WorldLine: Action-Driven Visual Simulation for Robotic Manipulation

arXiv 2026 42.8 method, system

TLDR

WorldLine is an action-driven visual simulator for robotic manipulation that learns transferable dynamics from action-free videos and grounds actions across embodiments, improving prediction and success accuracy.

Reasoning

The paper presents a novel decoupling of dynamics learning from action grounding, leveraging over 10,000 hours of action-free videos and 2,000 hours of action trajectories across embodiments. Strong empirical gains in robot-mask IoU and success prediction demonstrate effectiveness, though the abstract lacks architectural details and explicit limitations.

Read-first score

Read-first score 42.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 54.

Recency 5%

100

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

Methodology quality 15%

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

Topical relevance 25%

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

Reproducibility 15%

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

Graph prestige 15%

5.7

PageRank-style prestige over the local citation graph built from available OpenAlex references. local_pagerank=0.0016

Citation impact 15%

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

Citation velocity 10%

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: 577.

Keyword Scores

world simulator

world dynamics prediction

world model

generative world model

video world model

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Decouples transferable dynamics learning from heterogeneous action grounding, enabling large-scale training on action-free robot videos and separate embodiment-specific action grounding.
Uses an image-space action representation to provide a shared control interface across more than ten embodiments.
Introduces multi-view and failure-enriched training with relational regularization to improve interaction-sensitive prediction.
Uses robot-focused few-step distillation to enable efficient causal rollout while preserving action-critical motion.

Methodology

WorldLine learns manipulation dynamics from over 10,000 hours of action-free robot videos and grounds them using over 2,000 hours of action trajectories across more than ten embodiments. It employs an image-space action representation as a shared control interface, multi-view and failure-enriched training with relational regularization, and robot-focused few-step distillation for efficient causal rollout. Evaluation is conducted in held-out and out-of-domain settings, comparing against baselines on metrics such as robot-mask IoU, trajectory success accuracy, and task success improvement over direct policy execution.

Key Results

WorldLine maintains strong visual quality and robot-motion agreement across held-out and out-of-domain settings. It improves robot-mask IoU on failed trajectories by 0.1626 over the strongest baseline, predicts trajectory success with 74% mean accuracy across RoboTwin and AgiBot, and improves task success by up to 21.4 percentage points over direct policy execution without RoboTwin training or adaptation.

Resources

Paper