands of hours of real-world robot data spanning teleoperation, contact-rich interaction, and on-robot policy deployment to improve action-following fidelity and trajectory coverage
State expert module that decodes proprioceptive state from video latents to support next-chunk prediction by downstream VLA policies
World judge module that scores generated rollouts against task instructions, providing machine-verifiable success signals and rewards
Acceleration framework delivering a 25-frame rollout in 2.3 seconds on a single H100 with up to 4× frame skipping for long-horizon evaluation

Methodology

GE-Sim 2.0 is built on the action-conditioned video generation framework of Genie Envisioner and re-trained on thousands of hours of real-world robot data including teleoperation, contact-rich interaction, and on-robot policy deployment. Three new modules are added: a state expert to decode proprioceptive state from video latents, a world judge to score rollouts against task instructions, and an acceleration framework for fast inference. The model is evaluated on the WorldArena leaderboard and compared against dedicated robotic world models and closed-source general video generators.

Key Results

GE-Sim 2.0 tops the public WorldArena leaderboard at only 2B parameters, outperforming both dedicated robotic world models and closed-source general video generators, and policies trained against its rollouts and rewards translate into measurable real-world gains.

Resources

Paper