AI Paper Atlas

Research map / Video / Video generation

Learning 4D Embodied World Models

Haoyu Zhen, Qiao Sun, Hongxin Zhang, Junyan Li, Siyuan Zhou, Yilun Du, Chuang Gan

ICCV 2025 · Video › Video generation · approach: World-model video generation · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
predict depth and normal jointly with RGB instead of post-estimating them from generated videoCogVideoX · 4D scene generation (real)Chamfer L1: 0.2191 -> 0.2030Improved
predict depth and normal jointly with RGB instead of post-estimating themCogVideoX · 4D scene generation (synthetic)Chamfer L1: 0.2884 -> 0.0811Improved
predict depth and normal jointly with RGBCogVideoX · 4D scene generation (real)depth AbsRel: 26.17 -> 22.07Improved
predict depth and normal jointly with RGBCogVideoX · 4D scene generation (real)FVD: 20.64 -> 21.59Hurt
predict depth and normal jointly with RGBCogVideoX · 4D scene generation (synthetic)normal mean angular error: 20.36 -> 14.75Improved
use 4D video world model over 4D point cloud diffusion (4D Point-E)TesserAct · 4D scene generation (real)Chamfer L1: 0.2211 -> 0.2030Improved
use 4D point clouds with inverse dynamics instead of 2D inverse dynamics (UniPi*)4DWM · RLBench close boxsuccess rate: 81 -> 88Improved
use 4D point clouds with inverse dynamics instead of 2D inverse dynamics (UniPi*)4DWM · RLBench open drawersuccess rate: 67 -> 80Improved
use 4D point clouds instead of 2D inverse dynamics (UniPi*)4DWM · RLBench open microwavesuccess rate: 72 -> 70Hurt
use 4D point clouds instead of 2D inverse dynamics (UniPi*)4DWM · RLBench weighing offsuccess rate: 68 -> 62Hurt
use 4D point clouds instead of 2D inverse dynamics (UniPi*)4DWM · RLBench sweep to dustpansuccess rate: 49 -> 56Improved
use RGB-DN 4D prediction instead of Gaussian-splatting reconstruction (Shape of Motion) for novel viewsTesserAct · RLBench novel view synthesisPSNR: 10.94 -> 12.99Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy