mlsanityData to 7 Oct 2026

Positional Encoding Field

arXiv 2510.20385 ↗First posted 23 Oct 2025cs.CVcited by 20

On the map: Diffusion models → Image generation, text-to-image and editing

Abstract (arXiv)

Diffusion Transformers (DiTs) have emerged as the dominant architecture for visual generation, powering state-of-the-art image and video models. By representing images as patch tokens with positional encodings (PEs), DiTs combine Transformer scalability with spatial and temporal inductive biases. In this work, we revisit how DiTs organize visual content and discover that patch tokens exhibit a surprising degree of independence: even when PEs are perturbed, DiTs still produce globally coherent outputs, indicating that spatial coherence is primarily governed by PEs. Motivated by this finding, we introduce the Positional Encoding Field (PE-Field), which extends positional encodings from the 2D plane to a structured 3D field. PE-Field incorporates depth-aware encodings for volumetric reasoning and hierarchical encodings for fine-grained sub-patch control, enabling DiTs to model geometry directly in 3D space. Our PE-Field-augmented DiT achieves state-of-the-art performance on single-image novel view synthesis and generalizes to controllable spatial image editing.

Key facts. Cited by 20 papers in the map (data through 7 Oct 2026). Placed in Image generation, text-to-image and editing, one of 14 directions in Diffusion models. Builds on 45 of its 50 references that are in the map. First posted on arXiv 23 Oct 2025.

Builds on

45 of 50 references.

References · 45

  1. 2017Attention Is All You Needcited by 59,932
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  4. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,141
  5. 2022Scalable Diffusion Models with Transformerscited by 5,931
  6. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,581
  7. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,094
  8. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,550
  9. 2025Wan: Open and Advanced Large-Scale Video Generative Modelscited by 2,536
  10. 2024CogVideoX: Text-to-Video Diffusion Models with An Expert Transformercited by 2,198
  11. 2020pixelNeRF: Neural Radiance Fields from One or Few Imagescited by 1,574
  12. 2023Zero-1-to-3: Zero-shot One Image to 3D Objectcited by 1,483
33 more
  1. 2025VGGT: Visual Geometry Grounded Transformercited by 1,475
  2. 2025Qwen-Image Technical Reportcited by 973
  3. 2023LRM: Large Reconstruction Model for Single Image to 3Dcited by 834
  4. 2023pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstructioncited by 668
  5. 2023DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Visioncited by 514
  6. 2024CameraCtrl: Enabling Camera Control for Text-to-Video Generationcited by 448
  7. 2024CAT3D: Create Anything in 3D with Multi-View Diffusion Modelscited by 425
  8. 2024ViewCrafter: Taming Video Diffusion Models for High-fidelity Novel View Synthesiscited by 392
  9. 2019SynSin: End-to-end View Synthesis from a Single Imagecited by 364
  10. 2025GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Controlcited by 325
  11. 2023ReconFusion: 3D Reconstruction with Diffusion Priorscited by 317
  12. 2020Single-View View Synthesis with Multiplane Imagescited by 263
  13. 2023LucidDreamer: Domain-free Generation of 3D Gaussian Splatting Scenescited by 228
  14. 2025Stable Virtual Camera: Generative View Synthesis with Diffusion Modelscited by 179
  15. 2024LVSM: A Large View Synthesis Model with Minimal 3D Inductive Biascited by 178
  16. 2024CAT4D: Create Anything in 4D with Multi-View Video Diffusion Modelscited by 160
  17. 2024DimensionX: Create Any 3D and 4D Scenes from a Single Image with Controllable Video Diffusioncited by 139
  18. 2023Text2NeRF: Text-Driven 3D Scene Generation with Neural Radiance Fieldscited by 122
  19. 2024RealmDreamer: Text-Driven 3D Scene Generation with Inpainting and Depth Diffusioncited by 109
  20. 2021Geometry-Free View Synthesis: Transformers and no 3D Priorscited by 99
  21. 2024Wonderland: Navigating 3D Scenes from a Single Imagecited by 98
  22. 2021PixelSynth: Generating a 3D-Consistent Experience from a Single Imagecited by 82
  23. 2024NVS-Solver: Video Diffusion Model as Zero-Shot Novel View Synthesizercited by 79
  24. 2024GenWarp: Single Image to Novel Views with Semantic-Preserving Generative Warpingcited by 72
  25. 2024You See it, You Got it: Learning 3D Creation on Pose-Free Videos at Scalecited by 65
  26. 2022Single-View View Synthesis in the Wild with Learned Adaptive Multiplane Imagescited by 59
  27. 2025FlexWorld: Progressively Expanding 3D Scenes for Flexiable-View Synthesiscited by 37
  28. 2024MVGenMaster: Scaling Multi-View Generation from Any Image via 3D Priors Enhanced Diffusion Modelcited by 31
  29. 2025SpatialCrafter: Unleashing the Imagination of Video Diffusion Models for Scene Reconstruction from Limited Observationscited by 16
  30. 2025Taming Video Models for 3D and 4D Generation via Zero-Shot Camera Controlcited by 14
  31. 2022Bridging Implicit and Explicit Geometric Transformation for Single-Image View Synthesiscited by 11
  32. 2025Vid-CamEdit: Video Camera Trajectory Editing with Generative Rendering from Estimated Geometrycited by 7
  33. 2025CausNVS: Autoregressive Multi-view Diffusion for Flexible 3D Novel View Synthesiscited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 18.

20 papers.

Sort:

By area:

  1. 2026MosaicMem: Hybrid Spatial Memory for Controllable Video World Modelscited by 23
  2. 2025ReDirector: Creating Any-Length Video Retakes with Rotary Camera Encodingcited by 10
  3. 2026RayRoPE: Projective Ray Positional Encoding for Multi-view Attentioncited by 10
  4. 2026Matrix-Game 3.5: Enhancing Real-Time Streaming Interactive World Models with Patch Memorycited by 7
  5. 2026OpenWorldLib: A Unified Codebase and Definition of Advanced World Modelscited by 3
  6. 2026SCoPE: Sightline-Coordinate Positional Encoding for Video Diffusion Transformerscited by 3
  7. 2025ContextDrag: Precise Drag-Based Image Editing via Context-Preserving Token Injection and Position-Aligned Attentioncited by 1
  8. 2026GSCompleter: A Distillation-Free Plugin for Metric-Aware 3D Gaussian Splatting Completion in Secondscited by 1
  9. 2026Geometry without Position? When Positional Embeddings Help and Hurt Spatial Reasoningcited by 0
  10. 2026One View Is Enough: In-the-Wild Monocular Pretraining for Novel View Generationcited by 0
  11. 2026Token Warping Helps MLLMs Look from Nearby Viewpointscited by 0
  12. 2026UniGeo: Unifying Geometric Guidance for Camera-Controllable Image Editing via Video Modelscited by 0
8 more
  1. 2026Pose-ICL: 3D-Aware In-Context Learning for Pose-Controllable Subject Customizationcited by 0
  2. 2026RoPEMover: Depth-Aware Object Relocation via Positional Embeddingscited by 0
  3. 2026MetaView: Monocular Novel View Synthesis with Scale-Aware Implicit Geometry Priorscited by 0
  4. 2026HRDiT: Training-Free High-Resolution Image Generation with Off-the-Shelf Diffusion Transformer Modelscited by 0
  5. 2026ReX-Shot: Single-Image Rephotography via Geometry- and Camera-Grounded Generationcited by 0
  6. 2026DirectUV: Image-Conditioned UV Texture Generation with Surface-Aware Positional Encodingcited by 0
  7. 2026Geometry as Address: Routing Attention to Visual Memory for Long-Horizon Camera-Controlled Video Generationcited by 0
  8. 2026Texture Space Material Diffusioncited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026PE-Field 4D: Video Generation Models as Canvas
  2. 2025Stitch: Training-Free Position Control in Multimodal Diffusion Transformers
  3. 2026Registers Matter for Pixel-Space Diffusion Transformers
  4. 2025ResDiT: Evoking the Intrinsic Resolution Scalability in Diffusion Transformers
  5. 2025Massive Activations are the Key to Local Detail Synthesis in Diffusion Transformers
  6. 2026Appearance Pointers -- Multimodal Region Control of Diffusion Transformers
  7. 2025Unified Camera Positional Encoding for Controlled Video Generation
  8. 2024DiT4Edit: Diffusion Transformer for Image Editing