Pith. sign in

REVIEW 5 major objections 3 minor 81 references

Detecting Children with Autism Spectrum Disorder based on Script-Centric Behavior Understanding with Emotional Enhancement

T0 review · 5 major / 3 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read This paper claims an LLM pipeline can detect autism spectrum disorder from text-scripted behavior with an F1 of 95.24%, beating the supervised baseline without training on ASD samples.

desk verdict The pipeline is genuinely new and worth discussing, but the headline F1 is optimistic: the emotion threshold and prompt components were chosen on the evaluation set. read the letter →

arxiv 2411.09413 v2 pith:W6AOFHCF submitted 2024-11-14 cs.CV cs.AI

Wenxing Liu , Yueran Pan , Dong Zhang , Hongzhu Deng , Xiaobing Zou , Ming Li This is my paper · ORCID
classification cs.CVcs.AI
keywords autismspectrumdisorderzero-shotdetectionlargelanguagemodelsbehaviortextualizationemotionaudio-visualanalysispromptengineeringmulti-agent
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that the diagnostic signal for autism spectrum disorder in young children can be written down as a time-stamped behavioral script, and that a general-purpose large language model reading that script can classify the child without ever seeing labeled ASD data. The proposed pipeline converts audio-visual recordings of six structured assessment paradigms into text, adds textual descriptions of emotional dynamics, injects clinical knowledge through prompts, and lets one or several LLMs produce a diagnosis with an explanation. On a clinical dataset of 95 children with an average age of two years, the best multi-LLM version reaches an F1-score of 95.24%, surpassing the supervised behavior-signal baseline of 92.20%. If correct, this means the main barrier to automated ASD screening shifts from scarce labeled training samples to the quality of the script and the prompt.

What carries the argument

The central object is the Script-Centric Behavior Understanding (SCBU) pipeline, which converts audio-visual recordings into text scripts that LLMs read. It has four load-bearing parts: a behavior transcription module that extracts positions, gestures, head poses, gaze, speech, and facial expressions from video; a script transcription module with a hand-coded response parser and textualization table that maps observed events into fixed natural-language descriptions per paradigm; an emotion textualization module that detects moments of large valence change and adds a video-LLM description of the emotional segment; and a domain prompt module that injects DSM-V criteria and clinical experience. The multi-LLM variant, SCBU-Agents, adds a re-analysis, collaborative discussion, and final decision stage to reduce individual model hallucination.

What would settle it

Replace the fixed textual response descriptions with plain, faithful transcripts of the raw behavioral logs, keeping the same language model and prompt; if the detection score does not stay near the reported 95%, the hand-built script mapping, not the language model, carries the diagnosis.

Watch

Extended reading notes

Core claim

The central claim is that a zero-shot LLM framework can detect ASD from audio-visual behavioral data as accurately as, or better than, a supervised classifier that was trained on the same paradigms. The authors report that their Script-Centric Behavior Understanding (SCBU) method, which turns behavioral logs into structured text scripts and enriches them with emotion-related descriptions, reaches an F1-score of 95.24% when several LLMs discuss and jointly decide, and that even single LLMs approach or exceed the supervised baseline. They also report that the LLMs generate interpretable detection rationales that align with DSM-V diagnostic criteria, and that few-shot learning with a small number of script examples continues to improve performance beyond the supervised baseline.

Load-bearing premise

The hand-coded rules that turn raw behavioral logs into fixed textual responses keep all the clinically relevant information; if those rules drop the subtle timing or emotional cues that separate ASD from typical children, the reported accuracy cannot be reproduced no matter how capable the language model is.

Editorial extensions

If this is right

  • A general-purpose LLM can match or exceed a supervised behavior-signal classifier on ASD screening using no ASD-labeled training data, so the diagnostic bottleneck shifts from labeled samples to script design and prompt engineering.
  • Encoding emotional dynamics as text improves detection for most tested LLMs and raises the best multi-LLM system's F1 from 91.78% to 95.24%, indicating that valence-change descriptions carry clinically relevant signal.
  • Few-shot runs on an open-source model keep improving up to 20 script examples and surpass the supervised baseline, suggesting the method degrades gracefully when labeled data are scarce.
  • The same framework produces free-text rationales that align with DSM-V criteria, giving clinicians a reason to trust or audit each prediction.
  • Scripts generated by a generic audio-visual video understanding model instead of the hand-crafted transcriptor collapse specificity to near zero, showing the textualization design, not just LLM size, carries the result.

Reading between the lines

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

  • If the central claim holds, a testable extension would run the same prompts on recordings from other clinics or other paradigm sets to see whether the scripts, not the site-specific recording setup, are what generalizes.
  • Because response latency, response duration, and valence dynamics differ significantly between groups in the raw signals, a direct head-to-head experiment stripping timestamps or emotional descriptions from the scripts would isolate how much diagnostic signal survives the transcription.
  • If the fixed textual response descriptions implicitly encode the clinicians' expectations, the method may be reusing human expertise at the script-construction stage; a further test would generate scripts automatically from raw logs without the fixed response table and compare detection scores.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 3 minor

Summary. The paper proposes a zero-shot ASD detection framework, SCBU, that converts audio-visual behavioral recordings into structured text scripts through a behavioral transcription module and a script transcription module, then augments those scripts with emotion descriptions derived from valence dynamics, and finally queries LLMs with domain-specific prompts to obtain an ASD/TD classification and an interpretable rationale. The method is evaluated on a clinical database of 95 children (71 ASD, 24 TD) under leave-one-out cross-validation. The authors report that SCBU-Agents with emotion descriptions reaches an F1-score of 95.24%, surpassing the supervised baseline of 92.20% from Cheng et al. [15], and they provide additional few-shot results and ablations of domain knowledge, human experience, and emotional descriptions.

Significance. If the reported results are valid, the work would be a meaningful contribution to automated, interpretable ASD screening in low-data clinical settings, and the combination of hand-crafted behavior scripts with LLM reasoning is a plausible direction for affective computing. The paper's strengths include the use of a clinically collected multimodal dataset, explicit comparisons against a supervised baseline, a multi-LLM fusion mechanism, and a public release of behavior scripts, which supports reproducibility. However, the central quantitative claim of a 95.24% zero-shot F1-score is not supported by the current evaluation protocol: the emotion threshold and several prompt components are selected using the same leave-one-out test labels, and no confidence intervals, significance tests, or external validation are provided. Because the headline result is an optimistic point estimate with selection on the evaluation set, the paper needs substantial revision before the claimed performance can be credited.

major comments (5)
  1. [§IV-E, Fig. 11 and Algorithm 1] The emotion threshold is selected using the evaluation set itself. Algorithm 1 specifies a threshold of 0.2, but Section IV-E reports scanning Fig. 11 over thresholds and states that "the emotion threshold for all experiments in this study is set to 0.175." Under the leave-one-out protocol used for Table III, this is a global choice informed by the held-out subjects' labels, so the reported improvements from emotion descriptions, including the headline F1 of 95.24% for SCBU-Agents (w/ emotion), are optimistically biased estimates of the true zero-shot performance. To repair this, the authors must either fix all hyperparameters and prompt choices before seeing test labels, or perform model selection in an inner loop on training folds only and report the resulting unbiased estimate.
  2. [§IV-E, Table VIII and Section III-B2] The same selection-on-test-data pattern applies to the prompt components and the background/gender descriptions. Table VIII ablates domain knowledge, human experience, and emotion description based on leave-one-out accuracy, and the text in Section III-B2 states that "adding background and gender descriptions before the instruction description can improve diagnostic accuracy." These are not described as pre-registered choices or as validated on held-out folds. For a genuine zero-shot claim, every component included in the final pipeline must be justified independently of the test labels, or selected with a nested validation split.
  3. [Table III and Section IV-B] No confidence intervals, variance estimates, or significance tests are reported, although the evaluation is over only 95 subjects (71 ASD, 24 TD) and the LLM temperature is set to 0.7, which makes the outputs stochastic across runs. The differences among the reported F1 values (e.g., 95.24% vs. 92.20% for the baseline, or 94.04% vs. 91.55% for single LLMs) could be within sampling or stochastic noise. The authors should report repeated-run statistics, exact binomial or bootstrap confidence intervals, and a significance test against the supervised baseline.
  4. [§III-B and §IV-A] The "zero-shot" claim is substantially narrower than the framework description suggests. The script transcription module uses hand-coded response descriptions (Table II) and predefined response events that were designed for the same paradigm and dataset as the supervised system [15], so the pipeline encodes substantial clinical prior knowledge before any LLM inference. The paper does not validate that these hand-coded scripts preserve the discriminative information present in the raw audio-visual data or in the features used by [15]; the ablation in Table V only shows that generic video-SALMONN descriptions are much worse, which is not the same as demonstrating that the script mapping is lossless. The authors should clarify what exactly is zero-shot (the LLM's role) and provide an analysis of information retention, for example by comparing the script-derived features against the raw behavioral features from [15].
  5. [§IV-A and §IV-F] All experiments are conducted on a single dataset of 95 participants with an imbalanced class ratio (71 ASD, 24 TD), and no external validation is reported. Given the small sample size and the well-known instability of leave-one-out estimates on small imbalanced data, the authors should either validate on an independent ASD audio-visual dataset, report cross-validated confidence intervals, or explicitly state the generalizability limits of the current evidence. This is not a presentation detail; it is necessary to support the paper's claim that the method provides a reliable assisted-diagnosis tool.
minor comments (3)
  1. [Algorithm 1] The condition "dn > 0.2 or dn < 0.2" appears to be a typo: it should presumably read "dn > 0.2 or dn < -0.2", matching the description in Section III-B3.
  2. [Fig. 2 caption] The caption contains typos: "Script-Centeric" should be "Script-Centric", and "anwser" should be "answer".
  3. [Table III and its discussion in Section IV-B] The text refers to "SCBU-DeepSeekR1-4O (w/ emotion)" while the corresponding table row is for DeepSeekR1-671B; also, one row reads "× w" instead of "× w/o". These inconsistencies should be corrected.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline 95.24% F1 is not a clean zero-shot prediction: the emotion threshold was selected from the evaluation-set metrics, making part of the reported gain a fitted value.

  1. fitted input called prediction [Section IV-E ('Emotion thresholds'), Fig. 11; Algorithm 1 in Section III-B3]
    "Fig 11 presents the evaluation metrics for different emotion thresholds.Thus, the emotion threshold for all experiments in this study is set to 0.175."

    Algorithm 1 originally fixes the emotional dynamic threshold at 0.2, but the final configuration changes it to 0.175 after inspecting Fig. 11, which plots ACC/F1/SN/SP as functions of the threshold on the same leave-one-out evaluation used for the reported results. The paper describes no inner validation loop or held-out selection of this threshold, so under the LOO protocol the threshold choice uses the held-out subjects' labels. The emotion descriptions, and therefore the 95.24% F1 of SCBU-Agents (w/ emotion), are produced with a label-informed parameter; the reported zero-shot F1 is an optimistic fitted estimate rather than a configuration fixed before seeing test labels.

full rationale

The paper is an empirical evaluation rather than a derivation, so the main classification output is not definitionally equivalent to its inputs. The script transcription mapping, hand-written Table II responses, and LLM prompts are not fit to ASD labels and constitute independent content. However, the central quantitative claim is compromised by test-set selection: the emotion dynamic threshold was changed from 0.2 (Algorithm 1) to 0.175 after plotting evaluation metrics over thresholds on the same LOO evaluation (Fig. 11). Since no nested validation is reported, the emotion-enhanced result that reaches 95.24% F1 is partially a fitted result, not a purely zero-shot prediction. The same pattern is visible in the ad hoc adoption of background/gender descriptions ('we found that adding background and gender descriptions ... can improve diagnostic accuracy') and in the ablation-driven selection of prompt components in Table VIII, although those are less precisely quantified. The citations to the authors' prior work [15] are for the dataset and perceptual modules; they are published external support and are not a circular foundation for the LLM evaluation itself. Thus the circularity is partial and localized to evaluation-protocol leakage rather than to the method definition.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claim rests mainly on domain assumptions about the clinical validity of the paradigm-based script representation and on one numerically fitted threshold. No new physical or mathematical entities are introduced; the framework is a composition of existing models and hand-crafted text templates.

free parameters (2)
  • Emotion dynamic threshold alpha = 0.175
    Selected by evaluating LOOCV F1 across candidate thresholds (Fig 11), then fixed for all experiments. Algorithm 1 initially states 0.2, creating an inconsistency.
  • Emotion segment margin = 1 second (0.5s before and after dynamic point)
    Fixed in Algorithm 1 to define the segment around each dynamic point; not tuned, but arbitrary.
assumptions (5)
  • domain assumption The six paradigms and five response events in Table I are valid clinical ASD markers.
    Taken from clinical literature and prior work [15] without independent validation in this paper.
  • domain assumption The hand-coded response descriptions in Table II preserve the discriminative information in the behavioral logs.
    No evidence that the script text retains all cues that separate ASD from TD.
  • domain assumption Valence derivative magnitude captures clinically meaningful emotional dynamics.
    Algorithm 1 uses a threshold on the valence derivative; no comparison to human annotation.
  • domain assumption video-SALMONN's emotional descriptions are accurate for the selected segments.
    Only anecdotal examples in Fig 12; no systematic evaluation or error analysis.
  • domain assumption LLM zero-shot answers are reliable enough for screening when ensembled.
    No hallucination analysis, calibration, or failure case review.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Detecting Children with Autism Spectrum Disorder based on Script-Centric Behavior Understanding with Emotional Enhancement." pith.science (2026). https://pith.science/paper/W6AOFHCF

@misc{pith2026241109413,
  author       = {Pith},
  title        = {Pith review of: Detecting Children with Autism Spectrum Disorder based on Script-Centric Behavior Understanding with Emotional Enhancement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W6AOFHCF}},
  note         = {Machine review of arXiv:2411.09413}
}
read the original abstract

The early diagnosis of autism spectrum disorder (ASD) is critically dependent on systematic observation and analysis of children's social behaviors. While current methodologies predominantly utilize supervised learning approaches, their clinical adoption faces two principal limitations: insufficient ASD diagnostic samples and inadequate interpretability of the detection outcomes. This paper presents a novel zero-shot ASD detection framework based on script-centric behavioral understanding with emotional enhancement, which is designed to overcome the aforementioned clinical constraints. The proposed pipeline automatically converts audio-visual data into structured behavioral text scripts through computer vision techniques, subsequently capitalizing on the generalization capabilities of large language models (LLMs) for zero-shot/few-shot ASD detection. Three core technical contributions are introduced: (1) A multimodal script transcription module transforming behavioral cues into structured textual representations. (2) An emotion textualization module encoding emotional dynamics as the contextual features to augment behavioral understanding. (3) A domain-specific prompt engineering strategy enables the injection of clinical knowledge into LLMs. Our method achieves an F1-score of 95.24\% in diagnosing ASD in children with an average age of two years while generating interpretable detection rationales. This work opens up new avenues for leveraging the power of LLMs in analyzing and understanding ASD-related human behavior, thereby enhancing the accuracy of assisted autism diagnosis.

Figures

Figures reproduced from arXiv: 2411.09413 by the authors.

Figure 1
Figure 1. Pipeline comparing (a) behavioral signal processing [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The overview of our proposed Script-Centeric Behavior Understanding(SCBD) framework. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. A two-stage pipeline in the behavioral transcription module. (1) [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: The textualization process of the Response to Name [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: The emotion textualization process of the Response to Name paradigm. The blue line indicates the valence value. The [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: The details of the script description. The blue part is [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: This is the framework for the joint detection of multiple LLMs. The process is divided into three stages: (1) Re-analysis [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Statistics for ASD and TD groups in the clinical [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: Examples of LLM detection for children with ASD and TD [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: Comparison in the ranges of valence and arousal [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: The impact of different emotional thresholds on [PITH_FULL_IMAGE:figures/full_fig_p012_11.png]
Figure 12
Figure 12. Figure 12: An example description of an emotional dynamic [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

81 extracted references · 63 canonical work pages

  1. [55]

    Hear me, see me, understand me: Audio-visual autism behavior recognition,

    S. Deng et al., “Hear me, see me, understand me: Audio-visual autism behavior recognition,” IEEE Trans. Multimedia , p. 1–12, Dec. 2024, early access, doi: 10.1109/TMM.2024.3521838

    arXiv 2024

  2. [15]

    Computer-aided autism spectrum disorder diagnosis with behavior signal processing,

    M. Cheng et al. , “Computer-aided autism spectrum disorder diagnosis with behavior signal processing,” IEEE Trans. Affect. Comput. , vol. 14, no. 4, pp. 2982–3000, Oct. 2023

    work page 2023

  3. [1]

    Autism spectrum disorder,

    C. Lord, M. Elsabbagh, G. Baird, and J. Veenstra-Vanderweele, “Autism spectrum disorder,” The lancet, vol. 392, no. 10146, pp. 508–520, Aug. 2018

    2018

  4. [2]

    The dsm-5: Classification and criteria changes,

    D. A. Regier, E. A. Kuhl, and D. J. Kupfer, “The dsm-5: Classification and criteria changes,” World psychiatry, vol. 12, no. 2, pp. 92–98, Jun. 2013

    work page 2013

  5. [3]

    The lancet commission on the future of care and clinical research in autism,

    C. Lord et al., “The lancet commission on the future of care and clinical research in autism,” The Lancet, vol. 399, no. 10321, pp. 271–334, Jan. 2022

    work page 2022

  6. [4]

    Overview of meta-analyses on early intensive behavioral intervention for young children with autism spectrum disorders,

    B. Reichow, “Overview of meta-analyses on early intensive behavioral intervention for young children with autism spectrum disorders,” J. Autism Dev. Disord., vol. 42, pp. 512–520, 2012

    work page 2012

  7. [5]

    Autism diagnostic interview- revised: A revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders,

    C. Lord, M. Rutter, and A. Le Couteur, “Autism diagnostic interview- revised: A revised version of a diagnostic interview for caregivers of individuals with possible pervasive developmental disorders,” J. Autism Dev. Disord., vol. 24, no. 5, p. 659–685, Oct. 1994

    work page 1994

  8. [6]

    Austism diagnostic observation schedule: A standardized observation of communicative and social behavior,

    C. Lord et al., “Austism diagnostic observation schedule: A standardized observation of communicative and social behavior,” J. Autism Dev. Disord., vol. 19, no. 2, p. 185–212, Jun. 1989

    work page 1989

Show all 81 references
  1. [7]

    Deep learning with image-based autism spectrum disorder analysis: A systematic review,

    M. Z. Uddin, M. A. Shahriar, M. N. Mahamood, F. Alnajjar, M. I. Pramanik, and M. A. R. Ahad, “Deep learning with image-based autism spectrum disorder analysis: A systematic review,” Eng. Appl. Artif. Intell., vol. 127, p. 107185, Jan. 2024

    2024

  2. [8]

    Identifying autism spectrum disorder from resting-state fmri using deep belief network,

    Z.-A. Huang, Z. Zhu, C. H. Yau, and K. C. Tan, “Identifying autism spectrum disorder from resting-state fmri using deep belief network,” IEEE Trans. Neural Netw. Learn. Syst. , vol. 32, no. 7, p. 2847–2861, Jul. 2021

    2021

  3. [9]

    A residual graph convolutional network with spatio-temporal features for autism classification from fmri brain images,

    K.-W. Park and S.-B. Cho, “A residual graph convolutional network with spatio-temporal features for autism classification from fmri brain images,” Appl. Soft. Comput. , vol. 142, p. 110363, Jul. 2023

    2023

  4. [10]

    Spatial–temporal co-attention learning for diagnosis of mental disorders from resting-state fmri data,

    R. Liu, Z.-A. Huang, Y . Hu, Z. Zhu, K.-C. Wong, and K. C. Tan, “Spatial–temporal co-attention learning for diagnosis of mental disorders from resting-state fmri data,” IEEE Trans. Neural Netw. Learn. Syst. , vol. 35, no. 8, p. 10591–10605, Aug. 2024

    2024

  5. [11]

    Diagnosis of autism disorder based on deep network trained by augmented eeg signals,

    H. A. Ardakani, M. Taghizadeh, and F. Shayegh, “Diagnosis of autism disorder based on deep network trained by augmented eeg signals,” Int. J. Neural Syst. , vol. 32, no. 11, Aug. 2022

    2022

  6. [12]

    Resting-state eeg power differences in autism spectrum disorder: a systematic review and meta- analysis,

    W. S. Neo, D. Foti, B. Keehn, and B. Kelleher, “Resting-state eeg power differences in autism spectrum disorder: a systematic review and meta- analysis,” Transl. Psychiatr., vol. 13, no. 1, p. 389, 2023

    2023

  7. [13]

    Developmental trajectories of eeg aperiodic and periodic components in children 2–44 months of age,

    C. L. Wilkinson et al., “Developmental trajectories of eeg aperiodic and periodic components in children 2–44 months of age,” Nat. Commun. , vol. 15, no. 1, p. 5788, Jul. 2024

    2024

  8. [14]

    Discriminative few shot learning of facial dynamics in interview videos for autism trait classification,

    N. Zhang, M. Ruan, S. Wang, L. Paul, and X. Li, “Discriminative few shot learning of facial dynamics in interview videos for autism trait classification,” IEEE Trans. Affect. Comput. , vol. 14, no. 2, p. 1110–1124, Apr. 2023

    2023

  9. [17]

    Gaze patterns in children with autism spectrum disorder to emotional faces: Scanpath and similarity,

    W. Zhou, M. Yang, J. Tang, J. Wang, and B. Hu, “Gaze patterns in children with autism spectrum disorder to emotional faces: Scanpath and similarity,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 32, p. 865–874, 2024

    2024

  10. [18]

    Automatic classification of asd children using appearance-based features from videos,

    J. Li, Z. Chen, G. Li, G. Ouyang, and X. Li, “Automatic classification of asd children using appearance-based features from videos,” Neuro- computing, vol. 470, pp. 40–50, Jan. 2022

    2022

  11. [19]

    Using 2d video-based pose estimation for automated prediction of autism spectrum disorders in young children,

    N. Kojovic, S. Natraj, S. P. Mohanty, T. Maillart, and M. Schaer, “Using 2d video-based pose estimation for automated prediction of autism spectrum disorders in young children,” Sci Rep, vol. 11, no. 1, p. 15069, Jul. 2021

    2021

  12. [20]

    Benchmarking children’s asr with supervised and self-supervised speech foundation models,

    R. Fan, N. Balaji Shankar, and A. Alwan, “Benchmarking children’s asr with supervised and self-supervised speech foundation models,” in Proc. Annu. Conf. Int. Speech. Commun. Assoc. (INTERSPEECH) , Sep. 2024, p. 5173–5177

    2024

  13. [21]

    An automated assessment framework for atypical prosody and stereotyped idiosyncratic phrases related to autism spectrum disor- der,

    M. Li et al., “An automated assessment framework for atypical prosody and stereotyped idiosyncratic phrases related to autism spectrum disor- der,” Comput. Speech Lang. , vol. 56, p. 80–94, Jul. 2019

    2019

  14. [22]

    Computer vision-based assessment of autistic children: Analyzing interactions, emotions, human pose, and life skills,

    V . G. Prakash et al. , “Computer vision-based assessment of autistic children: Analyzing interactions, emotions, human pose, and life skills,” IEEE Access, vol. 11, pp. 47 907–47 929, 2023

    2023

  15. [23]

    Activity-based early autism diagnosis using a multi-dataset supervised contrastive learning approach,

    A. Rani and Y . Verma, “Activity-based early autism diagnosis using a multi-dataset supervised contrastive learning approach,” in Proc.- IEEE Winter Conf. Appl. Comput. Vis. (WACV) , Jan. 2024, pp. 7788–7797

    2024

  16. [24]

    Identifying children with autism spectrum disorder via transformer-based representation learning from dynamic facial cues,

    C. Xia, H. Chen, J. Han, D. Zhang, and K. Li, “Identifying children with autism spectrum disorder via transformer-based representation learning from dynamic facial cues,” IEEE Trans. Affect. Comput. , vol. 16, no. 1, p. 83–97, Jan. 2025

    2025

  17. [25]

    Can gpt-4v (ision) serve medical applications? case studies on gpt-4v for multimodal medical diagnosis,

    C. Wu, J. Lei, Q. Zheng, W. Zhao, W. Lin, X. Zhang, X. Zhou, Z. Zhao, Y . Zhang, Y . Wanget al., “Can gpt-4v (ision) serve medical applications? case studies on gpt-4v for multimodal medical diagnosis,” arXiv preprint arXiv:2310.09909, 2023

    arXiv

  18. [26]

    A generalist vision–language foundation model for diverse biomedical tasks,

    K. Zhang, R. Zhou, E. Adhikarla, Z. Yan, Y . Liu, J. Yu, Z. Liu, X. Chen, B. D. Davison, H. Ren et al., “A generalist vision–language foundation model for diverse biomedical tasks,” Nature Medicine, pp. 1–13, 2024

    2024

  19. [27]

    Large language models to identify social determinants of health in electronic health records,

    M. Guevara, S. Chen, S. Thomas, T. L. Chaunzwa, I. Franco, B. H. Kann, S. Moningi, J. M. Qian, M. Goldstein, S. Harper et al. , “Large language models to identify social determinants of health in electronic health records,” NPJ digital medicine , vol. 7, no. 1, p. 6, 2024

    2024

  20. [28]

    A survey of large language models in medicine: Progress, application, and challenge,

    H. Zhou et al. , “A survey of large language models in medicine: Progress, application, and challenge,” 2023, arXiv:2311.05112

    2023 arXiv

  21. [29]

    Large language models in medicine,

    A. J. Thirunavukarasu, D. S. J. Ting, K. Elangovan, L. Gutierrez, T. F. Tan, and D. S. W. Ting, “Large language models in medicine,” Nat. Med., vol. 29, no. 8, p. 1930–1940, Jul. 2023

    1930

  22. [30]

    Large language models in medicine: The potentials and pitfalls: A narrative review,

    J. A. Omiye, H. Gui, S. J. Rezaei, J. Zou, and R. Daneshjou, “Large language models in medicine: The potentials and pitfalls: A narrative review,” Ann. Intern. Med. , vol. 177, no. 2, p. 210–220, Feb. 2024

    2024

  23. [31]

    Gpt-4o system card,

    A. Hurst et al., “Gpt-4o system card,” 2024, arXiv:2410.21276

    2024 arXiv

  24. [32]

    Claude 3.5 sonnet,

    Anthropic, “Claude 3.5 sonnet,” 2024, [Online]. Available: https://www. anthropic.com/news/claude-3-5-sonnet

    2024

  25. [33]

    The llama 3 herd of models,

    A. Dubey et al., “The llama 3 herd of models,” 2024, arXiv:2407.21783

    2024 arXiv

  26. [34]

    Qwen2 technical report,

    A. Yang et al., “Qwen2 technical report,” 2024, arXiv:2407.10671

    2024 arXiv

  27. [35]

    Available: https://kimi.moonshot.cn/

    Moonshot, “kimi,” 2024, [Online]. Available: https://kimi.moonshot.cn/

    2024

  28. [36]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning,

    D. Guo et al., “Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning,” 2025, arXiv:2501.12948

    2025 arXiv

  29. [37]

    BLIP-2: Bootstrapping language- image pre-training with frozen image encoders and large language models,

    J. Li, D. Li, S. Savarese, and S. Hoi, “BLIP-2: Bootstrapping language- image pre-training with frozen image encoders and large language models,” in Proc. Int. Conf. Mach. Learn. (ICML) , vol. 202, Jul. 2023, pp. 19 730–19 742. IEEE TRANSACTIONS ON AFFECTIVE COMPUTING 14

    2023

  30. [38]

    Video-LLaMA: An instruction- tuned audio-visual language model for video understanding,

    H. Zhang, X. Li, and L. Bing, “Video-LLaMA: An instruction- tuned audio-visual language model for video understanding,” 2023, arXiv:2306.02858

    2023 arXiv

  31. [39]

    A prompt pattern catalog to enhance prompt engineering with chatgpt,

    J. White et al., “A prompt pattern catalog to enhance prompt engineering with chatgpt,” 2023, arXiv:2302.11382

    2023 arXiv

  32. [40]

    Emotional competence in children with autism: Diagnostic criteria and empirical evidence,

    S. Begeer, H. M. Koot, C. Rieffe, M. M. Terwogt, and H. Stegge, “Emotional competence in children with autism: Diagnostic criteria and empirical evidence,” Dev. Rev., vol. 28, no. 3, pp. 342–369, Sep. 2008

    2008

  33. [41]

    The role of emotion regulation in autism spectrum disorder,

    C. A. Mazefsky et al. , “The role of emotion regulation in autism spectrum disorder,” J. Am. Acad. Child Adolesc. Psychiatr. , vol. 52, no. 7, pp. 679–688, Jul. 2013

    2013

  34. [42]

    Technology-assisted emotion recognition for autism spectrum disorder (ASD) children: a systematic literature review,

    M. A. Rashidan et al. , “Technology-assisted emotion recognition for autism spectrum disorder (ASD) children: a systematic literature review,” IEEE Access, vol. 9, pp. 33 638–33 653, Mar. 2021

    2021

  35. [43]

    Computer vision analysis for quantification of autism risk behaviors,

    J. Hashemi et al., “Computer vision analysis for quantification of autism risk behaviors,” IEEE Trans. Affect. Comput. , vol. 12, no. 1, pp. 215– 226, Jan. 2021

    2021

  36. [44]

    Vision- assisted recognition of stereotype behaviors for early diagnosis of autism spectrum disorders,

    F. Negin, B. Ozyer, S. Agahian, S. Kacdioglu, and G. T. Ozyer, “Vision- assisted recognition of stereotype behaviors for early diagnosis of autism spectrum disorders,” Neurocomputing, vol. 446, pp. 145–155, Jul. 2021

    2021

  37. [45]

    Discriminative few shot learning of facial dynamics in interview videos for autism trait classification,

    N. Zhang, M. Ruan, S. Wang, L. Paul, and X. Li, “Discriminative few shot learning of facial dynamics in interview videos for autism trait classification,” IEEE Trans. Affect. Comput. , vol. 14, no. 2, pp. 1110– 1124, Apr. 2023

    2023

  38. [46]

    Stratification of children with autism spectrum disorder through fusion of temporal information in eye-gaze scan-paths,

    A. Atyabi et al., “Stratification of children with autism spectrum disorder through fusion of temporal information in eye-gaze scan-paths,” ACM Trans. Knowl. Discov. Data , vol. 17, no. 2, pp. 1–20, Feb. 2023

    2023

  39. [47]

    Deep learning to interpret autism spectrum disorder behind the camera,

    S. Chen, M. Jiang, and Q. Zhao, “Deep learning to interpret autism spectrum disorder behind the camera,” IEEE Transactions on Cognitive and Developmental Systems , 2024

    2024

  40. [48]

    Classifying asd children with lstm based on raw videos,

    J. Li, Y . Zhong, J. Han, G. Ouyang, X. Li, and H. Liu, “Classifying asd children with lstm based on raw videos,” Neurocomputing, vol. 390, pp. 226–238, May 2020

    2020

  41. [49]

    Detecting a child’s stimming behaviours for autism spectrum disorder diagnosis using rgbpose-slowfast network,

    J. B. S, D. Pandian, S. S. Rajagopalan, and D. Jayagopi, “Detecting a child’s stimming behaviours for autism spectrum disorder diagnosis using rgbpose-slowfast network,” in Proc. Int. Conf. Image Process. (ICIP), Oct. 2022, pp. 3356–3360

    2022

  42. [50]

    Vision-based activity recognition in children with autism- related behaviors,

    P. Wei, D. Ahmedt-Aristizabal, H. Gammulle, S. Denman, and M. A. Armin, “Vision-based activity recognition in children with autism- related behaviors,” Heliyon, vol. 9, no. 6, p. 16763, Jun. 2023

    2023

  43. [51]

    Large language models encode clinical knowledge,

    K. Singhal et al., “Large language models encode clinical knowledge,” Nature, vol. 620, no. 7973, pp. 172–180, Jul. 2023

    2023

  44. [52]

    Huatuo: Tuning llama model with chinese medical knowledge,

    H. Wang et al. , “Huatuo: Tuning llama model with chinese medical knowledge,” 2023, arXiv:2304.06975

    2023 arXiv

  45. [53]

    Glm-130b: An open bilingual pre-trained model,

    A. Zeng et al., “Glm-130b: An open bilingual pre-trained model,” 2022, arXiv:2210.02414

    2022 arXiv

  46. [54]

    Can generalist foundation models outcompete special- purpose tuning? Case study in medicine,

    H. Nori et al. , “Can generalist foundation models outcompete special- purpose tuning? Case study in medicine,” 2022, arXiv:2311.16452

    2022 arXiv

  47. [56]

    Phys- iological detection of affective states in children with autism spectrum disorder,

    S. Sarabadani, L. C. Schudlo, A. A. Samadani, and A. Kushski, “Phys- iological detection of affective states in children with autism spectrum disorder,” IEEE Trans. Affect. Comput., vol. 11, no. 4, pp. 588–600, Oct. 2020

    2020

  48. [57]

    Effects of com- puterized emotional training on children with high functioning autism,

    S. Piana, C. Malagoli, M. C. Usai, and A. Camurri, “Effects of com- puterized emotional training on children with high functioning autism,” IEEE Trans. Affect. Comput. , vol. 12, no. 4, pp. 1045–1054, Oct. 2021

    2021

  49. [58]

    Stimuli video quantification based on valence- arousal elicitation in children with autism spectrum disorder (asd),

    M. A. Rashidan et al., “Stimuli video quantification based on valence- arousal elicitation in children with autism spectrum disorder (asd),”IEEE Access, vol. 11, pp. 111 142–111 153, 2023

    2023

  50. [59]

    Solov2: Dynamic and fast instance segmentation,

    X. Wang, R. Zhang, T. Kong, L. Li, and C. Shen, “Solov2: Dynamic and fast instance segmentation,” in Adv. neural inf. proces. syst. (NeurIPS) , vol. 33, Dec. 2020, pp. 17 721–17 732

    2020

  51. [60]

    Retinaface: Single-shot multi-level face localisation in the wild,

    J. Deng, J. Guo, E. Ververas, I. Kotsia, and S. Zafeiriou, “Retinaface: Single-shot multi-level face localisation in the wild,” inProc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR) , Jun. 2020, pp. 5203–5212

    2020

  52. [61]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR) , Jun. 2019, pp. 4690–4699

    2019

  53. [62]

    Bag of tricks and a strong baseline for deep person re-identification,

    H. Luo, Y . Gu, X. Liao, S. Lai, and W. Jiang, “Bag of tricks and a strong baseline for deep person re-identification,” in IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recogn. Workshops (CVPRW) , Jun. 2019

    2019

  54. [63]

    Accurate head pose estima- tion using image rectification and a lightweight convolutional neural network,

    X. Li, D. Zhang, M. Li, and D.-J. Lee, “Accurate head pose estima- tion using image rectification and a lightweight convolutional neural network,” IEEE Trans. Multimedia , vol. 25, p. 2239–2251, Jan. 2022

    2022

  55. [64]

    Joint training on multiple datasets with inconsistent labeling criteria for facial expression recognition,

    C. Yu, D. Zhang, W. Zou, and M. Li, “Joint training on multiple datasets with inconsistent labeling criteria for facial expression recognition,” IEEE Trans. Affect. Comput. , vol. 15, no. 3, p. 1812–1825, Jul. 2024

    2024

  56. [65]

    Deep high-resolution representa- tion learning for human pose estimation,

    K. Sun, B. Xiao, D. Liu, and J. Wang, “Deep high-resolution representa- tion learning for human pose estimation,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR) , Jun. 2019, pp. 5693–5703

    2019

  57. [66]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2016, pp. 770–778

    2016

  58. [67]

    The Kaldi speech recognition toolkit,

    D. Povey et al. , “The Kaldi speech recognition toolkit,” in IEEE Workshop Autom. Speech Recogn. Underst. (ASRU) , Dec. 2011

    2011

  59. [68]

    Failure to respond to name is indicator of possible autism spectrum disorder,

    J. G. Frohna, “Failure to respond to name is indicator of possible autism spectrum disorder,” J. Pediatrics, vol. 151, no. 3, pp. 327–328, Sep. 2007

    2007

  60. [69]

    Application of the still-face paradigm in early screening for high-risk autism spectrum disorder in infants and toddlers,

    N. Qiu et al., “Application of the still-face paradigm in early screening for high-risk autism spectrum disorder in infants and toddlers,” Front. Pediatr., vol. 8, p. 290, Jun. 2020

    2020

  61. [70]

    Early identification of autism spectrum disorders,

    L. Zwaigenbaum, S. Bryson, and N. Garon, “Early identification of autism spectrum disorders,” Behav. Brain Res. , vol. 251, pp. 133–146, Aug. 2013

    2013

  62. [71]

    Brief report: Preliminary feasibility of the tedi: A novel parent-administered telehealth assessment for autism spectrum disorder symptoms in the first year of life,

    M. R. Talbott et al. , “Brief report: Preliminary feasibility of the tedi: A novel parent-administered telehealth assessment for autism spectrum disorder symptoms in the first year of life,” J. Autism Dev. Disord. , vol. 50, pp. 3432–3439, Nov. 2020

    2020

  63. [72]

    Early social communication scales (ESCS),

    P. Mundy, C. Delgado, J. Block, M. Venezia, A. Hogan, and J. Seibert, “Early social communication scales (ESCS),” 2003, [On- line]. Available: https://d1wqtxts1xzle7.cloudfront.net/103595277/escs manual 2003 2013-libre.pdf

    2003

  64. [73]

    Hed: A computational model of affective adaptation and emotion dynamics,

    J. E. Steephen, “Hed: A computational model of affective adaptation and emotion dynamics,” IEEE Trans. Affect. Comput. , vol. 4, no. 2, p. 197–210, Apr. 2013

    2013

  65. [74]

    A multivariate statistical model for emotion dynamics

    T. Krone, C. J. Albers, P. Kuppens, and M. E. Timmerman, “A multivariate statistical model for emotion dynamics.” Emotion, vol. 18, no. 5, p. 739–754, Aug. 2018

    2018

  66. [75]

    Daily life affective dynamics as transdiagnostic predic- tors of mental health symptoms: An ecological momentary assessment study,

    X. Zhu et al., “Daily life affective dynamics as transdiagnostic predic- tors of mental health symptoms: An ecological momentary assessment study,” J. Affect. Disord., vol. 351, p. 808–817, Apr. 2024

    2024

  67. [76]

    Video-xl: Extra-long vision language model for hour-scale video understanding,

    Y . Shu et al., “Video-xl: Extra-long vision language model for hour-scale video understanding,” 2024, arXiv:2409.14485

    2024 arXiv

  68. [77]

    Qwen2.5-vl technical report,

    S. Bai et al. , “Qwen2.5-vl technical report,” 2025, arXiv:2502.13923. [Online]. Available: https://arxiv.org/abs/2502.13923

    2025 arXiv

  69. [78]

    video-salmonn: Speech-enhanced audio-visual large language models,

    G. Sun, W. Yu, C. Tang, X. Chen, T. Tan, W. Li, L. Lu, Z. Ma, Y . Wang, and C. Zhang, “video-salmonn: Speech-enhanced audio-visual large language models,” arXiv preprint arXiv:2406.15704 , 2024

    2024 arXiv

  70. [79]

    Assessing the social skills of children with autism spectrum disorder via language-image pre-training models,

    W. Liu et al. , “Assessing the social skills of children with autism spectrum disorder via language-image pre-training models,” in Lect. Notes Comput. Sci. (PRCV) , Dec. 2023, pp. 260–271

    2023

  71. [80]

    Medagents: Large language models as collaborators for zero-shot medical reasoning,

    X. Tang et al., “Medagents: Large language models as collaborators for zero-shot medical reasoning,” in Find. Assoc. Comput Linguist. (ACL) , Aug. 2024, p. 599–621

    2024

  72. [81]

    The relation between valence and arousal in subjective experience

    P. Kuppens, F. Tuerlinckx, J. A. Russell, and L. F. Barrett, “The relation between valence and arousal in subjective experience.” Psychol. Bull. , vol. 139, no. 4, p. 917–940, Jul. 2013

    2013

  73. [82]

    Using the circumplex model of affect to study valence and arousal ratings of emotional faces by children and adults with autism spectrum disorders,

    A. Tseng et al., “Using the circumplex model of affect to study valence and arousal ratings of emotional faces by children and adults with autism spectrum disorders,” J. Autism Dev. Disord., vol. 44, no. 6, p. 1332–1346, Nov. 2013. Wenxing Liu received the master’s degree in c...

    2013

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.