GeoState: Learning View-Decoupled Geometry States for 3D World Modeling

arXiv 2026 34.8 method

TLDR

GeoState introduces a view-decoupled geometry representation for 3D world modeling, separating scene-state evolution from camera motion and enabling autoregressive prediction with zero-shot generalization across autonomous driving datasets.

Reasoning

Strong contribution in disentangling view-dependent observations from scene-state dynamics, with empirical validation on Waymo and zero-shot generalization to KITTI and nuScenes. However, the abstract lacks details on evaluation metrics, baselines, and limitations, and it does not address interactive or reinforcement-learning settings.

Read-first score

Read-first score 34.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 30.

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=experiment,metric,result

Reproducibility 15%

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

Topical relevance 25%

42.9

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

Graph prestige 15%

10.9

PageRank-style prestige over the local citation graph built from available OpenAlex references. local_pagerank=0.0031

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: 369.

Keyword Scores

world model

world dynamics prediction

generative world model

world simulator

video world model

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Introduces GeoState, a view-decoupled geometry representation for predictive 3D world modeling that separates scene-state evolution from camera-dependent observation.
Uses a scene tokenizer to aggregate multi-view geometry foundation model features into a view-decoupled scene state.
Employs a geometry decoder that queries the scene state using camera rays, enabling decoding under predicted or externally specified cameras.
Uses an autoregressive world model to predict future scene states while modeling camera motion separately.
Maintains a single context-derived scale throughout rollout, avoiding per-frame scale alignment.

Methodology

GeoState tokenizes multi-view geometry foundation model features into a view-decoupled scene state via a scene tokenizer, then decodes geometry by querying this state with camera rays. An autoregressive world model predicts future scene states while camera motion is modeled separately, and predicted states are decoded under predicted or externally specified cameras. The approach is evaluated on Waymo and zero-shot on KITTI and nuScenes against prior 3D world models.

Key Results

GeoState outperforms prior 3D world models on Waymo, supports geometrically consistent cross-view querying, and generalizes zero-shot to KITTI and nuScenes.

Resources

Paper