mlsanityData to 14 Oct 2026

Geometry Forcing: Marrying Video Diffusion and 3D Representation for Consistent World Modeling

arXiv 2507.07982 ↗First posted 10 Jul 2025cs.CV cs.AIcited by 96

On the map: World models → Video-generation world models and simulators

also in 1 more
Abstract (arXiv)

Videos inherently represent 2D projections of a dynamic 3D world. However, our analysis suggests that video diffusion models trained solely on raw video data often fail to capture meaningful geometric-aware structure in their learned representations. To bridge the gap between video diffusion models and the underlying 3D nature of the physical world, we propose Geometry Forcing, a simple yet effective method that encourages video diffusion models to internalize 3D representations. Our key insight is to guide the model's intermediate representations toward geometry-aware structure by aligning them with features from a geometric foundation model. To this end, we introduce two complementary alignment objectives: Angular Alignment, which enforces directional consistency via cosine similarity, and Scale Alignment, which preserves scale-related information by regressing geometric features from normalized diffusion representations. We evaluate Geometry Forcing on both camera-view conditioned and action-conditioned video generation tasks. Experimental results demonstrate that our method substantially improves visual quality and 3D consistency over the baseline methods. Project page: https://GeometryForcing.github.io.

Key facts. Cited by 96 papers in the map (data through 14 Oct 2026). Placed in Video-generation world models and simulators, one of 11 directions in World models. Builds on 83 of its 100 references that are in the map. First posted on arXiv 10 Jul 2025.

Builds on

83 of 100 references.

Closest ancestors · 11

Same direction, newest first

  1. 2025Video World Models with Long-term Spatial Memory5 Jun
  2. 2025Long-Context State-Space Video World Models26 May
  3. 2025WorldMem: Long-term Consistent World Simulation with Memory16 Apr
  4. 2025MineWorld: a Real-Time and Open-Source Interactive World Model on Minecraft11 Apr
  5. 2025Wan: Open and Advanced Large-Scale Video Generative Models26 Mar
  6. 2024Diffusion Models Are Real-Time Game Engines27 Aug
  7. 2024CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer12 Aug
  8. 2024Genie: Generative Interactive Environments23 Feb
  9. 2022Scalable Diffusion Models with Transformers19 Dec
  10. 2022Flow Matching for Generative Modeling6 Oct
  11. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 72

  1. 2017Attention Is All You Needcited by 60,361
  2. 2020Language Models are Few-Shot Learnerscited by 32,419
  3. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  4. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  5. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  6. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  7. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,226
  8. 2023Adding Conditional Control to Text-to-Image Diffusion Modelscited by 5,503
  9. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,700
  10. 2021Vision Transformers for Dense Predictioncited by 1,803
  11. 2024HunyuanVideo: A Systematic Framework For Large Video Generative Modelscited by 1,660
  12. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,567
38 more
  1. 2025VGGT: Visual Geometry Grounded Transformercited by 1,539
  2. 2018World Modelscited by 1,513
  3. 2018Towards Accurate Generative Models of Video: A New Metric & Challengescited by 1,234
  4. 2025Cosmos World Foundation Model Platform for Physical AIcited by 864
  5. 2020GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fieldscited by 785
  6. 2024Structured 3D Latents for Scalable and Versatile 3D Generationcited by 770
  7. 2024Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Thinkcited by 703
  8. 2021DROID-SLAM: Deep Visual SLAM for Monocular, Stereo, and RGB-D Camerascited by 687
  9. 2024Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusioncited by 681
  10. 2024Movie Gen: A Cast of Media Foundation Modelscited by 644
  11. 2023GAIA-1: A Generative World Model for Autonomous Drivingcited by 637
  12. 2025Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusioncited by 625
  13. 2024VideoCrafter2: Overcoming Data Limitations for High-Quality Video Diffusion Modelscited by 579
  14. 2025Continuous 3D Perception Model with Persistent Statecited by 548
  15. 2023VideoCrafter1: Open Diffusion Models for High-Quality Video Generationcited by 535
  16. 2023VideoMAE V2: Scaling Video Masked Autoencoders with Dual Maskingcited by 528
  17. 2022All are Worth Words: A ViT Backbone for Diffusion Modelscited by 502
  18. 2023VideoPoet: A Large Language Model for Zero-Shot Video Generationcited by 492
  19. 2024CameraCtrl: Enabling Camera Control for Text-to-Video Generationcited by 460
  20. 2024ViewCrafter: Taming Video Diffusion Models for High-fidelity Novel View Synthesiscited by 403
  21. 2024UniDepth: Universal Monocular Metric Depth Estimationcited by 382
  22. 2025Fast3R: Towards 3D Reconstruction of 1000+ Images in One Forward Passcited by 367
  23. 2022Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videoscited by 345
  24. 2024Diffusion for World Modeling: Visual Details Matter in Ataricited by 300
  25. 2024Navigation World Modelscited by 272
  26. 2023DreamBooth3D: Subject-Driven Text-to-3D Generationcited by 235
  27. 2023LucidDreamer: Domain-free Generation of 3D Gaussian Splatting Scenescited by 231
  28. 2024MegaSaM: Accurate, Fast, and Robust Structure and Motion from Casual Dynamic Videoscited by 228
  29. 2025Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligencecited by 217
  30. 2023WHAM: Reconstructing World-grounded Humans with Accurate 3D Motioncited by 206
  31. 2024WonderWorld: Interactive 3D Scene Generation from a Single Imagecited by 202
  32. 20234D-fy: Text-to-4D Generation Using Hybrid Score Distillation Samplingcited by 185
  33. 2025FLARE: Feed-forward Geometry, Appearance and Camera Estimation from Uncalibrated Sparse Viewscited by 180
  34. 2024Splatt3R: Zero-shot Gaussian Splatting from Uncalibrated Image Pairscited by 178
  35. 2025History-Guided Video Diffusioncited by 171
  36. 2025VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstructioncited by 166
  37. 2025WorldScore: A Unified Evaluation Benchmark for World Generationcited by 156
  38. 2025VGGT-SLAM: Dense RGB SLAM Optimized on the SL(4) Manifoldcited by 136

50 most cited of 72 All 72 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 14; 2026: 82.

96 papers.

Sort:

By area:

  1. 2025Spatial Forcing: Implicit Spatial Representation Alignment for Vision-language-action Modelcited by 134
  2. 2025RELIC: Interactive Video World Model with Long-Horizon Memorycited by 96
  3. 2025What matters for Representation Alignment: Global Information or Spatial Structure?cited by 84
  4. 2025A Comprehensive Survey on World Models for Embodied AIcited by 68
  5. 2026Lyra 2.0: Explorable Generative 3D Worldscited by 44
  6. 2025FantasyWorld: Geometry-Consistent World Modeling via Unified Video and 3D Predictioncited by 38
  7. 2025Don't Blind Your VLA: Aligning Visual Representations for OOD Generalizationcited by 35
  8. 2026Gen3R: 3D Scene Generation Meets Feed-Forward Reconstructioncited by 31
  9. 2026Self-Supervised Flow Matching for Scalable Multi-Modal Synthesiscited by 31
  10. 2026Improved Baselines with Representation Autoencoderscited by 31
  11. 2026VGGRPO: Towards World-Consistent Video Generation with 4D Latent Rewardcited by 24
  12. 2026MultiWorld: Scalable Multi-Agent Multi-View Video World Modelscited by 23
38 more
  1. 2026S-VAM: Shortcut Video-Action Model by Self-Distilling Geometric and Semantic Foresightcited by 17
  2. 2026LIVE: Long-horizon Interactive Video World Modelingcited by 16
  3. 2026AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memoriescited by 15
  4. 2026Representation Alignment for Just Image Transformers is not Easier than You Thinkcited by 14
  5. 2025Reinforcement Learning with Inverse Rewards for World Model Post-trainingcited by 13
  6. 2026VGGT-World: Transforming VGGT into an Autoregressive Geometry World Modelcited by 13
  7. 2026Geometry-Aware Implicit Memory for Video World Modelscited by 13
  8. 2026EgoForge: Goal-Directed Egocentric World Simulatorcited by 12
  9. 2026Action Images: End-to-End Policy Learning via Multiview Video Generationcited by 12
  10. 2026World-R1: Reinforcing 3D Constraints for Text-to-Video Generationcited by 11
  11. 2026Prisma-World: Camera-Controllable Multi-Agent Video World Modelcited by 11
  12. 2025CAMEO: Correspondence-Attention Alignment for Multi-View Diffusion Modelscited by 10
  13. 2026GEM-4D: Geometry-Enhanced Video World Models for Robot Manipulationcited by 8
  14. 2026Towards Interactive Video World Modeling: Frontiers, Challenges, Benchmarks, and Future Trendscited by 8
  15. 2026CineScene: Implicit 3D as Effective Scene Representation for Cinematic Video Generationcited by 7
  16. 2026DreamWorld: Unified World Modeling in Video Generationcited by 7
  17. 2026VideoWeave: Unlocking Geometric Consistency in Video Generation via Joint Geometry-Video Modelingcited by 7
  18. 2026Feed-Forward 3D Scene Modeling: A Problem-Driven Perspectivecited by 6
  19. 2025Enhancing Physical Consistency in Lightweight World Modelscited by 5
  20. 2025Captain Safari: A World Engine with Pose-Aligned 3D Memorycited by 5
  21. 2025WorldReel: 4D Video Generation with Consistent Geometry and Motion Modelingcited by 5
  22. 2025View-Consistent Diffusion Representations for 3D-Consistent Video Generationcited by 4
  23. 2025Real2Edit2Real: Generating Robotic Demonstrations via a 3D Control Interfacecited by 4
  24. 2026MemCam: Memory-Augmented Camera Control for Consistent Video Generationcited by 4
  25. 2026TORA: Topological Representation Alignment for 3D Shape Assemblycited by 4
  26. 2026Learning Visual Feature-Based World Models via Residual Latent Actioncited by 4
  27. 2026PixWorld: Unifying 3D Scene Generation and Reconstruction in Pixel Spacecited by 4
  28. 2026Efficient Camera-Controlled Video Generation of Static Scenes via Sparse Diffusion and 3D Renderingcited by 3
  29. 2026GeoNVS: Geometry Grounded Video Diffusion for Novel View Synthesiscited by 3
  30. 2026CoInteract: Physically-Consistent Human-Object Interaction Video Synthesis via Spatially-Structured Co-Generationcited by 3
  31. 2026Geo-Align: Video Generation Alignment via Metric Geometry Rewardcited by 3
  32. 2026RECAP-Forcing: Retaining Content Appearances for Long Video Generationcited by 3
  33. 2026SpatialCrafter: Single Image World Modeling with Generative 3D Proxiescited by 3
  34. 2026Test-Time Conditioning with Representation-Aligned Visual Featurescited by 2
  35. 2026OneWorld: Taming Scene Generation with 3D Unified Representation Autoencodercited by 2
  36. 2026GeoFlow: Enforcing Implicit Geometric Consistency in Video Generationcited by 2
  37. 2026Trust It or Not: Evidential Uncertainty for Feed-Forward 3D Reconstruction with Trust3Rcited by 2
  38. 2026Drift-Resistant Navigation World Model with Anchored Epipolar Guidancecited by 2

50 most cited of 96 All 96 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025UniGeo: Taming Video Diffusion for Unified Consistent Geometry Estimation
  2. 2026VIGOR: VIdeo Geometry-Oriented Reward for Temporal Generative Alignment
  3. 2025Epipolar Geometry Improves Video Generation Models
  4. 2025NOVA3D: Normal Aligned Video Diffusion Model for Single Image to 3D Generation
  5. 2024V3D: Video Diffusion Models are Effective 3D Generators
  6. 2024Diffusion4D: Fast Spatial-temporal Consistent 4D Generation via Video Diffusion Models
  7. 2026VideoGPA: Distilling Geometry Priors for 3D-Consistent Video Generation
  8. 2026World Reconstruction From Inconsistent Views