Pith. sign in

REVIEW 2 major objections 5 minor 78 references

MICAS: Multi-grained In-Context Adaptive Sampling for 3D Point Cloud Processing

T0 review · 2 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read MICAS claims that replacing fixed farthest-point sampling with task- and query-adaptive sampling is what lets a single in-context model handle reconstruction, denoising, registration, and part segmentation at once.

desk verdict Useful incremental paper on adaptive sampling for point cloud ICL, but the main gain attribution isn't fully pinned down—backbone freeze unstated, baseline numbers inconsistent. read the letter →

arxiv 2411.16773 v2 pith:EOCJYBT2 submitted 2024-11-25 cs.CV

Feifei Shao , Ping Liu , Zhao Wang , Yawei Luo , Hongwei Wang , Jun Xiao This is my paper · ORCID
classification cs.CV
keywords pointcloudprocessingin-contextlearningadaptivesamplingGumbel-softmaxpromptselectionmulti-taskpartsegmentationregistration
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

This paper argues that the fixed sampling strategies used by current in-context learning models for 3D point clouds are the bottleneck: the same sampling rule behaves poorly across different tasks (inter-task sensitivity) and across different prompt choices for one task (intra-task sensitivity). It proposes MICAS, a framework that adds two adaptive components to an existing in-context point cloud model: a task-adaptive point sampler that reads task cues from the prompt and selects central points via differentiable Gumbel sampling, and a query-specific prompt sampler that ranks candidate prompts by predicted performance and feeds the best one to the model. On the in-context point cloud benchmark used in this paper, covering reconstruction, denoising, registration, and part segmentation, MICAS improves part-segmentation mIOU from 79.0 to 87.9 over the baseline and lowers registration Chamfer distance from 14.1 to 9.8, with consistent gains elsewhere. The paper's core claim is that adaptive sampling at both point and prompt levels, rather than a new backbone, is what unlocks better multi-task in-context point cloud processing.

What carries the argument

The load-bearing mechanism is a pair of sampling modules. Task-adaptive point sampling concatenates a task feature extracted from the prompt (via a PointNet encoder) with per-point features of the query cloud, maps the combination through a fully connected layer to sampling weights, and relaxes discrete selection with Gumbel-softmax so the chosen central points are differentiable projections of the original cloud. Query-specific prompt sampling fuses the query cloud with candidate prompt pairs, runs a PointNet 'prompt sampler' to produce a probability per candidate, generates pseudo-labels by running the in-context model on each candidate, and trains with a list-wise ranking loss so the top-ranked prompt is the one expected to perform best. Together they make both the point-level tokenization and the prompt-level context adaptive to the query and task.

What would settle it

Retrain the full MICAS pipeline with the in-context backbone unfrozen in the first stage and compare the result with the paper's numbers: if the gap is large, adaptive sampling alone contributes less than claimed. Alternatively, replace the learned prompt sampler with random prompt selection and measure registration and reconstruction Chamfer distance: if performance does not drop, the ranking mechanism is not doing the work.

Watch

Extended reading notes

Core claim

MICAS establishes that the performance ceiling of point cloud in-context learning is set by non-adaptive sampling: replacing farthest-point sampling with a prompt-conditioned, Gumbel-softmax point sampler and adding query-conditioned prompt selection yields large, consistent gains across four tasks while keeping the underlying in-context model unchanged. The method does not modify the transformer backbone or the masked-point reconstruction objective; it only changes which points become central tokens and which prompt is shown. Under the paper's step-wise training protocol, the base in-context model is kept fixed and only the two sampling modules are trained, so the reported improvements are attributed to the sampling decisions themselves.

Load-bearing premise

The reported gains assume that, during the first training stage, the base in-context model is kept frozen while the adaptive sampling modules are trained; if the backbone parameters are also updated, the improvement could come partly from additional fine-tuning rather than from adaptive sampling alone.

Editorial extensions

If this is right

  • The same backbone, with only sampling changed, improves part segmentation mIOU by 8.9 points and registration Chamfer distance from 14.1 to 9.8 on the benchmark used in this paper.
  • The two modules are complementary: point sampling helps denoising and segmentation, prompt sampling helps reconstruction and registration, and combining them gives the best overall result.
  • The approach works on both the concatenation and the separation style backbone variants, and the gains survive swapping the PointNet feature extractor for DGCNN, indicating robustness to the choice of encoder.
  • Inference cost grows roughly threefold (about 15.6 ms to 47.1 ms per query), a trade the authors accept for the accuracy gain.
  • MICAS outperforms multi-task models with shared backbones and separate heads on part segmentation by a wide margin, suggesting adaptive sampling reduces the need for task-specific heads.

Reading between the lines

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

  • Beyond the paper: the step-wise training protocol leaves open whether the backbone is truly frozen; if it were trained jointly, part of the gain could come from fine-tuning rather than adaptive sampling alone.
  • Beyond the paper: because the prompt sampler produces a ranking, the same list-wise loss could be applied to select an ordered set of multiple prompts per query instead of a single best prompt.
  • Beyond the paper: testing MICAS against other learned samplers would clarify whether task conditioning, rather than learnable sampling in general, is the active ingredient behind the gains.
  • Beyond the paper: the benchmark uses synthetic data, so whether these gains transfer to noisy real-world scans in robotics or augmented reality remains an untested extension.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper proposes MICAS, a multi-grained in-context adaptive sampling mechanism for 3D point cloud processing. It introduces two components: task-adaptive point sampling, which uses PointNet-based encoders and Gumbel-softmax to select central points conditioned on prompt-level task information, and query-specific prompt sampling, which retrieves the best prompt for each query by training a ranker with a listwise loss on pseudo-labels from the in-context learning model. The method is evaluated on the ShapeNet In-Context dataset across reconstruction, denoising, registration, and part segmentation, reporting consistent improvements over PIC baselines (e.g., part segmentation mIOU from 79.0 to 87.9 for PIC-Cat) and supported by ablations and additional robustness experiments in the supplementary material.

Significance. If the results hold, the paper makes a useful contribution by addressing a real limitation of in-context learning for point clouds: the inflexibility of fixed sampling and prompt selection. The two-level adaptation is simple and well motivated, and the experiments show large gains on several tasks, with an ablation suggesting that the two modules are complementary. The supplementary robustness analysis with DGCNN is a positive addition, and the explicit discussion of the inference-time limitation is honest. However, the paper provides no code or checkpoints, and the central gain-attribution argument is undermined by an underspecified training protocol and by inconsistent baseline numbers between the main comparison and the ablation.

major comments (2)
  1. [3.4] Equation (8) defines the sampling loss as L_sampling = L_cd(R_pred, G) + α·L_cd(C, X), and Section 3.4 states only that 'we train the task-adaptive point sampling module.' If the ICL backbone (PIC) parameters are not frozen, the first term will update the backbone during Phase 1, so the gains in Table 2 (e.g., part segmentation mIOU 79.9→87.9 for PIC-Cat) could partly come from fine-tuning the in-context model rather than from adaptive sampling alone. The text must explicitly state that the ICL model is frozen during Phase 1; the 'Frozen' label in Figure 3 is not an adequate training-configuration specification. Please provide this clarification and, if the backbone is indeed trainable, redo the ablation with a frozen backbone to isolate the sampling contribution.
  2. [Tables 1 and 2] The FPS-only baseline rows in Table 2 are inconsistent with the corresponding PIC-Cat and PIC-Sep rows in Table 1. For example, PIC-Cat reconstruction average CD is 4.9 in Table 2 but 4.3 in Table 1, and part segmentation mIOU is 79.9 vs 79.0; PIC-Sep registration average CD is 8.6 vs 10.3. Because Table 2 is the ablation that attributes performance gains to the two proposed modules, the baselines used there must be identical (or the discrepancies explained) for the ablation to be interpretable. Please reconcile these numbers and state clearly which configuration each row represents.
minor comments (5)
  1. [Figure 3 caption] The caption refers to the proposed architecture as 'MAL-ICL' instead of 'MICAS'; please correct this typo.
  2. [4.2] In the 'Comparison to In-context learning Models' paragraph, PIC-Cat and PIC-Sep are repeatedly cited as reference [42], but these baselines come from reference [11]; please fix the citation.
  3. [Abstract] The abstract reports a '4.1% improvement' in part segmentation, but the tables show much larger changes (e.g., 79.0→87.9 mIOU in Table 1); please specify whether this is an absolute or relative improvement and which baseline it refers to.
  4. [3.3, Eq. (12)] The listwise ranking loss in Eq. (12) uses the notation r(~y) without defining how the ranking order is computed; please clarify this notation and the exact summation ranges.
  5. [8.1 (supplementary)] The limitation discussion mentions an added inference time of about 25 ms, but it would help to report the end-to-end timing of the full pipeline and to specify the GPU used; the inference time in Table 2 is on 'three 1080ti GPUs,' which is unusual for per-query latency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: MICAS is an empirical method paper whose gains are measured on held-out test queries against external baselines; self-referential pseudo-labeling in the prompt sampler is standard supervision, not a derivation that reduces to its own inputs.

full rationale

MICAS is an empirical method paper rather than a formal derivation, and I could not find any step where a claimed prediction or first-principles result is equivalent to its inputs by construction. The task-adaptive point sampler is trained with the loss in Eq. 8, which combines the ICL model's Chamfer reconstruction error with a fidelity term between sampled centers and the input cloud; comparing the resulting sampler against FPS on the same ICL backbone is a legitimate empirical ablation. The fact that Section 3.4 does not explicitly state whether the PIC backbone is frozen (Figure 3 labels it 'Frozen') is a potential attribution confound, because simultaneous backbone fine-tuning could contribute to the reported gains, but this is a correctness/experimental-protocol concern, not circularity. The query-specific prompt sampler in Section 3.3 is trained on pseudo-labels generated by the ICL model itself via Eq. 9, which is self-referential supervision; however, the pseudo-labels are computed for training queries, the ranking loss in Eq. 12 fits a selector to those labels, and the method is then evaluated on held-out test queries, so the reported improvements are not true by construction. No load-bearing self-citation chain is present: PIC [11] and related point-in-context works are external prior art used as baselines and backbones, not as unverified uniqueness theorems or ansatze that force the authors' choices. Therefore the paper receives a circularity score of 0.

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

The central claim rests on standard ML assumptions: the learned sampling and prompt selection modules generalize from training to test queries, the Gumbel-softmax relaxation works, and the ICL backbone can be treated as a fixed feature extractor. There are no invented physical entities. Key free parameters are the loss weight alpha, the candidate prompt count K, and the Gumbel temperature tau, none of which are ablated.

free parameters (4)
  • alpha (sampling loss weight) = 0.5
    Hyperparameter in Lsampling = Lcd(Rpred,G) + alpha * Lcd(C,X); no sensitivity analysis is reported.
  • K (number of prompt candidates) = 8
    Number of randomly selected candidate prompts per query; the paper suggests in limitations that reducing K to 4 or 2 would lower computational cost.
  • Gumbel temperature tau = not reported
    Used in Gumbel-softmax in Eq. 6; the paper does not specify the value or annealing schedule.
  • Patch count N and neighborhood size = 64 patches, 32 points per patch
    Architecture choices inherited from PIC; they determine the granularity of sampling and are not tuned in this work.
assumptions (4)
  • domain assumption PointNet-based encoders capture sufficient task and point features for adaptive sampling
    The task encoder and point encoder are built on PointNet branches; a robustness study with DGCNN is provided, but the paper does not analyze whether stronger encoders would change results.
  • domain assumption Gumbel-softmax relaxation preserves gradients and yields valid point selections
    Eqs. 6-7 use Gumbel-softmax to generate central points as weighted sums of original points; this follows prior work but assumes the relaxation does not hurt training or final performance.
  • domain assumption The ICL model's outputs provide reliable pseudo-labels for prompt ranking
    Eq. 9 uses the frozen ICL model to score prompts; if these scores are noisy or mis-calibrated across tasks, the list-wise ranking loss may not transfer to test queries.
  • ad hoc to paper Max-min normalization of pseudo-labels per task is appropriate
    Section 3.3 applies max-min normalization to make performance comparable across tasks; the paper does not justify this choice over other normalization schemes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MICAS: Multi-grained In-Context Adaptive Sampling for 3D Point Cloud Processing." pith.science (2026). https://pith.science/paper/EOCJYBT2

@misc{pith2026241116773,
  author       = {Pith},
  title        = {Pith review of: MICAS: Multi-grained In-Context Adaptive Sampling for 3D Point Cloud Processing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EOCJYBT2}},
  note         = {Machine review of arXiv:2411.16773}
}
abstract

Point cloud processing (PCP) encompasses tasks like reconstruction, denoising, registration, and segmentation, each often requiring specialized models to address unique task characteristics. While in-context learning (ICL) has shown promise across tasks by using a single model with task-specific demonstration prompts, its application to PCP reveals significant limitations. We identify inter-task and intra-task sensitivity issues in current ICL methods for PCP, which we attribute to inflexible sampling strategies lacking context adaptation at the point and prompt levels. To address these challenges, we propose MICAS, an advanced ICL framework featuring a multi-grained adaptive sampling mechanism tailored for PCP. MICAS introduces two core components: task-adaptive point sampling, which leverages inter-task cues for point-level sampling, and query-specific prompt sampling, which selects optimal prompts per query to mitigate intra-task sensitivity. To our knowledge, this is the first approach to introduce adaptive sampling tailored to the unique requirements of point clouds within an ICL framework. Extensive experiments show that MICAS not only efficiently handles various PCP tasks but also significantly outperforms existing methods. Notably, it achieves a remarkable $4.1\%$ improvement in the part segmentation task and delivers consistent gains across various PCP applications.

Figures

Figures reproduced from arXiv: 2411.16773 by the authors.

Figure 1
Figure 1. Inter-task and intra-task sensitivities in in-context learning. The red and green points are sampled using Farthest Point Sampling [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Comparison between the proposed MICAS and the traditional in-context learning framework. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Overview of the proposed MAL-ICL. (a) Task-adaptive point sampling is designed to achieve better point-level sampling. (b) [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Qualitative experimental results compared with the PIC-Cat [ [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

78 extracted references · 70 canonical work pages

  1. [1]

    Visual prompting via image inpaint- ing

    Amir Bar, Yossi Gandelsman, Trevor Darrell, Amir Glober- son, and Alexei Efros. Visual prompting via image inpaint- ing. In NeurIPS, 2022. 1

    work page 2022

  2. [2]

    Adaptive neural networks for efficient inference

    Tolga Bolukbasi, Joseph Wang, Ofer Dekel, and Venkatesh Saligrama. Adaptive neural networks for efficient inference. In ICML, 2017. 6

    work page 2017

  3. [3]

    Lan- guage models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Sub- biah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakan- tan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Lan- guage models are few-shot learners. In NeurIPS, 2020. 1

    work page 2020

  4. [4]

    From ranknet to lambdarank to lambdamart: An overview

    Christopher JC Burges. From ranknet to lambdarank to lambdamart: An overview. Learning, 2010. 6

    work page 2010

  5. [5]

    Shapenet: An information-rich 3d model repository

    Angel X Chang, Thomas Funkhouser, Leonidas Guibas, Pat Hanrahan, Qixing Huang, Zimo Li, Silvio Savarese, Mano- lis Savva, Shuran Song, Hao Su, et al. Shapenet: An information-rich 3d model repository. arXiv, 2015. 2, 6

    work page 2015

  6. [6]

    Unsu- pervised learning of geometric sampling invariant represen- tations for 3d point clouds

    Haolan Chen, Shitong Luo, Xiang Gao, and Wei Hu. Unsu- pervised learning of geometric sampling invariant represen- tations for 3d point clouds. In ICCV, 2021. 2

    work page 2021

  7. [7]

    Pra-net: Point relation-aware network for 3d point cloud analysis

    Silin Cheng, Xiwu Chen, Xinwei He, Zhe Liu, and Xiang Bai. Pra-net: Point relation-aware network for 3d point cloud analysis. IEEE Transactions on Image Processing, 30:4436– 4448, 2021. 3

    work page 2021

  8. [8]

    Autoencoders as cross-modal teachers: Can pretrained 2d image transform- ers help 3d representation learning? In ICLR, 2023

    Runpei Dong, Zekun Qi, Linfeng Zhang, Junbo Zhang, Jian- jian Sun, Zheng Ge, Li Yi, and Kaisheng Ma. Autoencoders as cross-modal teachers: Can pretrained 2d image transform- ers help 3d representation learning? In ICLR, 2023. 7

    work page 2023

Show all 78 references
  1. [9]

    Learning to sam- ple

    Oren Dovrat, Itai Lang, and Shai Avidan. Learning to sam- ple. In CVPR, 2019. 3, 1

    2019

  2. [10]

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

    Haoqiang Fan, Hao Su, and Leonidas J Guibas. A point set generation network for 3d object reconstruction from a single image. In CVPR, 2017. 4, 6, 7

    2017

  3. [11]

    Explore in-context learning for 3d point cloud understanding

    Zhongbin Fang, Xiangtai Li, Xia Li, Joachim M Buhmann, Chen Change Loy, and Mengyuan Liu. Explore in-context learning for 3d point cloud understanding. InNeurIPS, 2024. 1, 2, 3, 4, 5, 6, 7, 8

    2024

  4. [12]

    Discriminatory analysis: nonparametric dis- crimination, consistency properties

    Evelyn Fix. Discriminatory analysis: nonparametric dis- crimination, consistency properties. USAF school of Avia- tion Medicine, 1985. 4

    1985

  5. [13]

    Flex-convolution: Million-scale point-cloud learning beyond grid-worlds

    Fabian Groh, Patrick Wieschollek, and Hendrik PA Lensch. Flex-convolution: Million-scale point-cloud learning beyond grid-worlds. In ACCV, 2018. 2, 3

    2018

  6. [14]

    Pct: Point cloud transformer

    Meng-Hao Guo, Jun-Xiong Cai, Zheng-Ning Liu, Tai-Jiang Mu, Ralph R Martin, and Shi-Min Hu. Pct: Point cloud transformer. Computational Visual Media, 2021. 7

    2021

  7. [15]

    Point-bind & point-llm: Aligning point cloud with multi-modality for 3d understand- ing, generation, and instruction following

    Ziyu Guo, Renrui Zhang, Xiangyang Zhu, Yiwen Tang, Xi- anzheng Ma, Jiaming Han, Kexin Chen, Peng Gao, Xi- anzhi Li, Hongsheng Li, et al. Point-bind & point-llm: Aligning point cloud with multi-modality for 3d understand- ing, generation, and instruction following. arXiv prepri...

    2023 arXiv

  8. [16]

    Segpoint: Segment any point cloud via large language model

    Shuting He, Henghui Ding, Xudong Jiang, and Bihan Wen. Segpoint: Segment any point cloud via large language model. In ECCV, 2025. 3

    2025

  9. [17]

    Randla-net: Efficient semantic segmentation of large-scale point clouds

    Qingyong Hu, Bo Yang, Linhai Xie, Stefano Rosa, Yulan Guo, Zhihua Wang, Niki Trigoni, and Andrew Markham. Randla-net: Efficient semantic segmentation of large-scale point clouds. In CVPR, 2020. 2, 3

    2020

  10. [18]

    Surface reconstruction from point clouds: A survey and a benchmark

    Zhangjin Huang, Yuxin Wen, Zihao Wang, Jinjuan Ren, and Kui Jia. Surface reconstruction from point clouds: A survey and a benchmark. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 1

    2024

  11. [19]

    Categorical repa- rameterization with gumbel-softmax

    Eric Jang, Shixiang Gu, and Ben Poole. Categorical repa- rameterization with gumbel-softmax. In ICLR, 2016. 2, 4, 5

    2016

  12. [20]

    Subjective and objective quality evaluation of 3d point cloud denoising algorithms

    Alireza Javaheri, Catarina Brites, Fernando Pereira, and Jo˜ao Ascenso. Subjective and objective quality evaluation of 3d point cloud denoising algorithms. In ICMEW, 2017. 4

    2017

  13. [21]

    Bayesian point cloud re- construction

    Philipp Jenke, Michael Wand, Martin Bokeloh, Andreas Schilling, and Wolfgang Straßer. Bayesian point cloud re- construction. In Computer graphics forum, 2006. 4

    2006

  14. [22]

    Dg- pic: Domain generalized point-in-context learning for point cloud understanding

    Jincen Jiang, Qianyu Zhou, Yuhang Li, Xuequan Lu, Meili Wang, Lizhuang Ma, Jian Chang, and Jian Jun Zhang. Dg- pic: Domain generalized point-in-context learning for point cloud understanding. In ECCV, 2025. 1, 3

    2025

  15. [23]

    Normalization matters in weakly supervised object localiza- tion

    Jeesoo Kim, Junsuk Choe, Sangdoo Yun, and Nojun Kwak. Normalization matters in weakly supervised object localiza- tion. In ICCV, 2021. 6

    2021

  16. [24]

    Oneformer3d: One transformer for unified point cloud segmentation

    Maxim Kolodiazhnyi, Anna V orontsova, Anton Konushin, and Danila Rukhovich. Oneformer3d: One transformer for unified point cloud segmentation. In CVPR, 2024. 1

    2024

  17. [25]

    Large-scale point cloud semantic segmentation with superpoint graphs

    Loic Landrieu and Martin Simonovsky. Large-scale point cloud semantic segmentation with superpoint graphs. In CVPR, 2018. 4

    2018

  18. [26]

    Samplenet: Differ- entiable point cloud sampling

    Itai Lang, Asaf Manor, and Shai Avidan. Samplenet: Differ- entiable point cloud sampling. In CVPR, 2020. 3, 1

    2020

  19. [27]

    Diverse demon- strations improve in-context compositional generalization

    Itay Levy, Ben Bogin, and Jonathan Berant. Diverse demon- strations improve in-context compositional generalization. In ACL, 2023. 3

    2023

  20. [28]

    Uni- fied demonstration retriever for in-context learning

    Xiaonan Li, Kai Lv, Hang Yan, Tianyang Lin, Wei Zhu, Yuan Ni, Guotong Xie, Xiaoling Wang, and Xipeng Qiu. Uni- fied demonstration retriever for in-context learning. In ACL,

  21. [29]

    Pointcnn: Convolution on x-transformed points

    Yangyan Li, Rui Bu, Mingchao Sun, Wei Wu, Xinhan Di, and Baoquan Chen. Pointcnn: Convolution on x-transformed points. In NeurIPS, 2018. 2, 3

    2018

  22. [30]

    What makes good in- context examples for gpt-3? In DeeLIO, 2022

    Jiachang Liu, Dinghan Shen, Yizhe Zhang, Bill Dolan, Lawrence Carin, and Weizhu Chen. What makes good in- context examples for gpt-3? In DeeLIO, 2022. 3

    2022

  23. [31]

    Point-in-context: Understanding point cloud via in-context learning

    Mengyuan Liu, Zhongbin Fang, Xia Li, Joachim M Buh- mann, Xiangtai Li, and Chen Change Loy. Point-in-context: Understanding point cloud via in-context learning. arXiv,

  24. [32]

    Relation-shape convolutional neural network for point cloud analysis

    Yongcheng Liu, Bin Fan, Shiming Xiang, and Chunhong Pan. Relation-shape convolutional neural network for point cloud analysis. In CVPR, 2019. 2

    2019

  25. [33]

    Sgdr: Stochastic gradient descent with warm restarts

    Ilya Loshchilov and Frank Hutter. Sgdr: Stochastic gradient descent with warm restarts. In ICLR, 2017. 7

    2017

  26. [34]

    Score-based point cloud denoising

    Shitong Luo and Wei Hu. Score-based point cloud denoising. In ICCV, 2021. 1, 4

    2021

  27. [35]

    Large language model and domain- specific model collaboration for smart education

    Yawei Luo and Yi Yang. Large language model and domain- specific model collaboration for smart education. Frontiers of Information Technology & Electronic Engineering, 25(3): 333–341, 2024. 1

    2024

  28. [36]

    Re- thinking network design and local geometry in point cloud: A simple residual mlp framework

    Xu Ma, Can Qin, Haoxuan You, Haoxi Ran, and Yun Fu. Re- thinking network design and local geometry in point cloud: A simple residual mlp framework. In ICLR, 2022. 2

    2022

  29. [37]

    Dense 3d point cloud reconstruction using a deep pyramid network

    Priyanka Mandikal and Venkatesh Babu Radhakrishnan. Dense 3d point cloud reconstruction using a deep pyramid network. In WACV, 2019. 1, 4

    2019

  30. [38]

    Step-by- step: Separating planning from realization in neural data-to- text generation

    Amit Moryossef, Yoav Goldberg, and Ido Dagan. Step-by- step: Separating planning from realization in neural data-to- text generation. arXiv preprint arXiv:1904.03396, 2019. 6

    1904 arXiv

  31. [39]

    Adaptive hierarchical down-sampling for point cloud classification

    Ehsan Nezhadarya, Ehsan Taghavi, Ryan Razani, Bingbing Liu, and Jun Luo. Adaptive hierarchical down-sampling for point cloud classification. In CVPR, 2020. 3

    2020

  32. [40]

    Point-e: A system for generat- ing 3d point clouds from complex prompts

    Alex Nichol, Heewoo Jun, Prafulla Dhariwal, Pamela Mishkin, and Mark Chen. Point-e: A system for generat- ing 3d point clouds from complex prompts. arXiv preprint arXiv:2212.08751, 2022. 3

    2022 arXiv

  33. [41]

    Cross-lingual retrieval augmented prompt for low- resource languages

    Ercong Nie, Sheng Liang, Helmut Schmid, and Hinrich Sch¨utze. Cross-lingual retrieval augmented prompt for low- resource languages. In ACL, 2023. 3

    2023

  34. [42]

    Masked autoencoders for point cloud self-supervised learning

    Yatian Pang, Wenxiao Wang, Francis EH Tay, Wei Liu, Yonghong Tian, and Li Yuan. Masked autoencoders for point cloud self-supervised learning. In ECCV, 2022. 7, 8

    2022

  35. [43]

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

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. In CVPR, 2017. 4, 6, 7, 1, 2

    2017

  36. [44]

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

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. In NeurIPS, 2017. 2, 3, 4

    2017

  37. [45]

    Contrast with reconstruct: Contrastive 3d representation learning guided by generative pretraining

    Zekun Qi, Runpei Dong, Guofan Fan, Zheng Ge, Xiangyu Zhang, Kaisheng Ma, and Li Yi. Contrast with reconstruct: Contrastive 3d representation learning guided by generative pretraining. In ICML, 2023. 7

    2023

  38. [46]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In ICML, 2021. 3

    2021

  39. [47]

    Learn- ing to retrieve prompts for in-context learning

    Ohad Rubin, Jonathan Herzig, and Jonathan Berant. Learn- ing to retrieve prompts for in-context learning. In NAACL,

  40. [48]

    Gpa-net: No-reference point cloud quality assessment with multi-task graph convolu- tional network

    Ziyu Shan, Qi Yang, Rui Ye, Yujie Zhang, Yiling Xu, Xi- aozhong Xu, and Shan Liu. Gpa-net: No-reference point cloud quality assessment with multi-task graph convolu- tional network. TVCG, 2023. 1

    2023

  41. [49]

    Active learning for point cloud semantic segmentation via spatial-structural diversity reasoning

    Feifei Shao, Yawei Luo, Ping Liu, Jie Chen, Yi Yang, Yulei Lu, and Jun Xiao. Active learning for point cloud semantic segmentation via spatial-structural diversity reasoning. In ACM MM, 2022. 1, 4

    2022

  42. [50]

    Minigpt-3d: Efficiently aligning 3d point clouds with large language models using 2d priors

    Yuan Tang, Xu Han, Xianzhi Li, Qiao Yu, Yixue Hao, Long Hu, and Min Chen. Minigpt-3d: Efficiently aligning 3d point clouds with large language models using 2d priors. In ACM MM, 2024. 3

    2024

  43. [51]

    Kpconv: Flexible and deformable convolution for point clouds

    Hugues Thomas, Charles R Qi, Jean-Emmanuel Deschaud, Beatriz Marcotegui, Franc ¸ois Goulette, and Leonidas J Guibas. Kpconv: Flexible and deformable convolution for point clouds. In ICCV, 2019. 2, 3

    2019

  44. [52]

    Llama: Open and efficient foundation language models

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timoth´ee Lacroix, Baptiste Rozi`ere, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023. 3

    2023 arXiv

  45. [53]

    Assessing normalization techniques for simple ad- ditive weighting method

    Nazanin Vafaei, Rita A Ribeiro, and Luis M Camarinha- Matos. Assessing normalization techniques for simple ad- ditive weighting method. Procedia Computer Science, 199: 1229–1236, 2022. 6

    2022

  46. [54]

    Images speak in images: A generalist painter for in-context visual learning

    Xinlong Wang, Wen Wang, Yue Cao, Chunhua Shen, and Tiejun Huang. Images speak in images: A generalist painter for in-context visual learning. In CVPR, 2023. 1

    2023

  47. [55]

    Seggpt: Segmenting ev- erything in context

    Xinlong Wang, Xiaosong Zhang, Yue Cao, Wen Wang, Chunhua Shen, and Tiejun Huang. Seggpt: Segmenting ev- erything in context. In ICCV, 2023. 1

    2023

  48. [56]

    Dynamic graph cnn for learning on point clouds

    Yue Wang, Yongbin Sun, Ziwei Liu, Sanjay E Sarma, Michael M Bronstein, and Justin M Solomon. Dynamic graph cnn for learning on point clouds. ToG, 2019. 7, 1, 2

    2019

  49. [57]

    Pathnet: Path-selective point cloud denoising

    Zeyong Wei, Honghua Chen, Liangliang Nan, Jun Wang, Jing Qin, and Mingqiang Wei. Pathnet: Path-selective point cloud denoising. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 1

    2024

  50. [58]

    Learnable skeleton-aware 3d point cloud sampling

    Cheng Wen, Baosheng Yu, and Dacheng Tao. Learnable skeleton-aware 3d point cloud sampling. In CVPR, 2023. 2, 3, 5, 1

    2023

  51. [59]

    Attention-based point cloud edge sampling

    Chengzhi Wu, Junwei Zheng, Julius Pfrommer, and J ¨urgen Beyerer. Attention-based point cloud edge sampling. In CVPR, 2023. 3

    2023

  52. [60]

    Indexsample: A learnable sampling network in point cloud classification

    Zhenyu Wu, Kun Li, Yuhu Wu, Xin Zhang, and Shengming Li. Indexsample: A learnable sampling network in point cloud classification. In SICE, 2021. 3

    2021

  53. [61]

    Pointllm: Empowering large lan- guage models to understand point clouds

    Runsen Xu, Xiaolong Wang, Tai Wang, Yilun Chen, Jiang- miao Pang, and Dahua Lin. Pointllm: Empowering large lan- guage models to understand point clouds. In ECCV, 2025. 3

    2025

  54. [62]

    In-context learning with retrieved demonstrations for lan- guage models: A survey

    Xin Xu, Yue Liu, Panupong Pasupat, Mehran Kazemi, et al. In-context learning with retrieved demonstrations for lan- guage models: A survey. arXiv, 2024. 1, 3, 4, 6 10

    2024

  55. [63]

    Pointasnl: Robust point clouds processing using nonlocal neural networks with adaptive sampling

    Xu Yan, Chaoda Zheng, Zhen Li, Sheng Wang, and Shuguang Cui. Pointasnl: Robust point clouds processing using nonlocal neural networks with adaptive sampling. In CVPR, 2020. 3, 1

    2020

  56. [64]

    Teaser: Fast and certifiable point cloud registration

    Heng Yang, Jingnan Shi, and Luca Carlone. Teaser: Fast and certifiable point cloud registration. IEEE Transactions on Robotics, 2020. 1, 4

    2020

  57. [65]

    Modeling point clouds with self-attention and gumbel subset sampling

    Jiancheng Yang, Qiang Zhang, Bingbing Ni, Linguo Li, Jinxian Liu, Mengdie Zhou, and Qi Tian. Modeling point clouds with self-attention and gumbel subset sampling. In CVPR, 2019. 3, 4, 5

    2019

  58. [66]

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

    Li Yi, Vladimir G Kim, Duygu Ceylan, I-Chao Shen, Mengyan Yan, Hao Su, Cewu Lu, Qixing Huang, Alla Shef- fer, and Leonidas Guibas. A scalable active framework for region annotation in 3d shape collections. ToG, 2016. 2, 6

    2016

  59. [67]

    Generate rather than retrieve: Large language models are strong context generators

    Wenhao Yu, Dan Iter, Shuohang Wang, Yichong Xu, Mingx- uan Ju, Soumya Sanyal, Chenguang Zhu, Michael Zeng, and Meng Jiang. Generate rather than retrieve: Large language models are strong context generators. In ICLR, 2023. 3

    2023

  60. [68]

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

    Xumin Yu, Lulu Tang, Yongming Rao, Tiejun Huang, Jie Zhou, and Jiwen Lu. Point-bert: Pre-training 3d point cloud transformers with masked point modeling. In CVPR, 2022. 7

    2022

  61. [69]

    Clip2: Contrastive language- image-point pretraining from real-world point cloud data

    Yihan Zeng, Chenhan Jiang, Jiageng Mao, Jianhua Han, Chaoqiang Ye, Qingqiu Huang, Dit-Yan Yeung, Zhen Yang, Xiaodan Liang, and Hang Xu. Clip2: Contrastive language- image-point pretraining from real-world point cloud data. In CVPR, 2023. 3

    2023

  62. [70]

    Pointclip: Point cloud understanding by clip

    Renrui Zhang, Ziyu Guo, Wei Zhang, Kunchang Li, Xu- peng Miao, Bin Cui, Yu Qiao, Peng Gao, and Hongsheng Li. Pointclip: Point cloud understanding by clip. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8552–8562, 2022. 3

    2022

  63. [71]

    Learning 3d representations from 2d pre-trained models via image-to-point masked autoencoders

    Renrui Zhang, Liuhui Wang, Yu Qiao, Peng Gao, and Hong- sheng Li. Learning 3d representations from 2d pre-trained models via image-to-point masked autoencoders. In CVPR,

  64. [72]

    A survey on multi-task learning

    Yu Zhang and Qiang Yang. A survey on multi-task learning. TKDE, 2021. 1

    2021

  65. [73]

    Svc: Sight view constraint for robust point cloud registration

    Yaojie Zhang, Weijun Wang, Tianlun Huang, Zhiyong Wang, and Wei Feng. Svc: Sight view constraint for robust point cloud registration. Image and Vision Computing , page 105315, 2024. 1

    2024

  66. [74]

    Automatic chain of thought prompting in large language models

    Zhuosheng Zhang, Aston Zhang, Mu Li, and Alex Smola. Automatic chain of thought prompting in large language models. In ICLR, 2023. 3

    2023

  67. [75]

    Robust multi-task learning network for com- plex lidar point cloud data preprocessing

    Luda Zhao, Yihua Hu, Xing Yang, Zhenglei Dou, and Lin- shuang Kang. Robust multi-task learning network for com- plex lidar point cloud data preprocessing. Expert Systems with Applications, 2024. 1 11 MICAS: Multi-grained In-Context Adaptive Sampling for 3D Point Cloud Processi...

    2024

  68. [76]

    Ablation Study: Robustness Analysis In the task-adaptive point sampling module and query- specific prompt sampling module of our proposed MICAS, we design the task encoder, point encoder, and prompt sampling module based on PointNet [43]. To evaluate the robustness of MICAS, w...

  69. [77]

    As shown in Figures A1 and A2, our proposed MICAS consistently selects higher-quality central points, delivering superior outcomes and overcom- ing the limitations of FPS

    More Qualitative Analysis To demonstrate the effectiveness of our proposed MICAS in central point sampling and prediction, we present a vi- sual comparison between our task-adaptive point sampling method and Farthest Point Sampling (FPS) used in PIC- Cat [11] and PIC-Sep [11]....

  70. [78]

    best-performing

    Discussion 8.1. Limitations While our proposed MICAS represents a pioneering effort to address inter-task and intra-task sensitivity challenges in point cloud in-context learning, it has a limitation. Specif- ically, in the query-specific prompt sampling, we prioritize selecti...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.