Research map / Video / Video generation
Geometry Forcing: Marrying Video Diffusion and 3D Representation for Consistent World Modeling
Haoyu Wu, Diankun Wu, Tianyu He, Junliang Guo, Yang Ye, Yueqi Duan, Jiang Bian
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Geometry Forcing (VGGT alignment) | DFoT · RealEstate10K (256 frames) | FVD: 364 -> 243 | Improved |
| add Geometry Forcing (VGGT alignment) | DFoT · RealEstate10K (16 frames) | FVD: 252 -> 193 | Improved |
| add Geometry Forcing to Next-Frame Diffusion | NFD · Minecraft | FVD-16: 216 -> 205 | Improved |
| add Geometry Forcing to Wan2.1 | Wan2.1-1.3B · text-conditioned generation | imaging quality: 0.56 -> 0.59 | Improved |
| align to VGGT instead of DINOv2 (ablation) | Geometry Forcing · RealEstate10K (256 frames) | FVD-256: 297 -> 243 | Improved |
| add DINOv2 alignment on top of VGGT (ablation) | Geometry Forcing · RealEstate10K (256 frames) | FVD-256: 243 -> 237 | Improved |
| add Scale Alignment to Angular Alignment (ablation) | Geometry Forcing · RealEstate10K (256 frames) | FVD-256: 253.0 -> 243.0 | Improved |
| use MSE alignment instead of Angular+Scale (ablation) | Geometry Forcing · RealEstate10K (256 frames) | FVD-256: 243.0 -> 1648.0 | Hurt |
| internal alignment instead of explicit rendered geometry conditioning (ablation) | Geometry Forcing · RealEstate10K (256 frames) | FVD-256: 280 -> 243 | Improved |
| lengthen alignment context from 4 to 16 frames (ablation) | Geometry Forcing · RealEstate10K (256 frames) | FVD: 261 -> 243 | Improved |
| align last 3 layers instead of a single mid layer (ablation) | Geometry Forcing · RealEstate10K (256 frames) | FVD: 243 -> 280 | Hurt |
| combine Geometry Forcing with REPA (ablation) | Geometry Forcing · RealEstate10K (256 frames) | FVD: 243 -> 237 | Improved |
Later papers that compared with this method
2 later accepted papers report Geometry Forcing as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Captain Safari: A World Engine with Pose-Aligned 3D Memory (CVPR 2026 · better on 7 of 7 shared results)
- RoMan-4D: Learning Robot Arm Manipulation from 4D World Models (ECCV 2026 · better on 3 of 3 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VideoREPA — VideoREPA: Learning Physics for Video Generation through Relational Alignment with Foundation Models (NeurIPS 2025) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Diffusion4D: Fast Spatial-temporal Consistent 4D generation via Video Diffusion Models (NeurIPS 2024 · similarity 0.90)
- GeoVideo: Introducing Geometric Regularization into Video Generation Model (NeurIPS 2025 · similarity 0.90)
- Towards Physical Understanding in Video Generation: A 3D Point Regularization Approach (NeurIPS 2025 · similarity 0.89)
- Diffusion$^2$: Dynamic 3D Content Generation via Score Composition of Video and Multi-view Diffusion Models (ICLR 2025 · similarity 0.89)
- PanoWorld: Geometry-Consistent Panoramic Video World Modeling (NeurIPS 2026 · similarity 0.89)
- SweetDreamer: Aligning Geometric Priors in 2D diffusion for Consistent Text-to-3D (ICLR 2024 · similarity 0.89)
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.
