DreamTrue: Action-Faithful Robot World Model with Counterfactual Post-Training

arXiv 2026 42.2 method, application

TLDR

DreamTrue is a multi-view, cross-embodiment robot world model for action-faithful video prediction, using geometric calibration and counterfactual post-training with an embodied reward model.

Reasoning

Strengths include novel counterfactual post-training, geometric action conditioning, and strong empirical results on AgiBot with reduced defect rates. Weaknesses include reliance on human-annotated feedback and focus on robot manipulation video prediction rather than general world simulation.

Read-first score

Read-first score 42.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 53.

Recency 5%

100

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

Topical relevance 25%

75.7

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 15%

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

Reproducibility 15%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=dataset,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: 559.

Keyword Scores

world model

video world model

generative world model

interactive world model

world dynamics prediction

world simulator

model-based reinforcement learning world model

Deep Analysis

Innovations

Multi-view, cross-embodiment robot world model for action-faithful and physically plausible video prediction
Image-space rendering of action trajectories with offline geometric calibration to improve action following
Counterfactual post-training that modifies recorded action trajectories to generate futures under wider action and contact configurations
Embodied video reward model trained on a human-annotated defect dataset covering robot, object, and interaction defects to guide reinforcement-learning post-training

Methodology

DreamTrue is a multi-view, cross-embodiment robot world model that renders action trajectories into image-space conditions and uses offline geometric calibration to align these conditions with target videos. It applies counterfactual post-training by modifying recorded actions to generate videos under a wider range of actions and contact configurations. A human-annotated video dataset covering robot, object, and interaction defects is used to train an embodied video reward model, whose scores guide reinforcement-learning post-training. The model is evaluated on AgiBot using action-following performance and human-assessed interaction defect rate.

Key Results

On AgiBot, DreamTrue achieves state-of-the-art action following and reduces the human-assessed interaction defect rate from 48.12% to 6.25%. It ranks first in the world model track of the AgiBot World Challenge 2026.

Limitations

The abstract identifies imprecise calibration as an obstacle that can impair action following when training on existing robot datasets.
The abstract identifies limited coverage of unsuccessful interactions as an obstacle that can bias predictions toward successful outcomes.

Resources

Paper