Pith. sign in

REVIEW 5 major objections 6 minor 60 references

ArtFormer: Controllable Generation of Diverse 3D Articulated Objects

T0 review · 5 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read ArtFormer claims that a transformer over tree-structured tokens, paired with a diffusion-trained SDF shape prior, can generate diverse 3D articulated objects from text with varying part counts and precise joints.

desk verdict A genuinely new method for text-guided 3D articulated object generation with a tree-of-tokens transformer and codebook-sampled SDF prior, but evaluation gaps (no error bars, no MeshArt comparison, no decoding-validity check) keep the SOTA claim from being fully proven. read the letter →

arxiv 2412.07237 v3 pith:IAGJ6AWM submitted 2024-12-10 cs.CV cs.AIcs.RO

classification cs.CVcs.AIcs.RO
keywords 3Darticulatedobjectstext-conditionedgenerationshapepriorsigneddistancefunctiontree-structuretransformerkinematicjointspartimage-conditioned
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

ArtFormer proposes a way to generate 3D articulated objects — objects made of rigid parts connected by joints — from a text description, with the number of parts and their geometry left free rather than fixed in advance. The paper's claim is that this approach resolves a flexibility-quality tradeoff that has split earlier methods: some generate the kinematics only by retrieving geometry from a fixed dataset, while others generate geometry but struggle with precise joints and part structure. ArtFormer instead represents each articulated object as a tree of tokens, one per sub-part, and uses a transformer to predict the tree while a signed-distance-function shape prior decodes each part's geometry from a sampled latent code. Evaluated on six categories, the framework reports lower part-overlap, better coverage of the data distribution, and higher human-judged alignment and diversity than the baselines, and it also works with image conditioning. If correct, the method supplies a path to create novel, usable articulated assets for simulation and robot learning without manual part assembly.

What carries the argument

The tree-of-tokens parameterization: an articulated object is written as a tree in which each node stores one sub-part's bounding box, geometry latent code, joint axis, and joint limits, plus its parent's index. The articulation transformer generates this tree autoregressively, using a GRU-derived tree position embedding so the model knows each token's place in the hierarchy, and iterative decoding that predicts child nodes for every open node until all nodes are terminal. The frozen SDF shape prior is the other half: it is trained as a variational autoencoder on point clouds and SDF values, with a diffusion model over the latent space, so that a compact code sampled by the transformer decodes into a high-quality mesh by querying signed distances and running marching cubes. Codebook Gumbel-Softmax sampling of the geometry condition enlarges the effective latent space and injects the stochasticity that yields diverse part shapes.

What would settle it

Train the full pipeline on an enlarged set of categories (or on objects with more than ten parts) and measure part-overlapping ratio and per-part Chamfer distance to the training set; the central claim fails if adding categories collapses reconstruction quality, or if generated sub-parts turn out to be no farther from the training set than typical training-set neighbors are from each other, showing novelty is illusory.

Watch

Extended reading notes

Core claim

The central discovery is that a transformer working on a tree-of-tokens parameterization can jointly predict a part hierarchy and its joint geometry, as long as the heavy lifting of geometry is handed to a separate diffusion-trained SDF shape prior. Each node of the tree is a token carrying a bounding box, a 768-dimensional geometry latent code, a joint axis, joint limits, and the parent index; iterative decoding grows the tree level by level until every node emits a terminal token. The transformer's tree position embedding, built by running a GRU along the root-to-node path, gives the model structural awareness that ordinary sequential positional encodings lack. The shape prior, trained on PartNet and PartNet-Mobility, lets the transformer output compact codes plus discrete codebook-sampled conditions that the SDF decoder turns into high-quality meshes. In comparison with retrieval-based and direct-geometry baselines, the paper reports that ArtFormer generates more diverse objects with precise kinematics and less interpenetration, and produces geometry not present in the training set.

Load-bearing premise

The load-bearing premise is that the frozen signed-distance shape prior, trained on PartNet and PartNet-Mobility, has a latent space rich and general enough that arbitrary codes picked by the transformer decode into valid, category-appropriate geometry; the paper's own limitations section notes that adding object categories degrades geometry reconstruction because the SDF model generalizes poorly beyond the six trained categories.

Editorial extensions

If this is right

  • Text descriptions can specify articulation behavior — drawers that slide, doors that swing — and the model produces a matching tree with the right number of parts rather than a fixed graph.
  • Generated geometry is sampled, not retrieved, so outputs can be shapes absent from the training dataset, which directly addresses the diversity ceiling of retrieval-based systems.
  • Because decoding is autoregressive and per-part, existing articulated objects can be edited: a sub-part can be removed and regenerated under a new text instruction.
  • Swapping the text encoder for an image encoder yields single-image-conditioned generation, indicating the same tree/token formulation supports multiple input modalities.
  • If the framework scales to larger part counts, it becomes a practical generator of simulation-ready articulated assets for robot learning and digital-twin applications.

Reading between the lines

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

  • The four-codebook construction suggests diversity should grow combinatorially with codebook size; a direct test would be to vary codebook temperature or size and measure part-shape coverage.
  • The paper's own limitation note links category scaling to SDF generalization; a natural extension is to pair the transformer with a more generalizable SDF backbone to widen the category range beyond six.
  • The same token-tree recipe could be applied to other hierarchical functional objects — mechanical assemblies, furniture kits, modular robots — wherever part structure and joint behavior must be generated together.
  • Part editing via iterative decoding hints at a controllable inpainting tool for robot simulators: specify which joints remain and let the model fill in the rest.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes ArtFormer, a transformer-based framework for conditional generation of 3D articulated objects from text or image descriptions. Articulated objects are represented as trees of tokens, each token containing a bounding box, a joint axis and limit, a parent index, and a 768-dimensional geometry latent code. A frozen SDF shape prior, trained in a Diffusion-SDF style, decodes the geometry from latent codes, while the articulation transformer predicts kinematic attributes and codebook logits used to sample geometry conditions. The method introduces a tree position embedding and an iterative decoding procedure to generate objects with variable part counts and structure. Experiments on PartNet-Mobility categories compare ArtFormer against modified NAP and CAGE baselines using MMD, COV, 1-NNA, POR, a human study, and CLIP-R, with additional ablations and image-guided generation results. The central claim is that ArtFormer enables more diverse generation of articulated objects with high-quality geometry and precise kinematic features.

Significance. If the claims are substantiated, the paper makes a useful contribution by decoupling articulation-structure generation from geometry decoding via a frozen shape prior, and by proposing a tree-token parameterization with a dedicated tree position embedding. The code release, the inclusion of a human study, and the ablation of the tree position embedding and shape prior are concrete strengths. However, the comparative evaluation currently has gaps that limit the strength of the performance claims: no error bars or significance tests are reported, the baselines are modified versions rather than the original methods, and the validity of geometry decoded from codebook-sampled codes is not directly measured. These issues are load-bearing for the claimed state-of-the-art results.

major comments (5)
  1. [Sec. 4.3-4.4, Table 1] No error bars or statistical significance tests are reported for any metric. The differences between Ours and NAP-768 (e.g., MMD 0.0292 vs 0.0632, COV 0.5213 vs 0.3723, 1-NNA 0.5266 vs 0.6543) are large, but without repeated generation runs, bootstrap intervals, or paired significance tests it is unclear whether these differences are stable. Please report mean and standard deviation over at least three independent generation runs and add significance tests (e.g., bootstrap or paired tests) for MMD, COV, 1-NNA, and the human-study scores.
  2. [Sec. 4.2, Table 1] The compared baselines are modified versions of NAP and CAGE that use the proposed shape prior or a modified retrieval scheme, while the original NAP and CAGE are not evaluated. Since the paper claims to outperform these methods, the comparison should include the original unmodified baselines, or a detailed justification for the modifications and evidence that the modifications do not disadvantage the baselines. In addition, MeshArt [8], the closest cited transformer-based baseline, is not compared; please discuss this omission or include it in the evaluation.
  3. [Sec. 3.2-3.3, Fig. 5, Limitation (4)] The frozen SDF shape prior is the sole source of generated geometry, and the articulation transformer only emits codebook logits P and semantic codes c_s that condition the diffusion decoder. No experiment directly measures whether codebook-sampled codes decode into valid, category-appropriate geometry: e.g., the fraction of watertight meshes, the rate of spurious disconnected components, or consistency between the decoded part geometry and the predicted part semantics. Limitation (4) explicitly concedes that increasing object categories degrades geometry reconstruction due to limited SDF generalization, so this validity check is necessary to support the claim of high-quality geometry and diversity. Please add a quantitative evaluation of decoded meshes as a function of the predicted codebook distributions.
  4. [Sec. 7.1, Fig. 10] The 'novel shape generation' experiment is largely circular: it measures Chamfer distance between generated sub-parts and training-set sub-parts, but any continuous latent sampler will produce points that are not exactly in the training set. This does not establish that the generated shapes are semantically novel, plausible, or category-appropriate. Please complement the distance-based evaluation with a human rating or a category-consistency metric.
  5. [Sec. 4.1, Table 1] The paper does not report the train/test split, the number of objects per category, or the number of generated samples used to compute each metric. This information is needed to interpret MMD, COV, and 1-NNA, since these metrics are sensitive to sample size and test-set composition. Please specify these details, and if the test set is small, temper the corresponding conclusions.
minor comments (6)
  1. [Eq. (5)] The index range in the Gumbel-Softmax equation is slightly confusing; please clarify that the sum runs over the N codebook entries. Also report the value of the temperature tau used and a short sensitivity analysis, since tau directly controls the diversity of the sampled geometry conditions.
  2. [Sec. 4.4, Fig. 6] The image-guided generation results are presented only qualitatively; please either add quantitative metrics for this setting or explicitly state that it is a proof-of-concept demonstration.
  3. [Sec. 6.3] The 26.10% failure rate of the GPT-4o description generation is nontrivial; please report how failures are distributed across categories and discuss whether this biases the text-conditioned training set.
  4. [Table 2] Removing the tree position embedding improves MMD while worsening POR and COV; one sentence explaining this non-monotonic effect would help readers interpret the ablation.
  5. [Sec. 4.2] The description of the baseline modifications is too brief for reproduction; please provide the modified architectures (e.g., where the cross-attention layer is inserted, training schedules, hyperparameters) in the supplementary material.
  6. [Supplementary Sec. 7.4] The CLIP-R experiment reports only R=10; reporting R=1, R=5, and R=10 with confidence intervals would make the alignment claim more robust.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the generation pipeline is trained on external data and evaluated against ground-truth instances, with no load-bearing self-citation or fitted-input-as-prediction step.

full rationale

The paper's derivation chain is self-contained and does not reduce to its inputs. The shape prior (Sec. 3.2) is trained on PartNet and PartNet-Mobility with VAE and diffusion objectives (Eqs. 1-2), and the articulation transformer is trained on six PartNet-Mobility categories with explicit losses on part-attribute tokens and codebook logits (Eqs. 8-11). The main generation-quality claims are evaluated against ground-truth instances using MMD, COV, 1-NNA, POR, and a human study (Secs. 4.3-4.4), i.e. against external test data rather than against fitted constants. The codebook sampling in Eqs. (3)-(5) is a standard stochastic-quantization mechanism, and P is trained by a KL divergence against a data-derived distance distribution, not defined in terms of any evaluation metric. There is no self-citation chain: the cited Diffusion-SDF, NAP, CAGE, and SINGAPO works have disjoint author sets from the present paper, and no 'uniqueness theorem' is invoked to force the design. The Sec. 7.1 'novel shape' experiment measures Chamfer distance from generated sub-parts to the training set; while this operationalizes novelty relative to the training set rather than an independent benchmark, it is not a circular derivation because the metric is computed on generated outputs and is not part of any training objective. At most it is a weak evidence test, not a circular step. No predicted quantity is fitted and then renamed as a prediction, and no equation is equivalent to its own output by construction.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central claim depends on the learned shape prior, the tree parameterization, the GPT-4o generated text conditions, and several hand-chosen hyperparameters. There are no invented physical entities. The most consequential assumptions are the generalization of the SDF prior and the adequacy of the ChatGPT-4o descriptions as training conditions.

free parameters (5)
  • Gumbel-Softmax temperature tau = not specified
    Controls diversity of codebook sampling in Eq. 5; the authors state it is not tuned in this work (Sec. 3.2).
  • VAE KL regularization weight beta = not specified
    Balances reconstruction and Gaussian prior in Eq. 1; chosen by hand, value not reported.
  • Loss weights beta_o and beta_P = not specified
    Balance terminal-token loss and codebook prior loss in Eq. 11; chosen by hand, values not reported.
  • Codebook size N and number of codebooks = N not stated; 4 codebooks
    Discrete latent capacity affects diversity since the paper claims capacity increases from 4N to N^4, yet N is never specified.
  • Latent dimensions = z=768, c_g=64, c_s=32
    Architecture dimensions chosen by hand and affect model capacity and geometry fidelity.
assumptions (6)
  • standard math Transformer, diffusion, and Gumbel-Softmax training converge under the reported schedules.
    Invoked throughout Sec. 3.2 and 3.3 without proof; relies on standard deep learning practice.
  • domain assumption PartNet-Mobility annotations for segmentation, joints, and limits are accurate and complete.
    The entire training pipeline in Sec. 4.1 uses these annotations as ground truth.
  • domain assumption The SDF shape prior trained on PartNet and PartNet-Mobility generalizes to all six target categories.
    Sec. 3.2 and 4.1; the Limitations section admits this generalization is fragile when categories are added.
  • domain assumption Every object of interest can be represented as a rooted tree with each part having at most one parent.
    Sec. 3.1 defines the parameterization as a tree, which excludes articulated structures with loops or multi-parent dependencies.
  • domain assumption ChatGPT-4o generated descriptions, after excluding a 26.10% failure rate, provide valid conditioning text.
    Supplementary Sec. 6.3 describes the data generation and failure exclusion, but the effect of this filtering on the conditional distribution is not analyzed.
  • ad hoc to paper Chunking the geometry condition into 4 codebook segments preserves geometric fidelity while increasing diversity.
    Sec. 3.2 introduces the 4-chunk codebook design as the diversity mechanism; its adequacy is empirically probed only through the one-codebook ablation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ArtFormer: Controllable Generation of Diverse 3D Articulated Objects." pith.science (2026). https://pith.science/paper/IAGJ6AWM

@misc{pith2026241207237,
  author       = {Pith},
  title        = {Pith review of: ArtFormer: Controllable Generation of Diverse 3D Articulated Objects},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IAGJ6AWM}},
  note         = {Machine review of arXiv:2412.07237}
}
read the original abstract

This paper presents a novel framework for modeling and conditional generation of 3D articulated objects. Troubled by flexibility-quality tradeoffs, existing methods are often limited to using predefined structures or retrieving shapes from static datasets. To address these challenges, we parameterize an articulated object as a tree of tokens and employ a transformer to generate both the object's high-level geometry code and its kinematic relations. Subsequently, each sub-part's geometry is further decoded using a signed-distance-function (SDF) shape prior, facilitating the synthesis of high-quality 3D shapes. Our approach enables the generation of diverse objects with high-quality geometry and varying number of parts. Comprehensive experiments on conditional generation from text descriptions demonstrate the effectiveness and flexibility of our method.

Figures

Figures reproduced from arXiv: 2412.07237 by the authors.

Figure 1
Figure 1. We present the Articulation TransFormer, for high-quality generation articulated objects. This figure illustrates controlled generation across random trials based on text descriptions. Notably, it can generate a diverse range of objects with varying numbers of sub-parts and different geometry features. Abstract This paper presents a novel framework for modeling and conditional generation of 3D articulated objects. T… view at source ↗
Figure 2
Figure 2. Training Pipeline of Shape Prior Mini encoder Eg compresses the geometry latent code z into cg, which is then processed by the embedding vectors of codebooks to form cˆg. cˆg is the condition for diffusion decoder ϵ. Each sub-part has a semantic label, such as ‘the lid of cup’ or ‘handle of box’. These labels, encoded by the pre-trained text encoder, pass through mini encoder Es. The resultant vector cs is then pass… view at source ↗
Figure 3
Figure 3. Articulation Transformer: In the tree structure, i-th node carries 4 attributes: bi, ji, li and zi, which respectively repre￾sent the bounding box, joint axis, limit, and geometry latent code. oˆ represents the logits indicating whether the current output token is a terminal token T (a special token). i-th node pi∈ R 1024 to represent the relative position: p_i=\operatorname {CAT}\left (\{a_{\rev {p_k}}\}_{k=K}^{\re… view at source ↗
Figures from the paper (12 more)
Figure 5
Figure 5. Figure 5: Qualitative comparison between ArtFormer and baselines (Ours-1CB will be discussed in Sec. 4.5). Our method is capable of generating high-quality geometry and accurate joint relations. A human study methodology (HS) is used to assess the alignment between generated obj…
Figure 6
Figure 6. Figure 6: The figure displays 3 pairs of image condition inputs [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Training pipline for PointNet Γ, q(z|f), p(f|z) and SDF decoder Ω. For brevity, we refer to Γ and q(z|f) collectively as SDF encoders. And, similarly, we refer to p(f|z) and SDF decoder Ω as SDF decoders. This is a type of [Storage Furniture/Bottle/Toilet...]. Please f…
Figure 8
Figure 8. Figure 8: Prompt for GPT-4o to generate text description for ob1 [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Prompt Used to Restrict the Length of Output. 1 [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: The first column (blue) shows shapes generated by [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]
Figure 11
Figure 11. Figure 11: The text in the upper part of the figure represents the input text given to the model, where the intensity of the colors indicates [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: The figure illustrates the process of editing existing articulated objects using ArtFormer. (1) For the first object, the cabinet [PITH_FULL_IMAGE:figures/full_fig_p015_12.png]
Figure 13
Figure 13. Figure 13: Continued on next page 4 [PITH_FULL_IMAGE:figures/full_fig_p016_13.png]
Figure 13
Figure 13. Figure 13: The figure presents 15 pairs of input text conditions and articulated objects generated by ArtFormer. For each pair, the text on [PITH_FULL_IMAGE:figures/full_fig_p017_13.png]
Figure 14
Figure 14. Figure 14: The figure presents 10 pairs of input images and articulated objects generated by ArtFormer. For each pair, the larger image on [PITH_FULL_IMAGE:figures/full_fig_p018_14.png]
Figure 15
Figure 15. Figure 15: We present 4 pair of photographs for articulated objects from the real world (shown on the left side of each pair). Using our [PITH_FULL_IMAGE:figures/full_fig_p019_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 50 canonical work pages

  1. [8]

    Me- shart: Generating articulated meshes with structure-guided transformers, 2024

    Daoyi Gao, Yawar Siddiqui, Lei Li, and Angela Dai. Me- shart: Generating articulated meshes with structure-guided transformers, 2024. 2

  2. [1]

    URDFormer: A pipeline for constructing articulated simulation environments from real-world images

    Zoey Chen, Aaron Walsman, Marius Memmel, Kaichun Mo, Alex Fang, Karthikeya Vemuri, Alan Wu, Dieter Fox, and Abhishek Gupta. URDFormer: A pipeline for constructing articulated simulation environments from real-world images. ArXiv, abs/2405.11656, 2024. 2

  3. [2]

    Gensdf: Two- stage learning of generalizable signed distance functions

    Gene Chou, Ilya Chugunov, and Felix Heide. Gensdf: Two- stage learning of generalizable signed distance functions. In Proc. of Neural Information Processing Systems (NeurIPS) ,

  4. [3]

    Diffusion-SDF: Conditional generative modeling of signed distance func- tions

    Gene Chou, Yuval Bahat, and Felix Heide. Diffusion-SDF: Conditional generative modeling of signed distance func- tions. In Proceedings of the IEEE/CVF International Con- ference on Computer Vision, pages 2262–2272, 2023. 2, 3, 1

  5. [4]

    Blender - a 3D modelling and rendering package

    Blender Online Community. Blender - a 3D modelling and rendering package. Blender Foundation, Stichting Blender Foundation, Amsterdam, 2018. 5, 8

  6. [5]

    Auto- mated creation of digital cousins for robust policy learning,

    Tianyuan Dai, Josiah Wong, Yunfan Jiang, Chen Wang, Cem Gokmen, Ruohan Zhang, Jiajun Wu, and Li Fei-Fei. Auto- mated creation of digital cousins for robust policy learning,

  7. [6]

    MolGAN: An implicit generative model for small molecular graphs

    Nicola De Cao and Thomas Kipf. MolGAN: An implicit generative model for small molecular graphs. arXiv preprint arXiv:1805.11973, 2018. 3

  8. [7]

    HyperDiffusion: Generating implicit neu- ral fields with weight-space diffusion

    Ziya Erkoc ¸, Fangchang Ma, Qi Shan, Matthias Nießner, and Angela Dai. HyperDiffusion: Generating implicit neu- ral fields with weight-space diffusion. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 14254–14264, 2023. 2

Show all 60 references
  1. [9]

    Guibas, and Jie Yang

    Lin Gao, Jia-Mu Sun, Kaichun Mo, Yu-Kun Lai, Leonidas J. Guibas, and Jie Yang. SceneHGN: Hierarchical graph net- works for 3d indoor scene generation with fine-grained ge- ometry. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(7):8902–8919, 2023. 3

  2. [10]

    Predicting ar- ticulated human motion from spatial processes.International Journal of Computer Vision, 94:317–334, 2011

    Søren Hauberg and Kim Steenstrup Pedersen. Predicting ar- ticulated human motion from spatial processes.International Journal of Computer Vision, 94:317–334, 2011. 2

  3. [11]

    CARTO: Category and joint agnostic reconstruction of articulated objects

    Nick Heppert, Muhammad Zubair Irshad, Sergey Zakharov, Katherine Liu, Rares Andrei Ambrus, Jeannette Bohg, Ab- hinav Valada, and Thomas Kollar. CARTO: Category and joint agnostic reconstruction of articulated objects. In Pro- ceedings of the IEEE/CVF Conference on Computer Vis...

  4. [12]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 2

  5. [13]

    Equivariant diffusion for molecule gener- ation in 3d

    Emiel Hoogeboom, Vıctor Garcia Satorras, Cl ´ement Vignac, and Max Welling. Equivariant diffusion for molecule gener- ation in 3d. In International Conference on Machine Learn- ing, pages 8867–8887. PMLR, 2022. 3

  6. [14]

    GraphGDP: Generative diffusion processes for permuta- tion invariant graph generation

    Han Huang, Leilei Sun, Bowen Du, Yanjie Fu, and Weifeng Lv. GraphGDP: Generative diffusion processes for permuta- tion invariant graph generation. In 2022 IEEE International Conference on Data Mining (ICDM), pages 201–210. IEEE,

  7. [15]

    ScrewNet: Category-independent articulation model estimation from depth images using screw theory

    Ajinkya Jain, Rudolf Lioutikov, Caleb Chuck, and Scott Niekum. ScrewNet: Category-independent articulation model estimation from depth images using screw theory. In IEEE International Conference on Robotics and Automation (ICRA), pages 13670–13677. IEEE, 2021. 2

  8. [16]

    Categorical repa- rameterization with gumbel-softmax

    Eric Jang, Shixiang Gu, and Ben Poole. Categorical repa- rameterization with gumbel-softmax. In International Con- ference on Learning Representations, 2017. 4

  9. [17]

    Score-based generative modeling of graphs via the system of stochastic differential equations

    Jaehyeong Jo, Seul Lee, and Sung Ju Hwang. Score-based generative modeling of graphs via the system of stochastic differential equations. In International Conference on Ma- chine Learning, pages 10362–10383. PMLR, 2022. 3

  10. [18]

    Rethinking positional encoding in language pre-training

    Guolin Ke, Di He, and Tie-Yan Liu. Rethinking positional encoding in language pre-training. In International Confer- ence on Learning Representations, 2020. 4

  11. [19]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics, 42 (4), 2023. 8

  12. [20]

    Variational graph auto- encoders

    Thomas N Kipf and Max Welling. Variational graph auto- encoders. arXiv preprint arXiv:1611.07308, 2016. 3

  13. [21]

    NAP: Neural 3D articulated object prior

    Jiahui Lei, Congyue Deng, Bokui Shen, Leonidas Guibas, and Kostas Daniilidis. NAP: Neural 3D articulated object prior. In Advances in Neural Information Processing Sys- tems, 2024. 2, 6

  14. [22]

    Mobility fitting using 4d ransac

    Hao Li, Guowei Wan, Honghua Li, Andrei Sharf, Kai Xu, and Baoquan Chen. Mobility fitting using 4d ransac. Com- puter Graphics Forum, 35(5):79–88, 2016. 2

  15. [23]

    BLIP-2: Bootstrapping language-image pre-training with frozen image encoders and large language models

    Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. BLIP-2: Bootstrapping language-image pre-training with frozen image encoders and large language models. In In- ternational Conference on Machine Learning, pages 19730– 19742. PMLR, 2023. 8, 3

  16. [24]

    Grains: Generative re- cursive autoencoders for indoor scenes, 2019

    Manyi Li, Akshay Gadi Patil, Kai Xu, Siddhartha Chaudhuri, Owais Khan, Ariel Shamir, Changhe Tu, Baoquan Chen, Daniel Cohen-Or, and Hao Zhang. Grains: Generative re- cursive autoencoders for indoor scenes, 2019. 3

  17. [25]

    Chang, Manolis Savva, and Ali Mahdavi-Amiri

    Jiayi Liu, Denys Iliash, Angel X. Chang, Manolis Savva, and Ali Mahdavi-Amiri. SINGAPO: Single image controlled generation of articulated parts in object. arXiv preprint arXiv:2410.16499, 2024. 2

  18. [26]

    Sur- vey on modeling of articulated objects

    Jiayi Liu, Manolis Savva, and Ali Mahdavi-Amiri. Sur- vey on modeling of articulated objects. arXiv preprint arXiv:2403.14937, 2024. 1

  19. [27]

    CAGE: Controllable articulation generation

    Jiayi Liu, Hou In Ivan Tam, Ali Mahdavi-Amiri, and Mano- lis Savva. CAGE: Controllable articulation generation. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 17880–17889, 2024. 2, 3, 6

  20. [28]

    Lorensen and Harvey E

    William E. Lorensen and Harvey E. Cline. Marching Cubes: A high resolution 3d surface construction algorithm. In Proceedings of the 14th Annual Conference on Computer Graphics and Interactive Techniques, page 163–169. Asso- ciation for Computing Machinery, 1987. 6 10

  21. [29]

    Decoupled weight de- cay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight de- cay regularization. In International Conference on Learning Representations, 2017. 1

  22. [30]

    Mol- CycleGAN: a generative model for molecular optimization

    Łukasz Maziarka, Agnieszka Pocha, Jan Kaczmarczyk, Krzysztof Rataj, Tomasz Danel, and Michał Warchoł. Mol- CycleGAN: a generative model for molecular optimization. Journal of Cheminformatics, 12(1):2, 2020. 3

  23. [31]

    Mitra, Yong-Liang Yang, Dong-Ming Yan, Wilmot Li, and Maneesh Agrawala

    Niloy J. Mitra, Yong-Liang Yang, Dong-Ming Yan, Wilmot Li, and Maneesh Agrawala. Illustrating how mechanical as- semblies work. ACM Trans. Graph., 29(4), 2010. 2

  24. [32]

    AutoSDF: Shape priors for 3d comple- tion, reconstruction and generation

    Paritosh Mittal, Yen-Chi Cheng, Maneesh Singh, and Shub- ham Tulsiani. AutoSDF: Shape priors for 3d comple- tion, reconstruction and generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 306–315, 2022. 2

  25. [33]

    Chang, Li Yi, Subarna Tripathi, Leonidas J

    Kaichun Mo, Shilin Zhu, Angel X. Chang, Li Yi, Subarna Tripathi, Leonidas J. Guibas, and Hao Su. PartNet: A large- scale benchmark for fine-grained and hierarchical part-level 3D object understanding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Rec...

  26. [34]

    AO-Grasp: Articulated object grasp generation

    Carlota Par ´es Morlans, Claire Chen, Yijia Weng, Michelle Yi, Yuying Huang, Nick Heppert, Linqi Zhou, Leonidas Guibas, and Jeannette Bohg. AO-Grasp: Articulated object grasp generation. arXiv, 2023. 2

  27. [35]

    Yuille, Nuno Vasconcelos, and Xiaolong Wang

    Jiteng Mu, Weichao Qiu, Adam Kortylewski, Alan L. Yuille, Nuno Vasconcelos, and Xiaolong Wang. A-SDF: learning disentangled signed distance functions for articulated shape representation. ICCV, pages 12981–12991, 2021. 2

  28. [36]

    Classifier-free graph diffusion for molecular property tar- geting

    Matteo Ninniri, Marco Podda, and Davide Bacciu. Classifier-free graph diffusion for molecular property tar- geting. In Joint European Conference on Machine Learn- ing and Knowledge Discovery in Databases, pages 318–335. Springer, 2024. 3

  29. [37]

    Benchmark for compositional text-to- image synthesis

    Dong Huk Park, Samaneh Azadi, Xihui Liu, Trevor Darrell, and Anna Rohrbach. Benchmark for compositional text-to- image synthesis. In Thirty-fifth Conference on Neural Infor- mation Processing Systems Datasets and Benchmarks Track (Round 1), 2021. 2

  30. [38]

    DeepSDF: Learning continuous signed distance functions for shape representa- tion

    Jeong Joon Park, Peter Florence, Julian Straub, Richard Newcombe, and Steven Lovegrove. DeepSDF: Learning continuous signed distance functions for shape representa- tion. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 165–174, 2019. 2

  31. [39]

    Integrating tree path in transformer for code representa- tion

    Han Peng, Ge Li, Wenhan Wang, YunFei Zhao, and Zhi Jin. Integrating tree path in transformer for code representa- tion. In Advances in Neural Information Processing Systems, pages 9343–9354, 2021. 3

  32. [40]

    Integrating tree path in transformer for code representa- tion

    Han Peng, Ge Li, Wenhan Wang, YunFei Zhao, and Zhi Jin. Integrating tree path in transformer for code representa- tion. In Advances in Neural Information Processing Systems, pages 9343–9354. Curran Associates, Inc., 2021. 4

  33. [41]

    Morgan Quigley, Brian Gerkey, and William D. Smart. Pro- gramming Robots with ROS: A Practical Introduction to the Robot Operating System. O’Reilly Media, Inc., 2015. 2, 3

  34. [42]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of Machine Learn- ing Research, 21(140):1–67, 2020. 3, 5, 8, 1

  35. [43]

    Hierarchical text-conditional image gener- ation with clip latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image gener- ation with clip latents. ArXiv, abs/2204.06125, 2022. 3

  36. [44]

    Novel positional encodings to enable tree-based transformers

    Vighnesh Shiv and Chris Quirk. Novel positional encodings to enable tree-based transformers. In Advances in Neural Information Processing Systems, pages 12058–12068, 2019. 3

  37. [45]

    Novel positional encod- ings to enable tree-based transformers

    Vighnesh Shiv and Chris Quirk. Novel positional encod- ings to enable tree-based transformers. In Advances in Neu- ral Information Processing Systems. Curran Associates, Inc.,

  38. [46]

    GraphV AE: Towards generation of small graphs using variational au- toencoders

    Martin Simonovsky and Nikos Komodakis. GraphV AE: Towards generation of small graphs using variational au- toencoders. In Artificial Neural Networks and Machine Learning–ICANN 2018: 27th International Conference on Artificial Neural Networks, Rhodes, Greece, October 4-7, 2018,...

  39. [47]

    A probabilistic framework for learning kinematic models of ar- ticulated objects

    J ¨urgen Sturm, Cyrill Stachniss, and Wolfram Burgard. A probabilistic framework for learning kinematic models of ar- ticulated objects. Journal of Artificial Intelligence Research, 41:477–526, 2011. 2

  40. [48]

    MuJoCo: A physics engine for model-based control

    Emanuel Todorov, Tom Erez, and Yuval Tassa. MuJoCo: A physics engine for model-based control. In 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 5026–5033. IEEE, 2012. 3

  41. [49]

    DiGress: Dis- crete denoising diffusion for graph generation

    Clement Vignac, Igor Krawczuk, Antoine Siraudin, Bohan Wang, V olkan Cevher, and Pascal Frossard. DiGress: Dis- crete denoising diffusion for graph generation. In Interna- tional Conference on Learning Representations, 2023. 3

  42. [50]

    Graph- GAN: Graph representation learning with generative adver- sarial nets

    Hongwei Wang, Jia Wang, Jialin Wang, Miao Zhao, Weinan Zhang, Fuzheng Zhang, Xing Xie, and Minyi Guo. Graph- GAN: Graph representation learning with generative adver- sarial nets. In Proceedings of the AAAI conference on artifi- cial intelligence, pages 2508–2515, 2018. 3

  43. [51]

    Self-supervised neural articulated shape and appearance models

    Fangyin Wei, Rohan Chabra, Lingni Ma, Christoph Lassner, Michael Zollhoefer, Szymon Rusinkiewicz, Chris Sweeney, Richard Newcombe, and Mira Slavcheva. Self-supervised neural articulated shape and appearance models. InProceed- ings of the IEEE/CVF Conference on Computer Vision ...

  44. [52]

    MarrNet: 3d shape recon- struction via 2.5d sketches

    Jiajun Wu, Yifan Wang, Tianfan Xue, Xingyuan Sun, Bill Freeman, and Josh Tenenbaum. MarrNet: 3d shape recon- struction via 2.5d sketches. In Advances in Neural Informa- tion Processing Systems, pages 540–550, 2017. 2

  45. [53]

    3D ShapeNets: A deep representation for volumetric shapes

    Zhirong Wu, Shuran Song, Aditya Khosla, Fisher Yu, Lin- guang Zhang, Xiaoou Tang, and Jianxiong Xiao. 3D ShapeNets: A deep representation for volumetric shapes. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 1912–1920, 2015. 2

  46. [54]

    Chang, Leonidas J

    Fanbo Xiang, Yuzhe Qin, Kaichun Mo, Yikuan Xia, Hao Zhu, Fangchen Liu, Minghua Liu, Hanxiao Jiang, Yifu Yuan, He Wang, Li Yi, Angel X. Chang, Leonidas J. Guibas, and Hao Su. SAPIEN: A simulated part-based interactive en- vironment. In Proceedings of the IEEE/CVF Conference 11 ...

  47. [55]

    Joint-aware manipulation of deformable models

    Weiwei Xu, Jun Wang, KangKang Yin, Kun Zhou, Michiel van de Panne, Falai Chen, and Baining Guo. Joint-aware manipulation of deformable models. ACM Trans. Graph., 28 (3), 2009. 2

  48. [56]

    D-V AE: A variational autoencoder for di- rected acyclic graphs

    Muhan Zhang, Shali Jiang, Zhicheng Cui, Roman Garnett, and Yixin Chen. D-V AE: A variational autoencoder for di- rected acyclic graphs. In Advances in Neural Information Processing Systems, pages 1586–1598, 2019. 3

  49. [57]

    Unifying gener- ation and prediction on graphs with latent graph diffusion

    Cai Zhou, Xiyuan Wang, and Muhan Zhang. Unifying gener- ation and prediction on graphs with latent graph diffusion. In Advances in Neural Information Processing Systems , 2024. 3

  50. [58]

    3M- Diffusion: Latent multi-modal diffusion for language-guided molecular structure generation

    Huaisheng Zhu, Teng Xiao, and Vasant G Honavar. 3M- Diffusion: Latent multi-modal diffusion for language-guided molecular structure generation. In First Conference on Lan- guage Modeling, 2024. 3 12 ArtFormer: Controllable Generation of Diverse 3D Articulated Objects Supplemen...

  51. [59]

    SDF Model As we describe in Sec

    Implementation Details 6.1. SDF Model As we describe in Sec. 3.2, we firstly stack the PointNet Γ, q(z|f ), p(f |z) and SDF Decoder Ω, which is shown in Fig. 7. This stacked network is used to generate the latent code z from the point cloud and decode the mesh from z. To stren...

  52. [60]

    You can describe it in detail with more sentences

    Additional Experiments and Results 7.1. Novel Shape Generation We conducted an experiment inspired by Diffusion-SDF to demonstrate that our shape prior, guided by an articulation transformer, can generate new geometry shapes that never appear in the dataset. We used our model ...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.