Pith. sign in

REVIEW 4 major objections 5 minor 118 references

Hunyuan3D-Buffalo 1.0: A Unified Multimodal Model for Scalable 3D Generation, Understanding, and Editing

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read One unified 3D model outdoes prior systems in generation and editing

desk verdict A serious industrial unified 3D model with a genuinely novel data engine, but the editing data pipeline's admitted interior-mask inconsistency is the load-bearing risk. read the letter →

arxiv 2608.02711 v1 pith:V6TDQVA3 submitted 2026-08-03 cs.CV

Junliang Ye , Kenkun Liu , Guocun Wang , Yang Li , Yansong Qu , Chunshi Wang , Jingwei Xu , Yunhan Yang
show 7 more authors
Zibo Zhao Jiachen Xu Jiaao Yu Lifu Wang Zhihao Liang Zhuo Chen Chunchao Guo
This is my paper
classification cs.CV
keywords unified3Dmultimodalmodeltext-to-3Dgenerationeditingunderstandingpartdiffusiontransformervision-languageNano3D-v2datapipeline
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

Hunyuan3D-Buffalo 1.0 claims that a single architecture can do 3D understanding, text-to-3D generation, instruction-guided 3D editing, and text-grounded part generation, and that training these tasks together makes each one stronger. The authors built an 87M-sample multimodal corpus, including 12M editing pairs produced by the Nano3D-v2 pipeline, to overcome the shortage of geometrically consistent editing data. On Edit3D-Bench the model reports an average Chamfer distance of 0.0091 and an F1 of 0.6515, far ahead of prior methods, and a human study prefers its text-to-3D outputs about twice as often as the strongest baseline. The paper's central interpretive claim is that better generation and better understanding both improve editing, so unified training is not merely convenient but causally useful.

What carries the argument

The load-bearing mechanism is the two-model architecture plus a large synthetic data engine. Hunyuan3D-VLM encodes a colored point cloud through separate structure and appearance paths, compresses the tokens with a Q-Former, and augments the language vocabulary with 133 special tokens for quantized 3D bounding boxes, giving the model explicit part-level grounding. Hunyuan3D DiT, initialized from Hunyuan3D-2.1, performs flow-matching diffusion conditioned on VLM hidden states through an MLP connector, and for editing and part generation receives the source object's latent directly as additional input so unedited geometry can be preserved. On the data side, Nano3D-v2 generates editing pairs by selecting an anchor view with a VLM, editing that 2D view, predicting a 3D bounding box with a learned autoregressive transformer, applying voxel-level editing inside the box while freezing outside tokens, then refining geometry with LATTICE and texture with NaTex, and finally filtering and re-annotating pairs with a VLM. This pipeline is what makes 12M editing pairs available; without it, the unified training would not have editing supervision.

What would settle it

Run the Nano3D-v2 pipeline on a random sample of source assets, and for each pair measure the Chamfer distance restricted to the voxels inside the predicted bounding box that the source-target difference mask marks as unedited; if a substantial fraction of pairs exceed the model's own reported error range (e.g., average CD above about 0.009 on those interior regions), the end-to-end editing scores are inflated by training on inconsistent pairs.

Watch

Extended reading notes

Core claim

Hunyuan3D-Buffalo 1.0 claims that a single model can read a 3D object, answer questions about it, generate a new object from text, edit an existing object following an instruction, and extract or generate a named part, all without switching systems. The claim is carried by coupling a 3D vision-language model with a diffusion transformer: the VLM supplies semantic, structural, and spatial reasoning, an MLP connector projects its hidden states into the DiT's conditioning space, and for editing and part generation the source object's representation is concatenated into the denoiser so unedited regions are preserved. The authors further claim that unified training creates measurable synergy: stronger text-to-3D generation produces more complete edited geometry, and stronger 3D understanding improves localization and instruction following in editing. If correct, the result would make separate specialist models for each 3D task unnecessary for most purposes.

Load-bearing premise

The whole editing result stands on the assumption that the automatically generated editing pairs are geometrically consistent enough to learn from, specifically that regions inside the predicted 3D box that were not meant to change really are unchanged, an assumption the paper itself flags as difficult to guarantee.

Editorial extensions

If this is right

  • If the synergy claim is right, adding cheaper text-to-3D data can substitute for expensive editing data: the paper shows that 1,000 additional chicken samples improved a chicken-head edit without any new editing pairs.
  • A user could describe an object, request a change, and get back a 3D asset whose unedited regions match the source, because the source-object conditioning is built into the same architecture.
  • 3D editing evaluation numbers would improve by roughly an order of magnitude on Chamfer distance compared with earlier feed-forward and optimization-based editors, an 86.7% relative reduction on average CD versus the strongest prior baseline.
  • Part generation becomes a by-product of instruction following, so segmenting "the wheels" and generating a new wheel use the same prompt interface as editing.
  • Scaling text-to-3D pretraining data is a direct lever on editing quality, since human preference rates rise monotonically from 8.4% to 57.5% as the corpus grows from 3M to 50M samples.

Reading between the lines

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

  • A practical recipe implied by the paper is to keep scaling text-to-3D data rather than hand-labeling editing pairs, since the paper shows editing quality follows generation quality and text-to-3D data is far cheaper to construct.
  • A testable next step would be to ablate the VLM conditioning while keeping the DiT fixed: if understanding genuinely helps editing, replacing the 3D-VLM with a weaker semantic encoder should degrade localization even when the generative backbone is unchanged.
  • One could measure whether the reported editing gains persist on out-of-distribution edit types such as texture-only edits or edits far outside the training mask distribution, since the paper notes that texture editing remains largely unexplored.
  • The paper's own caveat about interior mask inconsistency suggests that a more precise editing-region mask could yield further editing gains without changing the model architecture.
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 / 5 minor

Summary. The manuscript presents Hunyuan3D-Buffalo 1.0, a unified 3D multimodal model that combines a frozen Hunyuan3D-VLM semantic backbone with a 3D-DiT initialized from Hunyuan3D-2.1. The authors construct an 87M-sample corpus (25M understanding samples, 50M text-to-3D pairs, and 12M editing pairs produced by the proposed Nano3D-v2 pipeline) and train the model in four stages, branching into editing, text-to-3D, and part-generation paths. They report state-of-the-art or leading results on UniPart-Bench for 3D understanding, a 100-prompt human preference study for text-to-3D generation, and Edit3D-Bench for instruction-guided editing. They also claim two cross-task synergies: stronger text-to-3D generation improves editing, and stronger 3D understanding improves editing.

Significance. If the reported results hold, this is a meaningful step toward unified 3D multimodal models, with a large corpus, a reusable architecture, and an unusually candid discussion of limitations in Section 6. The paper deserves credit for evaluating on external benchmarks (UniPart-Bench, Edit3D-Bench) and for explicitly acknowledging the editing-data pipeline's weaknesses. However, the central editing claims rest on self-reported numbers and on training pairs whose interior-mask consistency is admitted to be unguaranteed; the headline human-preference results also lack basic statistical reporting. The significance of the contribution therefore hinges on whether the authors can validate the editing supervision and provide transparent evaluation details. With that validation, the contribution would be strong.

major comments (4)
  1. [§3.4 and §6, item 4] The editing-data pipeline's consistency guarantee applies only outside the predicted 3D bounding box; Section 6, item 4, states that the interior of the mask 'often contains non-edited regions whose consistency is difficult to guarantee' and that these inconsistencies 'propagate into the end-to-end model during training, degrading editing quality.' Since the 12M editing pairs are the sole supervision for the reported editing capability, this is a load-bearing limitation rather than a routine future-work item. The manuscript should quantify the fraction of affected pairs (e.g., through a human audit of a random sample before and after the Stage 5 filtering), ablate training with tighter interior consistency checks, or otherwise demonstrate that the Edit3D-Bench gains are not an artifact of learning to reproduce noisy interiors. As written, the claim in Section 3.4 that the corpus supplies 'geometrically consistent' editing pairs is in tension with the admitted limitation.
  2. [§5.3, Table 7] The Edit3D-Bench evaluation reports no error bars, no number of test samples, and no statistical significance tests; the 86.7% CD reduction over Omni123 is a single point estimate. Because the benchmark and the Nano3D-v2 pipeline may share similar geometric transformations, the manuscript should clarify how the Edit3D-Bench ground-truth edited meshes were constructed and whether any overlap exists with the training corpus or with the operations the model was trained to imitate. A small independent test set with human-verified edits would substantially strengthen the claim.
  3. [§5.2, Table 5] The human evaluation is described without participant count, recruitment or screening criteria, or statistical analysis. The statement that 'the ranking is consistent across every individual participant' is difficult to assess without per-participant data or at least a measure of agreement; with three baselines plus the proposed method, a preference rate of 55.2% should be accompanied by a confidence interval, inter-annotator statistic, or significance test. Please report these details.
  4. [§5.3, Fig. 10] The claimed synergy that stronger text-to-3D generation improves editing is supported by a single qualitative chicken-head example, which is not sufficient to establish a general capability-transfer effect. A systematic study varying the scale and diversity of text-to-3D data and measuring editing metrics on a fixed editing set would be needed to substantiate the 'clear direction' stated in the text.
minor comments (5)
  1. [§3.4 and §6, item 4] The earlier section calls the editing pairs 'geometrically consistent' while the conclusion acknowledges unguaranteed interior consistency; please align the wording to avoid an apparent contradiction.
  2. [References] References [90] and [91] are the same work (3DEditFormer / 'Towards scalable and consistent 3D editing') and should be merged; reference [24] is cited as 'Vision Banana' in Section 3.5 but as 'Vision-Banana' elsewhere.
  3. [§1 and §3.4] There are minor typos: 'an unified' in the Figure 1 caption and 'Editing Plaining' in the Stage 2 heading of Section 3.4.
  4. [Table 6] The units '300w/1500w/5000w' should be defined explicitly (e.g., w = 10^4) to avoid confusion with the earlier '50M' notation.
  5. [Table 2] The caption-tier descriptions would benefit from a one-sentence explanation of how the 'Paraphrase' tier is generated from the 'Simplified' tier.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: headline results are measured on external benchmarks and human preference, and self-cited components are building blocks rather than forced inputs.

full rationale

The paper's derivation chain is not circular. Headline claims are evaluated on external benchmarks and human preference: UniPart-Bench (Tables 3-4), Edit3D-Bench (Table 7), and a 100-prompt four-way user study (Table 5) with prompts described as diverse and independent of the training-corpus construction pipeline. The training corpora (25M understanding samples, 50M text-to-3D pairs, 12M editing pairs) are produced by pipelines whose outputs are not defined in terms of the quantities being predicted. Nano3D-v2 is a data-generation engine, not a fitted evaluator; the editing loss is flow-matching on 3D latents, and the Edit3D-Bench metrics (Chamfer Distance and F1 against ground-truth meshes) are not constructed from the model's own fitted parameters. The paper does rely substantially on self-authored components (Hunyuan3D-2.1 initialization, Part-X-MLLM vocabulary and data, the Nano3D family, and Universe3D as a baseline), but these are building blocks or comparison baselines, not definitions of the reported results. No claimed number is equivalent by construction to a cited statement or to a training objective. The paper's own admitted limitation in Section 6, item 4 -- 'the interior of the mask often contains non-edited regions whose consistency is difficult to guarantee. Such inconsistencies propagate into the end-to-end model during training, degrading editing quality' -- is a genuine data-quality risk that could affect editing performance, but it does not make any benchmark result true by definition, nor does it turn a prediction into a fit. Similarly, the cross-task synergy claims are supported only by one qualitative chicken-head example (Fig. 10) and a conditioning-variant comparison; this is weak evidence, but it is not a circularity. No fitted parameter is relabeled as a prediction, and no uniqueness theorem or ansatz is imported from self-citations to force the architecture. Overall, the central results are externally falsifiable and self-contained against the benchmarks, so the appropriate circularity score is 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central empirical claims rest on the consistency of automatically generated training data, the reliability of VLM-based filtering, and the validity of external benchmarks. Several training hyperparameters are chosen by hand and directly affect the reported capability balance. No new physical or mathematical entities are introduced.

free parameters (5)
  • text-to-3D vs editing+part-generation sampling ratio = 1:1
    Stage 3 omni pretraining; authors call this balance essential for editing without sacrificing generation.
  • editing and part-generation data repeat factor = 4x
    Repeats the smaller editing corpus to match text-to-3D data within each batch.
  • geometry-quality threshold tau = 10
    Filters the cleanest text-to-3D split; data quality threshold chosen by hand.
  • Q-Former 3D token count = 512
    Fixed-length compression of 3D tokens before VLM fusion; capacity choice.
  • boundary alpha-blending band = 7 voxels
    Texture editing stitch width in NaTex latent space; chosen to smooth transitions.
assumptions (4)
  • domain assumption Nano3D-v2 editing pairs are geometrically consistent enough to serve as training supervision.
    Core premise of the editing data engine; acknowledged in Section 6 item 4 as imperfect.
  • domain assumption VLM filtering and annotation correctly identify valid editing pairs and instructions.
    Stage 5 uses VLM judgments for integrity, instruction alignment, and region consistency; errors propagate to the model.
  • domain assumption External benchmarks measure the claimed capabilities.
    UniPart-Bench and Edit3D-Bench are used as ground truth; Edit3D-Bench is a recent benchmark from the cited literature.
  • domain assumption Pretrained components remain valid after joint fine-tuning.
    Qwen-VL, Hunyuan3D-2.1 DiT, TRELLIS, LATTICE, and NaTex priors are reused; degradation from new training is not assessed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hunyuan3D-Buffalo 1.0: A Unified Multimodal Model for Scalable 3D Generation, Understanding, and Editing." pith.science (2026). https://pith.science/paper/V6TDQVA3

@misc{pith2026260802711,
  author       = {Pith},
  title        = {Pith review of: Hunyuan3D-Buffalo 1.0: A Unified Multimodal Model for Scalable 3D Generation, Understanding, and Editing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V6TDQVA3}},
  note         = {Machine review of arXiv:2608.02711}
}
read the original abstract

Recent advances in image generation have demonstrated the potential of unified multimodal models that integrate understanding, generation, and editing. However, unified 3D modeling remains constrained by scarce multimodal data, particularly the lack of large-scale and geometrically consistent editing data. To address this limitation, we propose Hunyuan3D-Buffalo 1.0, a unified framework supporting 3D understanding, text-to-3D generation, instruction-guided 3D editing, and text-grounded part generation within a single architecture. To enable scalable training, we construct an 87M-scale 3D multimodal corpus, comprising 25M understanding samples, 50M text-to-3D pairs, and 12M editing pairs generated using Nano3D-v2. Architecturally, the framework combines Hunyuan3D-VLM for semantic, structural, and spatial understanding with Hunyuan3D DiT for high-fidelity 3D synthesis. The VLM provides multimodal semantic conditions for generation, while editing and part generation additionally condition the diffusion process on the source object representation to preserve its overall structure and unedited regions. Extensive experiments show that Hunyuan3D-Buffalo 1.0 achieves state-of-the-art or leading performance on text-to-3D generation and 3D editing benchmarks, while exhibiting strong understanding and part-generation capabilities. Our analysis further shows that both generation and understanding improve editing, demonstrating the effectiveness of unified 3D multimodal training. Project Page: https://tencent-hunyuan.github.io/Hunyuan3D-Buffalo1.0/

Figures

Figures reproduced from arXiv: 2608.02711 by the authors.

Figure 1
Figure 1. Hunyuan3D-Buffalo 1.0 is an unified 3D multimodal framework that combines autoregressive modeling with diffusion-based generation, enabling 3D understanding, text￾to-3D generation, 3D editing, and text-grounded part generation within a single architecture. 2 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Pipeline of constructing text-to-3D training corpus. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Pipeline of constructing 3D editing training corpus (Nano3D-v2). [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Examples of editing pairs in the training corpus created by Nano3D-v2. [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Examples of multi-round editing by Nano3D-v2. [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Hunyuan3D-Buffalo 1.0 pipeline. The framework unifies 3D QA and grounding, text-to-3D generation, and 3D editing through a shared Hunyuan3D-VLM backbone, which connects language, 3D representations, and generative Hunyuan3D DiT modules for multi￾modal understanding, ge…
Figure 7
Figure 7. Figure 7: Overview of the training pipeline. The shared trunk proceeds through 3D-VLM [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Qualitative text to 3D results. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_8.png]
Figure 9
Figure 9. Figure 9: Qualitative shape editing results. Our method significantly outperforms all base￾lines in both geometric consistency before and after editing, and responsiveness to editing instructions. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_9.png]
Figure 10
Figure 10. Figure 10: Scaling up text-to-3d data facilitates 3D editing. Model A is our base model; when instructed to edit an object by replacing its head with a chicken head, it fails to produce a satisfactory result. Model B is built upon Model A by adding only 1,000 additional chicken …
Figure 11
Figure 11. Figure 11: Qualitative part generation results. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_11.png]
Figure 12
Figure 12. Figure 12: Qualitative shape editing results. 26 [PITH_FULL_IMAGE:figures/full_fig_p026_12.png]
Figure 13
Figure 13. Figure 13: Qualitative text to 3D results. 27 [PITH_FULL_IMAGE:figures/full_fig_p027_13.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

118 extracted references · 16 canonical work pages

  1. [1]

    S. Bai, K. Chen, X. Liu, J. Wang, W. Ge, S. Song, K. Dang, P . Wang, S. Wang, J. Tang, et al. Qwen2. 5-vl technical report.arXiv preprint arXiv:2502.13923, 2025

    arXiv 2025

  2. [2]

    Bar-On, D

    R. Bar-On, D. Cohen-Bar, and D. Cohen-Or. Editp23: 3d editing via propagation of image prompts to multi-view, 2025. URLhttps://arxiv.org/abs/2506.20652

    arXiv 2025

  3. [3]

    Barda, M

    A. Barda, M. Gadelha, V . G. Kim, N. Aigerman, A. H. Bermano, and T. Groueix. Instant3dit: Multiview inpainting for fast editing of 3d objects, 2024. URL https: //arxiv.org/abs/2412.00518

    arXiv 2024

  4. [4]

    K. Bhat, N. Khanna, K. Channa, T. Zhou, Y. Zhu, X. Sun, C. Shang, A. Sudar- shan, M. Chu, D. Li, et al. Cube: A roblox view of 3d intelligence.arXiv preprint arXiv:2503.15475, 2025

    arXiv 2025

  5. [5]

    W. Cai, S. Fang, W. Ye, X. Dong, Y. Yang, X. Zhang, W. Cheng, Y. Cao, G. Yu, and T. Chen. Native 3d editing with full attention.arXiv preprint arXiv:2511.17501, 2025

    arXiv 2025

  6. [6]

    C. Cao, C. Yu, F. Wang, X. Xue, and Y. Fu. Mvinpainter: Learning multi-view consistent inpainting to bridge 2d and 3d editing, 2024. URL https://arxiv.org/abs/2408. 08000

    2024

  7. [7]

    S. Cao, H. Chen, P . Chen, Y. Cheng, Y. Cui, X. Deng, Y. Dong, K. Gong, T. Gu, X. Gu, et al. Hunyuanimage 3.0 technical report.arXiv preprint arXiv:2509.23951, 2025

    arXiv 2025

  8. [8]

    H. Chen, R. Shi, Y. Liu, B. Shen, J. Gu, G. Wetzstein, H. Su, and L. Guibas. Generic 3d diffusion adapter using controlled multi-view editing, 2024. URL https://arxiv. org/abs/2403.12032

    arXiv 2024

Show all 118 references
  1. [9]

    J. Chen, Z. Xu, X. Pan, Y. Hu, C. Qin, T. Goldstein, L. Huang, T. Zhou, S. Xie, S. Savarese, et al. Blip3-o: A family of fully open unified multimodal models-architecture, training and dataset.arXiv preprint arXiv:2505.09568, 2025

    2025 arXiv

  2. [10]

    J. Chen, L. Xue, Z. Xu, X. Pan, S. Yang, C. Qin, A. Yan, H. Zhou, Z. Chen, L. Huang, et al. Blip3o-next: Next frontier of native image generation.arXiv preprint arXiv:2510.15857, 2025

    2025

  3. [11]

    L. Chen, P . Wang, G. Zhang, Z. Ma, and L. Zhang. Omni-3dedit: Generalized versatile 3d editing in one-pass. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12640–12650, 2026

    2026

  4. [12]

    M. Chen, J. Xie, I. Laina, and A. Vedaldi. Shap-editor: Instruction-guided latent 3d editing in seconds, 2023. URLhttps://arxiv.org/abs/2312.09246

    2023 arXiv

  5. [13]

    R. Chen, Y. Chen, N. Jiao, and K. Jia. Fantasia3d: Disentangling geometry and ap- pearance for high-quality text-to-3d content creation. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 22246–22256, 2023

    2023

  6. [14]

    X. Chen, Z. Wu, X. Liu, Z. Pan, W. Liu, Z. Xie, X. Yu, and C. Ruan. Janus-pro: Unified multimodal understanding and generation with data and model scaling.arXiv preprint arXiv:2501.17811, 2025

    2025 arXiv

  7. [15]

    Y. Chen, T. He, D. Huang, W. Ye, S. Chen, J. Tang, X. Chen, Z. Cai, L. Yang, G. Yu, et al. Meshanything: Artist-created mesh generation with autoregressive transformers. arXiv preprint arXiv:2406.10163, 2024

    2024 arXiv

  8. [16]

    Y. Chen, Z. Li, Y. Wang, H. Zhang, Q. Li, C. Zhang, and G. Lin. Ultra3d: Efficient and high-fidelity 3d generation with part attention.arXiv preprint arXiv:2507.17745, 2025

    2025 arXiv

  9. [17]

    Y. Chen, Y. Wang, Y. Luo, Z. Wang, Z. Chen, J. Zhu, C. Zhang, and G. Lin. Me- shanything v2: Artist-created mesh generation with adjacent mesh tokenization. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 13922– 13931, 2025

    2025

  10. [18]

    Y. Chi, X. Li, Z. Huang, and J. M. Rehg. Vinedresser3d: Towards agentic text-guided 3d editing. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12673–12683, 2026

    2026

  11. [19]

    Y. Cui, H. Chen, H. Deng, X. Huang, X. Li, J. Liu, Y. Liu, Z. Luo, J. Wang, W. Wang, Y. Wang, C. Wang, F. Zhang, Y. Zhao, T. Pan, X. Li, Z. Hao, W. Ma, Z. Chen, Y. Ao, T. Huang, Z. Wang, and X. Wang. Emu3.5: Native multimodal models are world learners, 2025. URLhttps://arxiv.o...

    2025 arXiv

  12. [20]

    C. Deng, D. Zhu, K. Li, C. Gou, F. Li, Z. Wang, S. Zhong, W. Yu, X. Nie, Z. Song, G. Shi, and H. Fan. Emerging properties in unified multimodal pretraining.arXiv preprint arXiv:2505.14683, 2025

    2025 arXiv

  13. [21]

    N. A. Dinh, I. Lang, H. Kim, O. Stein, and R. Hanocka. Geometry in style: 3d stylization via surface normal deformation, 2025. URLhttps://arxiv.org/abs/2503.23241. 28 Tencent Hunyuan

    2025 arXiv

  14. [22]

    S. Dong, L. Ding, Z. Huang, Z. Wang, T. Xue, and D. Xu. Interactive3d: Create what you want by interactive 3d generation, 2024. URLhttps://arxiv.org/abs/2404.16510

    2024 arXiv

  15. [23]

    Erkoc ¸, C

    Z. Erkoc ¸, C. G ¨umeli, C. Wang, M. Nießner, A. Dai, P . Wonka, H.-Y. Lee, and P . Zhuang. Preditor3d: Fast and precise 3d shape editing, 2024. URL https://arxiv.org/abs/ 2412.06592

    2024 arXiv

  16. [24]

    Gabeur, S

    V . Gabeur, S. Long, S. Peng, P . Voigtlaender, S. Sun, Y. Bao, K. Truong, Z. Wang, W. Zhou, J. T. Barron, K. Genova, N. Kannen, S. Ben, Y. Li, M. Guo, S. Yogin, Y. Gu, H. Chen, O. Wang, S. Xie, H. Zhou, K. He, T. Funkhouser, J.-B. Alayrac, and R. Soricut. Image generators are...

    2026 arXiv

  17. [25]

    W. Gao, D. Wang, Y. Fan, A. Bozic, T. Stuyck, Z. Li, Z. Dong, R. Ranjan, and N. Sarafi- anos. 3d mesh editing using masked lrms, 2024. URL https://arxiv.org/abs/2412. 08641

    2024

  18. [26]

    Y. Gao, L. Gong, Q. Guo, X. Hou, Z. Lai, F. Li, L. Li, X. Lian, C. Liao, L. Liu, et al. Seedream 3.0 technical report.arXiv preprint arXiv:2504.11346, 2025

    2025 arXiv

  19. [27]

    I. Gat, D. Cohen-Bar, G. Levy, E. Richardson, and D. Cohen-Or. Shapeup: Scalable image-conditioned 3d editing.arXiv preprint arXiv:2602.05676, 2026

    2026 arXiv

  20. [28]

    Z. Hao, D. W. Romero, T.-Y. Lin, and M.-Y. Liu. Meshtron: High-fidelity, artist-like 3d mesh generation at scale.arXiv preprint arXiv:2412.09548, 2024

    2024 arXiv

  21. [29]

    Haque, M

    A. Haque, M. Tancik, A. A. Efros, A. Holynski, and A. Kanazawa. Instruct-nerf2nerf: Editing 3d scenes with instructions. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 19740–19750, 2023

    2023

  22. [30]

    He, Z.-X

    X. He, Z.-X. Zou, C.-H. Chen, Y.-C. Guo, D. Liang, C. Yuan, W. Ouyang, Y.-P . Cao, and Y. Li. Sparseflex: High-resolution and arbitrary-topology 3d shape modeling. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 14822– 14833, 2025

    2025

  23. [31]

    Hsiao, B.-K

    T.-F. Hsiao, B.-K. Ruan, Y.-L. Liu, and H.-H. Shuai. Vecset-edit: Unleashing pre-trained lrm for mesh editing from single image. InProceedings of the Special Interest Group on Computer Graphics and Interactive Techniques Conference Conference Papers, pages 1–12, 2026

    2026

  24. [32]

    S. Hu, Y. Wei, F. Zha, Y. Guo, and J. Zhang. Easy3e: Feed-forward 3d asset editing via rectified voxel flow. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12730–12740, 2026

    2026

  25. [33]

    Huang, D

    Z. Huang, D. Zheng, C. Zou, R. Liu, X. Wang, K. Ji, W. Chai, J. Sun, L. Wang, Y. Lv, T. Huang, J. Liu, Q. Guo, M. Yang, J. Chen, and J. Zhou. Ming-univision: Joint image understanding and generation with a unified continuous tokenizer.arXiv preprint arXiv:2510.06590, 2025

  26. [34]

    Hunyuan3D

    T. Hunyuan3D. Hunyuan3d studio: End-to-end ai pipeline for game-ready 3d asset generation.arXiv preprint arXiv:2509.12815, 2025

    2025

  27. [35]

    Hunyuan3D, B

    T. Hunyuan3D, B. Zhang, C. Guo, H. Liu, H. Yan, H. Shi, J. Huang, J. Yu, K. Li, P . Wang, et al. Hunyuan3d-omni: A unified framework for controllable generation of 3d assets. arXiv preprint arXiv:2509.21245, 2025

    2025

  28. [36]

    Hunyuan3D, :, B

    T. Hunyuan3D, :, B. Zhang, C. Guo, D. Guo, H. Liu, H. Yan, H. Shi, J. Yu, J. Xu, J. Huang, K. Li, L. Wang, Linus, P . Wang, Q. Lin, R. Tang, X. Yang, Y. Li, Y. Guan, Y. Zhao, Y. Yang, Z. Lai, Z. Liang, and Z. Zhao. Hy3d-bench: Generation of 3d assets,

  29. [37]

    J. Kim, Y. Lan, A. Fortes, Y. Chen, and X. Pan. Fastmesh: Efficient artistic mesh generation via component decoupling.arXiv preprint arXiv:2508.19188, 2025

    2025

  30. [38]

    B. F. Labs, S. Batifol, A. Blattmann, F. Boesel, S. Consul, C. Diagne, T. Dockhorn, J. English, Z. English, P . Esser, S. Kulal, K. Lacey, Y. Levi, C. Li, D. Lorenz, J. M¨uller, D. Podell, R. Rombach, H. Saini, A. Sauer, and L. Smith. Flux.1 kontext: Flow matching for in-conte...

    2025 arXiv

  31. [39]

    Z. Lai, Y. Zhao, H. Liu, Z. Zhao, Q. Lin, H. Shi, X. Yang, M. Yang, S. Yang, Y. Feng, et al. Hunyuan3d 2.5: Towards high-fidelity 3d assets generation with ultimate details. arXiv preprint arXiv:2506.16504, 2025

    2025 arXiv

  32. [40]

    Z. Lai, Y. Zhao, Z. Zhao, H. Liu, Q. Lin, J. Huang, C. Guo, and X. Yue. Lattice: Democratize high-fidelity 3d generation at scale.arXiv preprint arXiv:2512.03052, 2025

    2025

  33. [41]

    Z. Lai, Y. Zhao, Z. Zhao, X. Yang, X. Huang, J. Huang, X. Yue, and C. Guo. Natex: Seamless texture generation as latent color diffusion.arXiv preprint arXiv:2511.16317, 2025. 29 Tencent Hunyuan

    2025

  34. [42]

    L. Li, Z. Huang, H. Feng, G. Zhuang, R. Chen, C. Guo, and L. Sheng. Voxhammer: Training-free precise and coherent 3d editing in native 3d space.arXiv preprint arXiv:2508.19247, 2025

    2025 arXiv

  35. [43]

    P . Li, S. Ma, J. Chen, Y. Liu, C. Zhang, W. Xue, W. Luo, A. Sheffer, W. Wang, and Y. Guo. Cmd: Controllable multiview diffusion for 3d editing and progressive generation,

  36. [44]

    W. Li, J. Liu, H. Yan, R. Chen, Y. Liang, X. Chen, P . Tan, and X. Long. Craftsman3d: High-fidelity mesh generation with 3d native generation and interactive geometry refiner.arXiv preprint arXiv:2405.14979, 2024

    2024 arXiv

  37. [45]

    W. Li, A. Toisoul, T. Monnier, R. Shapovalov, R. Ranjan, P . Tan, and A. Vedaldi. Meshflow: Efficient artistic mesh generation via meshvae and flow-based diffusion transformer. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5849–5858, 2026

    2026

  38. [46]

    Li, Z.-X

    Y. Li, Z.-X. Zou, Z. Liu, D. Wang, Y. Liang, Z. Yu, X. Liu, Y.-C. Guo, D. Liang, W. Ouyang, et al. Triposg: High-fidelity 3d shape synthesis using large-scale rec- tified flow models.arXiv preprint arXiv:2502.06608, 2025

    2025 arXiv

  39. [47]

    Y. Li, V . Cheung, X. Liu, Y. Chen, Z. Luo, B. Lei, H. Weng, Z. Zhao, J. Huang, Z. Chen, and C. Guo. Auto-regressive surface cutting.arXiv preprint arXiv:2506.18017, 2026

    2026 arXiv

  40. [48]

    Z. Li, W. Li, T. Wang, Z. Wang, J. Wu, H. Wang, Y. Yang, Z. Huang, Y. Li, P . Liu, and C. Guo. Moca: Mixture-of-components attention for scalable compositional 3d generation.arXiv preprint arXiv:2512.07628, 2025

    2025

  41. [49]

    Z. Li, Y. Wang, H. Zheng, Y. Luo, and B. Wen. Sparc3d: Sparse representation and construction for high-resolution 3d shapes modeling.arXiv preprint arXiv:2505.14521, 2025

    2025 arXiv

  42. [50]

    S. Lim, S. Yoon, G. Koo, H. Yun, and C. D. Yoo. Tango: Training-free 3d editing via tangent-space guidance and optimization.arXiv preprint arXiv:2607.14927, 2026

    2026 arXiv

  43. [51]

    C.-H. Lin, J. Gao, L. Tang, T. Takikawa, X. Zeng, X. Huang, K. Kreis, S. Fidler, M.-Y. Liu, and T.-Y. Lin. Magic3d: High-resolution text-to-3d content creation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 300–309, 2023

    2023

  44. [52]

    F. Liu, H. Wang, W. Chen, H. Sun, and Y. Duan. Make-your-3d: Fast and consistent subject-driven 3d content generation, 2024. URL https://arxiv.org/abs/2403.09625

    2024 arXiv

  45. [53]

    F. Liu, J. Ye, Y. Wang, H. Wang, Z. Wang, J. Zhu, and Y. Duan. Dreamreward-x: Boosting high-quality 3d generation with human preference alignment.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    2025

  46. [54]

    H. Liu, Y. Lin, J. Guo, R. Chu, J. Wang, R. Li, and Y. Yang. Velocity-space 3d asset editing.arXiv preprint arXiv:2605.07385, 2026

    2026 arXiv

  47. [55]

    M. Liu, M. A. Uy, D. Xiang, H. Su, S. Fidler, N. Sharp, and J. Gao. Partfield: Learning 3d feature fields for part segmentation and beyond. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 9704–9715, 2025

    2025

  48. [56]

    Z. Luo, Y. Li, M. Zhang, S. Wang, H. Yan, X. Song, T. Shang, W. Mao, H. Li, X. Han, and P . Ji. Bag: Body-aligned 3d wearable asset generation.IEEE Transactions on Visualization and Computer Graphics, 2026

    2026

  49. [57]

    C. Ma, Y. Li, X. Yan, J. Xu, Y. Yang, C. Wang, Z. Zhao, Y. Guo, Z. Chen, and C. Guo. P3-sam: Native 3d part segmentation.arXiv preprint arXiv:2509.06784, 2025

    2025

  50. [58]

    Z. Ma, H. Chen, Y. Yue, and G. Gkioxari. Feedforward 3d editing via text-steerable image-to-3d, 2025. URLhttps://arxiv.org/abs/2512.13678

    2025

  51. [59]

    K. Mo, S. Zhu, A. X. Chang, L. Yi, S. Tripathi, L. J. Guibas, and H. Su. Partnet: A large- scale benchmark for fine-grained and hierarchical part-level 3d object understanding. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 909–918, 2019

    2019

  52. [60]

    X. Pan, S. N. Shukla, A. Singh, Z. Zhao, S. K. Mishra, J. Wang, Z. Xu, J. Chen, K. Li, F. Juefei-Xu, et al. Transfer between modalities with metaqueries.arXiv preprint arXiv:2504.06256, 2025

    2025 arXiv

  53. [61]

    Poole, A

    B. Poole, A. Jain, J. T. Barron, and B. Mildenhall. Dreamfusion: Text-to-3d using 2d diffusion.arXiv preprint arXiv:2209.14988, 2022

    2022 arXiv

  54. [62]

    Z. Qi, R. Dong, S. Zhang, H. Geng, C. Han, Z. Ge, L. Yi, and K. Ma. Shapellm: Universal 3d object understanding for embodied interaction. InEuropean Conference on Computer Vision, pages 214–238. Springer, 2024. 30 Tencent Hunyuan

    2024

  55. [63]

    Z. Qi, Y. Fang, Z. Sun, X. Wu, T. Wu, J. Wang, D. Lin, and H. Zhao. Gpt4point: A unified framework for point-language understanding and generation. InProceedings of the ieee/cvf conference on computer vision and pattern recognition, pages 26417–26427, 2024

    2024

  56. [64]

    Z. Qi, Y. Yang, M. Zhang, L. Xing, X. Wu, T. Wu, D. Lin, X. Liu, J. Wang, and H. Zhao. Tailor3d: Customized 3d assets editing and generation with dual-side images, 2024. URLhttps://arxiv.org/abs/2407.06191

    2024 arXiv

  57. [65]

    L. Qiu, G. Chen, X. Gu, Q. Zuo, M. Xu, Y. Wu, W. Yuan, Z. Dong, L. Bo, and X. Han. Richdreamer: A generalizable normal-depth diffusion model for detail richness in text-to-3d. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9914–9925, 2024

    2024

  58. [66]

    L. Qu, H. Zhang, Y. Liu, X. Wang, Y. Jiang, Y. Gao, H. Ye, D. K. Du, Z. Yuan, and X. Wu. Tokenflow: Unified image tokenizer for multimodal understanding and generation. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 2545–2555, 2025

    2025

  59. [67]

    Seedream, Y

    T. Seedream, Y. Chen, Y. Gao, L. Gong, M. Guo, Q. Guo, Z. Guo, X. Hou, W. Huang, Y. Huang, et al. Seedream 4.0: Toward next-generation multimodal image generation. arXiv preprint arXiv:2509.20427, 2025

    2025 arXiv

  60. [68]

    Sella, G

    E. Sella, G. Fiebelman, P . Hedman, and H. Averbuch-Elor. Vox-e: Text-guided voxel editing of 3d objects, 2023. URLhttps://arxiv.org/abs/2303.12048

    2023 arXiv

  61. [69]

    Sella, H

    E. Sella, H. Phung, N. Amiel, O. Litany, O. Patashnik, and H. Averbuch-Elor. Prox- e: Fine-grained 3d shape editing via primitive-based abstractions.arXiv preprint arXiv:2604.23774, 2026

    2026 arXiv

  62. [70]

    Y. Shi, P . Wang, J. Ye, M. Long, K. Li, and X. Yang. Mvdream: Multi-view diffusion for 3d generation.arXiv preprint arXiv:2308.16512, 2023

    2023 arXiv

  63. [71]

    J. Tang, J. Ren, H. Zhou, Z. Liu, and G. Zeng. Dreamgaussian: Generative gaussian splatting for efficient 3d content creation.arXiv preprint arXiv:2309.16653, 2023

    2023 arXiv

  64. [72]

    J. Tang, Z. Li, Z. Hao, X. Liu, G. Zeng, M.-Y. Liu, and Q. Zhang. Edgerunner: Auto- regressive auto-encoder for artistic mesh generation.arXiv preprint arXiv:2409.18114, 2024

    2024 arXiv

  65. [73]

    C. Team. Chameleon: Mixed-modal early-fusion foundation models.arXiv preprint arXiv:2405.09818, 2024. https://doi.org/10.48550/arXiv.2405.09818. URL https: //github.com/facebookresearch/chameleon

    2024 doi arXiv

  66. [74]

    S. Tong, D. Fan, J. Li, Y. Xiong, X. Chen, K. Sinha, M. Rabbat, Y. LeCun, S. Xie, and Z. Liu. Metamorph: Multimodal understanding and generation via instruction tuning. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 17001– 17012, 2025

    2025

  67. [75]

    Touvron, T

    H. Touvron, T. Lavril, G. Izacard, X. Martinet, M.-A. Lachaux, T. Lacroix, B. Rozi`ere, N. Goyal, E. Hambro, F. Azhar, et al. Llama: Open and efficient foundation language models.arXiv preprint arXiv:2302.13971, 2023

    2023 arXiv

  68. [76]

    Van Den Oord, O

    A. Van Den Oord, O. Vinyals, et al. Neural discrete representation learning.Advances in neural information processing systems, 30, 2017

    2017

  69. [77]

    C. Wang, J. Ye, Y. Yang, Y. Li, Z. Lin, J. Zhu, Z. Chen, Y. Luo, and C. Guo. Part-x-mllm: Part-aware 3d multimodal large language model.arXiv preprint arXiv:2511.13647, 2025

    2025

  70. [78]

    C. Wang, H. Weng, J. Ye, B. Lei, Y. Li, Z. Zhao, Z. Lai, K. Zhang, Y. Yang, Z. Chen, et al. Polyflow: Continuous topology embedding flow matching for artist-style mesh generation.arXiv preprint arXiv:2606.30673, 2026

    2026 arXiv

  71. [79]

    Wang, Y.-C

    H. Wang, Y.-C. Guo, Y.-T. Liu, Z.-X. Zou, B. Zhang, W. Quan, D. Liang, Y.-P . Cao, and D.-M. Yan. Face: A face-based autoregressive representation for high-fidelity and efficient mesh generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogniti...

    2026

  72. [80]

    P . Wang, Y. He, X. Lv, Y. Zhou, L. Xu, J. Yu, and J. Gu. PartNeXt: A next-generation dataset for fine-grained and hierarchical 3d part understanding. InAdvances in Neural Information Processing Systems, 2025. Datasets and Benchmarks Track

    2025

  73. [81]

    X. Wang, X. Zhang, Z. Luo, Q. Sun, Y. Cui, J. Wang, F. Zhang, Y. Wang, Z. Li, Q. Yu, et al. Emu3: Next-token prediction is all you need.arXiv preprint arXiv:2409.18869, 2024. 31 Tencent Hunyuan

    2024 arXiv

  74. [82]

    Z. Wang, C. Lu, Y. Wang, F. Bao, C. Li, H. Su, and J. Zhu. Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distillation.Advances in Neural Information Processing Systems, 36:8406–8441, 2023

    2023

  75. [83]

    H. Weng, Y. Wang, T. Zhang, C. Chen, and J. Zhu. Pivotmesh: Generic 3d mesh generation via pivot vertices guidance.arXiv preprint arXiv:2405.16890, 2024

    2024 arXiv

  76. [84]

    H. Weng, Z. Zhao, B. Lei, X. Yang, J. Liu, Z. Lai, Z. Chen, Y. Liu, J. Jiang, C. Guo, et al. Scaling mesh generation via compressive tokenization. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 11093–11103, 2025

    2025

  77. [85]

    J. Weng, S. Zhang, Z. Diao, P . Li, H. Zhang, J. Chen, and H. Zhao. Feedforward 3d editing learns from semantic-part transformation.arXiv preprint arXiv:2605.27351, 2026

    2026 arXiv

  78. [86]

    C. Wu, J. Li, J. Zhou, J. Lin, K. Gao, K. Yan, S.-m. Yin, S. Bai, X. Xu, Y. Chen, et al. Qwen-Image technical report.arXiv preprint arXiv:2508.02324, 2025

    2025 arXiv

  79. [87]

    C. Wu, P . Zheng, R. Yan, S. Xiao, X. Luo, Y. Wang, W. Li, X. Jiang, Y. Liu, J. Zhou, et al. Omnigen2: Exploration to advanced multimodal generation.arXiv preprint arXiv:2506.18871, 2025

    2025 arXiv

  80. [88]

    Y. Wu, Z. Zhang, J. Chen, H. Tang, D. Li, Y. Fang, L. Zhu, E. Xie, H. Yin, L. Yi, et al. Vila-u: a unified foundation model integrating visual understanding and generation. arXiv preprint arXiv:2409.04429, 2024

    2024 arXiv

  81. [89]

    Z. Wu, P . Zhou, X. Yi, X. Yuan, and H. Zhang. Consistent3d: Towards consistent high-fidelity text-to-3d generation with deterministic sampling prior. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9892–9902, 2024

    2024

  82. [91]

    R. Xia, Y. Tang, and P . Zhou. Towards scalable and consistent 3d editing.arXiv preprint arXiv:2510.02994, 2025

    2025

  83. [92]

    Xiang, Z

    J. Xiang, Z. Lv, S. Xu, Y. Deng, R. Wang, B. Zhang, D. Chen, X. Tong, and J. Yang. Structured 3d latents for scalable and versatile 3d generation. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 21469–21480, 2025

    2025

  84. [93]

    R. Xu, X. Wang, T. Wang, Y. Chen, J. Pang, and D. Lin. Pointllm: Empowering large language models to understand point clouds. InEuropean Conference on Computer Vision, pages 131–147. Springer, 2024

    2024

  85. [94]

    Y. Xu, H. Zhu, C. Liu, T. Wang, K. Chen, S. Xu, J. Yang, Q. Zhang, et al. Beyond voxel 3d editing: Learning from 3d masks and self-constructed data. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 635–646, 2026

    2026

  86. [95]

    H. Yan, Y. Li, Z. Wu, S. Chen, W. Sun, T. Shang, W. Liu, T. Chen, X. Dai, C. Ma, H. Li, and P . Ji. Frankenstein: Generating semantic-compositional 3d scenes in one tri-plane. InACM SIGGRAPH Asia Conference Proceedings, 2024

    2024

  87. [96]

    H. Yan, M. Zhang, Y. Li, C. Ma, and P . Ji. Phycage: Physically plausible compositional 3d asset generation from a single image, 2024

    2024

  88. [97]

    X. Yan, J. Xu, Y. Li, C. Ma, Y. Yang, C. Wang, Z. Zhao, Z. Lai, Y. Zhao, Z. Chen, and C. Guo. X-part: high fidelity and structure coherent shape decomposition.arXiv preprint arXiv:2509.08643, 2025

    2025

  89. [98]

    Y. Yang, Y. Zhou, Y.-C. Guo, Z.-X. Zou, Y. Huang, Y.-T. Liu, H. Xu, D. Liang, Y.-P . Cao, and X. Liu. Omnipart: Part-aware 3d generation with semantic decoupling and structural cohesion.arXiv preprint arXiv:2507.06165, 2025

    2025 arXiv

  90. [99]

    Y. Yang, C. Wang, J. Ye, Y. Li, Z. Chen, Z. Huang, Y. Mu, Z. Chen, C. Guo, and X. Liu. Physforge: Generating physics-grounded 3d assets for interactive virtual world.arXiv preprint arXiv:2605.05163, 2026

    2026 arXiv

  91. [100]

    C. Ye, C. Cao, C. Pan, Y. Hao, Y. Zhi, Y. Hu, and X. Han. Omni123: Exploring 3d native foundation models with limited 3d data by unifying text to 2d and 3d generation. arXiv preprint arXiv:2604.02289, 2026

    2026

  92. [101]

    J. Ye, F. Liu, Q. Li, Z. Wang, Y. Wang, X. Wang, Y. Duan, and J. Zhu. Dreamreward: Text-to-3d generation with human preference. InEuropean Conference on Computer Vision, pages 259–276. Springer, 2024

    2024

  93. [102]

    J. Ye, Z. Wang, R. Zhao, S. Xie, and J. Zhu. Shapellm-omni: A native multimodal llm for 3d generation and understanding.arXiv preprint arXiv:2506.01853, 2025. 32 Tencent Hunyuan

    2025 arXiv

  94. [103]

    J. Ye, S. Xie, R. Zhao, Z. Wang, H. Yan, W. Zu, L. Ma, and J. Zhu. Nano3d: A training- free approach for efficient 3d editing without masks.arXiv preprint arXiv:2510.15019, 2025

    2025

  95. [104]

    J. Ye, Z. Huang, Y. Qu, C. Wang, Y. Yang, Y. Li, Y. Luo, Z. Chen, S. Lu, J. Zhu, et al. Universe3d: Emerging properties of unified multimodal models in 3d understanding and generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6...

    2026

  96. [105]

    T. Yi, J. Fang, J. Wang, G. Wu, L. Xie, X. Zhang, W. Liu, Q. Tian, and X. Wang. Gaussian- dreamer: Fast generation from text to 3d gaussians by bridging 2d and 3d diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recogni- tion, pages 679...

    2024

  97. [106]

    Y. Yin, Y. Zhou, J. Wei, X. Yang, J. Zhang, J. Bai, J. Ye, W. Zhang, and G. Lin. Edit- verse3d: High-quality 3d object editing with region-aware learning.arXiv preprint arXiv:2607.07187, 2026

    2026 arXiv

  98. [107]

    Zhang, J

    B. Zhang, J. Tang, M. Niessner, and P . Wonka. 3dshape2vecset: A 3d shape representa- tion for neural fields and generative diffusion models.ACM Transactions On Graphics (TOG), 42(4):1–16, 2023

    2023

  99. [108]

    Zhang, Z

    L. Zhang, Z. Wang, Q. Zhang, Q. Qiu, A. Pang, H. Jiang, W. Yang, L. Xu, and J. Yu. Clay: A controllable large-scale generative model for creating high-quality 3d assets. ACM Transactions on Graphics (TOG), 43(4):1–20, 2024

    2024

  100. [109]

    Zhang, Z

    Y. Zhang, Z. Li, M. Zhou, S. Wu, and J. Wu. The scene language: Representing scenes with programs, words, and embeddings, 2025. URL https://arxiv.org/abs/2410. 16770

    2025

  101. [110]

    R. Zhao, J. Ye, Z. Wang, G. Liu, Y. Chen, Y. Wang, and J. Zhu. Deepmesh: Auto- regressive artist-mesh creation with reinforcement learning. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 10612–10623, 2025

    2025

  102. [111]

    R. Zhao, B. Li, Z. Liu, Y. Liang, J. Ye, F. Liu, D. Wu, Z. Wang, X. Yu, Y. Rao, et al. Gem: Generative supervision helps embodied intelligence.arXiv preprint arXiv:2605.28548, 2026

    2026 arXiv

  103. [112]

    T. Zhao, Y. Zhang, H. Long, J. Zhang, W. Li, Y. Yang, G. Zhang, J. Hladk`y, M. Nießner, and W. Yang. Lato: 3d mesh flow matching with structured topology preserving latents.arXiv preprint arXiv:2603.06357, 2026

    2026

  104. [113]

    Z. Zhao, W. Liu, X. Chen, X. Zeng, R. Wang, P . Cheng, B. Fu, T. Chen, G. Yu, and S. Gao. Michelangelo: Conditional 3d shape generation based on shape-image-text aligned latent representation.Advances in neural information processing systems, 36: 73969–73982, 2023

    2023

  105. [114]

    Zheng, M

    Y. Zheng, M. Huang, N. Chen, and Z. Mao. Pro3d-editor : A progressive-views perspective for consistent and precise 3d editing, 2025. URL https://arxiv.org/abs/ 2506.00512

    2025 arXiv

  106. [115]

    Z. Zhou, F. Ma, C. Gui, X. Xia, H. Fan, Y. Yang, and T.-S. Chua. Anchorflow: Training- free 3d editing via latent anchor-aligned flows. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, pages 14387–14397, 2026

    2026

  107. [116]

    Y. Zhu, K. Deng, J.-P . Fauconnier, I. Navarro, D. Li, A. Pun, Y. Zhang, P . Zhuang, X. Sun, M. Agrawala, K. Bhat, and T. Zhou. Cubepart: An open-vocabulary part-controllable 3d generator.arXiv preprint arXiv:2605.28763, 2026

    2026 arXiv

  108. [117]

    Zhuang, D

    J. Zhuang, D. Kang, Y.-P . Cao, G. Li, L. Lin, and Y. Shan. Tip-editor: An accurate 3d editor following both text-prompts and image-prompts, 2024. URL https://arxiv. org/abs/2401.14828. 33

    2024 arXiv

  109. [2025]

    URLhttps://arxiv.org/abs/2505.07003

    arXiv

  110. [2026]

    URLhttps://arxiv.org/abs/2602.03907

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.