Pith. sign in

REVIEW 4 major objections 5 minor 99 references

SpikePEFT is the first parameter-efficient fine-tuning framework for spiking point cloud models: it adapts a frozen spiking backbone by tuning membrane decay and firing thresholds and selectively correcting near-threshold silent states, rea

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 14:36 UTC pith:OXUF6BCQ

load-bearing objection A sensible first PEFT framework for spiking point cloud models, but the core SSDA mechanism rests on an untested premise and a missing supplement. the 4 major comments →

arxiv 2607.29048 v1 pith:OXUF6BCQ submitted 2026-07-31 cs.CV

Parameter-Efficient Fine-Tuning for Spiking Point Cloud Models

Zihao Guo , Jihua Zhu , Yiding Sun , Lin Chen , Danwei Wang This is my paper
classification cs.CV
keywords spiking neural networkspoint cloud analysisparameter-efficient fine-tuningmembrane potentialsilent-state disambiguationintrinsic dynamics tuningenergy-efficient computationevent-driven
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper argues that a pre-trained spiking neural network for point clouds can be adapted to new tasks without full fine-tuning, by changing only the intrinsic dynamics of its neurons and by recovering information that binary spikes normally discard. The proposed SpikePEFT framework achieves accuracy slightly better than full fine-tuning on three real-world ScanObjectNN splits, using only about 5% of the parameters and adding a small amount of energy consumption. If correct, this means that the knowledge in a spiking 3D backbone is largely carried by its synaptic weights, and that task-specific adaptation can be done by adjusting membrane decay, firing thresholds, and the sub-threshold states that sit just below the firing boundary. The paper also shows the approach works across classification, part segmentation, and semantic segmentation, and on two different pre-trained spiking backbones.

Core claim

SpikePEFT introduces two mechanisms for parameter-efficient adaptation of spiking point cloud models: Intrinsic Dynamics Tuning (IDT) and Silent-State Disambiguation Adaptation (SSDA). IDT keeps all synaptic transformations frozen and instead applies bounded residual adjustments to the membrane decay and firing threshold of spiking neurons, using a logit-space sigmoid for decay and a multiplicative exponential for the threshold. SSDA addresses the information loss caused by binary spike propagation: it identifies neurons whose membrane potential lies in a narrow band just below the firing threshold—silent yet potentially informative—and applies a lightweight, event-driven bottleneck correcti

What carries the argument

The core object is the near-threshold membrane state of a spiking neuron. SSDA computes a binary mask Q_t^l that selects membrane potentials between the adapted firing threshold and a learnable lower interval boundary (V_th^l - delta^l), then uses a sparse bottleneck: a down-projection matrix that accumulates only columns corresponding to active events, a binary spike threshold in the bottleneck, and a zero-initialized up-projection that adds corrections only at the selected positions. IDT is the second mechanism, providing bounded residuals to membrane decay (in logit space) and firing threshold (multiplicative), with a tanh bound to keep the dynamics valid. Together they allow adaptation w

Load-bearing premise

The central premise is that membrane potentials in a narrow band just below the firing threshold contain task-relevant information that a learned sparse correction can recover; the paper defers the theory to a supplementary appendix that is not included, so the claim rests on empirical support.

What would settle it

Take a trained SpikePEFT model and add noise to the membrane potentials of near-threshold silent states before SSDA is applied; if the accuracy does not drop, those states are not actually carrying task-relevant information. A more direct test is to compute the mutual information between the values in the selected band and the task labels—near-zero would falsify the claim that SSDA is recovering genuinely informative sub-threshold evidence.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Frozen spiking backbones can be tuned for downstream 3D tasks with only ~5% of the trainable parameters, reducing storage and memory overhead compared with full fine-tuning.
  • SpikePEFT outperforms full fine-tuning on the most challenging ScanObjectNN split (PB_T50_RS) by more than one percentage point on both tested backbones, suggesting less destructive adaptation of pre-trained knowledge.
  • Recovering near-threshold silent states reduces the need for long temporal windows, improving accuracy with fewer time steps and lower latency.
  • The framework extends beyond classification to part and semantic segmentation, with larger gains in the more difficult point-level tasks.
  • Because the added operations are mostly accumulation operations (AC) rather than multiply–accumulate (MAC), the energy advantage of spiking models over ANN-based models is retained.

Where Pith is reading between the lines

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

  • The same pre-spike membrane modulation principle could apply to other discrete-activation networks, such as binary or quantized neural networks, where thresholding similarly discards borderline information.
  • The learnable interval delta in SSDA could be interpreted as an adaptive threshold band; a testable extension is to probe whether the learned band width correlates with task difficulty or per-channel sensitivity.
  • The energy comparison rests on theoretical per-operation costs; actual neuromorphic hardware measurements could change the magnitude of the advantage, though the qualitative shift from MAC to AC is likely to remain.
  • The success of SSDA hints that pre-trained spiking models may be underfitting because binary spike generation removes useful sub-threshold evidence; this could motivate new spike-encoding schemes or training objectives that explicitly preserve near-threshold information.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. SpikePEFT proposes a parameter-efficient fine-tuning framework for frozen spiking point cloud backbones. It combines Intrinsic Dynamics Tuning (IDT), which applies bounded residuals to LIF membrane decay and firing thresholds, and Silent-State Disambiguation Adaptation (SSDA), which selects near-threshold silent membrane states and applies a sparse event-driven low-rank correction before spike firing. The method is evaluated on ModelNet40, ScanObjectNN, ShapeNetPart, and S3DIS using SPM and E-3DSNN-L/SVL backbones, reporting 92.4% on ModelNet40 and 85.6% on ScanObjectNN PB_T50_RS with about 5% trainable parameters, and claiming to outperform full fine-tuning by up to 1.4 percentage points while preserving SNN energy efficiency. The internal equations (Eqs. 4-17) are consistent, and Table 4 identifies SSDA as the main source of improvement.

Significance. If verified, this is a useful first contribution: it extends PEFT to spiking 3D models, preserves spike sparsity, and reduces per-task storage. The sparse event-driven adapter design is interesting, and the benchmark coverage (classification, part segmentation, semantic segmentation) is broad. The claims are not yet fully supported because the core SSDA mechanism lacks the deferred theoretical justification and a control ablation, and the headline gains are reported without variance. With those additions, this could be a solid paper.

major comments (4)
  1. [Method, SSDA subsection and Eq. (9)] The central premise that near-threshold silent states in (V_th^l - δ^l, V_th^l) carry task-relevant information is explicitly deferred to 'Supp. A', but no supplement is provided. Table 4 shows SSDA is the largest source of gain (SSDA-only 91.7/89.7/89.2/83.9 vs IDT-only 89.1/83.6/82.8/77.8). Without a control that replaces the near-threshold selector by a random or fixed-width sub-threshold selector with equal parameter count, the gains cannot be attributed to 'silent-state disambiguation' rather than added low-rank capacity. Please provide the missing derivations and this ablation.
  2. [Abstract/Conclusion and Table 6] The 'preserving energy efficiency' claim is computed from theoretical 32-bit FP constants (Horowitz 2014: 4.6 pJ/MAC, 0.9 pJ/AC), not measured hardware energy. The 6.22 mJ and '67.1% lower than PointMamba' figures are operation-count estimates under those constants. Please report measured energy (or wall-clock latency) on relevant neuromorphic/GPU hardware, or clearly label the numbers as idealized estimates and soften the claim.
  3. [Table 1] Headline gains over full fine-tuning are 0.1-1.4 pp with no error bars, seeds, or significance tests; 0.1-0.4 pp differences are within typical run-to-run variation. Please report mean±std over at least 3 seeds for the main comparisons (especially SPM on ModelNet40 and E-3DSNN-L on ModelNet40, where gains are ≤0.1), or specify fixed seeds and show the comparison is stable.
  4. [Experiments section] Implementation details and additional results are referenced as 'Supp. B' and 'Supp. C', and SSDA theory as 'Supp. A', but no supplementary file is included. Essential hyperparameters (e.g., ε_λ, ε_v, δ_max, r, ϑ_a) and the 5.3M segmentation-head parameter allocation in Table 2/3 are not given in the main text. The supplement must be submitted or these details moved into the main text.
minor comments (5)
  1. [Conclusion] 'achieves competitive than full fine-tuning' should be 'achieves competitive performance compared with full fine-tuning'.
  2. [Method, Eqs. (6)-(16)] The notation S_t^{l,0} is introduced for the preliminary spike, but Eq. (16) computes S_t^l without explicitly restating the relationship. Clarify the notation for readers.
  3. [Figure 5] The shaded region is not quantitatively linked to the learned δ; state which δ (or range of δ) is plotted, or add a legend.
  4. [References] Many references are dated 2026. If this is intentional for the submission timeline, verify all are publicly available; otherwise cite published or arXiv-stable versions.
  5. [Table 5] The row label 'PEFT' is ambiguous; use 'SpikePEFT (Ours)' and clarify in the caption that the ANN baseline is PointMamba.

Circularity Check

0 steps flagged

No circular derivation; SSDA's deferred theoretical motivation is a missing-support issue, not a circularity.

full rationale

I walked the claimed derivation chain: IDT and SSDA are defined by explicit equations (Eqs. 4-17), and their reported gains are empirical evaluations on external benchmarks (ModelNet40, ScanObjectNN, ShapeNetPart, S3DIS) rather than consequences of the method's own definitions. The learned interval delta^l and projection matrices A^l, B^l are trained on data and evaluated on held-out test sets, which is standard optimization rather than a fitted input renamed as a prediction. The ablation in Table 4 shows SSDA contributing large gains, but this is an empirical observation, not an identity forced by construction. The paper defers the theoretical justification for near-threshold state selection to a supplementary file that is not included ('The theoretical motivation and additional analysis of near-threshold state selection are detailed in (cf. Supp. A)'), and this is a legitimate missing-support concern for the central premise, but it does not make the derivation circular: the premise is empirically testable and the method is evaluated against external baselines. There are several self-citations (e.g., Guo et al. 2026, Sun et al. 2026a-d, Zhang et al. 2026a-c), but they appear in related-work/positioning contexts and are not load-bearing justifications for SpikePEFT's predictions or uniqueness. No equation reduces to its own input, no parameter is fit to a target and then reported as an independent prediction, and no uniqueness theorem is imported from the authors' prior work. Overall, the paper's central claims are empirically supported rather than circular, with a minor caveat about the missing supplementary theory.

Axiom & Free-Parameter Ledger

5 free parameters · 4 axioms · 2 invented entities

The central empirical claims rest on four domain assumptions and five hand-chosen hyperparameters. No new physical entity is introduced; the two invented entities are internal modules whose independent evidence is limited to in-paper ablations. The absent supplement prevents inspection of parameter values and the promised theoretical motivation for SSDA.

free parameters (5)
  • epsilon_lambda = not reported
    Bound on IDT residual for membrane decay (Eq. 4); hand-chosen hyperparameter controlling adaptation magnitude.
  • epsilon_v = not reported
    Bound on IDT residual for firing threshold (Eq. 4); hand-chosen.
  • delta_max = not reported
    Fixed upper bound of the SSDA silent-state interval (Eq. 8); hand-chosen and determines which states are 'near-threshold'.
  • r (bottleneck dimension) = not reported
    Rank of SSDA down/up projection matrices (Eqs. 10-14); hand-chosen capacity.
  • vartheta_a = not reported
    Fixed firing threshold of the SSDA bottleneck units (Eq. 12); hand-chosen and shared.
axioms (4)
  • domain assumption LIF neuron dynamics with Heaviside spike generation and hard reset are an adequate model of the pretrained spiking backbone.
    Invoked throughout §Preliminaries; standard in SNN literature but not justified here.
  • domain assumption Frozen pre-trained SPM and E-3DSNN-L+SVL backbones retain enough generalizable capacity that tuning only neuron-intrinsic and silent-state parameters can match full fine-tuning.
    The whole PEFT strategy depends on this; supported only by the reported experiments.
  • ad hoc to paper Near-threshold silent membrane states (V_th - δ, V_th) contain task-relevant information recoverable by a learned correction.
    Central premise of SSDA; theoretical motivation deferred to absent Supp. A, so it is an unverified postulate of this paper.
  • domain assumption Energy can be estimated from MAC/AC counts using fixed pJ costs (E_MAC=4.6 pJ, E_AC=0.9 pJ, Horowitz 2014) and the sparse event-driven implementation realizes these counts.
    Basis for Table 6 energy numbers and the 'preserving energy efficiency' claim; not a hardware measurement.
invented entities (2)
  • Near-threshold silent-state selector (learnable interval δ_l) no independent evidence
    purpose: Identifies neurons that did not fire but whose membrane potential is within δ_l of the threshold, and routes only those through the SSDA correction path (Eqs. 8-10).
    The existence of task-relevant information in this interval is asserted; the only evidence is in-paper ablation (Table 4) and the theory is deferred to missing Supp. A.
  • SSDA bottleneck spike units Z_l no independent evidence
    purpose: Provide a sparse binary pathway that computes corrections for selected silent states via down/up projections (Eqs. 11-14).
    Introduced by the method; no falsifiable prediction outside the paper beyond benchmark accuracy.

pith-pipeline@v1.3.0-daily-deepseek · 14015 in / 16448 out tokens · 156116 ms · 2026-08-03T14:36:27.566032+00:00 · methodology

0 comments
read the original abstract

Spiking Neural Networks (SNNs) offer energy-efficient solutions for point cloud analysis on resource-constrained devices through event-driven computation. However, existing pre-trained spiking point cloud models rely on full fine-tuning for downstream task adaptation, incurring substantial parameter and storage overhead. Furthermore, binary spike propagation suppresses task-relevant sub-threshold information. To address these issues, we propose SpikePEFT, the first parameter-efficient fine-tuning framework for spiking point cloud models. Specifically, Intrinsic Dynamics Tuning (IDT) adaptively modulates membrane decay and firing thresholds, enabling efficient neuron-intrinsic adaptation while keeping the pre-trained synaptic transformations frozen. Moreover, Silent-State Disambiguation Adaptation (SSDA) recovers task-relevant information from informative silent states, thereby providing richer evidence for downstream adaptation. Extensive experiments across multiple benchmarks demonstrate the effectiveness and efficiency of SpikePEFT. In particular, our method achieves 92.4% accuracy on ModelNet40 and 85.6\% on the most challenging classification split ScanObjectNN(PB\_T50\_RS) while updating only about 5% of the trainable parameters and preserving the energy efficiency of SNNs. This work provides a promising step toward parameter-efficient adaptation of neuromorphic vision models.

Figures

Figures reproduced from arXiv: 2607.29048 by Danwei Wang, Jihua Zhu, Lin Chen, Yiding Sun, Zihao Guo.

Figure 1
Figure 1. Figure 1: (a) Membrane potential distribution across the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of SpikePEFT. Raw point clouds are sampled by Farthest Point Sampling (FPS), grouped into local patches [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative analysis results for semantic segmenta [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative analysis results for part segmentation [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Normalized membrane potential distributions in the shallow, middle, and deep blocks of SPM under full fine-tuning [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

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

99 extracted references · 8 linked inside Pith

  1. [1]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation , author =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  2. [2]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  3. [3]

    The Tenth International Conference on Learning Representations (ICLR) , year =

    Rethinking Network Design and Local Geometry in Point Cloud: A Simple Residual MLP Framework , author =. The Tenth International Conference on Learning Representations (ICLR) , year =

  4. [4]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    PointNeXt: Revisiting PointNet++ with Improved Training and Scaling Strategies , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  5. [5]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

    Point Transformer , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

  6. [6]

    Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition (CVPR) , year=

    Point Transformer V3: Simpler, Faster, Stronger , author =. Proceedings of the IEEE/CVF conference on Computer Vision and Pattern Recognition (CVPR) , year=

  7. [7]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Point-BERT: Pre-Training 3D Point Cloud Transformers with Masked Point Modeling , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  8. [8]

    European Conference on Computer Vision (ECCV) , year =

    Masked Autoencoders for Point Cloud Self-Supervised Learning , author =. European Conference on Computer Vision (ECCV) , year =

  9. [9]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Point-M2AE: Multi-scale Masked Autoencoders for Hierarchical Point Cloud Pre-training , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  10. [10]

    Advances in Neural Information Processing Systems (NeurIPS) , year=

    Pointgpt: Auto-regressively generative pre-training from point clouds , author=. Advances in Neural Information Processing Systems (NeurIPS) , year=

  11. [11]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

    Towards Compact 3D Representations via Point Feature Enhancement Masked Autoencoders , author =. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

  12. [12]

    First conference on language modeling (COLM) , year =

    Mamba: Linear-time sequence modeling with selective state spaces , author=. First conference on language modeling (COLM) , year =

  13. [13]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    PointMamba: A Simple State Space Model for Point Cloud Analysis , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  14. [14]

    Proceedings of the 32nd ACM International Conference on Multimedia (ACM MM) , year =

    Mamba3D: Enhancing Local Features for 3D Point Cloud Analysis via State Space Model , author =. Proceedings of the 32nd ACM International Conference on Multimedia (ACM MM) , year =

  15. [15]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

    Point Cloud Mamba: Point Cloud Learning via State Space Model , author =. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

  16. [16]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

    Instance-aware Dynamic Prompt Tuning for Pre-trained Point Cloud Models , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

  17. [17]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Dynamic Adapter Meets Prompt Tuning: Parameter-Efficient Transfer Learning for Point Cloud Analysis , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  18. [18]

    Proceedings of the 42nd International Conference on Machine Learning (ICML) , year =

    Exploring Vision Semantic Prompt for Efficient Point Cloud Understanding , author =. Proceedings of the 42nd International Conference on Machine Learning (ICML) , year =

  19. [19]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    PointLoRA: Low-Rank Adaptation with Token Selection for Point Cloud Learning , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  20. [20]

    Proceedings of the 40th International Conference on Machine Learning (ICML) , year =

    Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining , author =. Proceedings of the 40th International Conference on Machine Learning (ICML) , year =

  21. [21]

    ACM Transactions on Graphics (TOG) , year=

    Dynamic Graph CNN for Learning on Point Clouds , author=. ACM Transactions on Graphics (TOG) , year=

  22. [22]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Mikaela Angelina Uy and Quang-Hieu Pham and Binh-Son Hua and Thanh Nguyen and Sai-Kit Yeung , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  23. [23]

    Advances in Neural Information Processing Systems (NeurIPS) , year=

    ZigzagPointMamba: Spatial-Semantic Mamba for Point Cloud Understanding , author=. Advances in Neural Information Processing Systems (NeurIPS) , year=

  24. [24]

    International conference on machine learning (ICML) , year=

    Parameter-efficient transfer learning for NLP , author=. International conference on machine learning (ICML) , year=

  25. [25]

    Edward J Hu and Yelong Shen and Phillip Wallis and Zeyuan Allen-Zhu and Yuanzhi Li and Shean Wang and Lu Wang and Weizhu Chen , booktitle=. Lo

  26. [26]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year=

    Point-PEFT: Parameter-efficient fine-tuning for 3D pre-trained models , author=. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year=

  27. [27]

    IEEE transactions on pattern analysis and machine intelligence (TPAMI) , year=

    Parameter-Efficient Fine-Tuning in Spectral Domain for Point Cloud Learning , author=. IEEE transactions on pattern analysis and machine intelligence (TPAMI) , year=

  28. [28]

    Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) , year=

    PMA: Towards Parameter-Efficient Point Cloud Understanding via Point Mamba Adapter , author=. Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) , year=

  29. [29]

    Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) , year=

    MoST: Efficient Monarch Sparse Tuning for 3D Representation Learning , author=. Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) , year=

  30. [30]

    Proceedings of the 42nd International Conference on Machine Learning (ICML) , year =

    Ai, Zixiang and Liu, Zichen and Lei, Yuanhang and Cui, Zhenyu and Zou, Xu and Zhou, Jiahuan , title =. Proceedings of the 42nd International Conference on Machine Learning (ICML) , year =

  31. [31]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    3D ShapeNets: A Deep Representation for Volumetric Shapes , author=. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=

  32. [32]

    International Conference on Computer Vision (ICCV) , year =

    Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Data , author =. International Conference on Computer Vision (ICCV) , year =

  33. [33]

    International Conference on Learning Representations (ICLR) , year =

    Efficiently Modeling Long Sequences with Structured State Spaces , author=. International Conference on Learning Representations (ICLR) , year =

  34. [34]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Point transformer V2: Grouped Vector Attention and Partition-based Pooling , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  35. [35]

    Advances in neural information processing systems (NeurIPS) , year=

    Pointcnn: Convolution on x-transformed points , author=. Advances in neural information processing systems (NeurIPS) , year=

  36. [36]

    The Eleventh International Conference on Learning Representations (ICLR) , year =

    Autoencoders as Cross-Modal Teachers: Can Pretrained 2D Image Transformers Help 3D Representation Learning? , author=. The Eleventh International Conference on Learning Representations (ICLR) , year =

  37. [37]

    Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence (IJCAI) , year=

    Joint-MAE: 2D-3D joint masked autoencoders for 3D point cloud pre-training , author=. Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence (IJCAI) , year=

  38. [38]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR) , year=

    Learning 3d representations from 2d pre-trained models via image-to-point masked autoencoders , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR) , year=

  39. [39]

    Advances in Neural Information Processing Systems (NeurIPS) , year=

    Adaptformer: Adapting vision transformers for scalable visual recognition , author=. Advances in Neural Information Processing Systems (NeurIPS) , year=

  40. [40]

    Proceedings of the 2021 conference on empirical methods in natural language processing (EMNLP) , year=

    The power of scale for parameter-efficient prompt tuning , author=. Proceedings of the 2021 conference on empirical methods in natural language processing (EMNLP) , year=

    2021

  41. [41]

    European conference on computer vision (ECCV) , year=

    Visual prompt tuning , author=. European conference on computer vision (ECCV) , year=

  42. [42]

    Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (ACL), Volume 2: Short Papers , year=

    P-tuning: Prompt tuning can be comparable to fine-tuning across scales and tasks , author=. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (ACL), Volume 2: Short Papers , year=

  43. [43]

    European Conference on Computer Vision (ECCV) , year=

    Parameter-efficient and memory-efficient tuning for vision transformer: a disentangled approach , author=. European Conference on Computer Vision (ECCV) , year=

  44. [44]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR) , year=

    Conditional prompt learning for vision-language models , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR) , year=

  45. [45]

    International journal of computer vision (IJCV) , year=

    Learning to prompt for vision-language models , author=. International journal of computer vision (IJCV) , year=

  46. [46]

    European Conference on Computer Vision (ECCV) , year=

    Masked discrimination for self-supervised learning on point clouds , author=. European Conference on Computer Vision (ECCV) , year=

  47. [47]

    arXiv preprint arXiv:1512.03012 , year=

    Shapenet: An information-rich 3d model repository , author=. arXiv preprint arXiv:1512.03012 , year=

    Pith/arXiv arXiv

  48. [48]

    Advances in neural information processing systems (NeurIPS) , year=

    Attention is all you need , author=. Advances in neural information processing systems (NeurIPS) , year=

  49. [49]

    Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (NAACL) , year=

    Bert: Pre-training of deep bidirectional transformers for language understanding , author=. Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (NAACL) , year=

    2019

  50. [50]

    Pattern Recognition (PR) , year =

    HyperPoint: Multimodal 3D foundation model in hyperbolic space , author =. Pattern Recognition (PR) , year =

  51. [51]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year=

    StruMamba3D: Exploring Structural Mamba for Self-supervised Point Cloud Representation Learning , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year=

  52. [52]

    Proceedings of the 33rd ACM International Conference on Multimedia (ACM MM) , year=

    HydraMamba: Multi-Head State Space Model for Global Point Cloud Learning , author=. Proceedings of the 33rd ACM International Conference on Multimedia (ACM MM) , year=

  53. [53]

    Pattern Recognition (PR) , year =

    OctMamba: Mamba-based octree context entropy model for point cloud geometry compression , author =. Pattern Recognition (PR) , year =

  54. [54]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year=

    CloudMamba: Grouped Selective State Spaces for Point Cloud Analysis , author=. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year=

  55. [55]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

    Point cloud pre-training with diffusion models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

  56. [56]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR) , year=

    Relation-shape convolutional neural network for point cloud analysis , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR) , year=

  57. [57]

    ACM Transactions on Graphics (TOG) , year=

    A scalable active framework for region annotation in 3D shape collections , author=. ACM Transactions on Graphics (TOG) , year=

  58. [58]

    Unsupervised Point Cloud Pre-Training via Occlusion Completion , author =

  59. [59]

    David Hilbert , booktitle =

  60. [60]

    Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) , year=

    Spectral informed mamba for robust point cloud processing , author=. Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR) , year=

  61. [61]

    Prefix-tuning: Optimizing continuous prompts for generation , author=. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) (ACL) , year=

  62. [62]

    Proceedings of the IEEE/CVF international conference on computer vision (ICCV) , year=

    Revisiting the parameter efficiency of adapters from the perspective of precision redundancy , author=. Proceedings of the IEEE/CVF international conference on computer vision (ICCV) , year=

  63. [63]

    Journal of Machine Learning Research (JMLR) , year =

    Laurens van der Maaten and Geoffrey Hinton , title =. Journal of Machine Learning Research (JMLR) , year =

  64. [64]

    arXiv preprint arXiv:2602.23945 , year=

    Pointcot: A multi-modal benchmark for explicit 3d geometric reasoning , author=. arXiv preprint arXiv:2602.23945 , year=

    arXiv

  65. [65]

    arXiv preprint arXiv:2603.23957 , year=

    PointRFT: Explicit Reinforcement Fine-tuning for Point Cloud Few-shot Learning , author=. arXiv preprint arXiv:2603.23957 , year=

    arXiv

  66. [66]

    2025 International Joint Conference on Neural Networks (IJCNN) , year=

    PointDico: Contrastive 3D Representation Learning Guided by Diffusion Models , author=. 2025 International Joint Conference on Neural Networks (IJCNN) , year=

    2025

  67. [67]

    and Jiang, Helen and Brilakis, Ioannis and Fischer, Martin and Savarese, Silvio , title =

    Armeni, Iro and Sener, Ozan and Zamir, Amir R. and Jiang, Helen and Brilakis, Ioannis and Fischer, Martin and Savarese, Silvio , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  68. [68]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Spiking PointNet: Spiking Neural Networks for Point Clouds , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  69. [69]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

    Spiking Point Transformer for Point Cloud Classification , author =. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

  70. [70]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

    Efficient 3D Recognition with Event-Driven Spike Sparse Convolution , author =. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

  71. [71]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

    Efficient Spiking Point Mamba for Point Cloud Analysis , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

  72. [72]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

    Primary Visual Cortex Inspired Point Cloud Analysis Framework , author =. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

  73. [73]

    Proceedings of the International Conference on Learning Representations (ICLR) , year =

    3DSMT: A Hybrid Spiking Mamba-Transformer for Point Cloud Analysis , author =. Proceedings of the International Conference on Learning Representations (ICLR) , year =

  74. [74]

    Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

    Point-to-Spike Residual Learning for Energy-Efficient 3D Point Cloud Classification , author =. Proceedings of the AAAI Conference on Artificial Intelligence (AAAI) , year =

  75. [75]

    Proceedings of the International Conference on Learning Representations (ICLR) , year =

    Spiking Discrepancy Transformer for Point Cloud Analysis , author =. Proceedings of the International Conference on Learning Representations (ICLR) , year =

  76. [76]

    arXiv preprint arXiv:2505.17674 , year=

    SVL: Spike-based Vision-language Pretraining for Efficient 3D Open-world Understanding , author=. arXiv preprint arXiv:2505.17674 , year=

    Pith/arXiv arXiv

  77. [77]

    Nature , year =

    Towards Artificial General Intelligence with Hybrid Tianjic Chip Architecture , author =. Nature , year =

  78. [78]

    arXiv preprint arXiv:2605.03438 , year=

    Mantis: Mamba-native tuning is efficient for 3d point cloud foundation models , author=. arXiv preprint arXiv:2605.03438 , year=

    Pith/arXiv arXiv

  79. [79]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages=

    Efficient converted spiking neural network for 3d and 2d classification , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages=

  80. [80]

    Science Advances (Sci

    SpikingJelly: An Open-Source Machine Learning Infrastructure Platform for Spike-Based Intelligence , author=. Science Advances (Sci. Adv.) , volume=

Showing first 80 references.