Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Missing sensors break the fusion interface, not just the data; filling every modality slot with a proxy restores it.

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 · grok-4.5

2026-07-13 14:01 UTC pith:QHNEYCHU

load-bearing objection Solid empirical methods paper: interface-complete fusion via directed proxies is a useful framing and the HAR results are broad and carefully controlled, with the main limits already disclosed. the 3 major comments →

arxiv 2604.02056 v2 pith:QHNEYCHU submitted 2026-04-02 cs.CV

COMPASS: Complete Multimodal Fusion via Proxy Tokens and Shared Spaces for Ubiquitous Sensing

Hao Wang , Yanyu Qian , Pengcheng Weng , Zixuan Xia , William Dan , Yangxin Xu , Fei Wang This is my paper
classification cs.CV
keywords multimodal sensinghuman activity recognitionmissing modalitiesfusion-interface completionproxy tokensshared latent spaceubiquitous sensing
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.

Multimodal sensing systems for human activity recognition train a fusion head on a fixed set of modality slots, but at deployment sensors drop out and that interface changes. Compass treats this as a structural problem: each modality keeps a permanent slot; observed sensors put real tokens in their slots, and missing ones get proxy tokens estimated from the sensors that remain. Source-specific proxies for the same missing slot are averaged into one filler so the same simple fusion operator always sees a complete N-slot input. Training simulates dropouts, forces proxies to stay compatible with real slots, and stabilizes the shared representation space so the fillers stay useful for recognition. On three HAR benchmarks the approach improves accuracy under single- and multi-missing patterns, especially when strong modalities are absent, and beats controlled imputation, distillation, and translation baselines. The claim is that restoring the fusion interface is a simple, portable principle for robust multimodal sensing.

Core claim

Missing modalities create a fusion-interface mismatch as well as information loss. Completing every canonical modality slot—with a real token when the sensor is present and with an aggregated source-to-target proxy token when it is not—lets one lightweight fusion head operate unchanged under arbitrary missingness and yields stronger HAR robustness than branch-skipping, feature imputation, distillation, or translation-style proxies.

What carries the argument

Interface-complete fusion via directed source-to-target proxy generators: each missing slot is filled by averaging single-layer Transformer estimates from every observed source into one proxy token, so fusion always receives a fixed N-slot layout.

Load-bearing premise

A single global token per modality, completed by simple averaging of directed maps and fused by plain summation, is enough to restore the interactions the fusion head needs.

What would settle it

On a HAR benchmark where a dominant modality already carries almost all the signal, or on a dense spatial task such as pose estimation, a cross-attention or sequence-level fusion baseline would outperform the completed-slot sum fusion under the same missingness patterns.

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

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

3 major / 6 minor

Summary. COMPASS frames missing-modality multimodal sensing as a fusion-interface mismatch problem and proposes restoring a fixed N-slot input before prediction. Observed modalities fill their slots with real projected tokens; missing slots are filled by directed source-to-target proxy generators in a shared latent space, with multi-source estimates mean-aggregated into one filler. The fusion head is a parameter-free sum over the completed slots. Training uses synthetic modality masking, slot-compatibility L2 supervision against held-out real slots, VICReg-style space stabilization, and an auxiliary per-proxy task loss. On XRF55, MM-Fi, and OctoNet HAR, a single model is evaluated across all non-empty modality subsets, with multi-seed means, ablations of proxies/masking/losses, and controlled re-implementations of SMIL-style imputation, PTA distillation, and CMPT-style translation under matched fine-tuning. Gains are largest in low-modality and strong-modality-absent regimes; dominant-modality cases where cross-attention (X-Fi) remains competitive are disclosed.

Significance. If the results hold, the paper offers a clean and practical design principle for ubiquitous multimodal sensing: keep the fusion interface fixed and push missingness handling into pre-fusion slot completion, rather than subset-dependent fusion or full reconstruction. The contribution is well scoped to HAR classification with global tokens, and the empirical package is stronger than typical missing-modality sensing papers: full-subset tables on three public benchmarks, multi-seed statistics, matched controlled baselines (Table 5), and ablations that separate zero-fill, proxy-only, and full objectives (Tables 6–7). The shared-generator scalability note and corrected OctoNet ToF comparison are also scientifically careful. Within HAR, this is a solid systems/methods contribution rather than a foundational theoretical advance, but the interface-completeness framing is useful and falsifiable.

major comments (3)
  1. §3.3–3.4 and Table 6: the central claim attributes robustness primarily to restoring a canonical N-slot interface, yet the largest single ablation drop on XRF55 is removing space stabilization (−2.3pp), while zero-fill already beats X-Fi’s 7-scenario average (76.9 vs 72.1). Please add a short attribution analysis that separates (i) fixed-layout fusion, (ii) proxy generation, and (iii) shared-space/training objectives—e.g., fixed-layout with zeros vs fixed-layout with proxies under identical losses—so readers can judge how much of the gain is truly “interface completeness” versus representation geometry and supervision.
  2. §4.2 / Table 2 (MM-Fi): X-Fi remains better on R-only, I+R, and D+R. The paper notes this as a dominant-modality boundary, but the main narrative still presents COMPASS as broadly superior. Please quantify and discuss this trade-off more systematically (e.g., when the strongest observed modality exceeds a unimodal accuracy threshold, does sum fusion + proxies underperform expressive cross-attention?), and state the operating regime of the method more precisely in the abstract/conclusion rather than only in §4.3.
  3. §3.4 training protocol: all supervision for slot completion assumes fully observed training samples with synthetic masking. This is standard, but the paper’s deployment motivation includes sensors that may also be missing in training data. Either evaluate a realistic incomplete-training setting (random missingness in the training set without real-slot targets for some samples) or explicitly limit the claim to “complete training, incomplete inference,” which is currently understated relative to the introduction’s broader framing.
minor comments (6)
  1. Throughout the manuscript PDF/source, many figure captions and section headings appear as garbled replacement characters (e.g., “�������”). Ensure the camera-ready text renders COMPASS and all figure labels correctly.
  2. §3.2, Eq. (13): uniform averaging is default; the text says learned weighting “does not consistently improve,” but no table is shown. A one-row ablation (mean vs attention/confidence weighting) would make this claim checkable.
  3. Table 8 and the shared-generator paragraph in §4.3: the shared O(1) generator result is important for the scalability claim but is only briefly reported. Consider promoting numbers (params, OctoNet/XRF55 averages) into a small table.
  4. Figure 3: the compactness–separation scatter is informative; please define the exact formulas for “cross-modal compactness” and “inter-class separation” in the caption or appendix so the plot is reproducible.
  5. Notation: V, O, M, and slot index conventions are clear, but ¯u_i vs ˜u_i vs u_{i→j} could be summarized in a short symbol table for readers skimming §3.
  6. Related work §2.1–2.2 is generally fair; a brief explicit contrast with prompt-based missing-modality methods (beyond CMPT) on whether they preserve a fixed fusion layout would sharpen positioning.

Circularity Check

0 steps flagged

No significant circularity: empirical HAR method with trained proxies evaluated on held-out public benchmarks; no prediction reduces by construction to its inputs.

full rationale

COMPASS is an empirical multimodal sensing method paper. The load-bearing chain is: (i) define fixed N-slot interface completion via directed source-to-target proxy generators and mean aggregation (Eqs. 5–16); (ii) train with synthetic masking plus task, slot-compatibility, VICReg-style space, and per-proxy losses (Eqs. 21–31) using real masked-slot representations as supervision targets; (iii) report top-1 accuracy on XRF55, MM-Fi, and OctoNet under held-out missingness patterns against independent and controlled baselines. None of these steps equates a claimed prediction to a fitted constant or definitional identity: slot-compatibility L2 (Eq. 22) is a training regularizer, not a test-time identity that forces accuracy; fusion is parameter-free sum of completed slots, not a fit renamed as prediction. Self-citations (XRF55 dataset [25], PTA baseline [27], geometry-related work [29], X-Fi protocol/backbones [3]) supply data, baselines, or related context; they do not import a uniqueness theorem that forbids alternatives or force the reported gains. Ablations (Tables 6–7), multi-seed means, and disclosed boundary cases (dominant-modality settings where X-Fi wins) further show the central claim is externally falsifiable rather than circular. Score 0 is appropriate.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 2 invented entities

The central claim rests on standard deep-learning practice plus a small set of design choices (fixed slots, directed generators, mean aggregation, sum fusion, synthetic masking, and three auxiliary losses with hand-chosen weights). No new physical entities; the 'proxy token' is an architectural construct whose value is measured only by downstream accuracy on the three HAR benchmarks.

free parameters (4)
  • p_drop (synthetic missingness probability) = 0.7
    Default 0.7 chosen after a small grid; affects how often completion maps are trained.
  • loss weights λ1, λ2, λ3 = 0.2, 0.3, 0.5
    Balance slot-compatibility, space stabilization, and per-proxy task losses; set to 0.2, 0.3, 0.5 without extensive search reported.
  • VICReg coefficients (μ_inv, μ_var, μ_cov) = (5, 25, 1)
    Hand-set to (5, 25, 1) for representation-space stabilization.
  • shared token length L and dimension d = L=32, d=512
    Architectural choices L=32, d=512 that define the slot space.
axioms (4)
  • ad hoc to paper A fusion head trained on a fixed canonical N-slot layout will generalize better if that layout is restored at inference than if the head is adapted to every observed subset.
    Core design hypothesis stated in Introduction and Section 3.1; tested empirically rather than derived.
  • domain assumption Mean pooling of projected encoder tokens yields a sufficient real slot representation for HAR classification.
    Section 3.2; common in global-token multimodal HAR but not necessary for all sensing tasks.
  • ad hoc to paper Uniform averaging of directed source-to-target estimates is an adequate aggregator for multi-source completion.
    Section 3.2; ablations claim learned weighting does not consistently help, but this remains a modeling choice.
  • domain assumption Synthetic random masking of fully observed training samples approximates real missingness patterns at test time.
    Section 3.4; standard missing-modality training assumption.
invented entities (2)
  • Target-slot proxy token (completed slot filler) no independent evidence
    purpose: Populate a missing canonical fusion slot with a task-relevant, fusion-compatible representation inferred from observed modalities.
    Architectural construct optimized for slot compatibility and recognition, not full modality reconstruction; independent evidence is only the reported accuracy gains.
  • Interface-complete fusion principle no independent evidence
    purpose: Frame missing-modality robustness as restoration of a fixed N-slot input layout rather than subset adaptation or pure imputation.
    Conceptual framing that organizes the method; falsifiable only via comparative experiments on the same benchmarks.

pith-pipeline@v1.1.0-grok45 · 21284 in / 3030 out tokens · 25241 ms · 2026-07-13T14:01:16.541832+00:00 · methodology

0 comments
read the original abstract

Missing modalities in multimodal sensing cause not only information loss but also a fusion-interface mismatch: a fusion head trained on a canonical set of modality slots must operate on changing observed subsets at inference time. We propose Compass, an interface-complete fusion framework that restores this canonical slot structure before prediction. Each modality is assigned a fixed fusion slot. Observed modalities populate their slots with real representations, while absent modalities are filled with target-slot completion representations estimated from the observed sources. Multiple source-specific estimates for the same missing slot are aggregated into a single slot filler, allowing the same lightweight fusion operator to be applied under arbitrary missing-modality patterns. Training uses synthetic modality masking, slot-compatibility supervision, and representation-space stabilization to make completed slots compatible with real modality representations and useful for downstream recognition. Across XRF55, MM-Fi, and OctoNet, Compass improves robustness under diverse single- and multiple-missing settings, including controlled comparisons against imputation, distillation, and translation-style baselines. These results suggest that preserving the fusion interface is a simple and effective principle for robust multimodal sensing.

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Diverse via bounded Agreement: Geometric Regularization for Multimodal Fusion

    cs.CV 2026-01 unverdicted novelty 5.0

    A regularization method enforces diverse intra-modal embeddings and bounded inter-modal drift to improve both multimodal fusion and unimodal robustness.

Reference graph

Works this paper leans on

41 extracted references · 4 canonical work pages · cited by 1 Pith paper

  1. [1]

    Adrien Bardes, Jean Ponce, and Yann LeCun. 2022. VICReg: Variance-Invariance- Covariance Regularization for Self-Supervised Learning. InThe Tenth Interna- tional Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenReview.net. https://openreview.net/forum?id=xm6YD62D1Ub

    2022

  2. [2]

    Xinyan Chen, He Huang, Chris Xiaoxuan Lu, Lihua Xie, Yuecong Xu, Jianfei Yang, Shenghai Yuan, Yunjiao Zhou, and Han Zou. 2023. MM-Fi: Multi-Modal Non-Intrusive 4D Human Dataset for Versatile Wireless Sensing. InNeurIPS (NeurIPS 2023). 18756–18768. doi:10.52202/075280-0822

    work page doi:10.52202/075280-0822 2023

  3. [3]

    Xinyan Chen and Jianfei Yang. 2025. X-Fi: A Modality-Invariant Foundation Model for Multimodal Human Sensing. InThe Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 24-28, 2025. OpenRe- view.net. https://openreview.net/forum?id=b42wmsdwmB

    2025

  4. [4]

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xi- aohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In9th Interna- tional Conference on Learning Representations, IC...

    2021

  5. [5]

    Rohit Girdhar, Alaaeldin El-Nouby, Zhuang Liu, Mannat Singh, Kalyan Vasudev Alwala, Armand Joulin, and Ishan Misra. 2023. ImageBind One Embedding Space to Bind Them All. In2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 15180–15190. doi:10.1109/cvpr52729.2023.01457

    doi:10.1109/cvpr52729.2023.01457 2023

  6. [6]

    Zirun Guo, Tao Jin, and Zhou Zhao. 2024. Multimodal prompt learning with missing modalities for sentiment analysis and emotion recognition. InProceedings of the 62nd annual meeting of the association for computational linguistics (volume 1: Long papers). 1726–1736

    2024

  7. [7]

    Devamanyu Hazarika, Roger Zimmermann, and Soujanya Poria. 2020. Misa: Modality-invariant and-specific representations for multimodal sentiment analy- sis. InProceedings of the 28th ACM international conference on multimedia. 1122– 1131

    2020

  8. [8]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep Residual Learning for Image Recognition. In2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 770–778. doi:10.1109/cvpr.2016.90

    doi:10.1109/cvpr.2016.90 2016

  9. [9]

    Hinton, Oriol Vinyals, and Jeffrey Dean

    Geoffrey E. Hinton, Oriol Vinyals, and Jeffrey Dean. 2015. Distilling the Knowl- edge in a Neural Network.CoRRabs/1503.02531 (2015). arXiv:1503.02531 http://arxiv.org/abs/1503.02531

    Pith/arXiv arXiv 2015

  10. [10]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. LoRA: Low-Rank Adaptation of Large Language Models. InThe Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenReview.net. https://openreview.net/forum?id=nZeVKeeFYf9

    2022

  11. [11]

    Jaehyuk Jang, Yooseung Wang, and Changick Kim. 2024. Towards robust multi- modal prompting with missing modalities. InICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 8070–8074

    2024

  12. [12]

    2022.Visual Prompt Tuning

    Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. 2022.Visual Prompt Tuning. Springer Nature Switzerland, 709–727. doi:10.1007/978-3-031-19827-4_41

    doi:10.1007/978-3-031-19827-4_41 2022

  13. [13]

    Yi-Lun Lee, Yi-Hsuan Tsai, Wei-Chen Chiu, and Chen-Yu Lee. 2023. Multimodal prompting with missing modalities for visual recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 14943–14952

    2023

  14. [14]

    Jinghui Liu, Daniel Capurro, Anthony Nguyen, and Karin Verspoor. 2023. Attention-based multimodal fusion with contrast for robust clinical prediction in the face of missing modalities.Journal of Biomedical Informatics145 (Sept. 2023), 104466. doi:10.1016/j.jbi.2023.104466

    doi:10.1016/j.jbi.2023.104466 2023

  15. [15]

    Ilya Loshchilov and Frank Hutter. 2019. Decoupled Weight Decay Regularization. In7th International Conference on Learning Representations, ICLR 2019, New Or- leans, LA, USA, May 6-9, 2019. OpenReview.net. https://openreview.net/forum? id=Bkg6RiCqY7

    2019

  16. [16]

    Mengmeng Ma, Jian Ren, Long Zhao, Davide Testuggine, and Xi Peng. 2022. Are Multimodal Transformers Robust to Missing Modality?. In2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 18156–18165. doi:10.1109/cvpr52688.2022.01764

    doi:10.1109/cvpr52688.2022.01764 2022

  17. [17]

    Mengmeng Ma, Jian Ren, Long Zhao, Sergey Tulyakov, Cathy Wu, and Xi Peng

  18. [18]

    doi:10.1609/aaai.v35i3.16330

    SMIL: Multimodal Learning with Severely Missing Modality.Proceedings of the AAAI Conference on Artificial Intelligence35, 3 (May 2021), 2302–2310. doi:10.1609/aaai.v35i3.16330

    doi:10.1609/aaai.v35i3.16330 2021

  19. [19]

    Hai Pham, Paul Pu Liang, Thomas Manzini, Louis-Philippe Morency, and Barn- abás Póczos. 2019. Found in Translation: Learning Robust Joint Representations by Cyclic Translations between Modalities. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 33. 6892–6899

    2019

  20. [20]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning Transferable Visual Models From Natural Language Supervision. InICML (Proceedings of Machine Learning Research), Marina Meila and Tong Zhang (Eds.). PMLR, 8...

    2021

  21. [21]

    Salman Asif

    Md Kaykobad Reza, Ameya Patil, Mashhour Solh, and M. Salman Asif. 2025. Robust Multimodal Learning via Cross-Modal Proxy Tokens.Trans. Mach. Learn. Res.2025 (2025). https://openreview.net/forum?id=Wtc6wvcYJ0

    2025

  22. [22]

    Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. 2016. Rethinking the Inception Architecture for Computer Vision. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2818–2826. doi:10.1109/cvpr.2016.308

    doi:10.1109/cvpr.2016.308 2016

  23. [23]

    Luan Tran, Xiaoming Liu, Jiayu Zhou, and Rong Jin. 2017. Missing Modali- ties Imputation via Cascaded Residual Autoencoder. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 4971–4980

    2017

  24. [24]

    Yao-Hung Hubert Tsai, Paul Pu Liang, Amir Zadeh, Louis-Philippe Morency, and Ruslan Salakhutdinov. 2018. Learning factorized multimodal representations. arXiv preprint arXiv:1806.06176(2018)

    Pith/arXiv arXiv 2018

  25. [25]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is All you Need. InAdvances in Neural Information Processing Systems 30: An- nual Conference on Neural Information Processing Systems 2017, December 4- 9, 2017, Long Beach, CA, USA, Isabelle Guyon, Ulrike von Luxbur...

    2017

  26. [26]

    Fei Wang, Yizhe Lv, Mengdie Zhu, Han Ding, and Jinsong Han. 2024. XRF55: A Radio Frequency Dataset for Human Indoor Action Analysis.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies8, 1 (March 2024), 1–34. doi:10.1145/3643543

    doi:10.1145/3643543 2024

  27. [27]

    Hu Wang, Yuanhong Chen, Congbo Ma, Jodie Avery, Louise Hull, and Gustavo Carneiro. 2023. Multi-Modal Learning with Missing Modality via Shared-Specific Feature Modelling. In2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 15878–15887. doi:10.1109/cvpr52729.2023.01524

    doi:10.1109/cvpr52729.2023.01524 2023

  28. [28]

    Pengcheng Weng, Yanyu Qian, Yangxin Xu, and Fei Wang. 2026. Purify-then- Align: Towards Robust Human Sensing under Modality Missing with Knowledge Distillation from Noisy Multimodal Teacher. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 7634–7643

    2026

  29. [29]

    Renjie Wu, Hu Wang, and Hsiang-Ting Chen. 2024. A Comprehensive Survey on Deep Multimodal Learning with Missing Modality.CoRRabs/2409.07825 (2024). arXiv:2409.07825 doi:10.48550/ARXIV.2409.07825

    Pith/arXiv arXiv doi:10.48550/arxiv.2409.07825 2024

  30. [30]

    Zixuan Xia, Hao Wang, Pengcheng Weng, Yanyu Qian, Yangxin Xu, William Dan, and Fei Wang. 2026. Diverse via bounded Agreement: Geometric Regularization for Multimodal Fusion. arXiv:2601.21670 [cs.CV] https://arxiv.org/abs/2601.21670

    Pith/arXiv arXiv 2026

  31. [31]

    Jamal Deen, and Jiannong Cao

    Jiang Xiao, Huichuwu Li, Minrui Wu, Hai Jin, M. Jamal Deen, and Jiannong Cao

  32. [32]

    Surveys55, 5 (Dec

    A Survey on Wireless Device-free Human Sensing: Application Scenarios, Current Solutions, and Open Issues.Comput. Surveys55, 5 (Dec. 2022), 1–35. doi:10.1145/3530682

    work page doi:10.1145/3530682 2022

  33. [33]

    Jianfei Yang, Xinyan Chen, Han Zou, Chris Xiaoxuan Lu, Dazhuo Wang, Sumei Sun, and Lihua Xie. 2023. SenseFi: A library and benchmark on deep-learning- empowered WiFi human sensing.Patterns4, 3 (2023)

    2023

  34. [34]

    Dongsheng Yuan, Xie Zhang, Weiying Hou, Sheng Lyu, Yuemin Yu, Luca Jiang- Tao Yu, Chengxiao Li, and Chenshu Wu. 2025. OctoNet: A Large-Scale Multi- Modal Dataset for Human Activity Understanding Grounded in Motion-Captured 3D Pose Labels. InAdvances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 202...

    2025

  35. [35]

    Jure Zbontar, Li Jing, Ishan Misra, Yann LeCun, and Stéphane Deny. 2021. Barlow Twins: Self-Supervised Learning via Redundancy Reduction. InICML (Proceedings of Machine Learning Research), Marina Meila and Tong Zhang (Eds.). PMLR, 12310–12320. http://proceedings.mlr.press/v139/zbontar21a.html

    2021

  36. [36]

    Kevin Zhou

    Yue Zhang, Chengtao Peng, Qiuli Wang, Dan Song, Kaiyan Li, and S. Kevin Zhou

  37. [37]

    IEEE Transactions on Medical Imaging44, 1 (Jan

    Unified Multi-Modal Image Synthesis for Missing Modality Imputation. IEEE Transactions on Medical Imaging44, 1 (Jan. 2025), 4–18. doi:10.1109/tmi. 2024.3424785 �������� �������� ���������� ������ ��� ����� ������ ��� ������ ������ ��� ���������� ������� �������������� ���� ����� ����������� ��� ���

    doi:10.1109/tmi 2025

  38. [38]

    Fei Zhao, Chengcui Zhang, and Baocheng Geng. 2024. Deep Multimodal Data Fusion.Comput. Surveys56, 9 (April 2024), 1–36. doi:10.1145/3649447

    work page doi:10.1145/3649447 2024

  39. [39]

    Hengshuang Zhao, Li Jiang, Jiaya Jia, Philip Torr, and Vladlen Koltun. 2021. Point Transformer. In2021 IEEE/CVF International Conference on Computer Vision (ICCV). IEEE, 16239–16248. doi:10.1109/iccv48922.2021.01595

    doi:10.1109/iccv48922.2021.01595 2021

  40. [40]

    Jinming Zhao, Ruichen Li, and Qin Jin. 2021. Missing Modality Imagination Net- work for Emotion Recognition with Uncertain Missing Modalities. InProceedings 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). 2608–2618. doi:10.18653/...

    work page doi:10.18653/v1/2021.acl-long.203 2021

  41. [41]

    Ce Zheng, Wenhan Wu, Chen Chen, Taojiannan Yang, Sijie Zhu, Ju Shen, Nasser Kehtarnavaz, and Mubarak Shah. 2023. Deep Learning-based Human Pose Esti- mation: A Survey.Comput. Surveys56, 1 (Aug. 2023), 1–37. doi:10.1145/3603618

    doi:10.1145/3603618 2023