pith. sign in

arxiv: 2606.12099 · v1 · pith:L37RNUW4new · submitted 2026-06-10 · 💻 cs.CV

ISAP-3D: Identity-Slot Aligned Part-Aware 3D Generation

Junlin Hao , Haoshuai Fu , Xibin Song , Wei Li , Ruigang Yang , Xinggong Zhang , Jinchuan Zhang This is my paper

Pith reviewed 2026-06-27 10:08 UTC · model grok-4.3

classification 💻 cs.CV
keywords part-aware 3D generationidentity-slot alignmentsemantic identity tokensone-to-one layout predictionstructured 3D objects3D shape synthesispart decomposition
0
0 comments X

The pith

Semantic identity tokens anchor parts to specific slots, eliminating permutation ambiguity in 3D generation.

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

The paper identifies identity-slot permutation freedom as the root of structural ambiguity in part-aware 3D generation, where multiple slot assignments can match the same supervision and produce inconsistent decompositions. It proposes that stable results require explicit identity-aligned one-to-one slot modelling rather than implicit inference of identities and layouts. ISAP-3D implements this by anchoring each part with semantic identity tokens, predicting layouts in a one-to-one manner conditioned on those identities, and then synthesizing geometry from the layouts while preserving alignment across stages. A new part-level dataset with a unified semantic protocol enables the learnable alignment. A sympathetic reader would care because the approach targets more reliable and controllable structured object synthesis without needing strong external layout conditions.

Core claim

We attribute this ambiguity to identity-slot permutation freedom: without explicit identity-slot alignment, the correspondence between semantic parts and generation slots is not identifiable during training, allowing multiple slot assignments to fit the same supervision and leading to inconsistent decomposition. Based on this insight, we argue that stable part-aware generation requires identity-aligned one-to-one slot modelling. We therefore propose an identity-slot aligned framework, ISAP-3D, which anchors each part with semantic identity tokens and performs identity-conditioned one-to-one layout prediction, followed by layout-conditioned geometry synthesis. Structured local-global conditio

What carries the argument

Semantic identity tokens that anchor each part to enable identity-conditioned one-to-one layout prediction before geometry synthesis.

If this is right

  • Part allocation becomes stable without slot swapping or merging.
  • Controllability improves because identities directly condition the layout stage.
  • Alignment is preserved from semantic tokens through spatial layout to final geometry.
  • A unified semantic protocol in the dataset supports consistent identity-slot mapping across objects.

Where Pith is reading between the lines

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

  • Similar explicit anchoring could address assignment ambiguities in other structured generation tasks such as scene layout or multi-object synthesis.
  • The dataset construction approach may serve as a template for standardizing part semantics in future 3D benchmarks.
  • If identity alignment reduces training variance, it could lower the data requirements for learning consistent part decompositions.

Load-bearing premise

That structural ambiguity is caused primarily by the lack of explicit identity-slot alignment during training.

What would settle it

Training an otherwise identical model without the identity tokens on the same part-level dataset and checking whether slot swapping or merging reappears in the outputs.

Figures

Figures reproduced from arXiv: 2606.12099 by Haoshuai Fu, Jinchuan Zhang, Junlin Hao, Ruigang Yang, Wei Li, Xibin Song, Xinggong Zhang.

Figure 1
Figure 1. Figure 1: Identity-slot aligned part-aware 3D generation. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of our identity-aligned part generation framework. Given a reference image and semantic part descriptions, we first construct identity-aware condition to￾kens encoding part identities and global context. An Identity-Conditioned transformer predicts spatial layouts for each part, which are then used by a Layout-Conditioned geometry synthesis network to generate part geometries. identity and spatial… view at source ↗
Figure 3
Figure 3. Figure 3: Identity-aware condition and layout prediction. (a) Multi-modal inputs are encoded into identity conditions and global context tokens. (b) An identity-conditioned transformer predicts part bounding boxes using identity-aligned slot queries. parts, the goal is to synthesize a set of per-part geometries G = {Gk} K k=1, where each geometry Gk is semantically aligned with its corresponding inputs. To facilitat… view at source ↗
Figure 4
Figure 4. Figure 4: Layout-conditioned geometry synthesis. Predicted bounding boxes partition the voxel prior into part layout voxels, which are processed by a geometry flow transformer with local and global attention (L-MHA and G-MHA) to preserve identity consistency across stages. Filtered voxels are then decoded to generate the final part geometries. Layout-conditioned geometry generation. As illustrated in [PITH_FULL_IMA… view at source ↗
Figure 5
Figure 5. Figure 5: Comparison with OmniPart under the same identity condition across view￾points. Colors represent generation slot IDs. OmniPart assigns the same semantic part to different slots across viewpoints, leading to inconsistent part identities. In contrast, our identity-aligned design preserves stable slot assignments, producing consistent part decompositions. eration. PartField [17] performs post-hoc part segmenta… view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative comparison with representative part-aware generation methods. Our identity-aligned model produces semantically coherent and well-separated parts, while prior methods often generate ambiguous or merged components. Qualitative Results. We first compare our method with OmniPart [36], the strongest baseline with explicit identity conditioning, as shown in [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Training condition ablation. Removing voxel, text, or mask conditions weakens spatial grounding or identity supervision, leading to degraded layout prediction to the lack of explicit identity modeling, leading to ambiguous part boundaries or merged components. In contrast, our identity-aligned formulation generates semantically coherent parts with clearer boundaries and more stable structures. Quantitative… view at source ↗
Figure 8
Figure 8. Figure 8: Semantic compositional control. Different textual identity combinations lead to different part granularities, demonstrating controllable generation enabled by iden￾tity–slot alignment. Input Image Text + Image Text + Image + Mask Voxel + Bbox [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Inference condition ablation. Our model remains robust under weak conditions (image + text) and supports finer controllability when stronger identity cues such as masks or ground-truth layouts are provided. Removing voxel features causes the largest performance drop, as the model loses reliable spatial anchors for layout prediction. Without text identity, the model must rely solely on visual masks, which c… view at source ↗
read the original abstract

Part-aware 3D generation aims to synthesize structured objects with semantically meaningful components, yet often suffers from structural ambiguity due to identity-layout entanglement. Existing methods either infer part identity and spatial layout implicitly, which can lead to unstable part allocation (e.g., slot swapping or part merging), or rely on strong layout conditions that are difficult to obtain in practice. We attribute this ambiguity to identity-slot permutation freedom: without explicit identity-slot alignment, the correspondence between semantic parts and generation slots is not identifiable during training, allowing multiple slot assignments to fit the same supervision and leading to inconsistent decomposition. Based on this insight, we argue that stable part-aware generation requires identity-aligned one-to-one slot modelling. We therefore propose an identity-slot aligned framework, ISAP-3D, which anchors each part with semantic identity tokens and performs identity-conditioned one-to-one layout prediction, followed by layout-conditioned geometry synthesis. Structured local-global conditioning maintains identity alignment across semantic, spatial, and geometric stages. We also construct a part-level dataset with a unified semantic protocol to enable learnable and consistent identity-slot alignment. Extensive experiments demonstrate improved structural stability, controllability, and robustness over state-of-the-art part-aware generation baselines.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 2 minor

Summary. The paper claims that structural ambiguity in part-aware 3D generation arises from identity-slot permutation freedom, which prevents identifiable correspondence between semantic parts and generation slots during training. It proposes the ISAP-3D framework that anchors parts using semantic identity tokens, performs identity-conditioned one-to-one layout prediction, and follows with layout-conditioned geometry synthesis under structured local-global conditioning to maintain alignment. A part-level dataset with unified semantic protocol is introduced to support learnable alignment, and extensive experiments are reported to show gains in structural stability, controllability, and robustness over baselines.

Significance. If the experimental claims hold, the work could be significant for structured 3D generation by offering an explicit mechanism to resolve permutation symmetry in slot-based models, a recurring issue in part-aware synthesis. The dataset construction with unified semantics is a concrete enabling contribution that may support future consistent part-level modeling.

minor comments (2)
  1. Abstract: the phrase 'extensive experiments demonstrate improved structural stability' is stated without any accompanying metrics, table references, or baseline comparisons; the full experimental section should include specific quantitative results (e.g., IoU, stability scores) to substantiate the central claim.
  2. The introduction of 'semantic identity tokens' and 'identity-conditioned one-to-one layout prediction' would benefit from an early equation or diagram in §3 showing how the tokens are embedded and how the one-to-one constraint is enforced in the loss.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the careful summary of our work and the positive assessment of its potential significance. We note the recommendation for minor revision. No specific major comments were provided in the report, so we have no individual points to address at this time. We remain available to incorporate any additional feedback.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper attributes structural ambiguity to identity-slot permutation freedom and proposes identity-aligned one-to-one slot modeling as the remedy. This is framed as an independent modeling insight and architectural choice (semantic anchoring followed by conditioned layout and geometry stages) rather than any derivation that reduces by construction to fitted inputs, self-definitions, or self-citation chains. No equations, fitted parameters renamed as predictions, or load-bearing self-citations appear in the provided text. The dataset construction is presented as an enabling step, not a hidden assumption that collapses the claim. The central pipeline remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the domain assumption that permutation freedom is the primary source of ambiguity and on the introduction of new modeling constructs (identity tokens, one-to-one conditioning) whose effectiveness is asserted without independent verification in the provided text.

axioms (1)
  • domain assumption Identity-slot permutation freedom is the cause of structural ambiguity (slot swapping, part merging) in part-aware 3D generation.
    Explicitly stated in the abstract as the attribution for the observed instability.
invented entities (1)
  • semantic identity tokens no independent evidence
    purpose: Anchor each part with a fixed semantic identity to enable one-to-one slot alignment.
    Introduced as a core component of the ISAP-3D framework; no independent evidence of existence or utility is supplied.

pith-pipeline@v0.9.1-grok · 5765 in / 1457 out tokens · 18960 ms · 2026-06-27T10:08:09.087814+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

41 extracted references · 9 linked inside Pith

  1. [1]

    arXiv preprint arXiv:2401.05293 (2024)

    Alldieck, T., Kolotouros, N., Sminchisescu, C.: Score distillation sampling with learned manifold corrective. arXiv preprint arXiv:2401.05293 (2024)

    arXiv 2024

  2. [2]

    In: CVPR

    Chen, M., Shapovalov, R., Laina, I., Monnier, T., Wang, J., Novotny, D., Vedaldi, A.: Partgen: Part-level 3d generation and reconstruction with multi-view diffusion models. In: CVPR. pp. 5881–5892 (June 2025)

    2025

  3. [3]

    arXiv preprint arXiv:2507.13346 (2025)

    Chen, M., Wang, J., Shapovalov, R., Monnier, T., Jung, H., Wang, D., Ranjan, R., Laina, I., Vedaldi, A.: Autopartgen: Autogressive 3d part generation and discovery. arXiv preprint arXiv:2507.13346 (2025)

    arXiv 2025

  4. [4]

    In: CVPR

    Chen, R., Zhang, J., Liang, Y., Luo, G., Li, W., Liu, J., Li, X., Long, X., Feng, J., Tan, P.: Dora: Sampling and benchmarking for 3d shape variational auto-encoders. In: CVPR. pp. 16251–16261 (June 2025) Abbreviated paper title 15

    2025

  5. [5]

    arXiv preprint arXiv:2307.05663 (2023)

    Deitke, M., Liu, R., Wallingford, M., Ngo, H., Michel, O., Kusupati, A., Fan, A., Laforte, C., Voleti, V., Gadre, S.Y., VanderBilt, E., Kembhavi, A., Vondrick, C., Gkioxari, G., Ehsani, K., Schmidt, L., Farhadi, A.: Objaverse-xl: A universe of 10m+ 3d objects. arXiv preprint arXiv:2307.05663 (2023)

    Pith/arXiv arXiv 2023

  6. [6]

    In: CVPR

    Deitke, M., Schwenk, D., Salvador, J., Weihs, L., Michel, O., VanderBilt, E., Schmidt, L., Ehsani, K., Kembhavi, A., Farhadi, A.: Objaverse: A universe of annotated 3d objects. In: CVPR. pp. 13142–13153 (June 2023)

    2023

  7. [7]

    arXiv preprint arXiv:2510.26140 (2025)

    Ding, L., Dong, S., Li, Y., Gao, C., Chen, X., Han, R., Kuang, Y., Zhang, H., Huang, B., Huang, Z., Wang, Z., Xu, D., Xue, T.: Fullpart: Generating each 3d part at full resolution. arXiv preprint arXiv:2510.26140 (2025)

    arXiv 2025

  8. [8]

    In: AAAI

    Ding, Y., Zhuang, S., Li, K., Yue, Z., Qiao, Y., Wang, Y.: Muses: 3d-controllable image generation via multi-modal agent collaboration. In: AAAI. vol. 39, pp. 2753– 2761 (2025)

    2025

  9. [9]

    In: ICCV (2025)

    Dong, S., Ding, L., Chen, X., Li, Y., WANG, Y., Wang, Y., WANG, Q., Kim, J., Gao, C., Huang, Z., Wang, Z., Xue, T., Xu, D.: From one to more: Contextual part latents for 3d generation. In: ICCV (2025)

    2025

  10. [10]

    arXiv preprint arXiv:2503.21732 (2025)

    He, X., Zou, Z.X., Chen, C.H., Guo, Y.C., Liang, D., Yuan, C., Ouyang, W., Cao, Y.P., Li, Y.: Sparseflex: High-resolution and arbitrary-topology 3d shape modeling. arXiv preprint arXiv:2503.21732 (2025)

    arXiv 2025

  11. [11]

    arXiv preprint arXiv:2512.09435 (2025)

    He, X., Wu, Y., Guo, X., Ye, C., Zhou, J., Hu, T., Han, X., Du, D.: Uni- part: Part-level 3d generation with unified 3d geom-seg latents. arXiv preprint arXiv:2512.09435 (2025)

    arXiv 2025

  12. [12]

    arXiv preprint arXiv:2506.16504 (2025)

    Lai, Z., Zhao, Y., Liu, H., Zhao, Z., Lin, Q., Shi, H., Yang, X., Yang, M., Yang, S., Feng, Y., Zhang, S., Huang, X., Luo, D., Yang, F., Yang, F., Wang, L., Liu, S., Tang, Y., Cai, Y., He, Z., Liu, T., Liu, Y., Jiang, J., Linus, Huang, J., Guo, C.: Hunyuan3d 2.5: Towards high-fidelity 3d assets generation with ultimate details. arXiv preprint arXiv:2506.1...

    Pith/arXiv arXiv 2025

  13. [13]

    arXiv preprint arXiv:2512.03052 (2025)

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

    arXiv 2025

  14. [14]

    arXiv preprint arXiv:2502.06608 (2025)

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

    Pith/arXiv arXiv 2025

  15. [15]

    arXiv preprint arXiv:2512.07628 (2025)

    Li, Z., Li, W., Wang, T., Wang, Z., Wu, J., Wang, H., Yang, Y., Huang, Z., Li, Y., Guo,C.,Liu,P.:Moca:Mixture-of-componentsattentionforscalablecompositional 3d generation. arXiv preprint arXiv:2512.07628 (2025)

    arXiv 2025

  16. [16]

    Lin, Y., Lin, C., Pan, P., Yan, H., Feng, Y., Mu, Y., Fragkiadaki, K.: Partcrafter: Structured 3d mesh generation via compositional latent diffusion transformers (2025)

    2025

  17. [17]

    In: ICCV (2025)

    Liu, M., Uy, M.A., Xiang, D., Su, H., Fidler, S., Sharp, N., Gao, J.: Partfield: Learning 3d feature fields for part segmentation and beyond. In: ICCV (2025)

    2025

  18. [18]

    arXiv preprint arXiv:2303.11328 (2023)

    Liu, R., Wu, R., Hoorick, B.V., Tokmakov, P., Zakharov, S., Vondrick, C.: Zero-1- to-3: Zero-shot one image to 3d object. arXiv preprint arXiv:2303.11328 (2023)

    arXiv 2023

  19. [19]

    arXiv preprint arXiv:2309.03453 (2023)

    Liu, Y., Lin, C., Zeng, Z., Long, X., Liu, L., Komura, T., Wang, W.: Syncdreamer: Generating multiview-consistent images from a single-view image. arXiv preprint arXiv:2309.03453 (2023)

    Pith/arXiv arXiv 2023

  20. [20]

    In: CVPR

    Long, X., Guo, Y.C., Lin, C., Liu, Y., Dou, Z., Liu, L., Ma, Y., Zhang, S.H., Habermann, M., Theobalt, C., Wang, W.: Wonder3d: Single image to 3d using cross-domain diffusion. In: CVPR. pp. 9970–9980 (June 2024) 16 Authors Suppressed Due to Excessive Length

    2024

  21. [21]

    arXiv preprint arXiv:2509.06784 (2025)

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

    arXiv 2025

  22. [22]

    arXiv preprint arXiv:2209.14988 (2022)

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

    Pith/arXiv arXiv 2022

  23. [23]

    arXiv preprint arXiv:2308.16512 (2023)

    Shi, Y., Wang, P., Ye, J., Mai, L., Li, K., Yang, X.: Mvdream: Multi-view diffusion for 3d generation. arXiv preprint arXiv:2308.16512 (2023)

    Pith/arXiv arXiv 2023

  24. [24]

    arXiv preprint arXiv:2506.09980 (2025)

    Tang, J., Lu, R., Li, Z., Hao, Z., Li, X., Wei, F., Song, S., Zeng, G., Liu, M.Y., Lin, T.Y.: Efficient part-level 3d object generation via dual volume packing. arXiv preprint arXiv:2506.09980 (2025)

    arXiv 2025

  25. [25]

    arXiv preprint arXiv:2501.12202 (2025)

    Team, T.H.: Hunyuan3d 2.0: Scaling diffusion models for high resolution textured 3d assets generation. arXiv preprint arXiv:2501.12202 (2025)

    Pith/arXiv arXiv 2025

  26. [26]

    arXiv preprint arXiv:2506.15442 (2025)

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

    Pith/arXiv arXiv 2025

  27. [27]

    In: CVPR

    Wang, H., Du, X., Li, J., Yeh, R.A., Shakhnarovich, G.: Score jacobian chaining: Lifting pretrained 2d diffusion models for 3d generation. In: CVPR. pp. 12619– 12629 (2023)

    2023

  28. [28]

    arXiv preprint arXiv:2602.22785 (2026)

    Wang, L., Guo, H.X., Wang, X., Sun, F., Sun, K., Liu, P., Xiao, H., Wang, Z., Fu, G., Li, E., Liu, Y., Wang, Y.: Scenetransporter: Optimal transport-guided com- positional latent diffusion for single-image structured 3d scene generation. arXiv preprint arXiv:2602.22785 (2026)

    arXiv 2026

  29. [29]

    arXiv preprint arXiv:2305.16213 (2023)

    Wang, Z., Lu, C., Wang, Y., Bao, F., Li, C., Su, H., Zhu, J.: Prolificdreamer: High- fidelity and diverse text-to-3d generation with variational score distillation. arXiv preprint arXiv:2305.16213 (2023)

    arXiv 2023

  30. [30]

    arXiv preprint arXiv:2505.17412 (2025)

    Wu, S., Lin, Y., Zhang, F., Zeng, Y., Yang, Y., Bao, Y., Qian, J., Zhu, S., Torr, P., Cao, X., Yao, Y.: Direct3d-s2: Gigascale 3d generation made easy with spatial sparse attention. arXiv preprint arXiv:2505.17412 (2025)

    arXiv 2025

  31. [31]

    In: CVPR

    Xiang, J., Lv, Z., Xu, S., Deng, Y., Wang, R., Zhang, B., Chen, D., Tong, X., Yang, J.: Structured 3d latents for scalable and versatile 3d generation. In: CVPR. pp. 21469–21480 (June 2025)

    2025

  32. [32]

    arXiv preprint arXiv:2509.08643 (2025)

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

    arXiv 2025

  33. [33]

    arXiv preprint arXiv:2511.10040 (2025)

    Yang, X., Lai, S., Lyu, J., Li, H., Pan, B., Li, Y., Guo, J., Zhou, Z., Guo, Y.: Log3d: Ultra-high-resolution 3d shape modeling via local-to-global partitioning. arXiv preprint arXiv:2511.10040 (2025)

    arXiv 2025

  34. [34]

    arXiv preprint arXiv:2511.18801 (2025)

    Yang, Y., Li, H., Zhu, H., Yang, L., Lei, G., Xu, S., Zhang, B.: Partdiffuser: Part- wise 3d mesh generation via discrete diffusion. arXiv preprint arXiv:2511.18801 (2025)

    Pith/arXiv arXiv 2025

  35. [35]

    arXiv preprint arXiv:2504.07943 (2025)

    Yang, Y., Guo, Y.C., Huang, Y., Zou, Z.X., Yu, Z., Li, Y., Cao, Y.P., Liu, X.: Holopart: Generative 3d part amodal segmentation. arXiv preprint arXiv:2504.07943 (2025)

    arXiv 2025

  36. [36]

    arXiv preprint arXiv:2507.06165 (2025)

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

    arXiv 2025

  37. [37]

    ACM Trans

    Zhang, B., Tang, J., Nießner, M., Wonka, P.: 3dshape2vecset: A 3d shape represen- tation for neural fields and generative diffusion models. ACM Trans. Graph.42(4) (jul 2023)

    2023

  38. [38]

    arXiv preprint arXiv:2406.13897 (2024) Abbreviated paper title 17

    Zhang, L., Wang, Z., Zhang, Q., Qiu, Q., Pang, A., Jiang, H., Yang, W., Xu, L., Yu, J.: Clay: A controllable large-scale generative model for creating high-quality 3d assets. arXiv preprint arXiv:2406.13897 (2024) Abbreviated paper title 17

    arXiv 2024

  39. [39]

    ACM Transactions on Graphics44(4), 1–21 (Jul 2025)

    Zhang, L., Zhang, Q., Jiang, H., Bai, Y., Yang, W., Xu, L., Yu, J.: Bang: Dividing 3d assets via generative exploded dynamics. ACM Transactions on Graphics44(4), 1–21 (Jul 2025)

    2025

  40. [40]

    In: NeurIPS (2023)

    Zhao, Z., Liu, W., Chen, X., Zeng, X., Wang, R., Cheng, P., FU, B., Chen, T., YU, G., Gao, S.: Michelangelo: Conditional 3d shape generation based on shape- image-text aligned latent representation. In: NeurIPS (2023)

    2023

  41. [41]

    In: CVPR

    Zhu, D., Di, Y., Gavranovic, S., Ilic, S.: Sealion: Semantic part-aware latent point diffusion models for 3d generation. In: CVPR. pp. 11789–11798 (2025)

    2025