GigaWorld-1: A Roadmap to Build World Models for Robot Policy Evaluation

arXiv 2026 54.8 benchmark, survey

TLDR

Systematic study of world models for robot policy evaluation using WMBench benchmark, analyzing 7 video world models and 324k rollouts.

Reasoning

The paper introduces a large-scale benchmark (WMBench) with real-robot data and extensive experiments, providing clear insights on the importance of long-horizon consistency over visual realism. However, the abstract is cut off, and the specific limitations or missing details are not fully visible.

Read-first score

Read-first score 54.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 55.

Methodology quality 18%

100

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

Recency 6%

100

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

Topical relevance 29%

78.6

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

Reproducibility 18%

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

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

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 798.

Keyword Scores

world model

video world model

world dynamics prediction

generative world model

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

WMBench: a benchmark for controlled comparisons of world models for robotic policy evaluation
Systematic study revealing key properties of world models for reliable policy assessment
GigaWorld-1: a world model specially optimized for policy evaluation, along with a design roadmap

Methodology

The study constructs WMBench from real-robot teleoperation data and matched policy rollouts across diverse manipulation tasks, then systematically evaluates 7 video world models with 4 action representation schemes using over 324,000 simulated rollouts paired with real robot executions, further enriched with community challenge submissions, synthetic trajectories, and 12,000+ hours of training videos.

Key Results

Evaluator quality depends more on long-horizon, action-faithful rollout consistency than short-term visual realism; pretraining gains require balancing general world knowledge with robot-specific controllability; and architectural choices like action encoding, memory design, and evaluator-focused post-training strongly determine alignment with real-world robot behavior.

Resources

Paper