Pith. sign in

REVIEW 4 major objections 5 minor 47 references

StrokeVision-Bench introduces a 1,000-video dataset of stroke patients performing the Box and Block Test, in raw frames and 2D skeletons, and reports that the best model, R3D, reaches 87.68% accuracy.

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 ↗

A new 1,000-video benchmark of stroke patients performing box-and-block sub-actions, with raw frames and 2D skeletons, establishes baseline action classification accuracy for seven models.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection Genuinely new stroke BBT dataset, but not yet a usable benchmark: count mismatch, no data release, no label or pose validation. the 4 major comments →

arxiv 2509.07994 v1 pith:4GPR5XE7 submitted 2025-09-02 eess.IV cs.CVcs.LG

STROKEVISION-BENCH: A Multimodal Video And 2D Pose Benchmark For Tracking Stroke Recovery

David Robinson , Animesh Gupta , Rizwan Quershi , Qiushi Fu , Mubarak Shah This is my paper
classification eess.IV cs.CVcs.LG
keywords stroke rehabilitationBox and Block Testaction recognition2D pose estimationclinical video datasetupper extremity assessmentskeleton-based action recognitionbenchmarking
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

StrokeVision-Bench tries to establish that a small, clinically structured video dataset of stroke patients can support objective, computer-vision-based tracking of upper-extremity recovery. It contributes 1,000 one-second videos, each labeled as one of four Box and Block Test sub-actions, with both raw frames and 17-joint 2D skeletons. The benchmark results show that frame-based 3D CNNs classify these sub-actions most accurately (R3D at 87.68%), that skeleton-based models such as MotionBERT come close (84.29%) while preserving privacy, and that the main remaining difficulty is distinguishing 'transport with block' from 'transport without block.' If the dataset and labels hold up, this gives clinicians an affordable, camera-only route toward progress monitoring in rehabilitation.

Core claim

The paper's central claim is that StrokeVision-Bench is the first dedicated dataset of stroke patients performing a clinically structured block-transfer task, and that standard action-recognition models can classify the four sub-actions well enough to be useful for automated rehabilitation assessment. Each of the 1,000 videos is labeled as Grasping, Non-task Movement, Transport with Object, or Transport without Object, and comes in two modalities: raw RGB frames and 2D skeletons extracted with the Sapiens pose estimator. Benchmarking seven models, the authors report that R3D achieves 87.68% accuracy, R2Plus1D 86.96%, and MotionBERT 84.29%, while vision transformers fall to the mid-to-high 70

What carries the argument

The central object is the four-class sub-action taxonomy of the Box and Block Test: Grasping, Non-task Movement, Transport with Object, and Transport without Object. Each video is cut to one second and 30 frames, and is delivered in two modalities: raw RGB frames and Sapiens-derived 2D skeletons with 17 keypoints. The taxonomy turns a standard clinical test into a video classification problem, while the dual modalities let the benchmark compare frame-based and pose-based pipelines. The pre/post-session recording design is intended to capture movement-speed changes that indicate recovery progress, making the dataset a tool for longitudinal monitoring rather than only single-session recognitio

Load-bearing premise

The load-bearing premise is that the manual labels separating the four sub-actions, especially 'transport with block' versus 'transport without block,' and the automatically extracted 2D skeletons are accurate on stroke-impaired movement, so the reported accuracies describe real clinical differences rather than label ambiguity or pose artifacts.

What would settle it

Have two independent clinicians label a held-out subset of the same 1,000 videos into the four sub-actions; if their agreement on 'transport with block' versus 'transport without block' is close to chance, the benchmark accuracies are inflated by label ambiguity. Separately, compare Sapiens 2D keypoints on stroke-impaired arms against manual joint annotations or motion capture; large keypoint error would make the skeleton-based results uninterpretable clinically.

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

If this is right

  • Clinics could use a single camera to automatically tag Box and Block Test sub-actions, since R3D reaches 87.68% accuracy without wearable sensors or specialized equipment.
  • For counting block transfers, frame-based CNNs are the best current option; for joint-level mobility assessment, skeleton-based MotionBERT is nearly as accurate and keeps raw video private.
  • Pretrained vision transformers are a weak fit for small clinical datasets, losing roughly ten points to CNNs and skeleton-based models.
  • The persistent confusion between 'transport with block' and 'transport without block' is the main technical barrier, and will need better modeling, more balanced data, or more refined labels before subtle recovery changes can be tracked reliably.

Where Pith is reading between the lines

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

  • The paper reports no inter-annotator agreement, so a natural next step is to measure how consistently human labelers separate the four sub-actions; if clinicians also struggle with the two transport classes, part of the benchmark confusion is label ambiguity rather than model failure.
  • No validation of the Sapiens skeletons on hemiparetic arms is reported; comparing those keypoints against manual annotation or motion capture on stroke-impaired movement would show whether skeleton-based accuracy reflects true motion or pose-estimation bias.
  • A direct extension the authors mention but do not build is regressing the standard Box and Block Test outcome (blocks transferred per minute) or shoulder-abdomen joint angles from the same videos, which would connect the benchmark to established clinical scores.
  • The pre/post-session recordings invite a longitudinal study of the same patients across weeks; the paper describes the dataset as enabling recovery tracking but does not evaluate whether model predictions change in ways that align with clinical improvement.
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 / 5 minor

Summary. The paper introduces StrokeVision-Bench, claimed to be the first dedicated dataset of 1,000 one-second videos of stroke patients performing the Box and Block Test, with each sample available as raw RGB frames and 2D skeleton keypoints extracted by Sapiens. The four action classes are Grasping, Non-task Movement, Transport with Object, and Transport without Object. The authors benchmark four video-based models (R3D, R2Plus1D, Video MViT, Video Swin Transformer) and three skeleton-based models (MotionBERT, PoseConv3D, MS-G3D), reporting that R3D achieves the highest accuracy of 87.68%.

Significance. If the dataset is released and the labels are trustworthy, StrokeVision-Bench addresses a real gap: existing stroke rehabilitation video datasets mostly cover daily activities or mix healthy and stroke-impaired subjects, while this work targets a standardized clinical test with a clinically structured sub-action taxonomy. The multimodal raw-frame + 2D-skeleton design is useful for privacy-preserving and joint-level analysis, and the benchmark numbers provide a reasonable first point of comparison for future methods. The paper makes falsifiable quantitative claims (e.g., R3D 87.68%) on a fixed split, so the empirical part is clear. However, the central artifact is a dataset, and the manuscript currently does not supply release access, annotation reliability evidence, clinical metadata, or pose-validation on the target population. These omissions are load-bearing because the benchmark's interpretability depends on the validity of the labels and the skeleton modality.

major comments (4)
  1. [Section 3.1 and Fig. 2b] The abstract and Section 3.1 state that StrokeVision-Bench contains 1,000 videos, but Fig. 2b sums to 976: training counts 148 + 372 + 100 + 159 = 779 and validation counts 38 + 93 + 26 + 40 = 197, giving 779 + 197 = 976. This discrepancy undermines the central inventory claim. Please correct the number or the figure, and state the exact split sizes.
  2. [Section 3.1 and Section 6] No dataset release mechanism is provided: there is no URL, download link, access protocol, or licensing information. For a benchmark paper, the dataset is the contribution, and without a release mechanism the reported results are not reproducible by the community. Also, the paper does not mention IRB/ethics approval or consent procedures for patient video data, which is essential for a clinical dataset.
  3. [Section 3.1 and Fig. 3] The annotation protocol is described only as 'Each video was manually annotated into four action classes.' There is no information about the number of annotators, their clinical background, inter-annotator agreement, or adjudication of disagreements. This is especially concerning because Fig. 3 shows that 'Transport with block' vs. 'Transport without block' is the most confused pair for every model (e.g., R3D confuses them 6.9% each way; Video MViT gets 57.1% on 'without block'). In 1-second clips, the distinction may rest on subtle object presence, and without label-reliability evidence the headline 87.68% cannot be interpreted as clinically meaningful discrimination. Additionally, no clinical metadata (number of patients, sessions, pre/post pairing, severity) is reported, so the 'tracking recovery' claim is not yet supported.
  4. [Section 3.2 and Table 1] The 2D skeleton modality is generated by Sapiens, but the paper does not validate pose accuracy on stroke-impaired upper-limb movements. Hemiparetic motion, compensatory trunk movements, and atypical postures can degrade keypoint estimation, yet no PCK/OKS-style evaluation or visual audit is reported. Moreover, the 17-keypoint skeleton does not include hand/object keypoints, so skeleton-based methods such as MotionBERT (84.29%) can only infer block presence through indirect kinematic cues. Without a pose-quality check on this population, the skeleton-based benchmark results are not yet interpretable as measuring the intended sub-actions. The limitations section should explicitly address this, but currently Section 6 omits it.
minor comments (5)
  1. [Section 1 (contributions)] The claim 'first-ever dedicated dataset' and 'first stroke rehabilitation benchmarking dataset that focuses exclusively on stroke patients' should be supported by a more systematic comparison with prior datasets, especially StrokeRehab [12] and [23], to clarify what exactly is new (BBT sub-actions vs. daily activities).
  2. [Section 3.1] The sentence 'By including recordings before and after the sessions, our dataset captures changes in movement speed' is vague. Are these paired recordings of the same patient? If yes, the patient-level split should be described; if no, the claim about tracking recovery is not directly supported.
  3. [Section 4] Implementation details are minimal: no learning rate, batch size, number of epochs, input resolution, or frame sampling strategy is reported. This makes the baselines hard to reproduce even if the dataset were available.
  4. [References] Reference [14] for MViTv2 is formatted as an arXiv preprint but lacks the arXiv ID; reference [15] cites the Swin Transformer paper and [16] the Video Swin Transformer paper, but in Section 5 the text cites [15] for Video Swin Transformer, which may confuse readers.
  5. [Fig. 3] The confusion matrices show percentages but it is not stated whether rows are normalized per true class; the caption says 'accuracy' but the percentages appear to be row-normalized. Please clarify the normalization in the caption.

Circularity Check

0 steps flagged

No significant circularity: benchmark accuracies are direct empirical measurements on an externally pretrained model suite; no derivation reduces to its inputs.

full rationale

StrokeVision-Bench is a dataset-and-benchmark paper, not a derivational one. The central claims—that the dataset contains 1,000 curated videos of BBT sub-actions with raw frames and 2D skeletons, and that R3D achieves 87.68% accuracy—are empirical statements about collected data and measured model performance. No quantity is fitted to the target outcome and then renamed a prediction; the reported accuracies come from a fixed train/validation split with models initialized from external pretrained weights (Kinetics-400, NTU RGB+D) and fine-tuned on the dataset. The 2D skeletons are produced by Sapiens, an externally published pose estimator, and are used as input features rather than as fitted parameters. Manual annotation defines the ground-truth labels; while the paper does not report inter-annotator agreement or validate labels on this population, that omission concerns reliability and external validity, not circularity. Similarly, ambiguity between 'Transport with block' and 'Transport without block' is an empirical confusion pattern reported in the confusion matrices, not a circular inference. The discrepancy between the stated 1,000 videos and the Fig. 2b sum of 976 is an internal consistency issue, not a reduction of a claimed result to its inputs. The few self-citations (e.g., [9], [26], [27]) are contextual references and are not load-bearing for the dataset or benchmark results. No equation is defined in terms of the target accuracy, and no uniqueness theorem or ansatz is imported from the authors' prior work to force the conclusions. The paper therefore exhibits no significant circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

Nothing is derived; the central claim is a data collection. The listed assumptions are the premises that make the dataset and benchmarks interpretable: label quality, pose quality, clip length, and dataset size consistency. No free parameters are fitted, and no new entities are postulated.

axioms (4)
  • domain assumption The four action classes (Grasping, Non-task Movement, Transport with Object, Transport without Object) are clinically meaningful and consistently annotated.
    The paper relies on manual annotation of sub-actions without reporting annotation protocol, annotator qualifications, or inter-annotator agreement (Sections 3.1, 5).
  • domain assumption Sapiens 2D pose estimation is accurate on stroke-impaired upper extremity movements.
    All skeleton benchmarks depend on Sapiens keypoints for stroke patients with atypical movement patterns; no validation of pose accuracy on this population is provided (Section 3.2).
  • domain assumption A 1-second, 30-frame clip is sufficient to discriminate the four sub-actions.
    The dataset is curated as 30-frame clips of 1 second duration; if sub-actions last longer or shorter than the window, the labels are not representative (Section 3.1).
  • ad hoc to paper The dataset contains exactly 1,000 videos as claimed.
    The text states 1,000 videos but Fig. 2b sums to 976 train+validation samples; the discrepancy is not explained, so the stated total is a loaded premise for the central claim.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of STROKEVISION-BENCH: A Multimodal Video And 2D Pose Benchmark For Tracking Stroke Recovery." pith.science (2026). https://pith.science/paper/4GPR5XE7

@misc{pith2026250907994,
  author       = {Pith},
  title        = {Pith review of: STROKEVISION-BENCH: A Multimodal Video And 2D Pose Benchmark For Tracking Stroke Recovery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4GPR5XE7}},
  note         = {Machine review of arXiv:2509.07994}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Despite advancements in rehabilitation protocols, clinical assessment of upper extremity (UE) function after stroke largely remains subjective, relying heavily on therapist observation and coarse scoring systems. This subjectivity limits the sensitivity of assessments to detect subtle motor improvements, which are critical for personalized rehabilitation planning. Recent progress in computer vision offers promising avenues for enabling objective, quantitative, and scalable assessment of UE motor function. Among standardized tests, the Box and Block Test (BBT) is widely utilized for measuring gross manual dexterity and tracking stroke recovery, providing a structured setting that lends itself well to computational analysis. However, existing datasets targeting stroke rehabilitation primarily focus on daily living activities and often fail to capture clinically structured assessments such as block transfer tasks. Furthermore, many available datasets include a mixture of healthy and stroke-affected individuals, limiting their specificity and clinical utility. To address these critical gaps, we introduce StrokeVision-Bench, the first-ever dedicated dataset of stroke patients performing clinically structured block transfer tasks. StrokeVision-Bench comprises 1,000 annotated videos categorized into four clinically meaningful action classes, with each sample represented in two modalities: raw video frames and 2D skeletal keypoints. We benchmark several state-of-the-art video action recognition and skeleton-based action classification methods to establish performance baselines for this domain and facilitate future research in automated stroke rehabilitation assessment.

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

47 extracted references · 38 canonical work pages · 3 internal anchors

  1. [1]

    STROKEVISION-BENCH: A Multimodal Video And 2D Pose Benchmark For Tracking Stroke Recovery

    INTRODUCTION Stroke is the leading cause of serious chronic physical disabil- ity in the United States, impacting millions of individuals each year [34]. Among stroke survivors, approximately 95% expe- rience upper extremity (UE) dysfunction [7], with 30 to 66% exhibiting a significantly impaired ability to use the affected arm [35]. Standardized routine ...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  2. [2]

    RELATED WORK Video-based action recognition has shown remarkable success in modeling complex human activities in everyday environ- ments [18, 19, 27]. Translating these techniques to medical domains, particularly stroke rehabilitation, demands curated datasets and models capable of handling clinically meaningful movements and patient variability. Expensiv...

    work page

  3. [3]

    Curation of StrokeVision We collected videos of patients, where each patient is perform- ing the task of moving a block from one box to another

    STROKEVISION BENCHMARK 3.1. Curation of StrokeVision We collected videos of patients, where each patient is perform- ing the task of moving a block from one box to another. We curated videos recorded both pre- and post-session, enabling evaluation of the network’s ability to understand motion dif- ferences. Ideally, a patient demonstrates improved mobilit...

    work page

  4. [4]

    EXPERIMENTS Models: We examined multiple CNN and transformer-based networks for action classification using raw frames. For CNNs, we evaluated residual-based network architectures introduced in [33], including 3D ResNet (referred to as R3D), and ResNet with (2+1)D convolutions (referred to as R(2+1)D). For vision transformers, we selected Video MViT [14],...

    work page

  5. [5]

    Non-task movement

    BENCHMARKING ON STROKEVISION-BENCH Overall Results: We evaluated four video-based mod- els on raw frames and three skeleton-based methods on StrokeVision-Bench, reporting the results in Table 1. First, convolutional networks outperform vision transformers on our small and challenging dataset. For instance, accuracy falls from 87 % for R3D [ 33] to 74% for...

    work page

  6. [6]

    Non-task movement

    LIMITATIONS While StrokeVision-Bench represents a significant ad- vance in benchmarking action recognition methods for stroke patients, it has several limitations. First, the dataset includes only the Box and Block Test, leaving other clinical assessments of patient mobility unexamined. Second, the current dataset Non task movement Grasping Transport with...

    work page

  7. [7]

    We introduce the first dataset exclusively for stroke-impaired patients per- forming the standard test

    CONCLUSION In this paper, we introduce StrokeVision-Bench, a new dataset of 1000 videos documenting stroke rehabilitation pa- tients performing the Box and Block Test. We introduce the first dataset exclusively for stroke-impaired patients per- forming the standard test. By carefully curating recordings made before and after rehabilitation sessions, we ca...

    work page

  8. [8]

    Robot-measured performance metrics in stroke rehabilitation

    Sivakumar Balasubramanian, Ruihua Wei, Richard Herman, and Jiping He. Robot-measured performance metrics in stroke rehabilitation. In 2009 ICME International Conference on Complex Medical Engineering, pages 1–6. IEEE, 2009. 2

    work page 2009

  9. [9]

    Exter- nal validation of a dynamic prediction model for upper limb function after stroke

    Iris C Brunner, Eleni-Rosalina Andrinopoulou, Ruud Selles, Camilla Biering Lundquist, and Asger Roer Pedersen. Exter- nal validation of a dynamic prediction model for upper limb function after stroke. Archives of Rehabilitation Research and Clinical Translation, 6(1):100315, 2024. 1

    work page 2024

  10. [10]

    Discriminant validity of a new measure of self-efficacy for reaching movements after stroke-induced hemiparesis

    Shuya Chen, Rebecca Lewthwaite, Nicolas Schweighofer, and Carolee J Winstein. Discriminant validity of a new measure of self-efficacy for reaching movements after stroke-induced hemiparesis. Journal of Hand Therapy, 26(2):116–123, 2013. 2

    work page 2013

  11. [11]

    An image is worth 16x16 words: Transformers for image recog- nition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa De- hghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. An image is worth 16x16 words: Transformers for image recog- nition at scale. arXiv preprint arXiv:2010.11929, 2020. 1

    Pith/arXiv arXiv 2010

  12. [12]

    Revisiting skeleton-based action recognition

    Haodong Duan, Yue Zhao, Kai Chen, Dahua Lin, and Bo Dai. Revisiting skeleton-based action recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2969–2978, 2022. 1, 2, 3, 4

    work page 2022

  13. [13]

    Quantitative assessment of limb position sense following stroke

    Sean P Dukelow, Troy M Herter, Kimberly D Moore, Mary Jo Demers, Janice I Glasgow, Stephen D Bagg, Kathleen E Nor- man, and Stephen H Scott. Quantitative assessment of limb position sense following stroke. Neurorehabilitation and neural repair, 24(2):178–187, 2010. 2

    work page 2010

  14. [14]

    Agonist and antagonist activity during voluntary upper-limb movement in patients with stroke

    Carolyn Gowland, Hubert DeBruin, John V Basmajian, Nancy Plews, and Ion Burcea. Agonist and antagonist activity during voluntary upper-limb movement in patients with stroke. Physi- cal therapy, 72:624–624, 1992. 1

    work page 1992

  15. [15]

    Data-Efficient Training of CNNs and Transformers with Coresets: A Stability Perspective

    Animesh Gupta, Irtiza Hasan, Dilip K Prasad, and Deepak K Gupta. Data-efficient training of cnns and transformers with coresets: A stability perspective. arXiv preprint arXiv:2303.02095, 2023. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  16. [16]

    From play to replay: Composed video re- trieval for temporally fine-grained videos

    Animesh Gupta, Jay Parmar, Ishan Rajendrakumar Dave, and Mubarak Shah. From play to replay: Composed video re- trieval for temporally fine-grained videos. arXiv preprint arXiv:2506.05274, 2025. 2

    arXiv 2025

  17. [17]

    Case report of modified box and blocks test with motion capture to measure prosthetic function

    Jacqueline S Hebert and Justin Lewicke. Case report of modified box and blocks test with motion capture to measure prosthetic function. Journal of Rehabilitation Research & Development , 49(8), 2012. 2

    work page 2012

  18. [18]

    Video action understanding

    Matthew S Hutchinson and Vijay N Gadepally. Video action understanding. IEEE Access, 9:134611–134637, 2021. 1

    work page 2021

  19. [19]

    Strokerehab: A benchmark dataset for sub- second action identification

    Aakash Kaku, Kangning Liu, Avinash Parnandi, Haresh Ren- garaj Rajamohan, Kannan Venkataramanan, Anita Venkatesan, Audre Wirtanen, Natasha Pandit, Heidi Schambra, and Carlos Fernandez-Granda. Strokerehab: A benchmark dataset for sub- second action identification. Advances in neural information processing systems, 35:1671–1684, 2022. 3

    work page 2022

  20. [20]

    Sapiens: Foundation for human vision models

    Rawal Khirodkar, Timur Bagautdinov, Julieta Martinez, Su Zhaoen, Austin James, Peter Selednik, Stuart Anderson, and Shunsuke Saito. Sapiens: Foundation for human vision models. arXiv preprint arXiv:2408.12569, 2024. 2, 3

    Pith/arXiv arXiv 2024

  21. [21]

    Mvitv2: Improved multiscale vision trans- formers for classification and detection

    Y Li, CY Wu, H Fan, K Mangalam, B Xiong, J Malik, and C Feichtenhofer. Mvitv2: Improved multiscale vision trans- formers for classification and detection. arxiv. arXiv preprint arXiv:2112.01526, 6(8), 2021. 2, 3, 4

    Pith/arXiv arXiv 2021

  22. [22]

    Swin transformer: Hier- archical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hier- archical vision transformer using shifted windows. In Proceed- ings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 1, 3, 4

    work page 2021

  23. [23]

    Video swin transformer

    Ze Liu, Jia Ning, Yue Cao, Yixuan Wei, Zheng Zhang, Stephen Lin, and Han Hu. Video swin transformer. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3202–3211, 2022. 2, 3, 4

    work page 2022

  24. [24]

    Disentangling and unifying graph convo- lutions for skeleton-based action recognition

    Ziyu Liu, Hongwen Zhang, Zhenghao Chen, Zhiyong Wang, and Wanli Ouyang. Disentangling and unifying graph convo- lutions for skeleton-based action recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 143–152, 2020. 1, 2, 3, 4

    work page 2020

  25. [25]

    Predictive stroke risk model with vision transformer-based doppler features

    Chung-Ming Lo and Peng-Hsiang Hung. Predictive stroke risk model with vision transformer-based doppler features. Medical Physics, 51(1):126–138, 2024. 2

    work page 2024

  26. [26]

    VURF: A General-purpose Reasoning and Self-refinement Framework for Video Understanding

    Ahmad Mahmood, Ashmal Vayani, Muzammal Naseer, Salman Khan, and Fahad Shahbaz Khan. Vurf: A general-purpose rea- soning and self-refinement framework for video understanding. arXiv preprint arXiv:2403.14743, 2024. 2

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  27. [27]

    Machine learning in action: stroke diagnosis and outcome pre- diction

    Shraddha Mainali, Marin E Darsie, and Keaton S Smetana. Machine learning in action: stroke diagnosis and outcome pre- diction. Frontiers in neurology, 12:734345, 2021. 1

    work page 2021

  28. [28]

    Sb-bench: Stereotype bias benchmark for large multimodal models

    Vishal Narnaware, Ashmal Vayani, Rohit Gupta, Sirnam Swetha, and Mubarak Shah. Sb-bench: Stereotype bias benchmark for large multimodal models. arXiv preprint arXiv:2502.08779, 2025. 1

    arXiv 2025

  29. [29]

    Wearable sensors improve prediction of post-stroke walking function following inpatient rehabilita- tion

    Megan K O’Brien, Sung Y Shin, Rushmin Khazanchi, Michael Fanton, Richard L Lieber, Roozbeh Ghaffari, John A Rogers, and Arun Jayaraman. Wearable sensors improve prediction of post-stroke walking function following inpatient rehabilita- tion. IEEE Journal of Translational Engineering in Health and Medicine, 10:1–11, 2022. 1

    work page 2022

  30. [30]

    Data-driven quantitation of movement abnormality after stroke

    Avinash Parnandi, Aakash Kaku, Anita Venkatesan, Natasha Pandit, Emily Fokas, Boyang Yu, Grace Kim, Dawn Nilsen, Carlos Fernandez-Granda, and Heidi Schambra. Data-driven quantitation of movement abnormality after stroke. Bioengi- neering, 10(6):648, 2023. 3

    work page 2023

  31. [31]

    Automatic differentiation in pytorch

    Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan, Edward Yang, Zachary DeVito, Zeming Lin, Alban Desmaison, Luca Antiga, and Adam Lerer. Automatic differentiation in pytorch. 2017. 4

    work page 2017

  32. [32]

    Outcome measures in neurological physical therapy practice: part i

    Kirsten Potter, George D Fulk, Yasser Salem, and Jane Sullivan. Outcome measures in neurological physical therapy practice: part i. making sound decisions. Journal of Neurologic Physical Therapy, 35(2):57–64, 2011. 1

    work page 2011

  33. [33]

    Thinking beyond tokens: From brain-inspired intelligence to cognitive foundations for artificial general intelligence and its societal impact

    Rizwan Qureshi, Ranjan Sapkota, Abbas Shah, Amgad Muneer, Anas Zafar, Ashmal Vayani, Maged Shoman, Abdelrahman Eldaly, Kai Zhang, Ferhat Sadak, et al. Thinking beyond tokens: From brain-inspired intelligence to cognitive foundations for artificial general intelligence and its societal impact. arXiv preprint arXiv:2507.00951, 2025. 1

    Pith/arXiv arXiv 2025

  34. [34]

    Who is responsible? the data, models, users or regulations? responsible generative ai for a sustainable future

    Shaina Raza, Rizwan Qureshi, Anam Zahid, Joseph Fioresi, Ferhat Sadak, Muhammad Saeed, Ranjan Sapkota, Aditya Jain, Anas Zafar, Muneeb Ul Hassan, et al. Who is responsible? the data, models, users or regulations? responsible generative ai for a sustainable future. arXiv preprint arXiv:2502.08650, 2025. 1, 2

    arXiv 2025

  35. [35]

    Beyond con- tent: How grammatical gender shapes visual representation in text-to-image models

    Muhammed Saeed, Shaina Raza, Ashmal Vayani, Muhammad Abdul-Mageed, Ali Emami, and Shady Shehata. Beyond con- tent: How grammatical gender shapes visual representation in text-to-image models. arXiv preprint arXiv:2508.03199, 2025. 3

    work page arXiv 2025

  36. [36]

    Prediction tools for stroke rehabilitation

    Cathy M Stinear, Marie-Claire Smith, and Winston D Byblow. Prediction tools for stroke rehabilitation. Stroke, 50(11):3314– 3322, 2019. 1

    work page 2019

  37. [37]

    Outcome measures in neurological physical therapy practice: part ii

    Jane E Sullivan, A Williams Andrews, Desiree Lanzino, Aimee Peron, and Kirsten A Potter. Outcome measures in neurological physical therapy practice: part ii. a patient-centered process. Journal of Neurologic Physical Therapy, 35(2):65–74, 2011. 1

    work page 2011

  38. [38]

    Mobillama: Towards accurate and lightweight fully transparent gpt

    Omkar Thawakar, Ashmal Vayani, Salman Khan, Hisham Cholakal, Rao M Anwer, Michael Felsberg, Tim Baldwin, Eric P Xing, and Fahad Shahbaz Khan. Mobillama: Towards accurate and lightweight fully transparent gpt. arXiv preprint arXiv:2402.16840, 2024. 1

    Pith/arXiv arXiv 2024

  39. [39]

    Training data-efficient image transformers & distillation through attention

    Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Herve Jegou. Training data-efficient image transformers & distillation through attention. In International Conference on Machine Learning , volume 139, pages 10347–10357, July 2021. 4

    work page 2021

  40. [40]

    A closer look at spatiotemporal convolutions for action recognition

    Du Tran, Heng Wang, Lorenzo Torresani, Jamie Ray, Yann LeCun, and Manohar Paluri. A closer look at spatiotemporal convolutions for action recognition. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 6450–6459, 2018. 2, 3, 4

    work page 2018

  41. [41]

    Heart disease and stroke statistics–2017 update

    AS Update. Heart disease and stroke statistics–2017 update. Circulation, 135:e146–e603, 2017. 1

    work page 2017

  42. [42]

    Forced use of the upper extremity in chronic stroke patients: results from a single-blind randomized clinical trial

    Johanna H Van der Lee, Robert C Wagenaar, Gustaaf J Lankhorst, Tanneke W V ogelaar, Walter L Devillé, and Lex M Bouter. Forced use of the upper extremity in chronic stroke patients: results from a single-blind randomized clinical trial. Stroke, 30(11):2369–2375, 1999. 1

    work page 1999

  43. [43]

    All languages matter: Evaluating lmms on culturally diverse 100 languages

    Ashmal Vayani, Dinura Dissanayake, Hasindri Watawana, Noor Ahsan, Nevasini Sasikumar, Omkar Thawakar, Henok Biadglign Ademtew, Yahya Hmaiti, Amandeep Kumar, Kartik Kuckreja, et al. All languages matter: Evaluating lmms on culturally diverse 100 languages. arXiv preprint arXiv:2411.16508, 2024. 1

    Pith/arXiv arXiv 2024

  44. [44]

    All languages matter: Evaluating lmms on culturally diverse 100 languages

    Ashmal Vayani, Dinura Dissanayake, Hasindri Watawana, Noor Ahsan, Nevasini Sasikumar, Omkar Thawakar, Henok Biadglign Ademtew, Yahya Hmaiti, Amandeep Kumar, Kartik Kukreja, et al. All languages matter: Evaluating lmms on culturally diverse 100 languages. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 19565–19575, 2025. 2

    work page 2025

  45. [45]

    Auto- mated administration of the wolf motor function test for post- stroke assessment

    Eric Wade, Avinash Rao Parnandi, and Maja J Mataric. Auto- mated administration of the wolf motor function test for post- stroke assessment. In 2010 4th International Conference on Pervasive Computing Technologies for Healthcare, pages 1–7. IEEE, 2010. 2

    work page 2010

  46. [46]

    Virtual reality and robotics for stroke rehabilitation: where do we go from here? Topics in stroke rehabilitation, 18(6):685–700, 2011

    Eric Wade and Carolee J Winstein. Virtual reality and robotics for stroke rehabilitation: where do we go from here? Topics in stroke rehabilitation, 18(6):685–700, 2011. 2

    work page 2011

  47. [47]

    Motionbert: A unified perspective on learning human motion representations

    Wentao Zhu, Xiaoxuan Ma, Zhaoyang Liu, Libin Liu, Wayne Wu, and Yizhou Wang. Motionbert: A unified perspective on learning human motion representations. In Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023. 1, 2, 3, 4

    work page 2023

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.