WristWorld: Generating Wrist-Views via 4D World Models for Robotic Manipulation

arXiv 25.10 2025 57.6 method, application

TLDR

WristWorld generates wrist-view videos from anchor views using a 4D world model, improving VLA manipulation performance.

Reasoning

The paper introduces a novel two-stage method combining geometric reconstruction and video generation to bridge the anchor-wrist view gap, with strong empirical results on real robotic datasets. However, the approach is specialized to wrist-view generation and relies on a specific geometric model (VGGT), limiting generality.

Read-first score

Read-first score 57.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 35.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Methodology quality 25%

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

Topical relevance 42%

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 25%

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 310.

Keyword Scores

world model

video world model

generative world model

world dynamics prediction

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

First 4D world model that generates wrist-view videos solely from anchor views, addressing the gap between abundant anchor views and scarce wrist views.
Spatial Projection Consistency (SPC) Loss to estimate geometrically consistent wrist-view poses and 4D point clouds.
Two-stage pipeline: Reconstruction (extending VGGT with SPC loss) and Generation (temporally coherent video synthesis from reconstructed perspective).

Methodology

WristWorld operates in two stages: (i) Reconstruction, which extends VGGT and incorporates a Spatial Projection Consistency (SPC) Loss to estimate geometrically consistent wrist-view poses and 4D point clouds; (ii) Generation, which employs a video generation model to synthesize temporally coherent wrist-view videos from the reconstructed perspective. The model is evaluated on Droid, Calvin, and Franka Panda datasets.

Key Results

WristWorld achieves state-of-the-art video generation with superior spatial consistency, and improves VLA performance by raising the average task completion length on Calvin by 3.81% and closing 42.4% of the anchor-wrist view gap.

Resources

Paper