Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

Talk is Not Always Cheap: Promoting Wireless Sensing Models with Text Prompts

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read WiTalk claims that a text branch with LLM-generated hierarchical prompts improves wireless sensing accuracy and temporal action localization with no architectural changes or new data collection.

desk verdict Useful idea, but the test-time text retrieval is never specified, so the headline gains may be label leakage; needs clarification before the results can be trusted. read the letter →

arxiv 2504.14621 v2 pith:EJT7HPTI submitted 2025-04-20 cs.CV

Zhenkui Yang , Zeyi Huang , Ge Wang , Han Ding , Tony Xiao Han , Fei Wang This is my paper · ORCID
classification cs.CV
keywords wirelesssensinghumanactionrecognitiontemporallocalizationtextpromptslargelanguagemodelsmultimodalfusionchannelstateinformation
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

WiTalk seeks to establish that the action labels already sitting in wireless sensing datasets can be converted into a free accuracy boost. The paper combines large-language-model-generated text prompts—label-only, brief context, and detailed descriptions—with an existing wireless sensing model by adding a lightweight text branch and weighted feature fusion. On XRF55, adding the text branch raises accuracy by 3.9% for Wi-Fi, 2.59% for RFID, and 0.46% for mmWave; on WiFiTAL, average mAP rises 4.98%; on XRFV2, average mAP gains range from 4.02% to 13.68%. If true, this gives a no-new-data, no-architecture-change upgrade path for Wi-Fi, RFID, and mmWave activity recognition and localization.

What carries the argument

The load-bearing mechanism is the text branch: a contextual template plus LLM-generated hierarchical prompts (label-only, brief context, and detailed description), a text encoder producing fixed embeddings cached in JSON, a multi-head self-attention layer, and weighted fusion of text features with wireless features at a 0.1 text weight. This branch carries the argument because it is the only change between the W and W+T runs; all sensing encoders, losses, and hyperparameters stay fixed.

What would settle it

Run the full WiTalk pipeline but at inference replace the ground-truth label's text embedding with one derived from the model's own predicted label or from a fixed neutral description; if the W+T gains largely disappear, label leakage explains them.

Watch

Extended reading notes

Core claim

We claim that the latent text already present in wireless sensing datasets—the action labels—can be turned into a cheap performance booster. WiTalk enriches each label with a large language model under three prompt strategies of increasing semantic richness, encodes the enriched text with a frozen text encoder, refines it with multi-head self-attention, and fuses it with the wireless feature vector at a text weight of 0.1. Across XRF55, WiFiTAL, and XRFV2, this addition raises accuracy or mAP by the reported margins without changing the sensing model's architecture or collecting any new signal data.

Load-bearing premise

At test time, the text feature for each sample must be obtained without peeking at the sample's ground-truth action label, but the paper does not specify how that retrieval is done.

Editorial extensions

If this is right

  • Any existing wireless sensing model can be upgraded with text semantics by adding the text branch, with no architectural changes.
  • The improvement transfers across modalities (Wi-Fi, RFID, mmWave) and across tasks (human action recognition and temporal action localization), so the effect is not specific to one signal type.
  • Because labels already exist in every dataset, the method promises a zero-data-collection accuracy gain for deployed systems.
  • Text encoders trained only on images and text—not on wireless signals—still provide useful semantic guidance for wireless features.

Reading between the lines

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

  • If the text feature at inference is chosen using the model's own predicted label rather than the ground-truth class name, the reported gains may shrink; a class-agnostic retrieval test would separate genuine semantic regularisation from label leakage.
  • The prompt strategy that works best appears to depend on modality: RFID benefits more from detailed descriptions, while Wi-Fi prefers the brief-context prompt, suggesting prompt design should be tuned per signal type.
  • The same branch could be applied to other RF tasks such as gesture recognition or pose estimation, where text labels are equally available and the fusion cost would be the same.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes WiTalk, a text-enhanced wireless sensing framework that augments CSI, mmWave, and RFID based action recognition and temporal action localization models with text embeddings derived from LLM-generated prompts. The method encodes action labels using hierarchical prompts (label-only, brief description, detailed description), embeds them with CLIP (or other text encoders), refines them with multi-head self-attention, and fuses the text features with wireless signal features at a fixed weight. The authors report accuracy gains on XRF55 (WiFi +3.9%, RFID +2.59%, mmWave +0.46%), average mAP gains on WiFiTAL (+4.98%), and average mAP gains on XRFV2 (+4.02% to +13.68%) across multiple TAL backbones, and include an ablation over text encoders and prompt strategies. The central empirical claim is that adding a text branch improves wireless sensing performance without architectural changes or extra data collection.

Significance. If validated, the idea that semantic text prompts can cheaply boost wireless sensing models is practically attractive, and the paper's scope across three datasets, multiple wireless modalities, and several TAL backbones is appropriate. The authors also provide a code repository and include ablations over prompt strategies and encoders. However, the paper's empirical evidence is currently compromised by the absence of a specified inference-time retrieval protocol, the post-hoc selection of the best configuration on the test set, and the lack of variance estimates. These issues affect every headline result, so the significance of the claimed gains cannot be assessed from the manuscript as written.

major comments (4)
  1. [§4.1, §4.2.3, Figure 3] The paper never specifies how a text feature is obtained for a test sample without using its ground-truth action label. Section 4.1 states that text labels are encoded once and stored as key-value pairs in JSON files for efficient retrieval, and Section 4.2.3 describes fusing the retrieved text embedding with wireless features, but the retrieval key for a test sample is not defined. If the key is the ground-truth label, the reported gains in Tables 1–3 are inflated by label leakage; if the key is a predicted label, the method is a two-stage pipeline that is not described and would require a matched baseline for the W+T comparison. Since every headline result depends on this protocol, the central claim is unsupported as written.
  2. [§5.5, Tables 4–6] The best encoder-prompt configuration (CLIP with TCE) is selected using the same test set on which the final gains are reported. Several alternative configurations perform substantially worse than baseline, e.g., Qwen TCE on XRFV2 drops average mAP from 52.61 to 9.36, and LLaMA TCE drops to 13.74. Without a validation split, a pre-registered selection rule, or multiple seeds, the reported gains are optimistic and not properly controlled.
  3. [§4.2.3] The fusion weight of 0.9 for wireless features and 0.1 for text features is introduced as a fixed choice without justification or sensitivity analysis. This weight directly controls the contribution of the text branch, and the reported improvements could depend heavily on this single hand-picked hyperparameter. The paper should present a sweep over the fusion weight or a principled tuning procedure.
  4. [§5.3, §5.4, Tables 1–3] No error bars, standard deviations, or multiple seeded runs are reported. For improvements as small as 0.46% on XRF55 mmWave and differences of a few mAP points in Tables 2 and 3, the absence of variance estimates makes it impossible to assess whether the observed gains are statistically meaningful or due to run-to-run variation.
minor comments (4)
  1. [§3, 'Principle of WiFi-Based Human Motion Sensing'] The phrase 'capturing each subcarrier’s frequency response of each subcarrier' contains a redundant repetition of 'each subcarrier'.
  2. [Table 5] The entry '84.23 81.9275.04' in the CLIP TDE row appears to be missing a space (likely '84.23 81.92 75.04').
  3. [§6] The limitation paragraph correctly notes that improvements are inconsistent across signal types and that the mechanism is unclear; this honesty is appreciated, but the limitations should also explicitly address the need to specify the inference-time retrieval protocol.
  4. [References] Several references contain incomplete or malformed bibliographic metadata (e.g., reference [8] begins with 'Student Member Ieee'), which should be cleaned up.

Circularity Check

1 steps flagged · score 8.0 of 10

The text branch is built from ground-truth motion labels that Figure 1 says are used 'for training and inference,' so the W+T gains are label injection, not semantic transfer.

  1. self definitional [Figure 1 caption; Section 4.1; Section 4.2.3]
    "WiTalk transforms raw motion labels into hierarchical semantic representations. By leveraging large language models (LLMs), these enriched annotations enhance the performance of wireless sensing models for training and inference, enabling robust action recognition through integrated contextual semantics. ... Notably, text labels are encoded once and stored as key-value pairs in JSON files for efficient retrieval."

    The text feature is a function of the ground-truth action label, and the paper states this enriched annotation is used for inference, not only training. Because the task is to predict exactly that label, fusing the retrieved label-derived text feature into the wireless feature (Section 4.2.3) gives the classifier a feature that by construction contains the test label. The reported gains on XRF55 (3.9%, 2.59%, 0.46%), WiFiTAL (4.98%), and XRFV2 (4.02% to 13.68%) are therefore the expected result of label leakage; no inference-time protocol is provided for obtaining the text feature without the true label. The claim that text prompts improve wireless sensing is not independent of the target it is supposed to predict.

full rationale

The paper's other components (LLM prompt generation, CLIP encoding, MHSA, weighted fusion) are not circular; they are standard machinery. Self-citations to XRF55, XRFV2, and WiFiTAL datasets are not load-bearing. However, the central empirical claim rests on the W+T inference protocol. The figure caption's 'for training and inference' plus the key-value JSON retrieval means the only complete specification of the text branch at test time is label-keyed lookup. This makes the comparison circular: the model receives a feature derived from the correct action label while being evaluated on predicting that same label. The additional post-hoc selection of CLIP/TCE from Tables 4-6 is a validity concern (selection on the test set), but it is not needed for the circularity finding. Overall score 8: the central claim reduces to feeding the answer into the model as a 0.1-weighted auxiliary feature.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The central improvement depends on two unevaluated choices: the fixed fusion weight and a test-time text retrieval procedure that is undocumented. No new physical entities are introduced.

free parameters (2)
  • Fusion weight alpha = 0.1 text, 0.9 wireless
    Hand-set in Section 4.2.3, never ablated; the balance between text and wireless features materially affects the fused representation.
  • Text encoder and prompt strategy = CLIP with TCE
    Selected based on best test-set performance across six encoder and strategy combinations (Tables 4-6); alternatives produce much worse results, indicating selection on the test set.
assumptions (3)
  • domain assumption Baseline models (DML, WiFiTAD, XRFMamba) are correctly re-implemented via the original code.
    Experiments rely on the original code and papers; no verification of baseline accuracy beyond reported numbers.
  • domain assumption CLIP text embeddings provide semantically useful features for wireless action labels despite no wireless pretraining.
    The method assumes frozen CLIP features transfer to RF, radar, and RFID semantics; the paper itself notes the mechanism is unclear.
  • ad hoc to paper An inference-time text feature can be computed without the ground-truth label.
    The paper never specifies how a test sample's text feature is selected (Section 4.2.3, Figure 3); the whole evaluation depends on this unstated procedure.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Talk is Not Always Cheap: Promoting Wireless Sensing Models with Text Prompts." pith.science (2026). https://pith.science/paper/EJT7HPTI

@misc{pith2026250414621,
  author       = {Pith},
  title        = {Pith review of: Talk is Not Always Cheap: Promoting Wireless Sensing Models with Text Prompts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EJT7HPTI}},
  note         = {Machine review of arXiv:2504.14621}
}
read the original abstract

Wireless signal-based human sensing technologies, such as WiFi, millimeter-wave (mmWave) radar, and Radio Frequency Identification (RFID), enable the detection and interpretation of human presence, posture, and activities, thereby providing critical support for applications in public security, healthcare, and smart environments. These technologies exhibit notable advantages due to their non-contact operation and environmental adaptability; however, existing systems often fail to leverage the textual information inherent in datasets. To address this, we propose an innovative text-enhanced wireless sensing framework, WiTalk, that seamlessly integrates semantic knowledge through three hierarchical prompt strategies-label-only, brief description, and detailed action description-without requiring architectural modifications or incurring additional data costs. We rigorously validate this framework across three public benchmark datasets: XRF55 for human action recognition (HAR), and WiFiTAL and XRFV2 for WiFi temporal action localization (TAL). Experimental results demonstrate significant performance improvements: on XRF55, accuracy for WiFi, RFID, and mmWave increases by 3.9%, 2.59%, and 0.46%, respectively; on WiFiTAL, the average performance of WiFiTAD improves by 4.98%; and on XRFV2, the mean average precision gains across various methods range from 4.02% to 13.68%. Our codes have been included in https://github.com/yangzhenkui/WiTalk.

Figures

Figures reproduced from arXiv: 2504.14621 by the authors.

Figure 1
Figure 1. WiTalk transforms raw motion labels into hier [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The figure illustrates three fundamental signal propagation and sensing mechanisms used in wireless sensing systems; [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Overview of the WiTalk. Wireless signals are pro [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Visualization examples of the performance of XRFMamba and WiFiTAD on the XRFV2 dataset, where results marked [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. FSTC-Encoder: Feature--Spatial--Temporal Correlation Learning for Generalizable RF Sensing

    cs.LG 2026-08 conditional novelty 6.0 of 10

    FSTC-Encoder factorizes RF sensing into feature, spatial, and temporal correlation stages and reports the best mean cross-domain accuracy on Widar3.0 (92.15%) and strong results on CSI-Bench and XRF55 with a single sh...

Reference graph

Works this paper leans on

54 extracted references · 42 canonical work pages · cited by 1 Pith paper

  1. [1]

    Adib, Zachary Kabelac, Dina Katabi, and Rob Miller

    Fadel M. Adib, Zachary Kabelac, Dina Katabi, and Rob Miller. 2014. 3D Tracking via Body Radio Reflections. In Symposium on Networked Systems Design and Implementation

    work page 2014

  2. [2]

    Mahmudur Rahman, and Jared Q

    Mohammad Arif Ul Alam, Md. Mahmudur Rahman, and Jared Q. Widberg. 2021. PALMAR: Towards Adaptive Multi-inhabitant Activity Recognition in Point- Cloud Technology. IEEE INFOCOM 2021 - IEEE Conference on Computer Commu- nications (2021), 1–10

    work page 2021

  3. [3]

    Liu, Wen Wang, and Muhammad Shahzad

    Kamran Ali, Alex X. Liu, Wen Wang, and Muhammad Shahzad. 2015. Keystroke Recognition Using WiFi Signals. Proceedings of the 21st Annual International Conference on Mobile Computing and Networking (2015)

    work page 2015

  4. [4]

    Qiming Cao, Hongfei Xue, Tianci Liu, Xingchen Wang, Haoyu Wang, Xincheng Zhang, and Lu Su. 2024. mmCLIP: Boosting mmWave-based Zero-shot HAR via Signal-Text Alignment. In ACM International Conference on Embedded Networked Sensor Systems

    work page 2024

  5. [5]

    Guo Chen, Yifei Huang, Jilan Xu, Baoqi Pei, Zhe Chen, Zhiqi Li, Jiahao Wang, Kun- chang Li, Tong Lu, and Limin Wang. 2024. Video mamba suite: State space model as a versatile alternative for video understanding. arXiv preprint arXiv:2403.09626 (2024)

    arXiv 2024

  6. [6]

    Han Ding, Lei Guo, Cui Zhao, Fei Wang, Ge Wang, Zhiping Jiang, Wei Xi, and Jizhong Zhao. 2020. RFnet: Automatic gesture recognition and human identifica- tion using time series RFID signals. Mobile Networks and Applications 25 (2020), 2240–2253

    work page 2020

  7. [7]

    Han Ding, Chen Qian, Jinsong Han, Ge Wang, Wei Xi, Kun Zhao, and Jizhong Zhao. 2017. RFIPad: Enabling Cost-Efficient and Device-Free In-air Handwrit- ing Using Passive Tags. 2017 IEEE 37th International Conference on Distributed Computing Systems (ICDCS) (2017), 447–457

    work page 2017

  8. [8]

    Student Member Ieee Yuqian Hu, Senior Member Ieee Beibei Wang, Senior Member Ieee Chenshu Wu, and Fellow Ieee K. J. Ray Liu. 2021. mmKey: Universal Virtual Keyboard Using A Single Millimeter-Wave Radio. IEEE Internet of Things Journal 9 (2021), 510–524

    work page 2021

Show all 54 references
  1. [9]

    Dr. V. Ilango and Anupama P Jose. 2018. Soli (Omnipresent Gesture Sensing with Millimeter Wave Radar)

    2018

  2. [10]

    Wenjun Jiang, Chenglin Miao, Fenglong Ma, Shuochao Yao, Yaqing Wang, Ye Yuan, Hongfei Xue, Chen Song, Xin Ma, Dimitrios Koutsonikolas, Wenyao Xu, and Lu Su. 2018. Towards Environment Independent Device Free Human Activity Recognition. Proceedings of the 24th Annual Internation...

    2018

  3. [11]

    Xuejing Jiang, Ye Liu, and Xiaolei Wang. 2009. An Enhanced Approach of Indoor Location Sensing Using Active RFID. 2009 W ASE International Conference on Information Engineering 1 (2009), 169–172

    2009

  4. [12]

    Sai Shashank Kalakonda, Shubh Maheshwari, and Ravi Kiran Sarvadevabhatla

  5. [13]

    Abdelwahed Khamis, Branislav Kusy, Chun Tung Chou, Mary-Louise McLaws, and Wen Hu. 2020. RFWash: a weakly supervised tracking of hand hygiene technique. Proceedings of the 18th Conference on Embedded Networked Sensor Systems (2020)

    2020

  6. [14]

    Bo Lan, Pei Li, Jiaxi Yin, Yunpeng Song, Ge Wang, Han Ding, Jinsong Han, and Fei Wang. 2025. XRF V2: A Dataset for Action Summarization with Wi-Fi Signals, and IMUs in Phones, Watches, Earbuds, and Glasses. arXiv preprint arXiv:2501.19034 (2025)

    2025 arXiv

  7. [15]

    Zehan Li, Xin Zhang, Yanzhao Zhang, Dingkun Long, Pengjun Xie, and Meishan Zhang. 2023. Towards general text embeddings with multi-stage contrastive learning. arXiv preprint arXiv:2308.03281 (2023)

    2023 arXiv

  8. [16]

    Zhendong Liu, Le Zhang, Bing Li, Yingjie Zhou, Zhenghua Chen, and Ce Zhu

  9. [17]

    Yizhe Lv, Tingting Zhang, Yunpeng Song, Han Ding, Jinsong Han, and Fei Wang

  10. [18]

    Seungwhan Moon, Andrea Madotto, Zhaojiang Lin, Aparajita Saraf, Amy Bear- man, and Babak Damavandi. 2023. IMU2CLIP: Language-grounded Motion Sensor Translation with Multimodal Contrastive Learning. In Conference on Empirical Methods in Natural Language Processing

    2023

  11. [19]

    Sauradip Nag, Xiatian Zhu, Yi-Zhe Song, and Tao Xiang. 2022. Zero-shot tempo- ral action detection via vision-language prompting. In European conference on computer vision. Springer, 681–697

    2022

  12. [20]

    arXiv preprint arXiv:2501.13805 (2025)

    EgoHand: Ego-centric Hand Pose Estimation and Gesture Recognition with Head-mounted Millimeter-wave Radar and IMUs. arXiv preprint arXiv:2501.13805 (2025)

    2025 arXiv

  13. [21]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learni...

    2021

  14. [22]

    Sai Deepika Regani, Chenshu Wu, Beibei Wang, Min Wu, and K. J. Ray Liu. 2021. mmWrite: Passive Handwriting Tracking Using a Single Millimeter-Wave Radio. IEEE Internet of Things Journal 8 (2021), 13291–13305

    2021

  15. [23]

    Jacopo Pegoraro, Jesús Omar Lacruz, Michele Rossi, and Joerg Widmer. 2022. SPARCS: A Sparse Recovery Approach for Integrated Communication and Human Sensing in mmWave Systems. 2022 21st ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN) (2022), 79–91

    2022

  16. [24]

    Srivastava

    Akash Deep Singh, Sandeep Singh Sandha, Luis Garcia, and Mani B. Srivastava

  17. [25]

    Xiaoqi Sun, Yanwen Wang, Chenwei Zhang, Zheng Wang, Xiaokang Shi, and Yuanqing Zheng. 2024. LoDiHAR: A Low-Cost Distributed Human Activity Recognition System Based on RFID. In 2024 21st Annual IEEE International Con- ference on Sensing, Communication, and Networking (SECON) . ...

    2024

  18. [26]

    Dingfeng Shi, Yujie Zhong, Qiong Cao, Lin Ma, Jia Li, and Dacheng Tao. 2023. Tridet: Temporal action detection with relative boundary modeling. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . 18857– 18866

    2023

  19. [27]

    Christian C. J. M. Tiberius and Edward Verbree. 2004. GNSS positioning accuracy and availability within Location Based Services: The advantages of combined GPS-Galileo positioning

    2004

  20. [28]

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yas- mine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhos- ale, et al. 2023. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288 (2023)

    2023 arXiv

  21. [29]

    Kindt, Alejandro Masrur, and Graziano Pravadelli

    Cristian Turetta, Florenc Demrozi, Philipp H. Kindt, Alejandro Masrur, and Graziano Pravadelli. 2022. Practical identity recognition using WiFi’s Chan- nel State Information. 2022 Design, Automation & Test in Europe Conference & Exhibition (DATE) (2022), 76–79

    2022

  22. [30]

    Tuan N Tang, Kwonyoung Kim, and Kwanghoon Sohn. 2023. Temporalmaxer: Maximize temporal context with only max pooling for temporal action localiza- tion. arXiv preprint arXiv:2303.09055 (2023)

    2023 arXiv

  23. [31]

    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 Technologies 8, 1 (2024), 1–34

    2024

  24. [32]

    Fei Wang, Tingting Zhang, Bintong Zhao, Libao Xing, Tiantian Wang, Han Ding, and Tony Xiao Han. 2025. A Survey on Wi-Fi Sensing Generalizability: Taxonomy, Techniques, Datasets, and Future Research Prospects. arXiv preprint arXiv:2503.08008 (2025)

    2025

  25. [33]

    Fei Wang, Sanping Zhou, Stanislav Panev, Jinsong Han, and Dong Huang. 2019. Person-in-WiFi: Fine-Grained Person Perception Using WiFi. 2019 IEEE/CVF International Conference on Computer Vision (ICCV) (2019), 5451–5460

    2019

  26. [34]

    Fei Wang, Yiao Gao, Bo Lan, Han Ding, Jingang Shi, and Jinsong Han. 2023. U-Shape Networks Are Unified Backbones for Human Action Understanding From Wi-Fi Signals. IEEE Internet of Things Journal 11, 6 (2023), 10020–10030

    2023

  27. [35]

    Yuheng Wang, Haipeng Liu, Kening Cui, Anfu Zhou, Wensheng Li, and Huadong Ma. 2021. m-Activity: Accurate and Real-Time Human Activity Recognition Via Millimeter Wave Radar. ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) (20...

    2021

  28. [36]

    Teng Wei and Xinyu Zhang. 2015. mTrack: High-Precision Passive Tracking Using Millimeter Wave Radios. Proceedings of the 21st Annual International Conference on Mobile Computing and Networking (2015)

    2015

  29. [37]

    Yuxuan Weng, Guoquan Wu, Tianyue Zheng, Yanbing Yang, and Jun Luo. 2024. Large Model for Small Data: Foundation Model for Cross-Modal RF Human Activity Recognition. ArXiv abs/2410.19766 (2024)

    2024 arXiv

  30. [38]

    Xijun Wang, Ruiqi Xian, Tianrui Guan, and Dinesh Manocha. 2023. PLAR: Prompt Learning for Action Recognition. arXiv e-prints (2023), arXiv–2305

    2023

  31. [39]

    Hongfei Xue, Yan Ju, Chenglin Miao, Yijiang Wang, Shiyang Wang, Aidong Zhang, and Lu Su. 2021. mmMesh: towards 3D real-time dynamic human mesh construction using millimeter-wave. Proceedings of the 19th Annual International Conference on Mobile Systems, Applications, and Servi...

    2021

  32. [40]

    Kangwei Yan, Fei Wang, Bo Qian, Han Ding, Jinsong Han, and Xing Wei. 2024. Person-in-wifi 3d: End-to-end multi-person 3d pose estimation with wi-fi. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 969–978

    2024

  33. [41]

    Jianfei Yang, He Huang, Yunjiao Zhou, Xinyan Chen, Yuecong Xu, Shenghai Yuan, Han Zou, Chris Xiaoxuan Lu, and Lihua Xie. 2023. MM-Fi: Multi-Modal Non- Intrusive 4D Human Dataset for Versatile Wireless Sensing.ArXiv abs/2305.10345 (2023)

    2023 arXiv

  34. [42]

    Kang Xia, Wenzhong Li, Shiwei Gan, and Sanglu Lu. 2023. TS2ACT: Few-Shot Human Activity Sensing with Cross-Modal Co-Learning. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 7 (2023), 188:1–188:22

    2023

  35. [43]

    Zheng Yang, Yi Zhang, and Qian Zhang. 2023. Rethinking Fall Detection With Wi-Fi. IEEE Transactions on Mobile Computing 22 (2023), 6126–6143. ACM XX 2025, June 03–05, 2025, Woodstock, NY Yang et al

    2023

  36. [44]

    Chen-Lin Zhang, Jianxin Wu, and Yin Li. 2022. Actionformer: Localizing moments of actions with transformers. InEuropean Conference on Computer Vision. Springer, 492–510

    2022

  37. [45]

    Haopeng Zhang, Yili Ren, Haohan Yuan, Jingzhe Zhang, and Yitong Shen. 2025. Wi-Chat: Large Language Model Powered Wi-Fi Sensing. ArXiv abs/2502.12421 (2025)

    2025 arXiv

  38. [46]

    Lei Yang, Yekui Chen, Xiangyang Li, Chaowei Xiao, Mo Li, and Yunhao Liu

  39. [47]

    Hongwei Zheng, Han Li, Bowen Shi, Wenrui Dai, Botao Wang, Yu Sun, Min Guo, and Hongkai Xiong. 2023. Actionprompt: Action-guided 3d human pose estimation with text and pose prompting. In 2023 IEEE International Conference on Multimedia and Expo (ICME) . IEEE, 2657–2662

    2023

  40. [48]

    Yunjiao Zhou, He Huang, Shenghai Yuan, Han Zou, Lihua Xie, and Jianfei Yang

  41. [51]

    Yi Zhang, Yue Zheng, Kun Qian, Guidong Zhang, Yunhao Liu, Chenshu Wu, and Zheng Yang. 2021. Widar3.0: Zero-Effort Cross-Domain Gesture Recognition With Wi-Fi. IEEE Transactions on Pattern Analysis and Machine Intelligence 44 (2021), 8671–8688

    2021

  42. [54]

    IEEE Internet of Things Journal 10 (2023), 14128– 14136

    MetaFi++: WiFi-Enabled Transformer-Based Human Pose Estimation for Metaverse Avatar Simulation. IEEE Internet of Things Journal 10 (2023), 14128– 14136

    2023

  43. [2014]

    Proceedings of the 20th annual international conference on Mobile computing and networking (2014)

    Tagoram: real-time tracking of mobile RFID tags to high precision using COTS devices. Proceedings of the 20th annual international conference on Mobile computing and networking (2014)

    2014

  44. [2019]

    Proceedings of the 3rd ACM Workshop on Millimeter-wave Networks and Sensing Systems (2019)

    RadHAR: Human Activity Recognition from Point Clouds Generated through a Millimeter-wave Radar. Proceedings of the 3rd ACM Workshop on Millimeter-wave Networks and Sensing Systems (2019)

    2019

  45. [2023]

    In 2023 IEEE International Conference on Multimedia and Expo (ICME)

    Action-gpt: Leveraging large-scale language models for improved and gen- eralized action generation. In 2023 IEEE International Conference on Multimedia and Expo (ICME). IEEE, 31–36

    2023

  46. [2025]

    In Proceedings of the AAAI Conference on Artificial Intelligence , Vol

    WiFi CSI Based Temporal Activity Detection via Dual Pyramid Network. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 39. 550–558

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.