pith. sign in

arxiv: 2606.14096 · v2 · pith:KH64GHG6new · submitted 2026-06-12 · 💻 cs.CV

A New Multi-Domain Benchmark for Micro-Action Recognition and Detection

Yanbin Hao , Pengyu Liu , Xing Wei , Xun Yang , Dan Guo , Meng Wang This is my paper

Pith reviewed 2026-06-27 05:14 UTC · model grok-4.3

classification 💻 cs.CV
keywords micro-action recognitionbenchmark datasetmulti-domain evaluationaction detectionemotion recognitionhuman behavior analysiscross-domain transfer
0
0 comments X

The pith

MMA-82 expands micro-action recognition to 82 categories across four real domains and links them to emotional states.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces MMA-82 to move micro-action analysis from limited lab settings to more varied real-world conditions. It grows an earlier collection to 82 fine-grained categories with 77,856 annotations drawn from laboratory interviews, street interviews, psychiatric interviews, and television footage. The authors set up recognition and multi-label detection tasks that include cross-domain, few-shot, and zero-shot tests. Experiments reveal that current models still have trouble with domain changes, uneven category sizes, and precise timing. They also report that micro-actions track emotional states and add information beyond facial micro-expressions.

Core claim

We introduce MMA-82, a large-scale multi-domain benchmark extending prior work with 82 fine-grained micro-action categories and 77,856 annotated instances from 454 subjects across laboratory interviews, street interviews, psychiatric patient interviews, and emotion-rich television videos. We establish Micro-Action Recognition and Multi-label Micro-Action Detection tasks, with protocols for in-domain, cross-domain, few-shot, and zero-shot evaluation. Experiments indicate that existing approaches still face difficulties with realistic micro-action understanding, particularly under domain shift, long-tailed distributions, and complex temporal localization. Additionally, micro-actions show stron

What carries the argument

The MMA-82 benchmark dataset, which supplies the expanded label space, four-domain coverage, and evaluation protocols that support tests of robustness and generalization.

If this is right

  • Recognition systems must improve handling of domain shifts and long-tailed category distributions to work in practical settings.
  • Micro-action cues can be combined with facial signals to raise accuracy in automated emotion recognition.
  • Few-shot and zero-shot protocols show the need for stronger transfer methods when moving between interview and video domains.
  • Multi-label detection requires better temporal localization techniques for overlapping or brief actions.

Where Pith is reading between the lines

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

  • The dataset could support new monitoring tools that track subtle behavioral changes in mental-health or security contexts.
  • Similar multi-domain designs may be useful for other fine-grained human-movement tasks such as micro-gestures.
  • Integrating the annotations with existing facial-expression datasets could produce joint models that capture both body and face signals.
  • Researchers could measure whether pre-training on MMA-82 improves performance on related video-understanding benchmarks.

Load-bearing premise

The four selected domains and the 77,856 annotations together represent the full variety of real-world micro-actions without large labeling errors or selection bias.

What would settle it

A controlled test in which leading models reach high accuracy on cross-domain recognition without extra adaptation, or an independent check showing no reliable statistical link between the micro-action labels and measured emotional states.

Figures

Figures reproduced from arXiv: 2606.14096 by Dan Guo, Meng Wang, Pengyu Liu, Xing Wei, Xun Yang, Yanbin Hao.

Figure 1
Figure 1. Figure 1: Overview of the proposed MMA-82 benchmark. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: MMA-82 comprises 7 Body-level and 82 Action-level micro-actions, covering the majority of common micro-action categories. collection process. Our goal is to capture micro-actions across different environments, subjects, and recording conditions, and to improve both category diversity and real-world realism. To this end, we combine multiple collection strategies and carefully construct several complementary… view at source ↗
Figure 3
Figure 3. Figure 3: Comprehensive statistics of the MMA-82 from multiple perspectives. (a) and (b) show the statistical information of MMA-82-Rec and MMA-82-Det [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example video clips and annotations for the MMA-82-Rec dataset. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Example video clips and annotations from the MMA-82-Det dataset. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Sankey-style visualization of the Top-5 micro-actions associated with [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Error analysis for the MAD task. 1) False Negative Profiling: As shown in [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Video clip examples from the Emotion-rich television video collection. [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Decision-tree-based visualization of emotion discrimination and the Top-5 micro-actions most strongly associated with each emotion category. For each [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
read the original abstract

Micro-actions are short-duration, low-amplitude subtle body movements at the whole-body level that can reveal latent intentions, involuntary reactions, and fine-grained affective changes. Our previous MA-52 benchmark has provided an important foundation for micro-action recognition, but it remains limited in scale, scene diversity, task coverage, and evaluation protocols. To advance micro-action analysis toward more realistic and comprehensive settings, we introduce MMA-82, a large-scale multi-domain extension of MA-52. MMA-82 expands the label space from 52 to 82 fine-grained micro-action categories and covers four distinct domains, including laboratory interviews, street interviews, psychiatric patient interviews, and emotion-rich television videos, resulting in 77,856 annotated instances from 454 subjects. Built upon MMA-82, we establish two core tasks: Micro-Action Recognition and Multi-label Micro-Action Detection. For recognition, we further define in-domain and cross-domain protocols, including few-shot and zero-shot settings, to evaluate model robustness, transferability, and generalization. Extensive experiments show that current methods still struggle with realistic micro-action understanding, especially under domain shift, long-tailed category distributions, and complex temporal localization. Beyond benchmarking, we investigate the relationship between micro-actions and emotion, showing that micro-actions are strongly associated with emotional states and provide complementary cues to facial micro-expressions for improved emotion recognition. These results demonstrate that MMA-82 serves as a comprehensive and challenging benchmark for realistic micro-action analysis and a valuable resource for human-centered AI. MMA-82 is available at https://lpynow.github.io/MMA-82-AIM/.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper introduces MMA-82, a multi-domain extension of the prior MA-52 benchmark for micro-action analysis. It expands the label space to 82 fine-grained categories with 77,856 annotated instances from 454 subjects across four domains (laboratory interviews, street interviews, psychiatric patient interviews, and emotion-rich TV videos). The work defines micro-action recognition (with in-domain, cross-domain, few-shot, and zero-shot protocols) and multi-label detection tasks, reports that existing methods struggle under domain shift and long-tailed distributions, and presents evidence that micro-actions correlate with emotional states and complement facial micro-expressions.

Significance. If annotation quality is validated, MMA-82 would provide a substantially larger and more diverse resource than MA-52 for studying subtle whole-body movements, enabling more realistic evaluation of recognition, detection, and transfer under domain shift. The emotion-association analysis could open avenues for affective computing that integrate body cues beyond faces. The public release of the dataset supports reproducibility in human-centered AI research.

major comments (1)
  1. [Dataset construction] Dataset construction section: No inter-annotator agreement statistics (e.g., Cohen's or Fleiss' kappa) or detailed annotation protocol (including guidelines for distinguishing the 82 subtle, low-amplitude categories) are reported. Because the central claims rest on these 77,856 instances serving as reliable ground truth for all recognition, detection, cross-domain, and emotion-correlation experiments, the absence of such validation leaves open the possibility of substantial label noise or bias, directly affecting the reported performance gaps and complementarity findings.
minor comments (1)
  1. The abstract and introduction could clarify the total video duration and number of source videos to better contextualize the scale of 77,856 instances relative to prior benchmarks.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive assessment of MMA-82's potential contribution and for the constructive major comment. We address the point on dataset validation below and will incorporate the requested details in the revised manuscript.

read point-by-point responses

  1. Referee: [Dataset construction] Dataset construction section: No inter-annotator agreement statistics (e.g., Cohen's or Fleiss' kappa) or detailed annotation protocol (including guidelines for distinguishing the 82 subtle, low-amplitude categories) are reported. Because the central claims rest on these 77,856 instances serving as reliable ground truth for all recognition, detection, cross-domain, and emotion-correlation experiments, the absence of such validation leaves open the possibility of substantial label noise or bias, directly affecting the reported performance gaps and complementarity findings.

    Authors: We agree that explicit reporting of annotation quality is essential. The current manuscript focuses on benchmark construction and experimental protocols but omits these details. In the revision we will add a dedicated subsection under Dataset Construction that (1) describes the multi-stage annotation pipeline, including the guidelines used to differentiate the 82 low-amplitude categories, (2) reports inter-annotator agreement (Fleiss' kappa) computed on a randomly sampled subset of videos annotated by multiple independent annotators, and (3) discusses quality-control steps such as adjudication of disagreements. These additions will directly support the reliability of the ground-truth labels used throughout the experiments. revision: yes

Circularity Check

0 steps flagged

No circularity: dataset benchmark paper with no derivations or self-referential predictions

full rationale

This is a data contribution paper introducing the MMA-82 benchmark. No mathematical derivations, equations, fitted parameters, or predictions are present that could reduce to inputs by construction. The central claims rest on the new dataset collection and standard evaluations of existing models; there are no self-citation load-bearing uniqueness theorems, ansatzes smuggled via citation, or renamings of known results as novel derivations. Annotation quality and domain representativeness are assumptions but not derived quantities. Score 0 is the appropriate finding for a self-contained benchmark release.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper contributes a new annotated dataset rather than a derivation or model; the ledger therefore records the background assumptions required for any large-scale human action annotation effort.

axioms (2)
  • domain assumption Human annotators can reliably identify and label 82 fine-grained micro-action categories from video across multiple domains.
    The benchmark construction depends on the accuracy and consistency of manual labeling of subtle movements.
  • domain assumption The selected four domains capture representative real-world variability in micro-action appearance and context.
    Cross-domain protocols and generalization claims rest on this coverage assumption.

pith-pipeline@v0.9.1-grok · 5830 in / 1471 out tokens · 26500 ms · 2026-06-27T05:14:59.170096+00:00 · methodology

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

48 extracted references · 4 canonical work pages · 1 internal anchor

  1. [1]

    UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild

    K. Soomro, A. R. Zamir, and M. Shah, “Ucf101: A dataset of 101 human actions classes from videos in the wild,”arXiv preprint arXiv:1212.0402, 2012

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  2. [2]

    Quo vadis, action recognition? a new model and the kinetics dataset,

    J. Carreira and A. Zisserman, “Quo vadis, action recognition? a new model and the kinetics dataset,” inproceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 6299–6308

    2017

  3. [3]

    The” something something

    R. Goyal, S. Ebrahimi Kahou, V . Michalski, J. Materzynska, S. West- phal, H. Kim, V . Haenel, I. Fruend, P. Yianilos, M. Mueller-Freitaget al., “The” something something” video database for learning and evaluat- ing visual common sense,” inProceedings of the IEEE international conference on computer vision, 2017, pp. 5842–5850

    2017

  4. [4]

    Slowfast networks for video recognition,

    C. Feichtenhofer, H. Fan, J. Malik, and K. He, “Slowfast networks for video recognition,” inProceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 6202–6211

    2019

  5. [5]

    Video swin transformer,

    Z. Liu, J. Ning, Y . Cao, Y . Wei, Z. Zhang, S. Lin, and H. Hu, “Video swin transformer,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 3202–3211

    2022

  6. [6]

    Gpt4ego: unleashing the potential of pre-trained models for zero-shot egocentric action recognition,

    G. Dai, X. Shu, W. Wu, R. Yan, and J. Zhang, “Gpt4ego: unleashing the potential of pre-trained models for zero-shot egocentric action recognition,”IEEE Transactions on Multimedia, vol. 27, pp. 401–413, 2024

    2024

  7. [7]

    Facial action units as a joint dataset training bridge for facial expression recognition,

    S. Mao, X. Li, F. Zhang, X. Peng, and Y . Yang, “Facial action units as a joint dataset training bridge for facial expression recognition,”IEEE Transactions on Multimedia, vol. 27, pp. 3331–3342, 2025

    2025

  8. [8]

    Benchmarking micro-action recognition: Dataset, methods, and applications,

    D. Guo, K. Li, B. Hu, Y . Zhang, and M. Wang, “Benchmarking micro-action recognition: Dataset, methods, and applications,”IEEE Transactions on Circuits and Systems for Video Technology, vol. 34, no. 7, pp. 6238–6252, 2024

    2024

  9. [9]

    Mmad: Multi-label micro-action detection in videos,

    K. Li, P. Liu, D. Guo, F. Wang, Z. Wu, H. Fan, and M. Wang, “Mmad: Multi-label micro-action detection in videos,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2025, pp. 13 225–13 236

    2025

  10. [10]

    Ma- bench: Towards fine-grained micro-action understanding,

    K. Li, J. Gu, F. Wang, Z. Wu, H. Fan, and D. Guo, “Ma- bench: Towards fine-grained micro-action understanding,”arXiv preprint arXiv:2603.26586, 2026

    work page arXiv 2026

  11. [11]

    Motion matters: Motion-guided modulation network for skeleton-based micro- action recognition,

    J. Gu, K. Li, F. Wang, Y . Wei, Z. Wu, H. Fan, and M. Wang, “Motion matters: Motion-guided modulation network for skeleton-based micro- action recognition,” inProceedings of the 33rd ACM International Conference on Multimedia, 2025, pp. 5461–5470

    2025

  12. [12]

    Prototypical calibrating ambiguous samples for micro-action recogni- tion,

    K. Li, D. Guo, G. Chen, C. Fan, J. Xu, Z. Wu, H. Fan, and M. Wang, “Prototypical calibrating ambiguous samples for micro-action recogni- tion,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 5, 2025, pp. 4815–4823

    2025

  13. [13]

    imigue: An identity-free video dataset for micro-gesture understanding and emotion analysis,

    X. Liu, H. Shi, H. Chen, Z. Yu, X. Li, and G. Zhao, “imigue: An identity-free video dataset for micro-gesture understanding and emotion analysis,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 10 631–10 642

    2021

  14. [14]

    Smg: A micro-gesture dataset towards spontaneous body gestures for emotional stress state analysis,

    H. Chen, H. Shi, X. Liu, X. Li, and G. Zhao, “Smg: A micro-gesture dataset towards spontaneous body gestures for emotional stress state analysis,”International Journal of Computer Vision, vol. 131, no. 6, pp. 1346–1366, 2023

    2023

  15. [15]

    Context-aware emotion recognition networks,

    J. Lee, S. Kim, S. Kim, J. Park, and K. Sohn, “Context-aware emotion recognition networks,” inProceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 10 143–10 152

    2019

  16. [16]

    Samm: A spontaneous micro-facial movement dataset,

    A. K. Davison, C. Lansley, N. Costen, K. Tan, and M. H. Yap, “Samm: A spontaneous micro-facial movement dataset,”IEEE transactions on affective computing, vol. 9, no. 1, pp. 116–129, 2016

    2016

  17. [17]

    A spontaneous micro-expression database: Inducement, collection and baseline,

    X. Li, T. Pfister, X. Huang, G. Zhao, and M. Pietik ¨ainen, “A spontaneous micro-expression database: Inducement, collection and baseline,” in 2013 10th IEEE International Conference and Workshops on Automatic face and gesture recognition (fg). IEEE, 2013, pp. 1–6

    2013

  18. [18]

    Casme database: A dataset of spontaneous micro-expressions collected from neutralized faces,

    W.-J. Yan, Q. Wu, Y .-J. Liu, S.-J. Wang, and X. Fu, “Casme database: A dataset of spontaneous micro-expressions collected from neutralized faces,” in2013 10th IEEE international conference and workshops on automatic face and gesture recognition (FG). IEEE, 2013, pp. 1–7

    2013

  19. [19]

    Cas (me) 2: a database for spontaneous macro-expression and micro-expression spotting and recognition,

    F. Qu, S.-J. Wang, W.-J. Yan, H. Li, S. Wu, and X. Fu, “Cas (me) 2: a database for spontaneous macro-expression and micro-expression spotting and recognition,”IEEE Transactions on Affective Computing, vol. 9, no. 4, pp. 424–436, 2017

    2017

  20. [20]

    Cas (me) 3: A third generation facial spontaneous micro- expression database with depth information and high ecological validity,

    J. Li, Z. Dong, S. Lu, S.-J. Wang, W.-J. Yan, Y . Ma, Y . Liu, C. Huang, and X. Fu, “Cas (me) 3: A third generation facial spontaneous micro- expression database with depth information and high ecological validity,” IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 3, pp. 2782–2800, 2022

    2022

  21. [21]

    Ntu rgb+ d: A large scale dataset for 3d human activity analysis,

    A. Shahroudy, J. Liu, T.-T. Ng, and G. Wang, “Ntu rgb+ d: A large scale dataset for 3d human activity analysis,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 1010–1019

    2016

  22. [22]

    Ntu rgb+ d 120: A large-scale benchmark for 3d human activity understanding,

    J. Liu, A. Shahroudy, M. Perez, G. Wang, L.-Y . Duan, and A. C. Kot, “Ntu rgb+ d 120: A large-scale benchmark for 3d human activity understanding,”IEEE transactions on pattern analysis and machine intelligence, vol. 42, no. 10, pp. 2684–2701, 2019

    2019

  23. [23]

    The epic-kitchens dataset: Collection, challenges and baselines,

    D. Damen, H. Doughty, G. M. Farinella, S. Fidler, A. Furnari, E. Kaza- kos, D. Moltisanti, J. Munro, T. Perrett, W. Price, and M. Wray, “The epic-kitchens dataset: Collection, challenges and baselines,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 11, pp. 4125–4141, 2021

    2021

  24. [24]

    Ego4d: Around the world in 3,000 hours of egocentric video,

    K. Grauman, A. Westbury, E. Byrne, Z. Chavis, A. Furnari, R. Girdhar, J. Hamburger, H. Jiang, M. Liu, X. Liuet al., “Ego4d: Around the world in 3,000 hours of egocentric video,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 18 995–19 012

    2022

  25. [25]

    To- wards student actions in classroom scenes: New dataset and baseline,

    Z. Tan, C. Gao, A. Qin, R. Chen, T. Song, F. Yang, and D. Meng, “To- wards student actions in classroom scenes: New dataset and baseline,” IEEE Transactions on Multimedia, 2025

    2025

  26. [26]

    Casme ii: An improved spontaneous micro-expression database and the baseline evaluation,

    W.-J. Yan, X. Li, S.-J. Wang, G. Zhao, Y .-J. Liu, Y .-H. Chen, and X. Fu, “Casme ii: An improved spontaneous micro-expression database and the baseline evaluation,”PloS one, vol. 9, no. 1, p. e86041, 2014

    2014

  27. [27]

    Opendatalab: Empowering general artificial intelligence with open datasets,

    C. He, W. Li, Z. Jin, C. Xu, B. Wang, and D. Lin, “Opendatalab: Empowering general artificial intelligence with open datasets,”arXiv preprint arXiv:2407.13773, 2024

    work page arXiv 2024

  28. [28]

    The epic- kitchens dataset: Collection, challenges and baselines,

    D. Damen, H. Doughty, G. M. Farinella, S. Fidler, A. Furnari, E. Kaza- kos, D. Moltisanti, J. Munro, T. Perrett, W. Priceet al., “The epic- kitchens dataset: Collection, challenges and baselines,”IEEE Transac- tions on Pattern Analysis and Machine Intelligence, vol. 43, no. 11, pp. 4125–4141, 2020

    2020

  29. [29]

    Enhancing micro-video understanding by harnessing external sounds,

    L. Nie, X. Wang, J. Zhang, X. He, H. Zhang, R. Hong, and Q. Tian, “Enhancing micro-video understanding by harnessing external sounds,” inProceedings of the 25th ACM international conference on Multimedia, 2017, pp. 1192–1200

    2017

  30. [30]

    Spatial temporal graph convolutional networks for skeleton-based action recognition,

    S. Yan, Y . Xiong, and D. Lin, “Spatial temporal graph convolutional networks for skeleton-based action recognition,” inProceedings of the AAAI conference on artificial intelligence, vol. 32, no. 1, 2018

    2018

  31. [31]

    Revisiting skeleton- based action recognition,

    H. Duan, Y . Zhao, K. Chen, D. Lin, and B. Dai, “Revisiting skeleton- based action recognition,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 2969–2978

    2022

  32. [32]

    Pyskl: Towards good practices for skeleton action recognition,

    H. Duan, J. Wang, K. Chen, and D. Lin, “Pyskl: Towards good practices for skeleton action recognition,” inProceedings of the 30th ACM International Conference on Multimedia, 2022, pp. 7351–7354

    2022

  33. [33]

    Group contextualization for video recognition,

    Y . Hao, H. Zhang, C.-W. Ngo, and X. He, “Group contextualization for video recognition,” inProceedings of the ieee/cvf conference on computer vision and pattern recognition, 2022, pp. 928–938

    2022

  34. [34]

    Tsm: Temporal shift module for efficient video understanding,

    J. Lin, C. Gan, and S. Han, “Tsm: Temporal shift module for efficient video understanding,” inProceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 7083–7093

    2019

  35. [35]

    Pointtad: Multi- label temporal action detection with learnable query points,

    J. Tan, X. Zhao, X. Shi, B. Kang, and L. Wang, “Pointtad: Multi- label temporal action detection with learnable query points,”Advances in Neural Information Processing Systems, vol. 35, pp. 15 268–15 280, 2022

    2022

  36. [36]

    End-to-end temporal action detection with 1b parameters across 1000 frames,

    S. Liu, C.-L. Zhang, C. Zhao, and B. Ghanem, “End-to-end temporal action detection with 1b parameters across 1000 frames,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2024, pp. 18 591–18 601

    2024

  37. [37]

    Videomae: Masked autoen- coders are data-efficient learners for self-supervised video pre-training,

    Z. Tong, Y . Song, J. Wang, and L. Wang, “Videomae: Masked autoen- coders are data-efficient learners for self-supervised video pre-training,” Advances in neural information processing systems, vol. 35, pp. 10 078– 10 093, 2022

    2022

  38. [38]

    The mastery of movement

    R. Laban and L. Ullmann, “The mastery of movement.” 1971

    1971

  39. [39]

    Emotion regulation through movement: unique sets of movement characteristics are associated with and enhance basic emotions,

    T. Shafir, R. P. Tsachor, and K. B. Welch, “Emotion regulation through movement: unique sets of movement characteristics are associated with and enhance basic emotions,”Frontiers in psychology, vol. 6, p. 2030, 2016

    2030

  40. [40]

    How do we recognize emotion from movement? specific motor components contribute to the recog- nition of each emotion,

    A. Melzer, T. Shafir, and R. P. Tsachor, “How do we recognize emotion from movement? specific motor components contribute to the recog- nition of each emotion,”Frontiers in psychology, vol. 10, p. 392097, 2019

    2019

  41. [41]

    Boosted lightface: A hybrid dnn and gbm model for boosted facial recognition,

    S. I. Serengil and A. Ozpinar, “Boosted lightface: A hybrid dnn and gbm model for boosted facial recognition,”Gazi University Journal of Science, vol. 39, no. 1, pp. 452–466, 2026. [Online]. Available: https://dergipark.org.tr/en/pub/gujs/article/1794891 JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 11

    work page arXiv 2026

  42. [42]

    Qsgd: Communication-efficient sgd via gradient quantization and encoding,

    D. Alistarh, D. Grubic, J. Li, R. Tomioka, and M. V ojnovic, “Qsgd: Communication-efficient sgd via gradient quantization and encoding,” Advances in neural information processing systems, vol. 30, 2017

    2017

  43. [43]

    Actionformer: Localizing moments of actions with transformers,

    C.-L. Zhang, J. Wu, and Y . Li, “Actionformer: Localizing moments of actions with transformers,” inEuropean Conference on Computer Vision. Springer, 2022, pp. 492–510

    2022

  44. [44]

    Diagnosing error in temporal action detectors,

    H. Alwassel, F. C. Heilbron, V . Escorcia, and B. Ghanem, “Diagnosing error in temporal action detectors,” inProceedings of the European Conference on Computer Vision, 2018, pp. 256–272

    2018

  45. [45]

    An information-theoretic perspective of tf–idf measures,

    A. Aizawa, “An information-theoretic perspective of tf–idf measures,” Information Processing & Management, vol. 39, no. 1, pp. 45–65, 2003. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 12 APPENDIXA BASELINEIMPLEMENTATIONDETAILS In this appendix, we provide the implementation details of all baselines evaluated on the MMA-82 benchmark. A. Imple...

    2003

  46. [46]

    The baseline achieves a low FN in most dimensions

    False Negative Profiling:As shown in Figure 7(a), we report the model’s false negatives (FN) across three dimensions: coverage, length, and instances. The baseline achieves a low FN in most dimensions. As coverage and length increase, the FN decreases; however, as the number of instances increases, the FN rises

  47. [47]

    Confusion Error

    False Positive Profiling:As shown in Figure 7(b), in the left figure, we analyze the false positives at tIoU=0.5. These errors fall into five major categories, with “Confusion Error” and “Wrong Label Error” accounting for the majority. Furthermore, in the right figure, we break down the impact of each error type on the average mAP. Removing “Wrong Label E...

  48. [48]

    Retracting and stretching arms

    Sensitivity Profiling:To evaluate the model’s robustness, we analyze the performance sensitivity across different scales, as shown in Figure 7(c). The model performs stably under varying coverage levels, with performance maximizing at 32.1% when coverage is set to M. Performance tends to improve as action duration increases, but decreases as action densit...

    2021