Pith. sign in

REVIEW 4 major objections 6 minor 111 references

WorldClaw: Agentic 3D Open-World Generation at Scale

T0 review · 4 major / 6 minor · reviewed 2026-08-08 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read WorldClaw claims that a text prompt can be turned into a large, freely explorable 3D world by building global terrain first, then generating and placing region-by-region objects as editable meshes.

desk verdict WorldClaw is a well-described integration of existing components into a global-to-regional pipeline for editable 3D worlds, but the central coherence claim rests on an unmeasured assumption about image-composition fidelity and the paper has no quantitative evaluation. read the letter →

arxiv 2608.05248 v1 pith:VXDGSX3G submitted 2026-08-05 cs.AI cs.CV

Chunchao Guo , Jinpeng Li , Yang Li , Zilong Huang This is my paper · ORCID
classification cs.AIcs.CV
keywords 3Dworldgenerationopen-worldtext-to-3Dagenticpipelinecoarse-to-fineproceduralterrainimage-to-3Dreconstructionsceneediting
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

WorldClaw claims that a single open-ended text prompt can drive the construction of a large, freely explorable 3D world without generating the whole scene at once. Its three-stage pipeline first turns the prompt into a structured plan, then builds a continuous region-aware terrain that anchors world scale and layout, and finally populates only selected regions with individually reconstructed objects. The bet is that global coherence can be established once and then preserved while local detail is added through render-based refinement. If the claim holds, text-to-3D generation stops being a single monolithic model and becomes an editable, reusable asset pipeline compatible with game-engine workflows.

What carries the argument

The load-bearing object is the global-to-regional decomposition plus the terrain-conditioned 2D composition as layout prior. WorldClaw deliberately does not generate the world in one pass; instead it creates a global terrain foundation from a semantic layout map and a region-weighted height field, then for each selected region renders that terrain into a 2D image, has an image-editing model insert objects into the render, segments the instances, reconstructs each as a mesh, and uses shared camera geometry to place each object with scale factor $s_i = Z_t f^o_i / (Z_o \hat{f}_i)$ and a translation that maps the object reference point onto a terrain anchor. That placement equation, together with the later render-based agentic refinement of contacts, is what translates a 2D editing operation into coherent 3D layout.

What would settle it

Render a synthetic terrain seeded with objects of known 3D size and position, run WorldClaw's region-composition step on that render, then compare every reconstructed object's scale, orientation, and ground contact against the known values; if the median 3D placement error is large, or the image-editing model visibly alters terrain contours that were already rendered, the central claim is not supported.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a coarse-to-fine agentic decomposition can produce explicit 3D worlds from open-ended prompts: the system first extracts explicit constraints and planning agents complete them into a structured scene specification; a terrain generation stage builds a height field from a semantic layout map, region-specific noise and geomorphic operators, assigned materials, and scattered reusable terrain assets; a regional stage renders each selected region's terrain, asks an image-editing model to compose objects into that render, segments the inserted instances, reconstructs each as a textured mesh, and recovers its placement with a ray-based scale-and-position formula that maps the object camera onto the terrain camera. The output is a composition of independently editable meshes with explicit terrain placements, which is what makes the scene editable and reusable rather than a fixed render.

Load-bearing premise

Everything depends on the image-editing model producing a 2D view of the region in which object sizes, overlaps, and the terrain underneath are faithful enough that later mesh reconstruction and ray-based placement can recover true 3D positions; if that image lies about geometry or scale, no downstream correction can fix it.

Editorial extensions

If this is right

  • If WorldClaw works as described, a single text prompt can yield an explicit 3D world whose terrain and objects remain independently editable meshes, not a baked render.
  • Because object generation is regional and condition-based, only detail-demanding regions pay the cost of image-to-3D reconstruction; open areas stay cheap procedural terrain.
  • The resulting worlds support free-viewpoint exploration, asset reuse, and standard game-engine workflows, which is the step from image-like artifact to persistent environment.
  • The agentic refinement loop is what lets the pipeline correct scale, pose, and object–terrain contact issues without regenerating the whole scene.
  • The output representation makes iterative creation possible: edit one asset or one terrain patch and re-run refinement instead of restarting generation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the 2D composition step is reliable, WorldClaw's design implies an interactive workflow the paper does not test: a user could edit the composition image directly and have the updated 3D layout recalculated from the same placement equations, making paint-to-place a plausible interaction mode.
  • The separation of global terrain from regional objects suggests the pipeline could become a data engine for simulation training, since each generated world comes with instance masks, depth, normals, and explicit transforms for free.
  • The dependence on bounding-box scale calibration points to a testable bottleneck: scenes with heavy occlusion or objects that partially leave the frame should produce systematic size errors, and the paper does not yet quantify this.
  • Because the paper's limitations describe foundation-model quality as the main constraint, WorldClaw's architecture has a scaling path: replacing any single underlying model with a better one improves the whole world without redesigning the pipeline.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. WorldClaw is an agentic, coarse-to-fine pipeline for generating explicit 3D open-world scenes from open-ended text prompts. It first plans a structured scene specification, constructs a region-aware procedural terrain, and then selectively populates regions by rendering the terrain, composing objects through an image-editing model, reconstructing them as meshes with SAM3D, and recovering their placements with ray-based perspective equations. A render-based agentic loop refines object quality, scale, pose, and object–terrain contact. The paper claims that, across diverse prompts, WorldClaw produces large-scale scenes with coherent spatial organization, visually rich local content, and editable instance-level assets while preserving a consistent global terrain structure. The evaluation consists of curated qualitative results, walk-through renderings, and qualitative comparisons against SynCity, Marble, MajutsuCity, WorldGen, GPT-5.6 Sol, and WorldClaw.

Significance. If the central claim were quantitatively substantiated, WorldClaw would be a practically valuable contribution: it combines global terrain generation with instance-level editable assets, and its placement equations in Section 2.3.2 are internally coherent and reusable. The paper is also commendably detailed about the pipeline's components and candid about its limitations in Section 5. However, the claim is currently supported only by selected qualitative examples. There are no numerical metrics, no user study, no ablations, no failure-rate analysis, and no release of code or data. The same Section 5 concedes that image-generation models 'frequently failed to produce usable semantic layout maps or to preserve object appearance and pose,' which makes the absence of systematic evaluation particularly load-bearing. The contribution is therefore plausible but not yet established to the standard of the paper's central assertion.

major comments (4)
  1. [Section 3, Figs. 4–8] The paper's central claim—coherent spatial organization, visual quality, and preserved global terrain structure across diverse prompts—is supported only by curated qualitative examples. Section 3 contains no quantitative metrics for layout coherence, no pose-error measurements, no contact-ratio statistics, no success/failure rates, and no user study. This is especially consequential because Section 5 states that underlying image-generation models 'frequently failed to produce usable semantic layout maps or to preserve object appearance and pose,' so the displayed scenes may not be representative. I ask for at least per-prompt success rates, placement error statistics, contact/footprint metrics, and an independent or user-based preference evaluation before the central claim can be accepted.
  2. [Section 2.3.2, Eqs. (9)–(13)] The placement chain assumes that the GPT-Image-2 composition I_r^comp preserves the terrain render's camera, local terrain geometry, and object scale, so that SAM3D segmentation and the ray-based placement recover correct 3D poses. No experiment checks this premise. For example, if the image model shifts an object's ground contact, changes slope under an object, or draws an object at a scale inconsistent with the terrain, the anchor ray and depth Z_t in Eq. (12) are computed against incorrect 2D evidence and the placement inherits the error. A concrete remedy is a re-render fidelity check: project the placed objects and terrain according to Eq. (13), re-render the region from the same camera, and measure alignment error, scale error, and contact error, reporting failure rates. Without such a check, the central coherence claim is not supported.
  3. [Section 2.3.2, Eq. (12)] The scale transfer formula relies on 'approximately square pixels and locally uniform perspective scaling.' Neither condition is validated for the actual terrain render, and the reference point P_o may be off-center where perspective distortion is largest. Single-view depth errors from SAM3D can also propagate directly into Z_o and thus into s_i. The paper should provide a sensitivity analysis or an ablation showing that placement accuracy is robust to these assumptions, particularly for sloped terrain and wide-angle views.
  4. [Section 3.3, Fig. 8] The qualitative comparison claims superiority over six baselines in terrain expressiveness, content richness, prompt alignment, and free-viewpoint appearance. However, the comparison uses only four walk views per method, the prompts are 'adapted when necessary' without listing the exact wording, and there is no standardized metric or independent rating protocol. The absence of exact prompts and per-method generation settings makes the comparison non-reproducible. Please include the exact input prompts, generation settings, and either independent human ratings or quantitative proxy metrics (e.g., number of editable assets, region-boundary consistency, terrain elevation range, per-prompt completion rate).
minor comments (6)
  1. [Section 2.2.2, p. 7] The sentence 'Both I_layout and I_asset are generated by GPT-Image-2 [62] Scene Layout Map.' lacks a clear sentence break and a period after the citation; it should be reformulated as two separate statements.
  2. [Section 2.3.3, Fig. 3] The notion of a 'report queue' and the contents of the 'status reports' are not formally specified; a brief definition of the report schema would make the refinement loop easier to reproduce.
  3. [Section 3.3] The comparison prompts are said to be 'adapted when necessary' to each method's input format, but the exact adaptations are not shown; this makes fairness difficult to assess and should be documented in an appendix.
  4. [Section 5] The limitations paragraph states that open-source models 'often struggled' and 'frequently failed,' but no counts or rates are given; reporting the observed failure frequencies across prompts would strengthen the discussion.
  5. [References] References [1] and [64] are informal web sources; please cite archival versions if they exist, and check that all preprints have stable identifiers.
  6. [Notation, Section 2.3.2] The notation b f_i and f^o_i is visually confusable; using a single font or subscript convention would improve readability, especially in Eqs. (11)–(13).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: WorldClaw is a construction pipeline whose placement equations are projective-geometry identities, and its self-citations are not load-bearing.

full rationale

WorldClaw is an engineering/system paper that makes no quantitative predictive claims; its evidence is qualitative scene renderings. The core derivation chain is a construction pipeline: planning agents produce structured specifications (Eq. 3), terrain is generated from a height field combining masks, noise, and geomorphic operators (Eq. 6), and regional objects are produced by image composition, segmentation, reconstruction, and placement. The placement mathematics (Eqs. 9-13) are projective-geometry identities: Eq. 10 states the cropped intrinsics are the affine transform of the terrain intrinsics, Eq. 11 is a scale-calibration loop that forces the projected bounding box to match the segmented image footprint, and Eqs. 12-13 compute the placement scale and transform from equal apparent size and ray-anchor correspondence. These equations enforce consistency between the 2D composition and 3D placement by construction; they are not fitted parameters used to predict a held-out quantity. No parameter is fitted to a subset of results and then reported as a prediction; no claim is justified by a self-citation chain; no uniqueness theorem is imported from the authors' prior work. The paper uses Hunyuan3D, GPT-Image-2, and SAM3D as external tools, and its Limitations section (Sec. 5) explicitly acknowledges the pipeline's dependence on their generalization, which is an honest external-dependency statement rather than a circular one. The absence of a quantitative fidelity check for the image-composition premise is a correctness/robustness gap, not circularity. Related-work self-citations to Hunyuan-ecosystem papers are tool and context citations and do not carry the central argument. Accordingly, no circular steps are identified.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim rests on procedural tuning choices and strong assumptions about the reliability of proprietary foundation models. The paper itself acknowledges this dependence in Section 5. No new physical entities or theoretical constructs are introduced.

free parameters (3)
  • Region terrain parameters (h_r, w_{r,k}, alpha_{r,j}) = LLM-chosen, no values reported
    In Eq. 6 these control base elevation, noise, and geomorphic operator weights for each region. They are set per prompt by the terrain planning agent, not learned or reported, so the height-field output is not reproducible from the paper alone.
  • Asymmetric calibration tolerances (epsilon_i^-, epsilon_i^+) = Unspecified
    Eq. 11 uses these tolerances to suppress oversized reconstructions. They are chosen by hand and no values are given, which affects placement quality but not the core scientific claim.
  • Asset scattering densities in p_asset = LLM-chosen, no values reported
    Set by the terrain planning agent per region and used in global terrain asset scattering. They are not reported, making content density non-reproducible.
assumptions (5)
  • standard math Pinhole camera model and locally uniform perspective scaling
    Invoked in Eqs. 10 through 13 for placement. This is a standard idealization but may fail for wide-angle terrain renders or strong perspective distortion.
  • domain assumption Terrain is representable as a single-valued height field
    Eq. 6 defines H(x) only as a height over a 2D domain, so overhangs, caves, and vertical cliffs cannot be represented. This limits the scope of the claimed world coherence.
  • ad hoc to paper GPT-Image-2 produces region-composition images that preserve terrain geometry and object scale
    The entire regional population stage in Section 2.3.2 relies on this generative model behaving as a faithful scene composer. No quantitative validation is provided, and the limitations section admits open-source image models often fail in this role.
  • domain assumption SAM3D single-view reconstruction is scale-consistent after calibration
    The scale calibration in Eq. 11 assumes that bounding-box projection ratio is sufficient to correct reconstruction scale. Single-view ambiguity may still produce pose or geometry errors that placement cannot fix.
  • ad hoc to paper Render-based agent checks provide reliable visual judgment
    The refinement loops in Section 2.2.3 and Section 2.3.3 rely on the agent correctly detecting artifacts such as floating, penetration, and scale mismatch. The paper provides no measure of detection accuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of WorldClaw: Agentic 3D Open-World Generation at Scale." pith.science (2026). https://pith.science/paper/VXDGSX3G

@misc{pith2026260805248,
  author       = {Pith},
  title        = {Pith review of: WorldClaw: Agentic 3D Open-World Generation at Scale},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VXDGSX3G}},
  note         = {Machine review of arXiv:2608.05248}
}
read the original abstract

Generating large-scale, freely explorable 3D worlds from open-ended text remains challenging because a system must jointly maintain global spatial coherence, rich local content, and explicit assets suitable for downstream editing and reuse. We present WorldClaw, a fully agentic, coarse-to-fine framework for open-world 3D scene generation. Planning agents translate a text prompt into a structured specification of regions, terrain, assets, materials, and spatial relations. WorldClaw then builds a globally coherent terrain foundation from semantic layouts, reusable assets, generative or procedural materials, and a region-aware height field. For detail-demanding regions, it generates terrain-conditioned compositions, reconstructs editable textured meshes, and recovers their placement on the terrain; render-based agents further refine terrain, objects, appearance, and contacts. Across diverse open-world prompts, WorldClaw produces large-scale scenes with coherent spatial organization, visually compelling local content, and editable instance-level assets while preserving a consistent global terrain structure.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

111 extracted references · 39 canonical work pages

  1. [1]

    Blender-MCP: Open-source MCP to use Blender with any LLM.https://github.com/ahujasid/ blender-mcp, 2025

    ahujasid. Blender-MCP: Open-source MCP to use Blender with any LLM.https://github.com/ahujasid/ blender-mcp, 2025

    2025

  2. [2]

    Claude Opus 4.8.https://www.anthropic.com/news/claude-opus-4-8, 2026

    Anthropic. Claude Opus 4.8.https://www.anthropic.com/news/claude-opus-4-8, 2026

    2026

  3. [3]

    Holodeck 2.0: Vision-language-guided 3d world generation with editing.arXiv preprint arXiv:2508.05899, 2025

    Zixuan Bian, Ruohan Ren, Yue Yang, and Chris Callison-Burch. Holodeck 2.0: Vision-language-guided 3d world generation with editing.arXiv preprint arXiv:2508.05899, 2025

    arXiv 2025

  4. [4]

    Scenefactor: Factored latent 3d diffusion for controllable 3d scene generation

    Aleksey Bokhovkin, Quan Meng, Shubham Tulsiani, and Angela Dai. Scenefactor: Factored latent 3d diffusion for controllable 3d scene generation. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 628–639, 2025

    2025

  5. [5]

    Diffdreamer: Towards consistent unsupervised single-view scene extrapolation with conditional diffusion models

    Shengqu Cai, Eric Ryan Chan, Songyou Peng, Mohamad Shahbazi, Anton Obukhov, Luc Van Gool, and Gordon Wetzstein. Diffdreamer: Towards consistent unsupervised single-view scene extrapolation with conditional diffusion models. InProceedings of the IEEE/CVF international conference on computer vision, pages 2139–2150, 2023

    2023

  6. [6]

    Sam 3: Segment anything with concepts

    Nicolas Carion, Laura Gustafson, Yuan-Ting Hu, Shoubhik Debnath, Ronghang Hu, Didac Suris, Chaitanya Ryali, Kalyan Vasudev Alwala, Haitham Khedr, Andrew Huang, et al. Sam 3: Segment anything with concepts. arXiv preprint arXiv:2511.16719, 2025

    arXiv 2025

  7. [7]

    I-design: Personalized llm interior designer

    Ata Çelen, Guo Han, Konrad Schindler, Luc Van Gool, Iro Armeni, Anton Obukhov, and Xi Wang. I-design: Personalized llm interior designer. InEuropean Conference on Computer Vision, pages 217–234. Springer, 2024

    2024

  8. [8]

    Persistent nature: A generative model of unbounded 3d worlds

    Lucy Chai, Richard Tucker, Zhengqi Li, Phillip Isola, and Noah Snavely. Persistent nature: A generative model of unbounded 3d worlds. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 20863–20874, 2023

    2023

Show all 111 references
  1. [9]

    Mansion: Multi-floor language-to-3d scene generation for long-horizon tasks

    Lirong Che, Shuo Wen, Shan Huang, Chuang Wang, Yuzhe Yang, Gregory Dudek, Xueqian Wang, and Jian Su. Mansion: Multi-floor language-to-3d scene generation for long-horizon tasks. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 37175–37...

    2026

  2. [10]

    Trellisworld: Training-free world generation from object generators

    Hanke Chen, Yuan Liu, and Minchen Li. Trellisworld: Training-free world generation from object generators. arXiv preprint arXiv:2510.23880, 2025

    2025

  3. [11]

    Flexworld: Progressively expanding 3d scenes for flexible-view exploration.Advances in Neural Information Processing Systems, 38:172270–172299, 2026

    Luxi Chen, Zihan Zhou, Min Zhao, Yikai Wang, Ge Zhang, Wenhao Huang, Hao Sun, Ji-Rong Wen, and Chongxuan Li. Flexworld: Progressively expanding 3d scenes for flexible-view exploration.Advances in Neural Information Processing Systems, 38:172270–172299, 2026

    2026

  4. [12]

    Sam 3d: 3dfy anything in images

    Xingyu Chen, Fu-Jen Chu, Pierre Gleize, Kevin J Liang, Alexander Sax, Hao Tang, Weiyao Wang, Michelle Guo, Thibaut Hardin, Xiang Li, et al. Sam 3d: 3dfy anything in images. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7220–7232, 2026

    2026

  5. [13]

    Luciddreamer: Domain-free generation of 3d gaussian splatting scenes.arXiv preprint arXiv:2311.13384, 2023

    Jaeyoung Chung, Suyoung Lee, Hyeongjin Nam, Jaerin Lee, and Kyoung Mu Lee. Luciddreamer: Domain-free generation of 3d gaussian splatting scenes.arXiv preprint arXiv:2311.13384, 2023

    2023 arXiv

  6. [14]

    Citycraft: A real crafter for 3d city generation.arXiv preprint arXiv:2406.04983, 2024

    Jie Deng, Wenhao Chai, Junsheng Huang, Zhonghan Zhao, Qixuan Huang, Mingyan Gao, Jianshu Guo, Shengyu Hao, Wenhao Hu, Jenq-Neng Hwang, et al. Citycraft: A real crafter for 3d city generation.arXiv preprint arXiv:2406.04983, 2024

    2024 arXiv

  7. [15]

    Realistic modeling and rendering of plant ecosystems

    Oliver Deussen, Pat Hanrahan, Bernd Lintermann, Radomír Měch, Matt Pharr, and Przemyslaw Prusinkiewicz. Realistic modeling and rendering of plant ecosystems. InProceedings of the 25th annual conference on Computer graphics and interactive techniques, pages 275–286, 1998

    1998

  8. [16]

    Hiscene: creating hierarchical 3d scenes with isometric view generation

    Wenqi Dong, Bangbang Yang, Zesong Yang, Yuan Li, Tao Hu, Hujun Bao, Yuewen Ma, and Zhaopeng Cui. Hiscene: creating hierarchical 3d scenes with isometric view generation. InProceedings of the 33rd ACM International Conference on Multimedia, pages 9783–9792, 2025

    2025

  9. [17]

    Latticeworld: A multimodal large language model-empowered framework for interactive complex world generation.arXiv preprint arXiv:2509.05263, 2025

    Yinglin Duan, Zhengxia Zou, Tongwei Gu, Wei Jia, Zhan Zhao, Luyi Xu, Xinzhu Liu, Yenan Lin, Hao Jiang, Kang Chen, et al. Latticeworld: A multimodal large language model-empowered framework for interactive complex world generation.arXiv preprint arXiv:2509.05263, 2025. 25 Tence...

    2025 arXiv

  10. [18]

    Syncity: Training- free generation of 3d worlds

    Paul Engstler, Aleksandar Shtedritski, Iro Laina, Christian Rupprecht, and Andrea Vedaldi. Syncity: Training- free generation of 3d worlds. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 27585–27595, 2025

    2025

  11. [19]

    Gen-searcher: Reinforcing agentic search for image generation.arXiv preprint arXiv:2603.28767, 2026

    Kaituo Feng, Manyuan Zhang, Shuang Chen, Yunlong Lin, Kaixuan Fan, Yilei Jiang, Hongyu Li, Dian Zheng, Chenyang Wang, and Xiangyu Yue. Gen-searcher: Reinforcing agentic search for image generation.arXiv preprint arXiv:2603.28767, 2026

    2026 arXiv

  12. [20]

    Automatic scene generation: State-of-the-art techniques, models, datasets, challenges, and future prospects.IEEE Access, 2025

    Awal Ahmed Fime, Saifuddin Mahmud, Arpita Das, Md Sunzidul Islam, and Jong-Hoon Kim. Automatic scene generation: State-of-the-art techniques, models, datasets, challenges, and future prospects.IEEE Access, 2025

    2025

  13. [21]

    Scenescape: Text-driven consistent scene generation

    Rafail Fridman, Amit Abecasis, Yoni Kasten, and Tali Dekel. Scenescape: Text-driven consistent scene generation. Advances in Neural Information Processing Systems, 36:39897–39914, 2023

    2023

  14. [22]

    3dcodebench: Benchmarking agentic procedural 3d modeling via code.arXiv preprint arXiv:2606.01057, 2026

    Yipeng Gao, Lei Shu, Genzhi Ye, Xi Xiong, Ameesh Makadia, Meiqi Guo, Laurent Itti, and Jindong Chen. 3dcodebench: Benchmarking agentic procedural 3d modeling via code.arXiv preprint arXiv:2606.01057, 2026

    2026 arXiv

  15. [23]

    Terrain generation using procedural models based on hydrology.ACM Transactions on Graphics (TOG), 32(4):1–13, 2013

    Jean-David Génevaux, Éric Galin, Eric Guérin, Adrien Peytavie, and Bedrich Benes. Terrain generation using procedural models based on hydrology.ACM Transactions on Graphics (TOG), 32(4):1–13, 2013

    2013

  16. [24]

    Mind-brush: Integrating agentic cognitive search and reasoning into image generation.arXiv preprint arXiv:2602.01756, 2026

    Jun He, Junyan Ye, Zilong Huang, Dongzhi Jiang, Chenjue Zhang, Leqi Zhu, Renrui Zhang, Xiang Zhang, and Weijia Li. Mind-brush: Integrating agentic cognitive search and reasoning into image generation.arXiv preprint arXiv:2602.01756, 2026

    2026

  17. [25]

    Procedural content generation for games: A survey.ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), 9(1):1–22, 2013

    Mark Hendrikx, Sebastiaan Meijer, Joeri Van Der Velden, and Alexandru Iosup. Procedural content generation for games: A survey.ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), 9(1):1–22, 2013

    2013

  18. [26]

    Text2room: Extracting textured 3d meshes from 2d text-to-image models

    Lukas Höllein, Ang Cao, Andrew Owens, Justin Johnson, and Matthias Nießner. Text2room: Extracting textured 3d meshes from 2d text-to-image models. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 7909–7920, 2023

    2023

  19. [27]

    Large scene generation with cube-absorb discrete diffusion

    Qianjiang Hu and Wei Hu. Large scene generation with cube-absorb discrete diffusion. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 25186–25196, 2025

    2025

  20. [28]

    Scenecraft: An llm agent for synthesizing 3d scenes as blender code

    Ziniu Hu, Ahmet Iscen, Aashi Jain, Thomas Kipf, Yisong Yue, David A Ross, Cordelia Schmid, and Alireza Fathi. Scenecraft: An llm agent for synthesizing 3d scenes as blender code. InForty-first International Conference on Machine Learning, 2024

    2024

  21. [29]

    Midi: Multi-instance diffusion for single image to 3d scene generation

    Zehuan Huang, Yuan-Chen Guo, Xingqiao An, Yunhan Yang, Yangguang Li, Zi-Xin Zou, Ding Liang, Xihui Liu, Yan-Pei Cao, and Lu Sheng. Midi: Multi-instance diffusion for single image to 3d scene generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern R...

    2025

  22. [30]

    Scene4u: Hierarchical layered 3d scene reconstruction from single panoramic image for your immerse exploration

    Zilong Huang, Jun He, Junyan Ye, Lihan Jiang, Weijia Li, Yiping Chen, and Ting Han. Scene4u: Hierarchical layered 3d scene reconstruction from single panoramic image for your immerse exploration. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 26...

    2025

  23. [31]

    Majutsucity: Language-driven aesthetic-adaptive city generation with controllable 3d assets and layouts

    Zilong Huang, Jun He, Xiaobin Huang, Ziyi Xiong, Yang Luo, Junyan Ye, Weijia Li, Yiping Chen, and Ting Han. Majutsucity: Language-driven aesthetic-adaptive city generation with controllable 3d assets and layouts. InProceedings of the IEEE/CVF Conference on Computer Vision and ...

    2026

  24. [32]

    Hunyuan3d 2.1: From images to high-fidelity 3d assets with production-ready pbr material.arXiv preprint arXiv:2506.15442, 2025

    Team Hunyuan3D, Shuhui Yang, Mingxin Yang, Yifei Feng, Xin Huang, Sheng Zhang, Zebin He, Di Luo, Haolin Liu, Yunfei Zhao, et al. Hunyuan3d 2.1: From images to high-fidelity 3d assets with production-ready pbr material.arXiv preprint arXiv:2506.15442, 2025

    2025 arXiv

  25. [33]

    Hy-world 2.0: A multi-modal world model for reconstructing, generating, and simulating 3d worlds.arXiv preprint arXiv:2604.14268, 2026

    Team HY-World, Chenjie Cao, Xuhui Zuo, Zhenwei Wang, Yisu Zhang, Junta Wu, Zhenyang Liu, Yuning Gong, Yang Liu, Bo Yuan, et al. Hy-world 2.0: A multi-modal world model for reconstructing, generating, and simulating 3d worlds.arXiv preprint arXiv:2604.14268, 2026

    2026 arXiv

  26. [34]

    Procedural generation of articulated simulation-ready assets

    Abhishek Joshi. Procedural generation of articulated simulation-ready assets. Master’s thesis, Princeton University, 2026

    2026

  27. [35]

    Worldflow3d: Flowing through 3d distributions for unbounded world generation.arXiv preprint arXiv:2603.29089, 2026

    Amogh Joshi, Julian Ost, and Felix Heide. Worldflow3d: Flowing through 3d distributions for unbounded world generation.arXiv preprint arXiv:2603.29089, 2026. 26 Tencent Hunyuan

    2026

  28. [36]

    Diffindscene: Diffusion- based high-quality 3d indoor scene generation

    Xiaoliang Ju, Zhaoyang Huang, Yijin Li, Guofeng Zhang, Yu Qiao, and Hongsheng Li. Diffindscene: Diffusion- based high-quality 3d indoor scene generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4526–4535, 2024

    2024

  29. [37]

    Simworld studio: Automatic environment generation with evolving coding agent for embodied agent learning.arXiv preprint arXiv:2605.09423, 2026

    Haoqiang Kang, Xiaokang Ye, Yuhan Liu, Siddhant Hitesh Mantri, Lingjun Mao, James Fleming, Drishti Regmi, and Lianhui Qin. Simworld studio: Automatic environment generation with evolving coding agent for embodied agent learning.arXiv preprint arXiv:2605.09423, 2026

    2026 arXiv

  30. [38]

    Hunyuan3d 2.5: Towards high-fidelity 3d assets generation with ultimate details

    Zeqiang Lai, Yunfei Zhao, Haolin Liu, Zibo Zhao, Qingxiang Lin, Huiwen Shi, Xianghui Yang, Mingxin Yang, Shuhui Yang, Yifei Feng, et al. Hunyuan3d 2.5: Towards high-fidelity 3d assets generation with ultimate details. arXiv preprint arXiv:2506.16504, 2025

    2025 arXiv

  31. [39]

    Nuiscene: Exploring efficient generation of unbounded outdoor scenes

    Han-Hung Lee, Qinghong Han, and Angel X Chang. Nuiscene: Exploring efficient generation of unbounded outdoor scenes. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 26509– 26518, 2025

    2025

  32. [40]

    Nuiworld: Exploring a scalable framework for end-to-end controllable world generation.arXiv preprint arXiv:2601.19048, 2026

    Han-Hung Lee, Cheng-Yu Yang, Yu-Lun Liu, and Angel X Chang. Nuiworld: Exploring a scalable framework for end-to-end controllable world generation.arXiv preprint arXiv:2601.19048, 2026

    2026

  33. [41]

    Semcity: Semantic scene generation with triplane diffusion

    Jumin Lee, Sebin Lee, Changho Jo, Woobin Im, Juhyeong Seon, and Sung-Eui Yoon. Semcity: Semantic scene generation with triplane diffusion. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 28337–28347, 2024

    2024

  34. [42]

    Worldgrow: Generating infinite 3d world

    Sikuang Li, Chen Yang, Jiemin Fang, Taoran Yi, Jia Lu, Jiazhong Cen, Lingxi Xie, Wei Shen, and Qi Tian. Worldgrow: Generating infinite 3d world. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 6433–6441, 2026

    2026

  35. [43]

    Scene- dreamer360: Text-driven 3d-consistent scene generation with panoramic gaussian splatting.arXiv preprint arXiv:2408.13711, 2024

    Wenrui Li, Fucheng Cai, Yapeng Mi, Zhe Yang, Wangmeng Zuo, Xingtao Wang, and Xiaopeng Fan. Scene- dreamer360: Text-driven 3d-consistent scene generation with panoramic gaussian splatting.arXiv preprint arXiv:2408.13711, 2024

    2024 arXiv

  36. [44]

    Infinitenature-zero: Learning perpetual view generation of natural scenes from single images

    Zhengqi Li, Qianqian Wang, Noah Snavely, and Angjoo Kanazawa. Infinitenature-zero: Learning perpetual view generation of natural scenes from single images. InEuropean conference on computer vision, pages 515–534. Springer, 2022

    2022

  37. [45]

    Wonderland: Navigating 3d scenes from a single image

    Hanwen Liang, Junli Cao, Vidit Goel, Guocheng Qian, Sergei Korolev, Demetri Terzopoulos, Konstantinos N Plataniotis, Sergey Tulyakov, and Jian Ren. Wonderland: Navigating 3d scenes from a single image. In Proceedings of the Computer Vision and Pattern Recognition Conference, p...

    2025

  38. [46]

    Cymbadiff: Structured spatial diffusion for sketch-based 3d semantic urban scene generation.Advances in Neural Information Processing Systems, 38:81114–81138, 2026

    Li Liang, Bo Miao, Xinyu Wang, Naveed Akhtar, Jordan Vice, and Ajmal Mian. Cymbadiff: Structured spatial diffusion for sketch-based 3d semantic urban scene generation.Advances in Neural Information Processing Systems, 38:81114–81138, 2026

    2026

  39. [47]

    Infinicity: Infinite-scale city synthesis

    Chieh Hubert Lin, Hsin-Ying Lee, Willi Menapace, Menglei Chai, Aliaksandr Siarohin, Ming-Hsuan Yang, and Sergey Tulyakov. Infinicity: Infinite-scale city synthesis. InProceedings of the IEEE/CVF international conference on computer vision, pages 22808–22818, 2023

    2023

  40. [48]

    Scenethesis: A language and vision agentic framework for 3d scene generation.arXiv preprint arXiv:2505.02836, 2025

    Lu Ling, Chen-Hsuan Lin, Tsung-Yi Lin, Yifan Ding, Yu Zeng, Yichen Sheng, Yunhao Ge, Ming-Yu Liu, Aniket Bera, and Zhaoshuo Li. Scenethesis: A language and vision agentic framework for 3d scene generation.arXiv preprint arXiv:2505.02836, 2025

    2025 arXiv

  41. [49]

    Infinite nature: Perpetual view generation of natural scenes from a single image

    Andrew Liu, Richard Tucker, Varun Jampani, Ameesh Makadia, Noah Snavely, and Angjoo Kanazawa. Infinite nature: Perpetual view generation of natural scenes from a single image. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 14458–14467, 2021

    2021

  42. [50]

    Simworlds: A multi-agent system for dynamic 3d scene creation.arXiv preprint arXiv:2607.01766, 2026

    Chunjiang Liu, Xiaoyuan Wang, Haoyu Chen, Yizhou Zhao, Ming-Hsuan Yang, and László A Jeni. Simworlds: A multi-agent system for dynamic 3d scene creation.arXiv preprint arXiv:2607.01766, 2026

    2026 arXiv

  43. [51]

    Agentic 3d scene generation with spatially contextualized vlms.arXiv preprint arXiv:2505.20129, 2025

    Xinhang Liu, Yu-Wing Tai, and Chi-Keung Tang. Agentic 3d scene generation with spatially contextualized vlms.arXiv preprint arXiv:2505.20129, 2025

    2025 arXiv

  44. [52]

    Worldmirror: Universal 3d world reconstruction with any-prior prompting.arXiv preprint arXiv:2510.10726, 2025

    Yifan Liu, Zhiyuan Min, Zhenwei Wang, Junta Wu, Tengfei Wang, Yixuan Yuan, Yawei Luo, and Chunchao Guo. Worldmirror: Universal 3d world reconstruction with any-prior prompting.arXiv preprint arXiv:2510.10726, 2025. 27 Tencent Hunyuan

    2025 arXiv

  45. [53]

    Landcraft: Designing the structured 3d landscapes via text guidance

    Zhihao Liu, Fang Liu, Weihao Xuan, and Naoto Yokoya. Landcraft: Designing the structured 3d landscapes via text guidance. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 7467–7475, 2026

    2026

  46. [54]

    Yo’city: Personalized and boundless 3d realistic city scene generation via self-critic expansion

    Keyang Lu, Sifan Zhou, Hongbin Xu, Gang Xu, Zhifei Yang, Yikai Wang, Zhen Xiao, Jieyi Long, and Ming Li. Yo’city: Personalized and boundless 3d realistic city scene generation via self-critic expansion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Re...

    2026

  47. [55]

    Ll3m: Large language 3d modelers.arXiv preprint arXiv:2508.08228, 2025

    Sining Lu, Guan Chen, Nam Anh Dinh, Itai Lang, Ari Holtzman, and Rana Hanocka. Ll3m: Large language 3d modelers.arXiv preprint arXiv:2508.08228, 2025

    2025 arXiv

  48. [56]

    Fastscene: Text-driven fast 3d indoor scene generation via panoramic gaussian splatting.arXiv preprint arXiv:2405.05768, 2024

    Yikun Ma, Dandan Zhan, and Zhi Jin. Fastscene: Text-driven fast 3d indoor scene generation via panoramic gaussian splatting.arXiv preprint arXiv:2405.05768, 2024

    2024 arXiv

  49. [57]

    Lt3sd: Latent trees for 3d scene diffusion

    Quan Meng, Lei Li, Matthias Nießner, and Angela Dai. Lt3sd: Latent trees for 3d scene diffusion. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 650–660, 2025

    2025

  50. [58]

    Procedural modeling of buildings

    Pascal Müller, Peter Wonka, Simon Haegler, Andreas Ulmer, and Luc Van Gool. Procedural modeling of buildings. InACM SIGGRAPH 2006 Papers, pages 614–623. 2006

    2006

  51. [59]

    The synthesis and rendering of eroded fractal terrains

    F Kenton Musgrave, Craig E Kolb, and Robert S Mace. The synthesis and rendering of eroded fractal terrains. ACM Siggraph Computer Graphics, 23(3):41–50, 1989

    1989

  52. [60]

    Wonderturbo: Generating interactive 3d world in 0.72 seconds

    Chaojun Ni, Xiaofeng Wang, Zheng Zhu, Weijie Wang, Haoyun Li, Guosheng Zhao, Jie Li, Wenkang Qin, Guan Huang, and Wenjun Mei. Wonderturbo: Generating interactive 3d world in 0.72 seconds. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 27423–27...

    2025

  53. [61]

    GPT-5.6 Sol.https://developers.openai.com/api/docs/models/gpt-5.6-sol, 2026

    OpenAI. GPT-5.6 Sol.https://developers.openai.com/api/docs/models/gpt-5.6-sol, 2026

    2026

  54. [62]

    GPT-Image-2.https://developers.openai.com/api/docs/models/gpt-image-2, 2026

    OpenAI. GPT-Image-2.https://developers.openai.com/api/docs/models/gpt-image-2, 2026

    2026

  55. [63]

    Procedural modeling of cities

    Yoav IH Parish and Pascal Müller. Procedural modeling of cities. InProceedings of the 28th annual conference on Computer graphics and interactive techniques, pages 301–308, 2001

    2001

  56. [64]

    Amit J. Patel. Making maps with noise functions, 2015. URL https://www.redblobgames.com/maps/ terrain-from-noise/

    2015

  57. [65]

    Scenesmith: Agentic generation of simulation-ready indoor scenes.arXiv preprint arXiv:2602.09153, 2026

    Nicholas Pfaff, Thomas Cohn, Sergey Zakharov, Rick Cory, and Russ Tedrake. Scenesmith: Agentic generation of simulation-ready indoor scenes.arXiv preprint arXiv:2602.09153, 2026

    2026 arXiv

  58. [66]

    Hsm: Hierarchical scene motifs for multi-scale indoor scene generation

    Hou In Derek Pun, Hou In Ivan Tam, Austin T Wang, Xiaoliang Huo, Angel X Chang, and Manolis Savva. Hsm: Hierarchical scene motifs for multi-scale indoor scene generation. In2026 International Conference on 3D Vision (3DV), pages 1356–1367. IEEE, 2026

    2026

  59. [67]

    Infinite photorealistic worlds using procedural generation

    Alexander Raistrick, Lahav Lipson, Zeyu Ma, Lingjie Mei, Mingzhe Wang, Yiming Zuo, Karhan Kayan, Hongyu Wen, Beining Han, Yihan Wang, et al. Infinite photorealistic worlds using procedural generation. InProceedings of the IEEE/CVF conference on computer vision and pattern reco...

    2023

  60. [68]

    Infinigen indoors: Photorealistic indoor scenes using procedural generation

    Alexander Raistrick, Lingjie Mei, Karhan Kayan, David Yan, Yiming Zuo, Beining Han, Hongyu Wen, Meenal Parakh, Stamatis Alexandropoulos, Lahav Lipson, et al. Infinigen indoors: Photorealistic indoor scenes using procedural generation. InProceedings of the IEEE/CVF Conference o...

    2024

  61. [69]

    Xcube: Large-scale 3d generative modeling using sparse voxel hierarchies

    Xuanchi Ren, Jiahui Huang, Xiaohui Zeng, Ken Museth, Sanja Fidler, and Francis Williams. Xcube: Large-scale 3d generative modeling using sparse voxel hierarchies. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4209–4219, 2024

    2024

  62. [70]

    L3dg: Latent 3d gaussian diffusion

    Barbara Roessle, Norman Müller, Lorenzo Porzi, Samuel Rota Bulò, Peter Kontschieder, Angela Dai, and Matthias Nießner. L3dg: Latent 3d gaussian diffusion. InSIGGRAPH Asia 2024 Conference Papers, pages 1–11, 2024

    2024

  63. [71]

    3d-re-gen: 3d reconstruction of indoor scenes with a generative framework

    Tobias Sautter, Jan-Niklas Dihlmann, and Hendrik P A Lensch. 3d-re-gen: 3d reconstruction of indoor scenes with a generative framework. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 528–537, 2026

    2026

  64. [72]

    Worldexplorer: Towards generating fully navigable 3d scenes

    Manuel-Andreas Schneider, Lukas Höllein, and Matthias Nießner. Worldexplorer: Towards generating fully navigable 3d scenes. InProceedings of the SIGGRAPH Asia 2025 Conference Papers, pages 1–11, 2025. 28 Tencent Hunyuan

    2025

  65. [73]

    Scenemaker: Open-set 3d scene generation with decoupled de-occlusion and pose estimation model

    Yukai Shi, Weiyu Li, Zihao Wang, Hongyang Li, Xingyu Chen, Ping Tan, and Lei Zhang. Scenemaker: Open-set 3d scene generation with decoupled de-occlusion and pose estimation model. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 27146–...

    2026

  66. [74]

    Realmdreamer: Text-driven 3d scene generation with inpainting and depth diffusion.arXiv preprint arXiv:2404.07199, 2024

    Jaidev Shriram, Alex Trevithick, Lingjie Liu, and Ravi Ramamoorthi. Realmdreamer: Text-driven 3d scene generation with inpainting and depth diffusion.arXiv preprint arXiv:2404.07199, 2024

    2024 arXiv

  67. [75]

    A survey on procedural modelling for virtual worlds

    Ruben M Smelik, Tim Tutenel, Rafael Bidarra, and Bedrich Benes. A survey on procedural modelling for virtual worlds. InComputer graphics forum, volume 33, pages 31–50. Wiley Online Library, 2014

    2014

  68. [76]

    3d-gpt: Procedural 3d modeling with large language models

    Chunyi Sun, Junlin Han, Weijian Deng, Xinlong Wang, Zishan Qin, and Stephen Gould. 3d-gpt: Procedural 3d modeling with large language models. In2025 International Conference on 3D Vision (3DV), pages 1253–1263. IEEE, 2025

    2025

  69. [77]

    Recent advances in 3d object and scene generation: A survey

    Xiang Tang, Ruotong Li, and Xiaopeng Fan. Recent advances in 3d object and scene generation: A survey. arXiv preprint arXiv:2504.11734, 2025

    2025

  70. [78]

    Hunyuanworld 1.0: Generating immersive, explorable, and interactive 3d worlds from words or pixels.arXiv preprint arXiv:2507.21809, 2025

    HunyuanWorld Team, Zhenwei Wang, Yuhao Liu, Junta Wu, Zixiao Gu, Haoyuan Wang, Xuhui Zuo, Tianyu Huang, Wenhuan Li, Sheng Zhang, et al. Hunyuanworld 1.0: Generating immersive, explorable, and interactive 3d worlds from words or pixels.arXiv preprint arXiv:2507.21809, 2025

    2025 arXiv

  71. [79]

    Gaussiangpt: Towards autore- gressive 3d gaussian scene generation.arXiv preprint arXiv:2603.26661, 2026

    Nicolas von Lützow, Barbara Rössle, Katharina Schmid, and Matthias Nießner. Gaussiangpt: Towards autore- gressive 3d gaussian scene generation.arXiv preprint arXiv:2603.26661, 2026

    2026 arXiv

  72. [80]

    Worldgen: From text to traversable and interactive 3d worlds

    Dilin Wang, Hyunyoung Jung, Tom Monnier, Kihyuk Sohn, Chuhang Zou, Xiaoyu Xiang, Yu-Ying Yeh, Di Liu, Zixuan Huang, Thu Nguyen-Phuoc, et al. Worldgen: From text to traversable and interactive 3d worlds. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern R...

    2026

  73. [81]

    Perf: Panoramic neural radiance field from a single panorama.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(10):6905–6918, 2024

    Guangcong Wang, Peng Wang, Zhaoxi Chen, Wenping Wang, Chen Change Loy, and Ziwei Liu. Perf: Panoramic neural radiance field from a single panorama.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(10):6905–6918, 2024

    2024

  74. [82]

    Scenetransporter: Optimal transport-guided compositional latent diffusion for single-image structured 3d scene generation.arXiv preprint arXiv:2602.22785, 2026

    Ling Wang, Hao-Xiang Guo, Xinzhou Wang, Fuchun Sun, Kai Sun, Pengkun Liu, Hang Xiao, Zhong Wang, Guangyuan Fu, Eric Li, et al. Scenetransporter: Optimal transport-guided compositional latent diffusion for single-image structured 3d scene generation.arXiv preprint arXiv:2602.22...

    2026

  75. [83]

    Scenecode: Executable world programs for editable indoor scenes with articulated objects.arXiv preprint arXiv:2605.19587, 2026

    Puyi Wang, Yuhao Wang, Linjie Li, Zhengyuan Yang, Kevin Qinghong Lin, Yangguang Li, and Yu Cheng. Scenecode: Executable world programs for editable indoor scenes with articulated objects.arXiv preprint arXiv:2605.19587, 2026

    2026 arXiv

  76. [84]

    Creation and rendering of realistic trees

    Jason Weber and Joseph Penn. Creation and rendering of realistic trees. InProceedings of the 22nd annual conference on Computer graphics and interactive techniques, pages 119–128, 1995

    1995

  77. [85]

    3d scene generation: A survey.arXiv preprint arXiv:2505.05474, 2025

    Beichen Wen, Haozhe Xie, Zhaoxi Chen, Fangzhou Hong, and Ziwei Liu. 3d scene generation: A survey.arXiv preprint arXiv:2505.05474, 2025

    2025 arXiv

  78. [86]

    Instant architecture.ACM Transactions on Graphics (TOG), 22(3):669–677, 2003

    Peter Wonka, Michael Wimmer, François Sillion, and William Ribarsky. Instant architecture.ACM Transactions on Graphics (TOG), 22(3):669–677, 2003

    2003

  79. [87]

    Marble.https://marble.worldlabs.ai/, 2026

    World Labs. Marble.https://marble.worldlabs.ai/, 2026. Accessed: 2026-07-17

    2026

  80. [88]

    Blockfusion: Expandable 3d scene generation using latent tri-plane extrapolation

    Zhennan Wu, Yang Li, Han Yan, Taizhang Shang, Weixuan Sun, Senbo Wang, Ruikai Cui, Weizhe Liu, Hiroyuki Sato, Hongdong Li, et al. Blockfusion: Expandable 3d scene generation using latent tri-plane extrapolation. ACM Transactions on Graphics (ToG), 43(4):1–17, 2024

    2024

  81. [89]

    Scenepainter: Semantically consistent perpetual 3d scene generation with concept relation alignment

    Chong Xia, Shengjun Zhang, Fangfu Liu, Chang Liu, Khodchaphun Hirunyaratsameewong, and Yueqi Duan. Scenepainter: Semantically consistent perpetual 3d scene generation with concept relation alignment. In Proceedings of the IEEE/CVF International Conference on Computer Vision, p...

    2025

  82. [90]

    Sage: Scalable agentic 3d scene generation for embodied ai

    Hongchi Xia, Xuan Li, Zhaoshuo Li, Qianli Ma, Jiashu Xu, Ming-Yu Liu, Yin Cui, Tsung-Yi Lin, Wei-Chiu Ma, Shenlong Wang, et al. Sage: Scalable agentic 3d scene generation for embodied ai. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, page...

    2026

  83. [91]

    Citydreamer: Compositional generative model of unbounded 3d cities

    Haozhe Xie, Zhaoxi Chen, Fangzhou Hong, and Ziwei Liu. Citydreamer: Compositional generative model of unbounded 3d cities. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9666–9675, 2024

    2024

  84. [92]

    Generative gaussian splatting for unbounded 3d city generation

    Haozhe Xie, Zhaoxi Chen, Fangzhou Hong, and Ziwei Liu. Generative gaussian splatting for unbounded 3d city generation. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 6111–6120, 2025

    2025

  85. [93]

    Sketch2scene: Automatic generation of interactive 3d game scenes from user’s casual sketches.arXiv preprint arXiv:2408.04567, 2024

    Yongzhi Xu, Yonhon Ng, Yifu Wang, Inkyu Sa, Yunfei Duan, Zhenhong Sun, Yang Li, Pan Ji, and Hongdong Li. Sketch2scene: Automatic generation of interactive 3d game scenes from user’s casual sketches.arXiv preprint arXiv:2408.04567, 2024

    2024

  86. [94]

    Frankenstein: Generating semantic-compositional 3d scenes in one tri-plane

    Han Yan, Yang Li, Zhennan Wu, Shenzhou Chen, Weixuan Sun, Taizhang Shang, Weizhe Liu, Tian Chen, Xiaqiang Dai, Chao Ma, et al. Frankenstein: Generating semantic-compositional 3d scenes in one tri-plane. In SIGGRAPH Asia 2024 Conference Papers, pages 1–11, 2024

    2024

  87. [95]

    Layerpano3d: Layered 3d panorama for hyper-immersive scene generation

    Shuai Yang, Jing Tan, Mengchen Zhang, Tong Wu, Gordon Wetzstein, Ziwei Liu, and Dahua Lin. Layerpano3d: Layered 3d panorama for hyper-immersive scene generation. InProceedings of the special interest group on computer graphics and interactive techniques conference conference p...

    2025

  88. [96]

    Sceneweaver: All-in-one 3d scene synthesis with an extensible and self-reflective agent.Advances in neural information processing systems, 38:140319–140351, 2026

    Yandan Yang, Baoxiong Jia, Shujie Zhang, and Siyuan Huang. Sceneweaver: All-in-one 3d scene synthesis with an extensible and self-reflective agent.Advances in neural information processing systems, 38:140319–140351, 2026

    2026

  89. [97]

    P3d-bench: Benchmarking mllms for parametric 3d generation and structural reasoning.arXiv preprint arXiv:2606.11152, 2026

    Yikang Yang, Zhanpeng Hu, Youtian Lin, Mengqi Zhou, Jingxi Xu, Feihu Zhang, Jiaheng Liu, and Yao Yao. P3d-bench: Benchmarking mllms for parametric 3d generation and structural reasoning.arXiv preprint arXiv:2606.11152, 2026

    2026 arXiv

  90. [98]

    Yixuan Yang, Zhen Luo, Tongsheng Ding, Junru Lu, Mingqi Gao, Jinyu Yang, Victor Sanchez, and Feng Zheng. Optiscene: Llm-driven indoor scene layout generation via scaled human-aligned data synthesis and multi-stage preference optimization.Advances in Neural Information Processi...

    2026

  91. [99]

    Holodeck: Language guided generation of 3d embodied ai environments

    Yue Yang, Fan-Yun Sun, Luca Weihs, Eli VanderBilt, Alvaro Herrasti, Winson Han, Jiajun Wu, Nick Haber, Ranjay Krishna, Lingjie Liu, et al. Holodeck: Language guided generation of 3d embodied ai environments. In Proceedings of the IEEE/CVF Conference on Computer Vision and Patt...

    2024

  92. [100]

    Cast: Component-aligned 3d scene reconstruction from an rgb image.ACM Transactions on Graphics (TOG), 44(4):1–19, 2025

    Kaixin Yao, Longwen Zhang, Xinhao Yan, Yan Zeng, Qixuan Zhang, Lan Xu, Wei Yang, Jiayuan Gu, and Jingyi Yu. Cast: Component-aligned 3d scene reconstruction from an rgb image.ACM Transactions on Graphics (TOG), 44(4):1–19, 2025

    2025

  93. [101]

    Genclaw: Code-driven agentic image generation.arXiv preprint arXiv:2605.30248, 2026

    Junyan Ye, Jun He, Zilong Huang, Dongzhi Jiang, Xuan Yang, Rui Chen, and Weijia Li. Genclaw: Code-driven agentic image generation.arXiv preprint arXiv:2605.30248, 2026

    2026 arXiv

  94. [102]

    Vision-as-inverse-graphics agent via interleaved multimodal reasoning.arXiv preprint arXiv:2601.11109, 2026

    Shaofeng Yin, Jiaxin Ge, Zora Zhiruo Wang, Chenyang Wang, Xiuyu Li, Michael J Black, Trevor Darrell, Angjoo Kanazawa, and Haiwen Feng. Vision-as-inverse-graphics agent via interleaved multimodal reasoning.arXiv preprint arXiv:2601.11109, 2026

    2026 arXiv

  95. [103]

    Extend3d: Town-scale 3d generation

    Seungwoo Yoon, Jinmo Kim, and Jaesik Park. Extend3d: Town-scale 3d generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5892–5901, 2026

    2026

  96. [104]

    Wonderjourney: Going from anywhere to everywhere

    Hong-Xing Yu, Haoyi Duan, Junhwa Hur, Kyle Sargent, Michael Rubinstein, William T Freeman, Forrester Cole, Deqing Sun, Noah Snavely, Jiajun Wu, et al. Wonderjourney: Going from anywhere to everywhere. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Rec...

    2024

  97. [105]

    Wonderworld: Interactive 3d scene generation from a single image

    Hong-Xing Yu, Haoyi Duan, Charles Herrmann, William T Freeman, and Jiajun Wu. Wonderworld: Interactive 3d scene generation from a single image. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 5916–5926, 2025

    2025

  98. [106]

    Scene splatter: Momentum 3d scene generation from single image with video diffusion model

    Shengjun Zhang, Jinzhao Li, Xin Fei, Hao Liu, and Yueqi Duan. Scene splatter: Momentum 3d scene generation from single image with video diffusion model. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 6089–6098, 2025

    2025

  99. [107]

    Cityx: Controllable procedural content generation for unbounded 3d cities.arXiv preprint arXiv:2407.17572, 2024

    Shougao Zhang, Mengqi Zhou, Yuxi Wang, Chuanchen Luo, Rongyu Wang, Yiwei Li, Zhaoxiang Zhang, and Junran Peng. Cityx: Controllable procedural content generation for unbounded 3d cities.arXiv preprint arXiv:2407.17572, 2024. 30 Tencent Hunyuan

    2024 arXiv

  100. [108]

    Holodreamer: Holistic 3d panoramic world generation from text descriptions.arXiv preprint arXiv:2407.15187, 2024

    Haiyang Zhou, Xinhua Cheng, Wangbo Yu, Yonghong Tian, and Li Yuan. Holodreamer: Holistic 3d panoramic world generation from text descriptions.arXiv preprint arXiv:2407.15187, 2024

    2024 arXiv

  101. [109]

    Articraft: An agentic system for scalable articulated 3d asset generation

    Matt Zhou, Ruining Li, Xiaoyang Lyu, Zhaomou Song, Zhening Huang, Chuanxia Zheng, Christian Rupprecht, Andrea Vedaldi, and Shangzhe Wu. Articraft: An agentic system for scalable articulated 3d asset generation. arXiv preprint arXiv:2605.15187, 2026

    2026 arXiv

  102. [110]

    Scenex: Procedural controllable large-scale scene generation

    Mengqi Zhou, Yuxi Wang, Jun Hou, Shougao Zhang, Yiwei Li, Chuanchen Luo, Junran Peng, and Zhaoxiang Zhang. Scenex: Procedural controllable large-scale scene generation. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 10806–10814, 2025

    2025

  103. [111]

    Dreamscene360: Unconstrained text-to-3d scene generation with panoramic gaussian splatting

    Shijie Zhou, Zhiwen Fan, Dejia Xu, Haoran Chang, Pradyumna Chari, Tejas Bharadwaj, Suya You, Zhangyang Wang, and Achuta Kadambi. Dreamscene360: Unconstrained text-to-3d scene generation with panoramic gaussian splatting. InEuropean Conference on Computer Vision, pages 324–342....

    2024

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.