, Depth Anything V2) that converts observed RGB frames into explicit geometry representations (depth maps, camera parameters, point clouds).
World model forecasts future 3D geometry directly without recursive rollouts
GeoWM uses a geometry foundation model and flow-matching transformer to predict depth, camera pose, and 3D point clouds at arbitrary horizons, outperforming video- and feature-based world models with drastically lower inference time.
Chinese Tech
Mehrdad Noori · Guile Wu · Sam Hosseini · Dongfeng Bai
Huawei Noah's Ark Lab
Research Digest··3 min read
Noori et al.
Why this paper
From Huawei Noah's Ark Lab
In one line
GeoWM directly predicts future 3D geometry at requested horizons, improving accuracy and cutting long-horizon inference time by avoiding recursive rollouts.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§