pith. sign in

arxiv: 2606.10395 · v1 · pith:VEWD2XXRnew · submitted 2026-06-09 · 💻 cs.CV

Efficient RWKV-based Representation Learning for 3D Point Clouds

Yun Liu , Xuefeng Yan , Liangliang Nan , Xianzhi Li , Peng Li , Zhe Zhu , Honghua Chen , Mingqiang Wei This is my paper

Pith reviewed 2026-06-27 14:17 UTC · model grok-4.3

classification 💻 cs.CV
keywords point cloudsRWKVrepresentation learningself-supervised learning3D visionefficient modelinglocal perceptionspatial enhancement
0
0 comments X

The pith

P-RWKV adapts RWKV for point clouds by adding local perception and spatial enhancement while retaining linear complexity.

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

The paper introduces the P-RWKV block to apply the efficient RWKV sequence model to irregular 3D point clouds. Standard RWKV handles sequential text with linear cost but misses local geometric structures and spatial relations in point data. Two added components expand contextual perception along sequences and strengthen spatial awareness. The authors stack these blocks into the PointER self-supervised encoder and test extensions to cross-modality use. If the claim holds, point cloud tasks gain competitive accuracy at reduced computation and faster inference than quadratic attention methods.

Core claim

The P-RWKV block bridges sequence modeling and irregular 3D geometry while preserving RWKV efficiency; it consists of a Local Perception Expansion component to expand contextual perception along the spatio-temporal sequence and a Spatial Context Enhancement component to strengthen spatial awareness. Stacked P-RWKV blocks form the encoder of the single-modality PointER self-supervised framework, with the same sub-modules integrated into cross-modality settings and multiple architectures to demonstrate plug-and-play flexibility.

What carries the argument

The P-RWKV block, built from Local Perception Expansion (LPE) to expand contextual perception and Spatial Context Enhancement (SCE) to strengthen spatial awareness.

If this is right

  • Stacked P-RWKV blocks serve as an effective encoder for self-supervised point cloud representation learning in the PointER framework.
  • P-RWKV extends to cross-modality settings while keeping its core sub-modules.
  • The sub-modules integrate as plug-and-play components into multiple existing architectures.
  • The approach delivers competitive task performance at lower computational cost and inference latency than attention-based alternatives.

Where Pith is reading between the lines

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

  • The linear scaling could support processing denser point clouds on hardware with limited memory.
  • Similar local and spatial additions might adapt RWKV-style models to other irregular structures such as meshes or graphs.
  • The plug-and-play property suggests the block could replace attention layers in existing 3D pipelines with minimal redesign.

Load-bearing premise

Adding the Local Perception Expansion and Spatial Context Enhancement components lets RWKV capture local geometric structures and spatial dependencies in point clouds without losing its linear complexity advantage.

What would settle it

A benchmark run on large point clouds where P-RWKV exhibits quadratic scaling in time or memory, or where accuracy falls below standard attention baselines without the claimed latency reduction.

Figures

Figures reproduced from arXiv: 2606.10395 by Honghua Chen, Liangliang Nan, Mingqiang Wei, Peng Li, Xianzhi Li, Xuefeng Yan, Yun Liu, Zhe Zhu.

Figure 1
Figure 1. Figure 1: Efficiency comparison under identical settings. We evalu￾ate transformer-based Point-MAE [1], mamba-based PointMamba [4], and the proposed RWKV-based PointER in terms of floating-point operations (‘FLOPs’), as well as per-epoch pre-training and inference times (‘Pre-train E-T’ and ‘Inference E-T’), with group numbers set to 64 and 128. We further analyze the FLOPs of the inference models of Point-MAE and P… view at source ↗
Figure 2
Figure 2. Figure 2: Visualization the details of the RWKV block from [6]. In the time-mixing module, the linear projection vectors R, K, and V are obtained through linear interpolation be￾tween the current and previous timestep inputs, enabling token shift–based information exchange. The WKV operator is then employed to compute attention scores and update hidden states, where a channel-wise vector W is adjusted by relative po… view at source ↗
Figure 4
Figure 4. Figure 4: Overall architecture of our PointER. Embedding: Given an input point cloud P = {pi |1 ≤ i ≤ N} ∈ R N×3 , where pi = (xi , yi , zi), we first apply Furthest Point Sampling (FPS) to select g representative centers C ∈ R g×3 . Subsequently, K-Nearest Neighbors (KNN) is employed to retrieve the neighboring points of the centers C, forming the patches Gs ∈ R g×k×3 around the centers. To enhance convergence, all… view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of the completion results on ShapeNet55 with frozen pre-trained models. Visualization of completion on ModelNet40. To fur￾ther evaluate the cross-dataset completion capability of our pre-trained model, we perform a completion task on Mod￾elNet40 [44] instances using the model pre-trained on ShapeNet55. As shown in [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of the completion results on ModelNet40 with frozen pre-trained models. Quantitative results of completion on ShapeNet55 and ModelNet40. We randomly sample partial point sets from ShapeNet55 and ModelNet40 and perform completion with the frozen pre-trained models. Chamfer Distance (CD) and Earth Mover’s Distance (EMD) are employed as quantitative metrics to objectively assess reconstruction f… view at source ↗
Figure 7
Figure 7. Figure 7: Visualization of the part segmentation results on the test set of ShapeNetPart [50]. Mesh Point Cloud PointER Mamba3D [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Visualization of part segmentation results on our collected dataset. amples are shown in [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Visualization of the ViG block from [10] (top) and the proposed P-GLA block (bottom) integrating the LPE and SCE modules. The bidi￾rectional GLA (BiGLA) layer proposed in ViG [10] introduces a directional gating mechanism to adaptively select global contextual information from different directions. Most parameters are shared between the forward and backward directions, enabling efficient bidirectional depe… view at source ↗
Figure 10
Figure 10. Figure 10: Overall architecture of PTv3 [24]. As shown in [PITH_FULL_IMAGE:figures/full_fig_p011_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Overall architecture of the proposed PTv3-RWKV based on PTv3 [24]. We replace the attention and MLP modules in PTv3 with the proposed spatial-mix and channel-mix modules, respec￾tively, with minimal modification, resulting in PTv3-RWKV, as illustrated in [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
read the original abstract

The recent receptance weighted key value (RWKV) model combines RNN-style recurrence, offering a linear-complexity alternative to Transformers' quadratic self-attention for modeling global dependencies. However, when directly applied to point clouds, RWKV, originally developed for sequential text, struggles to capture local geometric structures and model spatial dependencies effectively. To address this, we propose the \textbf{P-RWKV} block, which bridges the gap between sequence modeling and irregular 3D geometry while preserving the efficiency advantages of RWKV. It consists of a Local Perception Expansion (LPE) component to expand contextual perception along the spatio-temporal sequence and a Spatial Context Enhancement (SCE) component to strengthen spatial awareness. To validate the effectiveness of P-RWKV for point cloud understanding, we construct PointER, a single-modality self-supervised representation learning framework whose encoder is composed of stacked P-RWKV blocks. Furthermore, we extend P-RWKV to a cross-modality setting and integrate the proposed core sub-modules into multiple architectures, demonstrating strong plug-and-play flexibility and architectural generality. Extensive experiments show that the P-RWKV block and its key sub-modules achieve competitive performance across various tasks with lower computational cost and inference latency. Code will be released upon acceptance.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The paper proposes the P-RWKV block as an adaptation of the RWKV model for 3D point clouds. It introduces Local Perception Expansion (LPE) to expand contextual perception along spatio-temporal sequences and Spatial Context Enhancement (SCE) to strengthen spatial awareness, while claiming to preserve RWKV's linear complexity. The work constructs the PointER self-supervised single-modality framework with stacked P-RWKV blocks and extends the core modules to cross-modality settings, asserting competitive performance on various tasks with lower computational cost and inference latency.

Significance. If the performance and efficiency claims are substantiated by experiments, the work could provide a practical linear-complexity alternative to quadratic-attention models for irregular 3D data, with plug-and-play flexibility as a notable strength for architectural generality in point cloud understanding.

major comments (2)
  1. [Abstract] Abstract: the central claims of 'competitive performance across various tasks with lower computational cost and inference latency' and effective bridging of sequence modeling with irregular 3D geometry rest entirely on experimental validation, yet the abstract supplies no quantitative metrics, baselines, ablation results, or error analysis, rendering it impossible to assess whether the data support the stated claims.
  2. [§3 (P-RWKV block description)] The assumption that LPE and SCE additions preserve linear complexity while capturing local geometric structures and spatial dependencies in point clouds is load-bearing for the efficiency advantage; without an explicit complexity analysis or proof that no quadratic terms are introduced (e.g., in the spatio-temporal sequence handling), the efficiency claim cannot be verified.
minor comments (2)
  1. [Abstract] The abstract would benefit from at least one key quantitative result (e.g., mIoU or latency comparison) to ground the performance claims.
  2. [Method] Notation for the LPE and SCE components should be introduced with explicit equations or pseudocode early in the method section for clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the two major comments below and will revise the manuscript accordingly to strengthen the presentation.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claims of 'competitive performance across various tasks with lower computational cost and inference latency' and effective bridging of sequence modeling with irregular 3D geometry rest entirely on experimental validation, yet the abstract supplies no quantitative metrics, baselines, ablation results, or error analysis, rendering it impossible to assess whether the data support the stated claims.

    Authors: We agree that the abstract would be strengthened by including quantitative support for the claims. In the revised version, we will incorporate key experimental results (e.g., task accuracies, FLOPs reductions, and latency comparisons versus baselines) directly into the abstract while keeping it concise. revision: yes

  2. Referee: [§3 (P-RWKV block description)] The assumption that LPE and SCE additions preserve linear complexity while capturing local geometric structures and spatial dependencies in point clouds is load-bearing for the efficiency advantage; without an explicit complexity analysis or proof that no quadratic terms are introduced (e.g., in the spatio-temporal sequence handling), the efficiency claim cannot be verified.

    Authors: We acknowledge the value of an explicit complexity breakdown. The design of LPE and SCE intentionally reuses the linear-time recurrence of RWKV and avoids any quadratic operations (e.g., no dense pairwise interactions). In the revision we will add a dedicated complexity subsection in §3 that derives the O(N) cost for each component, including the spatio-temporal sequence handling, to make the linear-complexity claim fully verifiable. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The manuscript text provided (abstract plus placeholder for full content) contains no equations, parameter fits, derivation chains, or self-citations that reduce any claimed result to its own inputs by construction. The P-RWKV proposal, LPE/SCE components, and PointER framework are described as architectural additions without any visible self-definitional loops, fitted-input predictions, or load-bearing self-citations. The central claims therefore remain independent of the patterns that would trigger a positive circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no equations, training details, or architectural specifications sufficient to identify free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5770 in / 1100 out tokens · 16738 ms · 2026-06-27T14:17:38.793606+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

57 extracted references · 9 canonical work pages · 4 internal anchors

  1. [1]

    Masked autoencoders for point cloud self-supervised learning,

    Y . Pang, W. Wang, F. E. H. Tay, W. Liu, Y . Tian, and L. Yuan, “Masked autoencoders for point cloud self-supervised learning,” inComputer Vision - ECCV, 2022, pp. 604–621

    2022

  2. [2]

    Point-m2ae: Multi-scale masked autoencoders for hierarchical point cloud pre-training,

    R. Zhang, Z. Guo, P. Gao, R. Fang, B. Zhao, D. Wang, Y . Qiao, and H. Li, “Point-m2ae: Multi-scale masked autoencoders for hierarchical point cloud pre-training,” inNeurIPS, 2022, pp. 27 061–27 074

    2022

  3. [3]

    Point- gpt: Auto-regressively generative pre-training from point clouds,

    G. Chen, M. Wang, Y . Yang, K. Yu, L. Yuan, and Y . Yue, “Point- gpt: Auto-regressively generative pre-training from point clouds,” in NeurIPS, 2023

    2023

  4. [4]

    Pointmamba: A simple state space model for point cloud analysis,

    D. Liang, X. Zhou, W. Xu, X. Zhu, Z. Zou, X. Ye, X. Tan, and X. Bai, “Pointmamba: A simple state space model for point cloud analysis,” in NeurIPS, 2024

    2024

  5. [5]

    Mamba: Linear-Time Sequence Modeling with Selective State Spaces

    A. Gu and T. Dao, “Mamba: Linear-time sequence modeling with selective state spaces,”CoRR, vol. abs/2312.00752, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  6. [6]

    RWKV: reinventing rnns for the transformer era,

    B. Peng, E. Alcaide, Q. Anthony, A. Albalak, S. Arcadinho, S. Bider- man, H. Cao, X. Cheng, M. Chung, L. Derczynski, X. Du, M. Grella, K. K. GV , X. He, H. Hou, P. Kazienko, J. Kocon, J. Kong, B. Koptyra, H. Lau, J. Lin, K. S. I. Mantri, F. Mom, A. Saito, G. Song, X. Tang, J. S. Wind, S. Wozniak, Z. Zhang, Q. Zhou, J. Zhu, and R. Zhu, “RWKV: reinventing r...

    2023

  7. [7]

    Mamba3d: Enhancing local features for 3d point cloud analysis via state space model,

    X. Han, Y . Tang, Z. Wang, and X. Li, “Mamba3d: Enhancing local features for 3d point cloud analysis via state space model,” inACM International Conference on Multimedia, 2024, pp. 4995–5004

    2024

  8. [8]

    Vmamba: Visual state space model,

    Y . Liu, Y . Tian, Y . Zhao, H. Yu, L. Xie, Y . Wang, Q. Ye, J. Jiao, and Y . Liu, “Vmamba: Visual state space model,” inNeurIPS, 2024

    2024

  9. [9]

    Vision mamba: Efficient visual representation learning with bidirectional state space model,

    L. Zhu, B. Liao, Q. Zhang, X. Wang, W. Liu, and X. Wang, “Vision mamba: Efficient visual representation learning with bidirectional state space model,” inInternational Conference on Machine Learning. Open- Review.net, 2024

    2024

  10. [10]

    Vig: Linear- complexity visual sequence learning with gated linear attention,

    B. Liao, X. Wang, L. Zhu, Q. Zhang, and C. Huang, “Vig: Linear- complexity visual sequence learning with gated linear attention,” inAAAI Conference on Artificial Intelligence, 2025, pp. 5182–5190

    2025

  11. [11]

    Eagle and Finch: RWKV with Matrix-Valued States and Dynamic Recurrence

    B. Peng, D. Goldstein, Q. Anthony, A. Albalak, E. Alcaide, S. Biderman, E. Cheah, X. Du, T. Ferdinan, H. Hou, P. Kazienko, K. K. GV , J. Kocon, B. Koptyra, S. Krishna, R. M. Jr., N. Muennighoff, F. Obeid, A. Saito, G. Song, H. Tu, S. Wozniak, R. Zhang, B. Zhao, Q. Zhao, P. Zhou, J. Zhu, and R. Zhu, “Eagle and finch: RWKV with matrix-valued states and dyna...

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  12. [12]

    Mamba or RWKV: exploring high-quality and high-efficiency segment anything model,

    H. Yuan, X. Li, L. Qi, T. Zhang, M. Yang, S. Yan, and C. C. Loy, “Mamba or RWKV: exploring high-quality and high-efficiency segment anything model,”CoRR, vol. abs/2406.19369, 2024

    work page arXiv 2024

  13. [13]

    RWKV- CLIP: A robust vision-language representation learner,

    T. Gu, K. Yang, X. An, Z. Feng, D. Liu, W. Cai, and J. Deng, “RWKV- CLIP: A robust vision-language representation learner,” inEMNLP, 2024, pp. 4799–4812

    2024

  14. [14]

    Video rwkv:video action recognition based RWKV,

    Z. Yin, C. Li, and X. Dong, “Video rwkv:video action recognition based RWKV,”CoRR, vol. abs/2411.05636, 2024

    work page arXiv 2024

  15. [15]

    Vision-rwkv: Efficient and scalable visual perception with rwkv-like architectures,

    Y . Duan, W. Wang, Z. Chen, X. Zhu, L. Lu, T. Lu, Y . Qiao, H. Li, J. Dai, and W. Wang, “Vision-rwkv: Efficient and scalable visual perception with rwkv-like architectures,” inInternational Conference on Learning Representations, 2025, pp. 1–17. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13

    2025

  16. [16]

    Pointrwkv: Efficient rwkv-like model for hierarchical point cloud learning,

    Q. He, J. Zhang, J. Peng, H. He, X. Li, Y . Wang, and C. Wang, “Pointrwkv: Efficient rwkv-like model for hierarchical point cloud learning,” inAAAI Conference on Artificial Intelligence, 2025, pp. 3410– 3418

    2025

  17. [17]

    Pointpillars: Fast encoders for object detection from point clouds,

    A. H. Lang, S. V ora, H. Caesar, L. Zhou, J. Yang, and O. Beijbom, “Pointpillars: Fast encoders for object detection from point clouds,” in IEEE Conference on Computer Vision and Pattern Recognition. IEEE, 2019, pp. 12 697–12 705

    2019

  18. [18]

    V oxnet: A 3d convolutional neural network for real-time object recognition,

    D. Maturana and S. A. Scherer, “V oxnet: A 3d convolutional neural network for real-time object recognition,” inIEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2015, pp. 922– 928

    2015

  19. [19]

    Pointnet: Deep learning on point sets for 3d classification and segmentation,

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,” inIEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 77–85

    2017

  20. [20]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space,

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas, “Pointnet++: Deep hierarchical feature learning on point sets in a metric space,” inNeurIPS, 2017, pp. 5099–5108

    2017

  21. [21]

    Dynamic graph CNN for learning on point clouds,

    Y . Wang, Y . Sun, Z. Liu, S. E. Sarma, M. M. Bronstein, and J. M. Solomon, “Dynamic graph CNN for learning on point clouds,”ACM Trans. Graph., vol. 38, no. 5, pp. 146:1–146:12, 2019

    2019

  22. [22]

    Point trans- former,

    H. Zhao, L. Jiang, J. Jia, P. H. S. Torr, and V . Koltun, “Point trans- former,” inIEEE International Conference on Computer Vision, 2021, pp. 16 239–16 248

    2021

  23. [23]

    Point transformer V2: grouped vector attention and partition-based pooling,

    X. Wu, Y . Lao, L. Jiang, X. Liu, and H. Zhao, “Point transformer V2: grouped vector attention and partition-based pooling,” inNeurIPS, 2022

    2022

  24. [24]

    Point transformer V3: simpler, faster, stronger,

    X. Wu, L. Jiang, P. Wang, Z. Liu, X. Liu, Y . Qiao, W. Ouyang, T. He, and H. Zhao, “Point transformer V3: simpler, faster, stronger,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2024, pp. 4840–4851

    2024

  25. [25]

    Point-bert: Pre- training 3d point cloud transformers with masked point modeling,

    X. Yu, L. Tang, Y . Rao, T. Huang, J. Zhou, and J. Lu, “Point-bert: Pre- training 3d point cloud transformers with masked point modeling,” in IEEE Conference on Computer Vision and Pattern Recognition, 2022, pp. 19 291–19 300

    2022

  26. [26]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” inNeurIPS, 2017, pp. 5998–6008

    2017

  27. [27]

    BERT: pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M. Chang, K. Lee, and K. Toutanova, “BERT: pre-training of deep bidirectional transformers for language understanding,” in Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2019, pp. 4171–4186

    2019

  28. [28]

    Language models are unsupervised multitask learners,

    A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, I. Sutskeveret al., “Language models are unsupervised multitask learners,”OpenAI blog, vol. 1, no. 8, p. 9, 2019

    2019

  29. [29]

    An image is worth 16x16 words: Trans- formers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Trans- formers for image recognition at scale,” inInternational Conference on Learning Representations, 2021, pp. 1–22

    2021

  30. [30]

    Blackmamba: Mixture of experts for state-space models,

    Q. Anthony, Y . Tokpanov, P. Glorioso, and B. Millidge, “Blackmamba: Mixture of experts for state-space models,”CoRR, vol. abs/2402.01771, 2024

    work page arXiv 2024

  31. [31]

    Moe- mamba: Efficient selective state space models with mixture of experts,

    M. Pi ´oro, K. Ciebiera, K. Kr ´ol, J. Ludziejewski, and S. Jaszczur, “Moe- mamba: Efficient selective state space models with mixture of experts,” CoRR, vol. abs/2401.04081, 2024

    work page arXiv 2024

  32. [32]

    Mobilemamba: Lightweight multi-receptive visual mamba network,

    H. He, J. Zhang, Y . Cai, H. Chen, X. Hu, Z. Gan, Y . Wang, C. Wang, Y . Wu, and L. Xie, “Mobilemamba: Lightweight multi-receptive visual mamba network,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition. IEEE, 2025, pp. 4497–4507

    2025

  33. [33]

    Mambaad: Exploring state space models for multi- class unsupervised anomaly detection,

    H. He, Y . Bai, J. Zhang, Q. He, H. Chen, Z. Gan, C. Wang, X. Li, G. Tian, and L. Xie, “Mambaad: Exploring state space models for multi- class unsupervised anomaly detection,” inNeurIPS, 2024

    2024

  34. [34]

    Vivim: a video vision mamba for medical video object segmentation,

    Y . Yang, Z. Xing, and L. Zhu, “Vivim: a video vision mamba for medical video object segmentation,”CoRR, vol. abs/2401.14168, 2024

    work page arXiv 2024

  35. [35]

    Point cloud mamba: Point cloud learning via state space model,

    T. Zhang, H. Yuan, L. Qi, J. Zhang, Q. Zhou, S. Ji, S. Yan, and X. Li, “Point cloud mamba: Point cloud learning via state space model,” in AAAI Conference on Artificial Intelligence, vol. 39, no. 10, 2025, pp. 10 121–10 130

    2025

  36. [36]

    Onboard deep lossless and near-lossless predictive coding of hyperspectral images with line-based attention,

    D. Valsesia, T. Bianchi, and E. Magli, “Onboard deep lossless and near-lossless predictive coding of hyperspectral images with line-based attention,”IEEE Trans. Geosci. Remote. Sens., vol. 62, pp. 1–14, 2024

    2024

  37. [37]

    Pytorch: An imperative style, high- performance deep learning library,

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. K ¨opf, E. Z. Yang, Z. DeVito, M. Raison, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala, “Pytorch: An imperative style, high- performance deep learning library,” inNeurIPS, 2019, pp. 8024–8035

    2019

  38. [38]

    Layer Normalization

    L. J. Ba, J. R. Kiros, and G. E. Hinton, “Layer normalization,”CoRR, vol. abs/1607.06450, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  39. [39]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inIEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 770–778

    2016

  40. [40]

    A point set generation network for 3d object reconstruction from a single image,

    H. Fan, H. Su, and L. J. Guibas, “A point set generation network for 3d object reconstruction from a single image,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2017, pp. 2463–2471

    2017

  41. [41]

    Take-a-photo: 3d- to-2d generative pre-training of point cloud models,

    Z. Wang, X. Yu, Y . Rao, J. Zhou, and J. Lu, “Take-a-photo: 3d- to-2d generative pre-training of point cloud models,” inIEEE/CVF International Conference on Computer Vision, 2023, pp. 5617–5627

    2023

  42. [42]

    Relation-shape convolutional neural network for point cloud analysis,

    Y . Liu, B. Fan, S. Xiang, and C. Pan, “Relation-shape convolutional neural network for point cloud analysis,” inIEEE Conference on Computer Vision and Pattern Recognition, 2019, pp. 8895–8904

    2019

  43. [43]

    ShapeNet: An Information-Rich 3D Model Repository

    A. X. Chang, T. A. Funkhouser, L. J. Guibas, P. Hanrahan, Q. Huang, Z. Li, S. Savarese, M. Savva, S. Song, H. Su, J. Xiao, L. Yi, and F. Yu, “Shapenet: An information-rich 3d model repository,”CoRR, vol. abs/1512.03012, 2015

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  44. [44]

    3d shapenets: A deep representation for volumetric shapes,

    Z. Wu, S. Song, A. Khosla, F. Yu, L. Zhang, X. Tang, and J. Xiao, “3d shapenets: A deep representation for volumetric shapes,” inIEEE Conference on Computer Vision and Pattern Recognition, 2015, pp. 1912–1920

    2015

  45. [45]

    Pointnext: Revisiting pointnet++ with improved training and scaling strategies,

    G. Qian, Y . Li, H. Peng, J. Mai, H. Hammoud, M. Elhoseiny, and B. Ghanem, “Pointnext: Revisiting pointnet++ with improved training and scaling strategies,” inNeurIPS, 2022

    2022

  46. [46]

    Rethinking network design and local geometry in point cloud: A simple residual MLP framework,

    X. Ma, C. Qin, H. You, H. Ran, and Y . Fu, “Rethinking network design and local geometry in point cloud: A simple residual MLP framework,” inInternational Conference on Learning Representations, 2022, pp. 1– 14

    2022

  47. [47]

    Masked discrimination for self-supervised learning on point clouds,

    H. Liu, M. Cai, and Y . J. Lee, “Masked discrimination for self-supervised learning on point clouds,” inComputer Vision - ECCV, vol. 13662, 2022, pp. 657–675

    2022

  48. [48]

    Revisiting point cloud classification: A new benchmark dataset and classification model on real-world data,

    M. A. Uy, Q. Pham, B. Hua, D. T. Nguyen, and S. Yeung, “Revisiting point cloud classification: A new benchmark dataset and classification model on real-world data,” inIEEE International Conference on Com- puter Vision. IEEE, 2019, pp. 1588–1597

    2019

  49. [49]

    Unsuper- vised point cloud pre-training via occlusion completion,

    H. Wang, Q. Liu, X. Yue, J. Lasenby, and M. J. Kusner, “Unsuper- vised point cloud pre-training via occlusion completion,” inIEEE/CVF International Conference on Computer Vision, 2021, pp. 9762–9772

    2021

  50. [50]

    A scalable active framework for region annotation in 3d shape collections,

    L. Yi, V . G. Kim, D. Ceylan, I. Shen, M. Yan, H. Su, C. Lu, Q. Huang, A. Sheffer, and L. J. Guibas, “A scalable active framework for region annotation in 3d shape collections,”ACM Trans. Graph., vol. 35, no. 6, pp. 210:1–210:12, 2016

    2016

  51. [51]

    3d semantic parsing of large-scale indoor spaces,

    I. Armeni, O. Sener, A. R. Zamir, H. Jiang, I. K. Brilakis, M. Fischer, and S. Savarese, “3d semantic parsing of large-scale indoor spaces,” in IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 1534–1543

    2016

  52. [52]

    Gated linear attention transformers with hardware-efficient training,

    S. Yang, B. Wang, Y . Shen, R. Panda, and Y . Kim, “Gated linear attention transformers with hardware-efficient training,” inForty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024. OpenReview.net, 2024

    2024

  53. [53]

    Training data-efficient image transformers & distillation through attention,

    H. Touvron, M. Cord, M. Douze, F. Massa, A. Sablayrolles, and H. J ´egou, “Training data-efficient image transformers & distillation through attention,” inProceedings of the 38th International Conference on Machine Learning, ICML, ser. Proceedings of Machine Learning Research, vol. 139, 2021, pp. 10 347–10 357

    2021

  54. [54]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” inIEEE/CVF International Conference on Computer Vision. IEEE, 2021, pp. 9992–10 002

    2021

  55. [55]

    Towards large-scale 3d representation learning with multi-dataset point prompt training,

    X. Wu, Z. Tian, X. Wen, B. Peng, X. Liu, K. Yu, and H. Zhao, “Towards large-scale 3d representation learning with multi-dataset point prompt training,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024. IEEE, 2024, pp. 19 551–19 562

    2024

  56. [56]

    Sonata: Self-supervised learning of reliable point representations,

    X. Wu, D. DeTone, D. P. Frost, T. Shen, C. Xie, N. Yang, J. J. Engel, R. A. Newcombe, H. Zhao, and J. Straub, “Sonata: Self-supervised learning of reliable point representations,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025, Nashville, TN, USA, June 11-15, 2025. Computer Vision Foundation / IEEE, 2025, pp. 22 193–22 204

    2025

  57. [57]

    Scannet: Richly-annotated 3d reconstructions of indoor scenes,

    A. Dai, A. X. Chang, M. Savva, M. Halber, T. A. Funkhouser, and M. Nießner, “Scannet: Richly-annotated 3d reconstructions of indoor scenes,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2017, pp. 2432–2443

    2017