Genie Sim PanoWorld: An Infinite Indoor 3D World Generation Pipeline via Panoramic Scene Modeling and Simulation

arXiv 2026 34.8 method

TLDR

A two-stage feed-forward pipeline generating trajectory-controllable panoramic video from a single 360° panorama and reconstructing it into a navigable 3D Gaussian scene for embodied AI simulation.

Reasoning

The paper presents a clear methodology combining latent video diffusion with 3D reconstruction, and reports experimental comparisons and zero-shot generalization. Weaknesses include limited evidence of interactive or RL-specific capabilities, as the abstract focuses on generation and reconstruction rather than embodied interaction.

Read-first score

Read-first score 34.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 39.

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,experiment,metric

Topical relevance 25%

55.7

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

Reproducibility 15%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

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: 290.

Keyword Scores

video world model

generative world model

world model

world dynamics prediction

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Two-stage feed-forward pipeline that bridges generation and reconstruction via an explicit, trajectory-controllable panoramic video from a single 360° panorama.
NavMesh-planned SE(3) roaming trajectory injection into a latent video diffusion model through dense geometry-warped conditioning.
Long-short trajectory mixed training and a self-consistency objective based on shortcut models, enabling high-fidelity video in four CFG-free denoising steps.
Feed-forward panoramic reconstructor that lifts generated panoramic video into a 3D Gaussian scene supporting real-time, free-viewpoint roaming as a simulation-ready asset for embodied AI.

Methodology

Genie Sim PanoWorld uses a two-stage feed-forward pipeline: first, a latent video diffusion model generates trajectory-controllable panoramic video from a single 360° panorama, conditioned on NavMesh-planned SE(3) trajectories and dense geometry-warped information, and trained with long-short trajectory mixing and a shortcut-model self-consistency objective. Second, a feed-forward panoramic reconstructor converts the generated video into a 3D Gaussian scene. The method is evaluated against geometry-conditioned baselines on panoramic video generation and downstream 3D reconstruction, including zero-shot unseen indoor scenes.

Key Results

Genie Sim PanoWorld outperforms geometry-conditioned baselines in both panoramic video generation and downstream 3D reconstruction, and generalizes zero-shot to unseen indoor scenes while generating video in four CFG-free denoising steps.

Resources

Paper