PanoWorld: Geometry-Consistent Panoramic Video World Modeling

arXiv 2026 63.7 method

TLDR

PanoWorld generates geometry-consistent 360° video from a single image and caption using depth and trajectory consistency losses.

Reasoning

The paper introduces a novel approach to panoramic video generation that explicitly enforces geometric consistency, supported by a new dataset and lightweight regularizers. Strengths include clear problem framing and empirical validation; weaknesses include reliance on a pre-trained model and limited scope to panoramic video.

Read-first score

Read-first score 63.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 35.

Recency 6%

100

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

Reproducibility 18%

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

Methodology quality 18%

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

Citation impact 18%

79.3

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.79300995

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 velocity 12%

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

Field roles

FoundationFrontierBridgeMethodology anchorReproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 515.

Keyword Scores

video world model

world model

generative world model

world dynamics prediction

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Framing panoramic video generation as a geometry- and dynamics-consistent latent state modeling problem rather than pure visual synthesis
Depth consistency loss against pseudo ground-truth panoramic depth
Trajectory consistency loss that supervises 3D world-frame positions of tracked points across time
Spherical-geometry-aware adaptation to conditioning and positional encoding
PanoGeo dataset with consistent depth, trajectory, and prompt annotations from diverse real and synthetic sources

Methodology

PanoWorld builds on a pre-trained perspective video world model and introduces two lightweight regularizers: a depth consistency loss using pseudo ground-truth panoramic depth, and a trajectory consistency loss that supervises 3D world-frame positions of tracked points over time. It further applies spherical-geometry-aware adaptation to the conditioning and positional encoding, and uses the newly introduced PanoGeo dataset for training and stratified evaluation.

Key Results

PanoWorld improves geometric consistency over prior panoramic generation methods while maintaining competitive visual realism.

Resources

Paper Code