Pith. sign in

REVIEW 4 major objections 4 minor 33 references

RAG-HAR+ shows that LLM-based activity recognition can keep accuracy while cutting online LLM calls by 89.3–99.9% and per-sample latency by 5.7–44.5×, by moving the LLM into an offline design stage and a small online fallback role.

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 ↗

RAG-HAR+ adapts retrieval features offline via an LLM agent and routes only ambiguous windows to the LLM, cutting online LLM cost by ~90–99.9% with competitive accuracy on six HAR benchmarks.

T0 review reviewed 2026-08-01 challenge →

load-bearing objection A practical, clever cost-reduction extension of RAG-HAR with a strong cost story, but the validation-tuned design and single-split evaluation mean the headline numbers need more evidence before I'd trust them across datasets. the 4 major comments →

arxiv 2607.26631 v1 pith:J6ZBM2BA submitted 2026-07-29 cs.LG cs.IR

RAG-HAR+: Towards Cost-Efficient LLM-Based Human Activity Recognition for Edge Deployment

Hansi Karunarathna , Nirhoshan Sivaroopan , Chamara Madarasingha , Anura Jayasumana , Kanchana Thilakarathna This is my paper
classification cs.LG cs.IR
keywords Human Activity RecognitionRetrieval-Augmented GenerationLarge Language ModelsEdge DeploymentCost OptimizationFeature Group SelectionMajority VotingWearable Sensors
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.

The reading

RAG-HAR+ is a training-free extension of retrieval-augmented human activity recognition that tries to make LLM-based HAR cheap enough for continuous edge deployment. Its central idea is to stop calling an LLM for every sensor window. Instead, an offline Retrieval Designer Agent uses the LLM once per dataset to choose compact, dataset-specific feature groups for similarity search, and online the system classifies most windows directly by majority vote over retrieved labeled neighbors, routing only undecided cases to an Ambiguity Resolver Agent. Across six HAR benchmarks the paper claims competitive or improved recognition accuracy while reducing online LLM tokens by 89.3–99.9% and per-sample latency by 5.7–44.5×, with a smartphone prototype showing similar gains in a live mobile setting. A sympathetic reader would see this as a practical route to retrieval-first, LLM-assisted HAR that scales with ambiguity rather than data volume.

Core claim

The paper claims that the dominant barrier to deploying LLM-based HAR is not accuracy but the per-sample online LLM call, and that this barrier can be removed by changing when and where the LLM is used. RAG-HAR+ moves the LLM upstream into a one-time offline Retrieval Designer Agent that selects three complementary feature groups from a pool of statistical, temporal, spectral, and signal-shape descriptors, tuned on a validation split. Online, it indexes training windows as multi-vector records, retrieves top-q neighbors under validation-tuned group weights, and classifies by parameter-free majority voting whenever one label has a unique plurality; only genuinely ambiguous ties are sent to th

What carries the argument

The load-bearing mechanism is the Retrieval Designer Agent, an offline LLM prompted iteratively with validation feedback to design three compact, non-overlapping feature groups from a 67-feature candidate pool per dataset. These groups define a multi-vector index where each window is represented by three normalized numerical vectors; during inference, weighted multi-vector search (with weights grid-searched on validation) merges and re-ranks candidates, and a deterministic majority-voting routing rule decides whether a query is answered directly from retrieval or deferred to the Ambiguity Resolver Agent. The separation of a numerical-vector retriever from a segment-wise textual prompt repres

Load-bearing premise

The claimed accuracy and cost savings depend on the assumption that the feature groups, group weights, K=3, q=10, and routing rule chosen on the validation split of one dataset — mostly USC-HAD — transfer to the test split and to the other five datasets without overfitting or dataset-specific bias.

What would settle it

A decisive check is to re-run the Retrieval Designer Agent with a different validation split (or hold out entire subjects not used anywhere in design) and compare the resulting feature groups, online LLM call rate, and F1 to the reported numbers; if the 89.3–99.9% token reductions and competitive F1 scores collapse when the validation split changes, the cost-efficiency result is an artifact of validation tuning rather than a property of the retrieval-first design.

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

If this is right

  • Online inference cost under RAG-HAR+ scales with the number of ambiguous windows rather than the total number of test samples, making continuous streaming HAR practical on phones and edge gateways.
  • Because the offline design cost is amortized over all later inference, new datasets can be handled without training a classifier or fine-tuning an LLM, requiring only a one-time feature-group design pass.
  • The reported per-sample latency of 0.41–2.04 s across benchmarks is compatible with near-real-time activity recognition on mobile devices, as demonstrated by the smartphone prototype.
  • The framework keeps a small but useful role for LLM reasoning: the Ambiguity Resolver Agent adds 0.1–3.3 percentage points of F1 on top of retrieval voting, justifying its retention for confusing activity pairs.
  • The design suggests that retrieval quality, not LLM reasoning, is the primary levers for both accuracy and cost in retrieval-augmented sensor classification.

Where Pith is reading between the lines

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

  • The same two-stage pattern—offline LLM design of a retrieval representation, online majority-vote routing with selective fallback—could transfer to other time-series classification tasks beyond HAR, such as gesture recognition, sleep staging, or anomaly detection, whenever labeled exemplars are plentiful.
  • A natural extension the paper leaves implicit is replacing the tie-based routing rule with a soft uncertainty threshold (e.g., vote-margin or label entropy), which could reduce fallback calls further or recover accuracy on datasets where ties are rare but near-miss majorities are common.
  • Since the Retrieval Designer Agent's groups and weights are chosen on a validation split of each dataset, the approach would only be truly 'zero-shot' for a new dataset if those design choices transfer without re-running the offline stage; testing this transfer directly would clarify how much per-dataset design cost remains.
  • The privacy framing is deliberately scoped out: routing only ambiguous windows to a cloud LLM reduces raw sensor data exposure compared with sending every window, but the ambiguous windows still carry sensitive motion information, so quantifying that residual exposure would be a valuable follow-up.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. RAG-HAR+ extends the authors' prior RAG-HAR framework for training-free, LLM-based human activity recognition. It introduces an offline Retrieval Designer Agent that iteratively selects three dataset-specific feature groups from a motion-descriptor pool using validation-set feedback, followed by online multi-vector retrieval, weighted re-ranking, and majority-vote classification over the top-q neighbors. Only tie cases are sent to an LLM Ambiguity Resolver Agent. The paper evaluates on six public HAR datasets and reports competitive or improved accuracy relative to RAG-HAR while reducing online LLM token usage by 89.3–99.9% and per-sample latency by 5.7–44.5×, together with a smartphone prototype demonstrating the feasibility of the approach in a mobile setting.

Significance. If the reported accuracy–cost trade-offs are stable, this is a meaningful step toward practical LLM-based HAR: it shifts the LLM from a per-sample online classifier to a one-time offline retrieval designer plus a selective fallback, and it provides a careful decomposition of the factors that make the retriever effective. The paper's strengths include a broad ablation study (feature-group selection, number of groups, group assignment, group weights, retrieval depth q, and prompt representation), a clear separation of offline and online costs, and an actual mobile prototype with real network round-trip latency. These strengths are undermined by the absence of repeated evaluations, error bars, significance tests, and released code/data/prompt logs; as presented, the central cost-reduction claim is not yet shown to be a stable property of the method rather than of the specific validation-selected configuration.

major comments (4)
  1. [§III-B2, §IV-E1d, Table IV] The headline 89.3–99.9% token-reduction claim is not yet stable because the retrieval configuration is selected on validation splits and then evaluated on a single test split. The Retrieval Designer Agent iteratively proposes feature groups and keeps the best validation configuration; weights (w1,w2,w3) are grid-searched on validation; K=3 and q=10 are chosen from USC-HAD validation; and the Ambiguity Resolver prompt representation is also chosen on USC-HAD validation. Since the fallback rate is exactly the rate at which top-q voting ties, these choices directly determine the token reductions in Table IV. The validation signals are weak for several datasets (two held-out subjects; a 20% window-level split for Skoda), and no number of design rounds or candidate configurations is reported. Please report the search effort, repeat the full protocol over multiple subject splits or bootstrap r
  2. [§IV-A, Table II] The claim that RAG-HAR+ maintains 'competitive or improved' performance is not statistically supported on the two datasets where it is below RAG-HAR. Table II reports GOTOV F1 of 75.97 versus 79.92 for RAG-HAR, and PAMAP2 F1 of 90.60 versus 91.12, with no error bars, no confidence intervals, and no significance tests. Because each dataset is evaluated once, and the LLM fallback is stochastic, the observed gaps could be within run-to-run noise. The abstract's 'competitive' interpretation needs explicit support for these datasets, for example repeated runs with different seeds, paired significance tests, or at least confidence intervals around the reported point estimates.
  3. [§IV-B, Table III] The amortization argument for the offline design cost is under-specified. Table III reports only the final total of 132,176 tokens across six datasets, but not how many Retrieval Designer Agent rounds were executed, how many candidate feature-group configurations were evaluated, or the stopping criterion used. This matters both for the 'one-time cost' claim and for the overfitting concern: the number of validation-maximizing candidates is the effective number of comparisons behind the selected G*. Please report the per-dataset round count, the validation-score trajectory, and the number of candidates considered, and incorporate those numbers into the cost comparison.
  4. [Reproducibility] No code, data, exact prompt templates, prompt logs, or design-round traces are released. Because the Retrieval Designer Agent's outputs drive every downstream result, the paper should at least release the exact prompts, the per-round feature-group proposals, the selected feature groups, and the grid-search implementation, or provide a detailed reproducibility appendix. Without these artifacts, the validation-selection process and the reported token reductions cannot be independently recomputed or audited.
minor comments (4)
  1. [§I, final bullet] Typo: 'Ambuiguity' should be 'Ambiguity'.
  2. [§IV-F] The heading 'F . Discussion' has an unwanted space after the initial letter.
  3. [§IV-E1d] The validation-split protocol differs across datasets: most use two held-out subjects, while Skoda uses a random 20% window-level split. The paper should justify this difference, since window-level splits can create subject or temporal leakage that subject-level splits avoid.
  4. [Table IV] The token counts are reported for gpt-5-mini, but the paper does not state whether these counts include the fixed system-prompt overhead or only the sample-dependent content. A clear statement of what is included in 'tokens per sample' would make the comparison with RAG-HAR easier to interpret.

Circularity Check

0 steps flagged

No significant circularity: the central accuracy/cost claims are measured on held-out test splits, and the RAG-HAR self-citation is a comparator, not a load-bearing derivation step.

full rationale

No circular step is present in the claimed derivation chain. RAG-HAR+ is an empirical evaluation: the Retrieval Designer Agent (Section III-B2), feature-group weights (Section IV-E1d), number of groups (Section IV-E1b), and retrieval depth q (Section IV-E3a) are selected on validation splits constructed from the training split, while the reported accuracy, F1, token usage, and latency are measured on test splits. The accuracy-cost result is therefore not constructed to equal the validation objective; it is a generalization measurement. The online token reduction is an explicit and acknowledged consequence of the deterministic routing rule in Section III-C4, not a hidden prediction that reduces to an input by definition. The paper's use of its own prior RAG-HAR [7] as a baseline and as the source of preprocessing/windowing protocols is a self-citation, but it is used as an external comparator and inherited experimental setup, not as a uniqueness theorem or as the justification for the central claim. The claimed improvements are independently checkable against the cited external baselines and held-out data. No equation or fitted parameter is renamed as a prediction.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 0 invented entities

No new physical or mathematical entities are introduced. The Retrieval Designer Agent and Ambiguity Resolver Agent are software components, not postulated entities. The main burden on the reader is the large number of validation-fitted choices: feature groups, weights, K, q, and prompt representation, all selected on validation data, mostly on USC-HAD, and then applied broadly. The central claim relies on the assumption that this validation tuning transfers to the test splits and other datasets.

free parameters (5)
  • Number of feature groups K = 3
    Set from USC-HAD validation ablation (Fig. 6b, Section IV-E1b), then fixed for all datasets.
  • Retrieval depth q = 10
    Set from USC-HAD validation ablation (Fig. 7, Section IV-E3a), then fixed for all datasets.
  • Feature-group weights (w1, w2, w3) = Per-dataset grid-search result, e.g., USC-HAD 0.2/0.6/0.2
    Selected via grid search on validation retrieval F1 (Section IV-E1d, Table V); this is a direct fit to validation data.
  • Selected feature groups G* = Listed per dataset in Table I
    The LLM Retrieval Designer Agent iteratively proposes feature groups, evaluates them on a validation split, and the best round is selected (Section III-B). This is a high-dimensional validation-fitted configuration.
  • Ambiguity Resolver prompt representation = Segment-wise basic statistics
    Chosen from USC-HAD validation ablation (Fig. 8, Section IV-E3b); the selected representation is the one that performed best on validation.
axioms (5)
  • domain assumption An LLM can design effective retrieval feature groups from feature names/descriptions and validation feedback.
    Core mechanism of the Retrieval Designer Agent (Section III-B); the paper demonstrates this empirically on validation sets but does not prove it generally.
  • domain assumption Validation-set retrieval metrics used by the agent and weight grid search are predictive of test-set performance.
    Standard ML assumption underlying all hyperparameter selection in Section III-B and Section IV-E; load-bearing for the reported test-set numbers.
  • domain assumption A unique majority among retrieved neighbor labels is a reliable signal for direct classification; ties are a good proxy for ambiguity.
    Routing rule in Section III-C4; the paper itself notes more adaptive uncertainty strategies as future work (Section IV-F).
  • domain assumption Statistical, temporal, spectral, and signal-shape features over sliding windows capture activity-discriminative information.
    Foundational to the feature pool inherited from TSfel (Section III-B1); not derived in this paper.
  • domain assumption Z-score normalization estimated from the training split and applied to the test split avoids leakage.
    Standard practice described in Section III-C1; assumed to be correctly implemented and consistent with prior HAR protocols.

reviewed 2026-08-01 · how reviews work

0 comments
Cite this review

Pith. "Pith review of RAG-HAR+: Towards Cost-Efficient LLM-Based Human Activity Recognition for Edge Deployment." pith.science (2026). https://pith.science/paper/J6ZBM2BA

@misc{pith2026260726631,
  author       = {Pith},
  title        = {Pith review of: RAG-HAR+: Towards Cost-Efficient LLM-Based Human Activity Recognition for Edge Deployment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/J6ZBM2BA}},
  note         = {Machine review of arXiv:2607.26631}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Human Activity Recognition (HAR) from wearable sensors supports applications in healthcare, rehabilitation, fitness tracking, and smart environments. Yet, existing deep learning approaches require dataset-specific training, large labeled corpora, and repeated adaptation to new sensor settings or activity taxonomies. Retrieval-Augmented Generation for Human Activity Recognition (RAG-HAR) addresses this by framing HAR as a training-free, retrieval-augmented task, in which statistical descriptions of sensor windows are used to retrieve similar labeled examples that guide LLM-based classification. We introduce RAG-HAR+, a retrieval-first and cost-optimized extension that strengthens retrieval while reducing dependence on LLM-based inference. RAG-HAR+ uses an offline Retrieval Designer Agent to design dataset-specific feature groups from a diverse pool of motion descriptors, enabling sensor windows to be compared using features better aligned with dataset-specific activity patterns. During inference, RAG-HAR+ uses majority voting over retrieved neighbors for samples with strong retrieval evidence and defers only uncertain cases to an LLM-based Ambiguity Resolver Agent. Across six HAR benchmarks, RAG-HAR+ maintains competitive or improved performance while reducing LLM usage, token consumption, and inference time. We further extend the RAG-HAR mobile prototype to demonstrate the practical feasibility of retrieval-first, LLM-assisted HAR in mobile sensing scenarios.

Figures

Figures reproduced from arXiv: 2607.26631 by Anura Jayasumana, Chamara Madarasingha, Hansi Karunarathna, Kanchana Thilakarathna, Nirhoshan Sivaroopan.

Figure 1
Figure 1. Figure 1: Overview of RAG-HAR+. (a) Offline Retrieval Designer Agent for selecting dataset-specific feature groups. (b) Feature indexing and inference [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Prompt structure used by the Retrieval Designer Agent. (a) Initial prompt used to generate the first feature-group configuration. (b) Improvement [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (a) Average online LLM token usage. (b) Average inference latency [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of RAG-HAR and RAG-HAR+. For RAG-HAR, the blue [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: t-SNE analysis of RAG-HAR vs. RAG-HAR+ retriever performance [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Ablation of the offline retrieval design. (a) Held-out USC-HAD test performance of the basic-statistics, all-candidate, mRMR-selected, and Retrieval [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: USC-HAD validation-set ablation of the textual representation [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: RAG-HAR+ widget embedded within a mobile fitness-tracking application: (a) activity prediction, (b) activity insights and history, (c) labeled motion [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Per-activity F1-scores of the mobile-integrated RAG-HAR+ prototype [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

33 extracted references · 2 linked inside Pith

  1. [1]

    A tutorial on human activity recognition using body-worn inertial sensors,

    A. Bulling, U. Blanke, and B. Schiele, “A tutorial on human activity recognition using body-worn inertial sensors,”ACM Computing Surveys, vol. 46, no. 3, pp. 1–33, 2014

    2014

  2. [2]

    A survey on human activity recognition using wearable sensors,

    O. D. Lara and M. A. Labrador, “A survey on human activity recognition using wearable sensors,”IEEE Communications Surveys & Tutorials, vol. 15, no. 3, pp. 1192–1209, 2013

    2013

  3. [3]

    Activity recognition from on-body sensors: Accuracy- power trade-off by dynamic sensor selection,

    P. Zappi, C. Lombriser, T. Stiefmeier, E. Farella, D. Roggen, L. Benini, and G. Tr ¨oster, “Activity recognition from on-body sensors: Accuracy- power trade-off by dynamic sensor selection,” inProceedings of the European Conference on Wireless Sensor Networks. Springer, 2008, pp. 17–33

    2008

  4. [4]

    Deep convolutional neural networks on multichannel time series for human activity recognition,

    J. Yang, M. N. Nguyen, P. P. San, X. Li, and S. Krishnaswamy, “Deep convolutional neural networks on multichannel time series for human activity recognition,” inProceedings of the International Joint Conference on Artificial Intelligence, 2015, pp. 3995–4001

    2015

  5. [5]

    Deep convolutional and lstm recurrent neural networks for multimodal wearable activity recognition,

    F. J. Ord ´o˜nez and D. Roggen, “Deep convolutional and lstm recurrent neural networks for multimodal wearable activity recognition,”Sensors, vol. 16, no. 1, p. 115, 2016

    2016

  6. [6]

    Metier: A deep multi-task learning-based activity and user recognition model using wearable sensors,

    L. Chen, Y . Zhang, and L. Peng, “Metier: A deep multi-task learning-based activity and user recognition model using wearable sensors,”Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 4, no. 1, Mar. 2020. [Online]. Available: https://doi.org/10.1145/3381012

    doi:10.1145/3381012 2020

  7. [7]

    Rag-har: Retrieval augmented generation-based human activity recognition,

    N. Sivaroopan, H. Karunarathna, C. Madarasingha, A. Jayasumana, and K. Thilakarathna, “Rag-har: Retrieval augmented generation-based human activity recognition,” in2026 IEEE International Conference on Pervasive Computing and Communications (PerCom). IEEE, 2026, pp. 1–12

    2026

  8. [8]

    Large language models are zero-shot time series forecasters,

    N. Gruver, M. Finzi, S. Qiu, and A. G. Wilson, “Large language models are zero-shot time series forecasters,”Advances in Neural Information Processing Systems, vol. 36, 2023

    2023

  9. [9]

    Timegpt- 1,

    A. Garza, C. Challu, and M. Mergenthaler-Canseco, “Timegpt- 1,”arXiv preprint arXiv:2310.03589, 2023. [Online]. Available: https://arxiv.org/abs/2310.03589

    Pith/arXiv arXiv 2023

  10. [10]

    Time-llm: Time series forecasting by reprogramming large language models,

    M. Jin, S. Wang, L. Ma, Z. Chu, J. Y . Zhang, X. Shi, P.-Y . Chen, Y . Liang, Y .-F. Li, S. Pan, and Q. Wen, “Time-llm: Time series forecasting by reprogramming large language models,” inInternational Conference on Learning Representations, 2024

    2024

  11. [11]

    Sensorllm: Aligning large language models with motion sensors for human activity recognition,

    Z. Li, S. Deldari, L. Chen, H. Xue, and F. D. Salim, “Sensorllm: Aligning large language models with motion sensors for human activity recognition,” inProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, 2025, pp. 354–379

    2025

  12. [12]

    Llm4har: Generalizable on-device human activity recognition with pretrained llms,

    Z. Hong, Y . Song, Z. Li, A. Yu, S. Zhong, Y . Ding, T. He, and D. Zhang, “Llm4har: Generalizable on-device human activity recognition with pretrained llms,” inProc of the 31st ACM SIGKDD conf on Knowledge Discovery and Data Mining V . 2, 2025

    2025

  13. [13]

    A personal health large language model for sleep and fitness coaching,

    J. Khasentino, A. Belyaeva, X. Liuet al., “A personal health large language model for sleep and fitness coaching,”Nature Medicine, vol. 31, pp. 3394–3403, 2025. [Online]. Available: https://www.nature.com/articles/s41591-025-03888-0

    2025

  14. [14]

    Contrastive self- supervised learning for sensor-based human activity recognition,

    B. Khaertdinov, E. Ghaleb, and S. Asteriadis, “Contrastive self- supervised learning for sensor-based human activity recognition,” in 2021 IEEE International Joint Conference on Biometrics (IJCB), 2021, pp. 1–8

    2021

  15. [15]

    Towards llm- powered ambient sensor based multi-person human activity recognition,

    X. Chen, J. Cumin, F. Ramparany, and D. Vaufreydaz, “Towards llm- powered ambient sensor based multi-person human activity recognition,” in2024 IEEE 30th International Conference on Parallel and Distributed Systems (ICPADS), 2024, pp. 609–616

    2024

  16. [16]

    Unsupervised human activity recognition through two-stage prompting with chatgpt,

    Q. Xia, T. Maekawa, and T. Hara, “Unsupervised human activity recognition through two-stage prompting with chatgpt,”arXiv preprint arXiv:2306.02140, 2023. [Online]. Available: https://arxiv.org/abs/2306. 02140

    Pith/arXiv arXiv 2023

  17. [17]

    Hargpt: Are llms zero-shot human activity recognizers?

    S. Ji, X. Zheng, and C. Wu, “Hargpt: Are llms zero-shot human activity recognizers?” in2024 IEEE International Workshop on Foundation Models for Cyber-Physical Systems & Internet of Things (FMSys), 2024, pp. 38–43

    2024

  18. [18]

    Tsfel: Time series feature extraction library,

    M. Barandas, D. Folgado, L. Fernandes, S. Santos, M. Abreu, P. Bota, H. Liu, T. Schultz, and H. Gamboa, “Tsfel: Time series feature extraction library,”SoftwareX, vol. 11, p. 100456, Jan. 2020. [Online]. Available: http://dx.doi.org/10.1016/j.softx.2020.100456

    arXiv 2020

  19. [19]

    Deep triplet networks with attention for sensor-based human activity recognition,

    B. Khaertdinov, E. Ghaleb, and S. Asteriadis, “Deep triplet networks with attention for sensor-based human activity recognition,” in2021 IEEE International Conference on Pervasive Computing and Communi- cations (PerCom), 2021, pp. 1–10

    2021

  20. [20]

    Adversarial deep feature extraction network for user independent human activity recognition,

    S. Suh, V . F. Rey, and P. Lukowicz, “Adversarial deep feature extraction network for user independent human activity recognition,” in2022 IEEE Percom, 2022

    2022

  21. [21]

    Alae-tae-cutmix+: Beyond the state-of-the- art for human activity recognition using wearable sensors,

    N. Ahmad and H.-f. Leung, “Alae-tae-cutmix+: Beyond the state-of-the- art for human activity recognition using wearable sensors,” in2023 IEEE International Conference on Pervasive Computing and Communications (PerCom), 2023, pp. 222–231

    2023

  22. [22]

    Investigating enhancements to contrastive predictive coding for human activity recognition,

    H. Haresamudram, I. Essa, and T. Pl ¨otz, “Investigating enhancements to contrastive predictive coding for human activity recognition,” in 2023 IEEE International Conference on Pervasive Computing and Communications (PerCom), 2023, pp. 232–241

    2023

  23. [23]

    Usc-had: A daily activity dataset for ubiquitous activity recognition using wearable sensors,

    M. Zhang and A. A. Sawchuk, “Usc-had: A daily activity dataset for ubiquitous activity recognition using wearable sensors,” inProceedings of the ACM Conference on Ubiquitous Computing, 2012, pp. 1036–1043

    2012

  24. [24]

    Introducing a new benchmarked dataset for activity monitoring,

    A. Reiss and D. Stricker, “Introducing a new benchmarked dataset for activity monitoring,” inProceedings of the International Symposium on Wearable Computers. IEEE, 2012, pp. 108–109

    2012

  25. [25]

    mhealthdroid: A novel framework for agile development of mobile health applications,

    O. Banos, R. Garcia, J. A. Holgado-Terriza, M. Damas, H. Pomares, I. Rojas, A. Saez, and C. Villalonga, “mhealthdroid: A novel framework for agile development of mobile health applications,” inProceedings of the International Work-Conference on Ambient Assisted Living. Springer, 2014, pp. 91–98

    2014

  26. [26]

    Activity recognition using wearable sensors for tracking the elderly,

    S. Paraschiakos, R. Cachucho, M. Moed, D. van Heemst, S. Mooijaart, E. P. Slagboom, A. Knobbe, and M. Beekman, “Activity recognition using wearable sensors for tracking the elderly,”User Modeling and User-Adapted Interaction, 2020

    2020

  27. [27]

    Smart devices are different: Assessing and mitigating mobile sensing heterogeneities for activity recognition,

    A. Stisen, H. Blunck, S. Bhattacharya, T. S. Prentow, M. B. Kjærgaard, A. K. Dey, T. Sonne, and M. M. Jensen, “Smart devices are different: Assessing and mitigating mobile sensing heterogeneities for activity recognition,” inProceedings of the ACM Conference on Embedded Networked Sensor Systems, 2015, pp. 127–140

    2015

  28. [28]

    Evaluating large language models as virtual annotators for time-series physical sensing data,

    A. Hota, S. Chatterjee, and S. Chakraborty, “Evaluating large language models as virtual annotators for time-series physical sensing data,”ACM Transactions on Intelligent Systems and Technology, 2025

    2025

  29. [29]

    GPT-5 mini Model,

    OpenAI, “GPT-5 mini Model,” https://developers.openai.com/api/docs/ models/gpt-5-mini, 2026, accessed: 2026-07-02

    2026

  30. [30]

    Large language models meet wearable sensing: A demo of training-free har with rag-har,

    H. Karunarathna, N. Sivaroopan, C. Madarasingha, A. Jayasumana, and K. Thilakarathna, “Large language models meet wearable sensing: A demo of training-free har with rag-har,” in2026 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops), 2026, pp. 470–472

    2026

  31. [31]

    Flutter: Build apps for any screen,

    Google, “Flutter: Build apps for any screen,” https://flutter.dev

  32. [32]

    Milvus vector database,

    Milvus, “Milvus vector database,” https://milvus.io//

  33. [33]

    A comprehensive survey on synthetic network traffic generation,

    N. Sivaroopan, K. Silva, C. Madarasingha, T. Dahanayaka, G. Jourjon, A. Jayasumana, and K. Thilakarathna, “A comprehensive survey on synthetic network traffic generation,”IEEE Communications Surveys & Tutorials, vol. 28, pp. 5949–5983, 2026

    2026

This paper was first reviewed by deepseek-v4-flash on August 1, 2026.