Driving · 2026-10-07

GeoWM: Efficient Direct World Modeling in Explicit Geometry

Mehrdad Noori, Guile Wu, Sam Hosseini, Dongfeng Bai

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

TL;DR

GeoWM is a geometry world model that directly forecasts future 3D scene geometry at specified horizons, without recursive rollout. It turns observed RGB frames into a geometric history using a frozen geometry foundation model, predicts future camera motion, projects the last observed depth into the predicted viewpoint as an anchor, and uses a flow-matching diffusion transformer to predict future depth directly for each horizon.

Why it matters

The authors argue that typical world-model paradigms predict pixels or latent features and then recover geometry, often requiring recursive rollouts that increase error accumulation and computational cost at long horizons. GeoWM instead forecasts explicit geometry directly at the requested horizon, conditioning on predicted camera viewpoint, and the paper reports improved depth, camera pose, and 3D geometry forecasting with substantially reduced inference time at longer horizons.

Method

Use a frozen geometry foundation model to extract depth maps and camera poses from the K most recent RGB frames; build a geometric history.
Predict future camera motion for any requested horizon with a lightweight camera-motion predictor; warp the last observed depth into the predicted viewpoint as a geometric anchor.
Forecast the target depth map at t+h directly with a flow-matching diffusion transformer conditioned on observed depth history and the horizon/anchor; each horizon is queried independently (no recursive rollout).

Limitation

The paper does not explicitly state a limitation in the provided text.

Abstract (from arXiv)

Modeling 3D scene geometry and its evolution over time is essential for autonomous driving and robotics. A common paradigm is to use world models to predict future images or latent representations of the environment and subsequently recover geometry from these predictions. However, this paradigm does not explicitly model geometric structure and typically relies on recursive rollouts to reach longer prediction horizons, leading to error accumulation and increasing computational cost. To address these limitations, we present GeoWM, a geometry world model that directly forecasts future scene geometry at specified future horizons without recursive rollout. The key idea is to leverage a geometry foundation model to transform observed RGB frames into a geometric history, which conditions a flow-matching transformer to predict the scene geometry at a specified future horizon. We further show that a lightweight camera-motion predictor can accurately estimate the future viewpoint, and that projecting the observed geometry into the predicted viewpoint provides an effective geometric prior for future geometry forecasting. Extensive experiments on four datasets spanning urban driving, aerial flight, and dynamic manipulation demonstrate that GeoWM outperforms the evaluated world models in forecasting depth, camera pose, and 3D scene geometry, while substantially reducing inference time at longer horizons.

Publication record

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

First public release: 2026-10-05 · Indexed by Pub-AI: 2026-10-07 UTC

preprint