Driving · 2026-10-16

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

Hongchang Liu, Jian Liu, Yujuan Ding, Liang Peng, Xun Jiang, Fei Ma, Xing Xu, Yi Bin

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

GeoState learns a view-decoupled geometry state for 3D world modeling. A scene tokenizer aggregates multi-view GFM features into a fixed-size GeoState in a shared scene coordinate system; a ray-conditioned geometry decoder queries it under a camera only at decoding time. A temporal model predicts future GeoStates while modeling camera evolution separately, enabling decoding under predicted or external cameras and avoiding per-frame scale alignment.

Why it matters

The authors address view dependence in geometry foundation model features when used as predictive states. By separating scene-state evolution from camera-induced observation variation, GeoState aims to make latent dynamics easier to predict over long horizons, support geometrically consistent cross-view querying, and transfer zero-shot to other driving datasets.

Method

Train a scene tokenizer to aggregate multi-view GFM outputs into a fixed-length view-decoupled GeoState in a shared scene coordinate system.
Use a query-conditioned geometry decoder that reconstructs depth from a GeoState using camera rays, so the same GeoState can be decoded under different cameras without re-encoding.
Train an autoregressive world model to predict future GeoStates using a sliding four-state context, while camera evolution is handled separately.

Limitation

The paper text does not state a specific limitation of GeoState.

Abstract (from arXiv)

Recent 3D world models increasingly use features from geometry foundation models (GFMs) as predictive world states, benefiting from their rich geometric priors. However, the view dependence of these features entangles scene dynamics with camera-induced variation, making the resulting state evolution harder to predict. We propose GeoState, a view-decoupled geometry representation for predictive 3D world modeling that separates scene-state evolution from camera-dependent observation. Specifically, a scene tokenizer aggregates multi-view GFM features into a view-decoupled scene state, which a geometry decoder queries using camera rays. An autoregressive world model then predicts future scene states, while camera motion is modeled separately. The predicted states can therefore be decoded under either predicted or externally specified cameras, while retaining a single context-derived scale throughout the rollout without per-frame scale alignment. Experiments show that GeoState outperforms prior 3D world models on Waymo, supports geometrically consistent cross-view querying, and generalizes zero-shot to KITTI and nuScenes. Project page: https://hongchang-liu.github.io/GeoState-project-page/

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-10-15 · Indexed by Pub-AI: 2026-10-16 UTC

preprint