pith. sign in

arxiv: 2606.31946 · v1 · pith:ABY24G6Ynew · submitted 2026-06-30 · 💻 cs.CV

World Narrative Model for Highly Controllable Video Generation: A Paradigm Shift from Pixel Sampling to Physical World Orchestration

Ye Chen , Xuanhong Chen , Yupeng Zhu , Liming Tan , Zhewen Wan , Yuxuan Xiong , Tielong Wang , Jinfan Liu
show 18 more authors
Wuze Zhang Xiongzhen Zhang Feifei Li Xianglin Luo Zhehan Zhao Zhifan Zhang Laisheng Kou Zhujing Liang Yugang Chen Muchun Chen Xu Miao Yijing Zhang Xiaojie Sheng Qiang Hu Jialiang Chen Weimin Zhang Wenjun Zhang Bingbing Ni
This is my paper

Pith reviewed 2026-07-01 05:28 UTC · model grok-4.3

classification 💻 cs.CV
keywords controllable video generation4D world representationphysical narrativeneural shadercollaborative agentspre-visualizationmultimodal control
0
0 comments X

The pith

The World Narrative Model decouples structured 4D physical narratives from pixel sampling to drive controllable video generation.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper claims existing video models lack controllability because they treat generation as direct pixel distribution sampling without modeling explicit instance-level 4D physical structure. It introduces the World Narrative Model where collaborative agents convert sparse inputs such as text, reference videos, and sketches into an editable world representation containing geometry, layouts, skeleton motions, trajectories, camera paths, and lighting. This representation serves as a deterministic blueprint that guides existing video foundation models to produce output footage. If correct, the approach replaces probabilistic trial-and-error with quantitative specification, allowing creators to set parameters in a filmmaking-aligned pipeline. The framework stays modular so individual components like the world model or agents can be refined independently.

Core claim

WNM replaces end-to-end black-box sampling with orchestrated 4D pre-visualization for media generation. Collaborative agents translate sparse multimodal inputs, including text, reference videos, and sketches, into a fully editable world representation with scene geometry, object layouts, character/animal skeleton motion, trajectories, camera motion, and lighting at quantitative, physically meaningful granularity. This representation acts as a deterministic structural blueprint that drives existing video foundation models, either frozen or lightly adapted, to render final footage, turning the base model into a faithful neural shader.

What carries the argument

The orchestrated 4D pre-visualization, which creates an explicit editable world representation that acts as a deterministic blueprint driving base video models as neural shaders.

If this is right

  • Creators can specify geometry, motion, camera parameters, and lighting in deterministic quantitative terms rather than through repeated random sampling.
  • The number of probabilistic generation attempts needed to achieve desired results decreases substantially.
  • The output videos follow creator-specified layout, motion, and cinematography more closely than direct sampling approaches.
  • The overall system supports automatic pre-visualization and human refinement steps that align with professional filmmaking workflows.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The modular separation could allow direct substitution of physics-based simulators into the world representation stage for improved physical accuracy.
  • Director consoles for refinement might lower the barrier for non-experts to produce videos with professional-level control.
  • Because the world representation is fully editable, downstream tasks such as consistent multi-shot sequences or interactive editing become feasible without regenerating from scratch.

Load-bearing premise

Collaborative agents can reliably translate sparse multimodal inputs into a fully editable world representation with quantitative, physically meaningful granularity.

What would settle it

Generate videos from a world representation with precisely specified object trajectories and camera paths, then measure whether output frames match those trajectories and paths within a small quantitative error bound.

Figures

Figures reproduced from arXiv: 2606.31946 by Bingbing Ni, Feifei Li, Jialiang Chen, Jinfan Liu, Laisheng Kou, Liming Tan, Muchun Chen, Qiang Hu, Tielong Wang, Weimin Zhang, Wenjun Zhang, Wuze Zhang, Xianglin Luo, Xiaojie Sheng, Xiongzhen Zhang, Xuanhong Chen, Xu Miao, Ye Chen, Yijing Zhang, Yugang Chen, Yupeng Zhu, Yuxuan Xiong, Zhehan Zhao, Zhewen Wan, Zhifan Zhang, Zhujing Liang.

Figure 1
Figure 1. Figure 1: Three key dimensions that indicate the controllability in video generation task, including [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Motivations of the proposed new video generation paradigm: from end-to-end generation towards two-phase task decoupling, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: An systematic overview of our proposed World Narrative Model. The model is based on a series of collaborating agentic workflows including: scene [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Module 1: scene layout generation agentic workflow. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Module 2: asset generation and placement agentic workflow. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Module 3: actor motion and trajectory generation agentic workflow. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Module 4: cinematography and lighting setting agentic workflow. [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: A diagram visualization of four director’s control panels, including scene, asset, motion and camera manipulations. [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visualization of six representative video generation results by using WNM as control. The upper rows illustrate the rendered video frames by Seedance [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
read the original abstract

The fundamental obstacle to industrial grade video generation is the lack of controllability: existing models treat video as a pixel distribution sampling problem, bypassing the explicit, instance level $4D$ $(3D + T)$ physical world. Consequently, content creators cannot specify geometry, motion, camera parameters, or lighting in a deterministic, quantitative way, leading to the infamous ''gacha'' loop that makes professional content creation prohibitively inefficient and expensive. To address this, we introduce the World Narrative Model (WNM), a paradigm that decouples what to render -- the structured physical narrative -- from how to render -- the pixel generation process. WNM replaces end-to-end black-box sampling with orchestrated $4D$ pre-visualization for media generation. Collaborative agents translate sparse multimodal inputs, including text, reference videos, and sketches, into a fully editable world representation with scene geometry, object layouts, character/animal skeleton motion, trajectories, camera motion, and lighting at quantitative, physically meaningful granularity. This representation acts as a deterministic structural blueprint that drives existing video foundation models, either frozen or lightly adapted, to render final footage, turning the base model into a faithful neural shader. Built on this engine, our human-AI platform supports automatic world generation and pre-visualization aligned with professional filmmaking pipelines, while director consoles enable seamless human refinement. Experiments show that WNM greatly reduces probabilistic ``gacha'' calls and produces videos whose layout, motion, and cinematography closely follow creator intent. The framework is open and modular, allowing each component, such as world representation, control agents, and adapters, to be independently improved. Project website: https://glassroom.sjtu.edu.cn/WNM/.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper claims that existing video generation models suffer from poor controllability due to treating video as pixel sampling, leading to inefficient 'gacha' iteration. It introduces the World Narrative Model (WNM) to decouple the structured 4D physical world narrative from pixel rendering: collaborative agents convert sparse multimodal inputs (text, videos, sketches) into a fully editable 4D representation (geometry, layouts, skeletons, trajectories, camera, lighting) at quantitative granularity; this blueprint then drives frozen or lightly adapted video foundation models as a 'neural shader'. The approach is said to enable professional filmmaking pipelines via a human-AI platform, with experiments purportedly showing reduced gacha calls and better intent following. The framework is presented as open and modular.

Significance. If the agent-driven 4D orchestration can be shown to deliver reliable quantitative representations, the decoupling could meaningfully advance controllable video synthesis by aligning generation with deterministic production workflows rather than probabilistic sampling. The explicit emphasis on modularity and openness (allowing independent refinement of world representation, agents, and adapters) is a constructive design choice that could support incremental progress.

major comments (2)
  1. [Abstract] Abstract (collaborative agents paragraph): the load-bearing claim that agents translate sparse inputs into a 4D blueprint 'with scene geometry, object layouts, character/animal skeleton motion, trajectories, camera motion, and lighting at quantitative, physically meaningful granularity' supplies no algorithm, architecture, loss functions, training procedure, or completeness metric; without this, the asserted decoupling from end-to-end sampling and elimination of gacha loops cannot be evaluated.
  2. [Abstract] Abstract (final sentence before website): the assertion 'Experiments show that WNM greatly reduces probabilistic ``gacha'' calls and produces videos whose layout, motion, and cinematography closely follow creator intent' is unsupported by any experimental section, datasets, quantitative metrics, baselines, or ablation results in the manuscript.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed review and the opportunity to clarify our contribution. We respond point-by-point to the major comments below.

read point-by-point responses

  1. Referee: [Abstract] Abstract (collaborative agents paragraph): the load-bearing claim that agents translate sparse inputs into a 4D blueprint 'with scene geometry, object layouts, character/animal skeleton motion, trajectories, camera motion, and lighting at quantitative, physically meaningful granularity' supplies no algorithm, architecture, loss functions, training procedure, or completeness metric; without this, the asserted decoupling from end-to-end sampling and elimination of gacha loops cannot be evaluated.

    Authors: We acknowledge that the abstract presents a high-level description without specifying algorithms, architectures, loss functions, training procedures, or completeness metrics for the collaborative agents. The manuscript outlines the overall paradigm but does not provide these implementation details. We will revise the manuscript to include the requested technical specifications for the agent system that generates the quantitative 4D representation. revision: yes

  2. Referee: [Abstract] Abstract (final sentence before website): the assertion 'Experiments show that WNM greatly reduces probabilistic ``gacha'' calls and produces videos whose layout, motion, and cinematography closely follow creator intent' is unsupported by any experimental section, datasets, quantitative metrics, baselines, or ablation results in the manuscript.

    Authors: We agree that the current manuscript contains no experimental section, datasets, quantitative metrics, baselines, or ablations to support the claim. The statement refers to qualitative demonstrations on the project website. We will revise the abstract to remove or qualify the experimental assertion and, if appropriate, add a preliminary experimental section with supporting evidence in the revised manuscript. revision: yes

Circularity Check

0 steps flagged

No significant circularity; conceptual proposal without derivations or fittings

full rationale

The paper introduces a conceptual paradigm (WNM) that decouples world representation from pixel rendering via collaborative agents, but supplies no equations, parameter fittings, uniqueness theorems, or derivation steps. The abstract states the agents' translation capability as a given without exhibiting any reduction of outputs to inputs by construction. No self-citations, ansatzes, or renamings of known results appear in the load-bearing claims. This is a standard non-finding for a high-level architectural proposal lacking mathematical content.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 3 invented entities

The paper introduces several new conceptual entities and a core domain assumption without external validation or independent evidence in the provided abstract.

axioms (1)
  • domain assumption Existing video foundation models can be driven by a structural 4D blueprint to act as neural shaders.
    This assumption underpins the claim that the world representation turns base models into faithful renderers.
invented entities (3)
  • World Narrative Model (WNM) no independent evidence
    purpose: Decouples structured physical narrative from pixel generation process
    Core new paradigm introduced to solve controllability
  • Collaborative agents no independent evidence
    purpose: Translate sparse multimodal inputs into editable 4D world representation
    Mechanism for automatic world generation
  • 4D pre-visualization blueprint no independent evidence
    purpose: Serves as deterministic structural input to video models
    The orchestrated output that replaces black-box sampling

pith-pipeline@v0.9.1-grok · 5949 in / 1457 out tokens · 38103 ms · 2026-07-01T05:28:36.530977+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

34 extracted references · 15 canonical work pages · 12 internal anchors

  1. [1]

    Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Models

    Y . Liu, K. Zhang, Y . Li, Z. Yan, C. Gao, R. Chen, Z. Yuan, Y . Huang, H. Sun, J. Gao, L. He, and L. Sun, “Sora: A review on background, technology, limitations, and opportunities of large vision models,”arXiv preprint arXiv:2402.17177, 2024. [Online]. Available: https://arxiv.org/abs/2402.17177

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  2. [2]

    Kling-MotionControl technical report,

    Kling Team, J. Chen, Y . Ding, Z. Fang, K. Gai, K. He, X. He, J. Hua, M. Lao, X. Li, H. Liu, J. Liu, X. Liu, F. Shi, X. Shi, P. Sun, S. Tang, P. Wan, T. Wen, Z. Wu, H. Zhang, R. Zhao, Y . Zhang, and Y . Zhou, “Kling-MotionControl technical report,” arXiv preprint arXiv:2603.03160, Mar. 2026. [Online]. Available: https://arxiv.org/abs/2603.03160

    work page arXiv 2026

  3. [3]

    Seedance 1.0: Exploring the Boundaries of Video Generation Models

    Y . Gao, H. Guo, T. Hoang, W. Huang, L. Jiang, F. Kong, H. Li, J. Li, L. Li, X. Li, X. Li, Y . Li, S. Lin, Z. Lin, J. Liu, S. Liu, X. Nie, Z. Qing, Y . Ren, L. Sun, Z. Tian, R. Wang, S. Wang, G. Wei, G. Wu, J. Wu, R. Xia, F. Xiao, X. Xiao, J. Yan, C. Yang, J. Yang, R. Yang, T. Yang, Y . Yang, Z. Ye, X. Zeng, Y . Zeng, H. Zhang, Y . Zhao, X. Zheng, P. Zhu,...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  4. [4]

    Seedance 1.5 pro: A Native Audio-Visual Joint Generation Foundation Model

    Team Seedanceet al., “Seedance 1.5 pro: A native audio-visual joint generation foundation model,” 2025, seedance 1.5 pro Technical Report. [Online]. Available: https://arxiv.org/abs/2512.13507

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    Seedance 2.0: Advancing Video Generation for World Complexity

    ——, “Seedance 2.0: Advancing video generation for world complexity,” Apr. 2026, seedance 2.0 Model Card. [Online]. Available: https: //arxiv.org/abs/2604.14148

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  6. [6]

    Wan: Open and Advanced Large-Scale Video Generative Models

    Team Wan, A. Wang, B. Ai, B. Wen, C. Mao, C.-W. Xie, D. Chen, F. Yu, H. Zhao, J. Yang, J. Zeng, J. Wang, J. Zhang, J. Zhou, J. Wang, J. Chen, K. Zhu, K. Zhao, K. Yan, L. Huang, M. Feng, N. Zhang, P. Li, P. Wu, R. Chu, R. Feng, S. Zhang, S. Sun, T. Fang, T. Wang, T. Gui, T. Weng, T. Shen, W. Lin, W. Wang, W. Wang, W. Zhou, W. Wang, W. Shen, W. Yu, X. Shi, ...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  7. [7]

    TapNow: Your agentic creative canvas,

    TapNow, “TapNow: Your agentic creative canvas,” Official website, 2026, accessed: 2026-06-28. [Online]. Available: https://www.tapnow.ai/

    2026

  8. [8]

    Lovart: The world’s first ai design agent,

    Lovart, “Lovart: The world’s first ai design agent,” Official website, 2026, accessed: 2026-06-28. [Online]. Available: https://www.lovart.ai/

    2026

  9. [9]

    Seko: World-class ai video generation platform,

    SenseTime, “Seko: World-class ai video generation platform,” Official website, 2026, accessed: 2026-06-28. [Online]. Available: https: //seko.sensetime.com/

    2026

  10. [10]

    CapCut: AI-Powered Photo and Video Editor for Everyone,

    CapCut, “CapCut: AI-Powered Photo and Video Editor for Everyone,” https://www.capcut.com/, 2026, accessed: 2026-06-29

    2026

  11. [11]

    Nano AI: Your Personal Super Agent,

    360 Group, “Nano AI: Your Personal Super Agent,” https://www.n.cn/, 2026, accessed: 2026-06-29

    2026

  12. [12]

    LibTV: Professional video creation platform,

    LiblibAI, “LibTV: Professional video creation platform,” Official website, 2026, accessed: 2026-06-28. [Online]. Available: https: //www.liblib.tv/

    2026

  13. [13]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” inAdvances in Neural Information Processing Systems, vol. 30. Curran Associates, Inc., 2017, pp. 5998–6008. [Online]. Available: https://proceedings.neurips.cc/paper/7181-attention-is-all-you-need

    2017

  14. [14]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” inAdvances in Neural Information Processing Systems, vol. 33. Curran Associates, Inc., 2020, pp. 6840–6851. [Online]. Available: https://proceedings.neurips.cc/paper/2020/hash/ 4c5bcfec8584af0d967f1ab10179ca4b-Abstract.html

    2020

  15. [15]

    Self-supervised learning from images with a joint-embedding predictive architecture,

    M. Assran, Q. Duval, I. Misra, P. Bojanowski, P. Vincent, M. Rab- bat, Y . LeCun, and N. Ballas, “Self-supervised learning from images with a joint-embedding predictive architecture,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2023, pp. 15 619–15 629

    2023

  16. [16]

    V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning

    M. Assran, A. Bardes, D. Fan, Q. Garrido, R. Howes, M. Komeili, M. Muckley, A. Rizvi, C. Roberts, K. Sinha, A. Zholus, S. Arnaud, A. Gejji, A. Martin, F. R. Hogan, D. Dugas, P. Bojanowski, V . Khalidov, P. Labatut, F. Massa, M. Szafraniec, K. Krishnakumar, Y . Li, X. Ma, S. Chandar, F. Meier, Y . LeCun, M. Rabbat, and N. Ballas, “V-JEPA 2: Self-supervised...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  17. [17]

    Marble: A multimodal world model,

    World Labs, “Marble: A multimodal world model,” Official product announcement, Nov. 2025, published November 12, 2025; accessed: 2026-06-28. [Online]. Available: https://www.worldlabs.ai/ blog/marble-world-model

    2025

  18. [18]

    WorldGen: From text to traversable and interactive 3D worlds,

    D. Wang, H. Jung, T. Monnier, K. Sohn, C. Zou, X. Xiang, Y .-Y . Yeh, D. Liu, Z. Huang, T. Nguyen-Phuoc, Y . Fan, S. Oprea, Z. Wang, R. Shapovalov, N. Sarafianos, T. Groueix, A. Toisoul, P. Dhar, X. Chu, M. Chen, G. Y . Park, M. Gupta, Y . Azziz, R. Ranjan, and A. Vedaldi, “WorldGen: From text to traversable and interactive 3D worlds,”arXiv preprint arXiv...

    work page arXiv 2025

  19. [19]

    Cosmos World Foundation Model Platform for Physical AI

    NVIDIA, “Cosmos world foundation model platform for physical AI,”arXiv preprint arXiv:2501.03575, Jan. 2025. [Online]. Available: https://arxiv.org/abs/2501.03575

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  20. [20]

    Cosmos 3: Omnimodal World Models for Physical AI

    ——, “Cosmos 3: Omnimodal world models for physical AI,” arXiv preprint arXiv:2606.02800, Jun. 2026. [Online]. Available: https://arxiv.org/abs/2606.02800

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  21. [21]

    INSPATIO-WORLD: A Real-Time 4D World Simulator via Spatiotemporal Autoregressive Modeling

    InSpatio Team, D. Shen, G. Zhang, H. Liu, H. Ji, H. Bao, H. Zhai, J. Liu, J. Guo, N. Wang, S. Pan, W. Pan, W. Xie, X. Liu, X. Xiang, X. Zhang, X. Chen, Y . Wang, Y . Chen, Z. Fan, Z. Le, Z. Ye, and Z. Zhao, “INSPATIO-WORLD: A real-time 4D world simulator via spatiotemporal autoregressive modeling,”arXiv preprint arXiv:2604.07209, 2026. [Online]. Available...

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  22. [22]

    LoRA: Low-rank adaptation of large language models,

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-rank adaptation of large language models,” in International Conference on Learning Representations, 2022. [Online]. Available: https://openreview.net/forum?id=nZeVKeeFYf9

    2022

  23. [23]

    Adding conditional control to text-to-image diffusion models,

    L. Zhang, A. Rao, and M. Agrawala, “Adding conditional control to text-to-image diffusion models,” inProceedings of the IEEE/CVF International Conference on Computer Vision, October 2023, pp. 3836–3847. [Online]. Available: https://openaccess.thecvf. com/content/ICCV2023/html/Zhang Adding Conditional Control to Text-to-Image Diffusion Models ICCV 2023 paper.html

    2023

  24. [24]

    SAM 3D: 3dfy anything in images,

    SAM 3D Team, X. Chen, F.-J. Chu, P. Gleize, K. J. Liang, A. Sax, H. Tang, W. Wang, M. Guo, T. Hardin, X. Li, A. Lin, J. Liu, Z. Ma, A. Sagar, B. Song, X. Wang, J. Yang, B. Zhang, P. Doll ´ar, G. Gkioxari, M. Feiszli, and J. Malik, “SAM 3D: 3dfy anything in images,” Nov

  25. [25]

    SAM 3D: 3Dfy Anything in Images

    [Online]. Available: https://arxiv.org/abs/2511.16624

    work page internal anchor Pith review Pith/arXiv arXiv

  26. [26]

    arXiv preprint arXiv:2602.15989 (2026)

    X. Yang, D. Kukreja, D. Pinkus, A. Sagar, T. Fan, J. Park, S. Shin, J. Cao, J. Liu, N. Ugrinovic, M. Feiszli, J. Malik, P. Doll ´ar, and K. Kitani, “SAM 3D Body: Robust full-body human mesh recovery,” Feb. 2026. [Online]. Available: https://arxiv.org/abs/2602.15989

    work page arXiv 2026

  27. [27]

    VGGT: Visual geometry grounded transformer,

    J. Wang, M. Chen, N. Karaev, A. Vedaldi, C. Rupprecht, and D. Novotny, “VGGT: Visual geometry grounded transformer,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2025, pp. 5294–5306. [Online]. Available: https://openaccess.thecvf.com/content/CVPR2025/html/Wang VGGT Visual Geometry Grounded Transformer CV...

    2025

  28. [28]

    J. Wang, M. Chen, S. Zhang, N. Karaev, J. Sch ¨onberger, P. Labatut, P. Bojanowski, D. Novotny, A. Vedaldi, and C. Rupprecht, “VGGT-ω,” 18 inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2026, pp. 21 486–21 499. [Online]. Available: https://openaccess.thecvf.com/content/CVPR2026/ html/Wang VGGT-ohm CVPR 202...

    2026

  29. [29]

    Depth Anything 3: Recovering the Visual Space from Any Views

    H. Lin, S. Chen, J. Liew, D. Y . Chen, Z. Li, G. Shi, J. Feng, and B. Kang, “Depth Anything 3: Recovering the visual space from any views,” Nov. 2025. [Online]. Available: https://arxiv.org/abs/2511.10647

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  30. [30]

    Native and compact structured latents for 3D generation,

    J. Xiang, X. Chen, S. Xu, R. Wang, Z. Lv, Y . Deng, H. Zhu, Y . Dong, H. Zhao, N. J. Yuan, and J. Yang, “Native and compact structured latents for 3D generation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Jun. 2026, pp. 14 419–14 429. [Online]. Available: https://openaccess.thecvf. com/content/CVPR2026/htm...

    2026

  31. [31]

    Autoregressive image generation using residual quantization,

    D. Lee, C. Kim, S. Kim, M. Cho, and W.-S. Han, “Autoregressive image generation using residual quantization,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Jun. 2022, pp. 11 523–11 532

    2022

  32. [32]

    SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features

    M. Tschannen, A. Gritsenko, X. Wang, M. F. Naeem, I. Alabdulmohsin, N. Parthasarathy, T. Evans, L. Beyer, Y . Xia, B. Mustafa, O. H ´enaff, J. Harmsen, A. Steiner, and X. Zhai, “SigLIP 2: Multilingual vision- language encoders with improved semantic understanding, localization, and dense features,”arXiv preprint arXiv:2502.14786, 2025. [Online]. Available...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  33. [33]

    Introducing Codex,

    OpenAI, “Introducing Codex,” https://openai.com/index/ introducing-codex/, May 2025

    2025

  34. [34]

    GPT-5.5 System Card,

    ——, “GPT-5.5 System Card,” OpenAI, Tech. Rep., Apr. 2026. [Online]. Available: https://openai.com/index/gpt-5-5-system-card/

    2026