Pith. sign in

REVIEW 4 major objections 6 minor 37 references

Towards Paradigm-General Suicide Risk Detection via Speech LLM

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A single speech model with routed low-rank experts can detect suicide risk across ten distinct speaking tasks more accurately than task-specific models, and its confidence scores are better calibrated.

desk verdict Credible MoDE cross-task accuracy with an unsupported 'unseen paradigms' claim in the abstract. read the letter →

arxiv 2509.22153 v2 pith:WS4G3MPZ submitted 2025-09-26 eess.AS

Jialun Li , Weitao Jiang , Ziyun Cui , Yinan Duan , Diyang Qu , Chao Zhang , Runsen Chen , Chang Lei
show 1 more author
Wen Wu
This is my paper · ORCID
classification eess.AS
keywords suicideriskdetectionspeechLLMmixtureofexpertsDoRAcross-tasklearningconfidencecalibrationadolescentmentalhealthelicitationparadigms
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 claims that speech-based suicide risk assessment does not require a separate model for each elicitation task. A single speech large language model, augmented with a mixture of weight-decomposed low-rank experts and a learned router, can jointly handle ten spontaneous-speech tasks—verbal fluency, reading, open-ended answers, face description, and divergent thinking—and outperforms both separate per-task models and ordinary joint training. On 1,223 adolescents, the average accuracy rises from 0.628 to 0.656, a 4.5 per cent relative gain, while expected calibration error drops from 0.099 to 0.061. The authors conclude that unifying paradigms improves detection accuracy, modelling efficiency, and reliability for medical deployment.

What carries the argument

The central object is MoDE, a mixture of DoRA experts grafted onto a speech LLM. DoRA (weight-decomposed low-rank adaptation) splits each pretrained weight into a magnitude scalar and a direction, and applies low-rank updates only to the direction; MoDE keeps ten such adapters and computes a softmax distribution over them from the Transformer's last hidden state. The router is trained on the frozen backbone before the experts are activated, a decoupling intended to keep routing stable. A KL-divergence load-balancing term prevents expert collapse, and temperature scaling on the router's softmax controls routing sparsity. Together these components let a single model act as a task-specific expert on some inputs and a shared generalist on others.

What would settle it

Re-run the comparison on the full 1,223-participant cohort without dropping partial completers, and replace self-report MINI-KID labels with clinician-administered structured assessment; if MoDE's average accuracy no longer beats separate tuning or its ECE no longer beats joint tuning, the claimed gains were artefacts of label noise or test-set selection.

Watch

Extended reading notes

Core claim

The central claim is that cross-task training through MoDE improves both accuracy and calibration over single-task models. The paper presents a single Qwen2.5-Omni-7B speech LLM whose transformer weights are updated by a weighted combination of ten DoRA experts, with a router deciding expert weights from the hidden states. In experiments, MoDE reaches 0.656 average accuracy across ten SSR tasks versus 0.628 for separate tuning and 0.635 for joint tuning; the largest task-level gain is on the 'manage distress' question (0.684 to 0.758). The model also records the best ECE (0.061), MCE (0.089), NLL, NCE, AUROC and AUPRC among compared configurations, and its accuracy rises as low-confidence predictions are rejected, while separate tuning's rejection curve is flat. The abstract further asserts that the model can generalise to unseen paradigms, meaning a routing pattern learned from known tasks should transfer to a new elicitation task.

Load-bearing premise

The whole evaluation leans on the self-report MINI-KID suicidality interview being a trustworthy binary ground truth for current suicide risk, and on the refined test set of participants who finished all ten tasks being representative of the full 1,223-person cohort.

Editorial extensions

If this is right

  • A deployment pipeline could serve all ten assessment tasks from one model checkpoint, replacing ten separately tuned detectors and simplifying maintenance.
  • Clinicians can use the model's confidence scores as a triage signal: the rejection analysis shows that withholding low-confidence predictions raises accuracy, so uncertain cases could be routed to human review.
  • The learned specialisation is inspectable: reading tasks mostly trigger one expert, face-description tasks another, and the divergent-thinking task a unique combination, so routing patterns can be checked per paradigm.
  • Because manual task-to-expert assignment performs worse than learned routing, forcing a priori grouping of tasks sacrifices useful cross-task cooperation.
  • If the claimed unseen-paradigm generalisation holds, cross-task training should carry over to new speech-based screening tasks without retraining a fresh model.

Reading between the lines

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

  • The calibration gain may be the most consequential claim for practice: if it holds outside this dataset, cross-task training could become a cheap way to make medical speech classifiers safer without a separate calibration stage.
  • The routing pattern hints that the router is discovering cognitive dimensions (phonetic reading, emotional face description, divergent thinking) rather than only task identity; one could test this by comparing expert activation matrices with text-only task embeddings.
  • A direct extension is to treat unseen-paradigm generalisation as a zero-shot transfer experiment: train MoDE on nine tasks, freeze router and experts, evaluate on the tenth, and report accuracy and calibration on that held-out task.
  • The dataset is roughly balanced, so accuracy is a meaningful headline metric, but real clinics face different base rates; calibration metrics like NLL should be rechecked under prevalence shift before deployment.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes MoDE, a mixture of DoRA experts on a speech LLM backbone (Qwen2.5-Omni-7B), to unify ten speech-based suicide risk (SSR) assessment tasks into a single model. It reports accuracy comparisons against separate tuning and conventional joint tuning on a dataset of 1,223 Chinese adolescents, ablations of the MoDE components, an analysis of expert-task specialization, and calibration/reject-option experiments. The central claims are that MoDE improves detection accuracy and confidence calibration over both single-task and joint-tuning baselines, and that the model 'can generalise to unseen paradigms.'

Significance. If the accuracy gains are taken at face value, the paper is a useful demonstration that a single speech LLM with lightweight routed adapters can serve multiple suicide-risk elicitation tasks, with potential practical efficiency benefits. The strengths include a relatively large clinically annotated dataset, ten diverse tasks, three-seed reporting with standard errors, and clearly presented tables. However, the paper's most distinctive claim—generalization to unseen paradigms—is not tested at all in the present evaluation, and the calibration comparison omits the joint-tuning control that would be needed to attribute the improvement to cross-task learning. As it stands, the demonstrated scope is substantially narrower than the abstract and conclusions claim.

major comments (4)
  1. [Abstract; Section 3.2; Section 4.2] The claim that MoDE 'can generalise to unseen paradigms' is not supported by any reported experiment. The data split in Section 3.2 is participant-level 8:1:1, so all ten tasks appear in both training and test; no task or paradigm is held out. A leave-one-task-out or similar held-out-paradigm evaluation is required to substantiate the generalization claim, or the claim must be removed from the abstract and conclusions.
  2. [Section 4.7; Table 4] The calibration comparison is made only between 'Separate' tuning and 'MoDE'. The conventional joint-tuning baseline (Qwen Omni joint tuning, already reported in Section 4.1/Figure 3) is not evaluated for ECE, MCE, NLL, NCE, AUROC, or AUPRC. Therefore the conclusion that 'cross-task learning improves confidence calibration' is not established; the improvement could be due to the MoDE-specific components (router, load balancing, router temperature) rather than joint training. Please add calibration metrics for the joint-tuning baseline.
  3. [Section 2.1; Section 3.2; Section 6] The MINI-KID label limitation is acknowledged, but the refinement of the test set to only participants who completed all ten tasks may introduce selection bias. The paper does not report how many participants were excluded, nor does it compare the demographic or clinical characteristics of included versus excluded participants. Because all reported numbers are measured against these labels and on this refined set, please quantify the refinement and discuss its impact on the conclusions, or report results on the full test set.
  4. [Section 1; References [5], [6]] The statement that this is 'the first work that investigates suicide risk detection using speech LLM' appears to conflict with the authors' own reference [5], which uses Whisper and large language models for spontaneous speech-based suicide risk detection, and with reference [6], the SpeechWellness Challenge that incorporates multiple speech tasks. Please clarify the precise novelty relative to these prior works, or soften the claim.
minor comments (6)
  1. [Title / Header] The manuscript header reads 'TOW ARDS CROSS-TASK SUICIDE RISK DETECTION VIA SPEECH LLM' while the submitted title is 'Towards Paradigm-General Suicide Risk Detection via Speech LLM'; 'cross-task' and 'cross-paradigm' are used interchangeably throughout, and this terminology should be unified.
  2. [Section 4.3; Table 3] The claim that removing load balancing 'collapses to a single DoRA expert' is not substantiated by any reported quantity; please report expert utilization metrics or router entropy for the ablation without load balancing.
  3. [Section 2.4] Temperature scaling is applied to the router softmax rather than to the final classification probabilities; the terminology is potentially confusing because temperature scaling is a standard post-hoc calibration method for output probabilities. Please clarify that this is a router-temperature hyperparameter and justify its role in the training objective.
  4. [Section 2.3; Section 3.2] The router architecture (e.g., number and size of hidden layers, which transformer layer provides the hidden states) and the exact DoRA parameterization per expert are not fully specified; additional implementation details would improve reproducibility.
  5. [Figures 3–6] Several figures contain rendering artifacts (e.g., Unicode sequences in captions and axis labels) and low text resolution; please re-export the figures at higher quality.
  6. [General] There are several typographical errors, including 'difference task' (Section 4.1), 'its not only' (Section 4.7), and 'TOW ARDS' in the title; these should be corrected in a final pass.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: MoDE's accuracy and calibration gains are measured on held-out test audio against independent baselines, and no derivation step reduces to its inputs by construction.

full rationale

The paper's central derivation is a standard empirical machine-learning pipeline, not a circular one. The proposed MoDE model is trained on an 80% split of the 1,223-participant dataset and evaluated on a held-out test set (Section 3.2: 'The dataset was split into train/dev/test sets at a ratio of 8:1:1'), and the reported accuracies in Table 2 compare MoDE with separate tuning and joint tuning on the same held-out test data. No fitted parameter is renamed as a prediction: temperature scaling and load balancing are ablated in Table 3 and their contributions are measured empirically, not asserted by construction. The calibration comparison in Table 4 is computed on the same held-out test data using standard ECE/MCE/NLL metrics, so the claimed calibration improvement is an empirical result rather than an artifact of the model definition. The reliance on the authors' own earlier dataset and SpeechWellness baselines ([5,6]) is substantial but not circular: the prior work supplies the collected speech corpus and label protocol (MINI-KID), while the present paper's contribution is a new architecture evaluated against those external data and against baselines trained in this paper. The authors' Section 6 limitation statement ('it is a self-report tool and cannot fully capture the multifaceted nature of suicide risk') is a candid acknowledgment of label-validity risk, not evidence of circular derivation. One overstatement exists: the abstract claims the model 'can generalise to unseen paradigms,' yet all ten SSR tasks appear in both training and test splits, so the claim is not experimentally tested. That is an evidence gap or overclaim, not circularity, because nothing in the derivation assumes the conclusion. No equation in the paper defines the output in terms of the target metric, and no prediction is statistically forced by construction. Accordingly, the appropriate circularity score is 0.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard machine-learning hyperparameters (number of experts, DoRA rank, alpha, load-balancing coefficient, temperature) and domain assumptions about speech as a signal and MINI-KID labels as ground truth. No new physical or conceptual entities are introduced.

free parameters (5)
  • Number of MoDE experts E = 10
    Set to match the ten SSR tasks; Figure 4 shows performance varies with E, so this choice affects results.
  • DoRA rank r = 32
    Hyperparameter for low-rank adaptation; affects model capacity.
  • DoRA alpha = 64
    Scaling hyperparameter for DoRA updates.
  • Load balancing coefficient lambda_LB = not reported
    Coefficient for the KL load-balancing loss in Eq. (3); not specified in the paper.
  • Temperature scaling parameter = not reported
    Applied to router softmax; value not specified.
assumptions (4)
  • domain assumption Speech contains measurable acoustic and lexical markers of suicide risk
    Relies on prior work [3,4]; no new evidence in this paper.
  • domain assumption MINI-KID suicidality module provides valid binary labels for current suicide risk
    Acknowledged in Section 6 as a self-report tool that cannot fully capture suicide risk.
  • ad hoc to paper Weighted combination of DoRA experts with router weights is a valid adaptation mechanism
    Eq. (2) is an architectural choice without theoretical justification beyond empirical results; adopted from MoE and DoRA literature.
  • domain assumption Test set participants who completed all ten tasks are representative of the full dataset
    Section 3.2 refines the test set; no analysis of dropouts is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Paradigm-General Suicide Risk Detection via Speech LLM." pith.science (2026). https://pith.science/paper/WS4G3MPZ

@misc{pith2026250922153,
  author       = {Pith},
  title        = {Pith review of: Towards Paradigm-General Suicide Risk Detection via Speech LLM},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WS4G3MPZ}},
  note         = {Machine review of arXiv:2509.22153}
}
read the original abstract

Suicide risk among adolescents remains a critical public health concern, and speech provides a non-invasive and scalable approach for its detection. Speech-based suicide risk assessment commonly relies on carefully designed speech elicitation paradigms (\textit{e.g.,} verbal fluency, reading, or question answering) to probe cognitive and affective states. Existing approaches, however, typically focus on one single paradigm at a time. This paper, for the first time, investigates cross-paradigm approaches that unify diverse speech elicitation paradigms within a single model. Specifically, we use a speech LLM as backbone with a mixture of DoRA experts (MoDE) to capture complementary cues across assessments dynamically, tested on 1,223 participants across ten speech elicitation paradigms. Results show that MoDE outperforms both paradigm-specific and conventional joint-learning models. Moreover, it can generalise to unseen paradigms and provide better confidence calibration.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 31 canonical work pages

  1. [5]

    This paper investigates unifying diverse speech tasks within a single model for cross-task suicide risk detection

    CONCLUSIONS For automatic suicide risk detection from spontaneous speech, par- ticipants are typically asked to complete multiple speech-based as- sessment tasks, and the model performs detection using the resulting recordings as input. This paper investigates unifying diverse speech tasks within a single model for cross-task suicide risk detection. A nov...

    work page

  2. [6]

    Though MINI-KID is widely used and often treated as a benchmark for adolescent assessment, it is a self-report tool and cannot fully capture the multifaceted nature of suicide risk

    LIMITATIONS AND ETHICAL CONSIDERATIONS Our conclusions are derived from the MINI-KID scale that gauges current suicide risk based on respondents’ immediate answers. Though MINI-KID is widely used and often treated as a benchmark for adolescent assessment, it is a self-report tool and cannot fully capture the multifaceted nature of suicide risk. Accordingl...

    work page

  3. [1]

    INTRODUCTION Suicide remains a leading cause of death among adolescents, making timely risk identification a public health imperative [1]. Detection of suicide risk is challenging not only because it lacks a uniform clinical profile but also because it often depends on patients’ will- ingness and ability to disclose subjective experiences [2]. Speech offe...

    work page

  4. [2]

    Towards Paradigm-General Suicide Risk Detection via Speech LLM

    METHODS 2.1. Datasets and SSR Tasks The dataset used in this study [5, 6] consists of voice recordings collected from 1,223 Chinese adolescents (aged 10–18). Data were obtained under controlled experimental conditions and paired with standardised suicide risk assessments administered via the Mini In- ternational Neuropsychiatric Interview for Children and...

    work page Pith review arXiv 2026

  5. [3]

    The speaker responded to the following question:{Task Description}. Based on the provided audio, assess the suicide risk of the speaker (Yes, No)

    EXPERIMENTAL SETUP 3.1. Backbone Model and Baselines Qwen2.5-Omni-7B1 [28], a multimodal large language model capa- ble of processing both speech and text inputs, was used as the back- bone model. To evaluate the effectiveness of speech LLMs on sui- cide risk detection, we compared Qwen-Omni with Whisper-Large- v32 [29], a widely used speech foundation mo...

    work page

  6. [4]

    RESULTS AND DISCUSSION 4.1. Using Speech LLM for Suicide Risk Detection We first compared Qwen-Omni with Whisper on both separate tun- ing and joint tuning settings to evaluate the effectiveness of speech LLM on suicide risk detection tasks. Results are shown in Fig. 3. It can be observed that, under the separate tuning setting, Qwen-Omni does not necessa...

    work page

  7. [7]

    Suicide and suicide attempts in adolescents,

    B. Shain, “Suicide and suicide attempts in adolescents,”Pedi- atrics, vol. 138, pp. e20161420, 2016

    work page 2016

  8. [8]

    A review of depression and suicide risk as- sessment using speech analysis,

    N. Cummins, S. Scherer, J. Krajewski, S. Schnieder, J. Epps, and T. Quatieri, “A review of depression and suicide risk as- sessment using speech analysis,”Speech Communication, vol. 71, pp. 10–49, 2015

    work page 2015

Show all 37 references
  1. [9]

    Linguistic features of suicidal thoughts and behaviors: A systematic review,

    S. Homan, M. Gabi, N. Klee, S. Bachmann, A.-M. Moser, S. Michel, A.-M. Bertram, A. Maatz, G. Seiler, E. Stark, et al., “Linguistic features of suicidal thoughts and behaviors: A systematic review,”Clinical psychology review, vol. 95, pp. 102161, 2022

    2022

  2. [10]

    Acoustic and language analysis of speech for suicidal ideation among US veterans,

    A. Belouali, S. Gupta, V . Sourirajan, J. Yu, N. Allen, A. Alaoui, M. Dutton, and M. Reinhard, “Acoustic and language analysis of speech for suicidal ideation among US veterans,”BioData Mining, vol. 14, pp. 11, 2021

    2021

  3. [11]

    Spontaneous speech-based suicide risk detection using whisper and large language models,

    Z. Cui, C. Lei, W. Wu, Y . Duan, D. Qu, J. Wu, R. Chen, and C. Zhang, “Spontaneous speech-based suicide risk detection using whisper and large language models,” inProc. Inter- speech, Kos Island, 2024

    2024

  4. [12]

    The 1st SpeechWellness Challenge: Detecting suicide risk among adolescents,

    W. Wu, Z. Cui, C. Lei, Y . Duan, D. Qu, J. Wu, B. Zhou, R. Chen, and C. Zhang, “The 1st SpeechWellness Challenge: Detecting suicide risk among adolescents,” inProc. Inter- speech, Rotterdam, 2025

    2025

  5. [13]

    A machine learning-based case–control study on suicide risk identification: Integrating acoustic and linguistic features un- der stress conditions,

    Q. Lin, J. Zhang, W. Wang, C. Tan, X. Wu, and J. Zhao, “A machine learning-based case–control study on suicide risk identification: Integrating acoustic and linguistic features un- der stress conditions,”Depression and Anxiety, vol. 2025, no. 1, pp. 1671972, 2025

    2025

  6. [14]

    Towards speech emotion recognition “in the wild

    J. Kim, G. Englebienne, K. Truong, and V . Evers, “Towards speech emotion recognition “in the wild”’ using aggregated corpora and deep multi-task learning,” inProc. Interspeech, Sockholm, 2017

    2017

  7. [15]

    Performance of machine learning algorithms for dementia assessment: Impacts of language tasks, recording media, and modalities,

    M. Parsapoor, M. Alam, and A. Mihailidis, “Performance of machine learning algorithms for dementia assessment: Impacts of language tasks, recording media, and modalities,”BMC Medical Informatics and Decision Making, vol. 23, pp. 45, 2023

    2023

  8. [16]

    Em- powering large language models for end-to-end speech trans- lation leveraging synthetic data,

    Y . Pu, X. Liu, G. Zhang, Z. Yan, W. Zhang, and X. Chen, “Em- powering large language models for end-to-end speech trans- lation leveraging synthetic data,” inProc. Interspeech, 2025, pp. 26–30

    2025

  9. [17]

    SALMONN: Towards generic hearing abilities for large language models,

    C. Tang, W. Yu, G. Sun, X. Chen, T. Tan, W. Li, L. Lu, Z. Ma, and C. Zhang, “SALMONN: Towards generic hearing abilities for large language models,” inProc. ICLR, Vienna, 2024

    2024

  10. [18]

    Audiopalm: A large language model that can speak and listen,

    K. Rubenstein, C. Asawaroengchai, D. Nguyen, A. Bapna, Z. Borsos, F. Quitry, P. Chen, D. Badawy, Wei Han, Eugene Kharitonov, et al., “Audiopalm: A large language model that can speak and listen,”arXiv preprint arXiv:2306.12925, 2023

    2023 arXiv

  11. [19]

    Mixture of experts: A literature survey,

    S. Masoudnia and R. Ebrahimpour, “Mixture of experts: A literature survey,”Artificial Intelligence Review, vol. 42, pp. 275–293, 2014

    2014

  12. [20]

    Locmoe: A low-overhead moe for large language model training,

    J. Li, Z. Sun, X. He, L. Zeng, Y . Lin, E. Li, B. Zheng, R. Zhao, and X. Chen, “Locmoe: A low-overhead moe for large language model training,”arXiv preprint arXiv:2401.13920, 2024

    2024 arXiv

  13. [21]

    Effi- cient large scale language modeling with mixtures of experts,

    M. Artetxe, S. Bhosale, N. Goyal, T. Mihaylov, M. Ott, S. Shleifer, X. Lin, J. Du, S. Iyer, R. Pasunuru, et al., “Effi- cient large scale language modeling with mixtures of experts,” arXiv preprint arXiv:2112.10684, 2021

    2021 arXiv

  14. [22]

    Moe-pruner: Pruning mixture-of-experts large language model using the hints from its router,

    Y . Xie, Z. Zhang, D. Zhou, C. Xie, Z. Song, X. Liu, Y . Wang, X. Lin, and A. Xu, “Moe-pruner: Pruning mixture-of-experts large language model using the hints from its router,”arXiv preprint arXiv:2410.12013, 2024

    2024 arXiv

  15. [23]

    Openmoe: An early effort on open mixture-of-experts lan- guage models,

    F. Xue, Z. Zheng, Y . Fu, J. Ni, Z. Zheng, W. Zhou, and Y . You, “Openmoe: An early effort on open mixture-of-experts lan- guage models,”arXiv preprint arXiv:2402.01739, 2024

    2024 arXiv

  16. [24]

    DoRA: Weight-decomposed low-rank adapta- tion,

    S. Liu, C. Wang, H. Yin, P. Molchanov, Y . Wang, K. Cheng, and M. Chen, “DoRA: Weight-decomposed low-rank adapta- tion,” inProc. ICML, Vienna, 2024

    2024

  17. [25]

    Reliability and validity of the mini international neuropsychiatric inter- view for children and adolescents (mini-kid),

    D. Sheehan, J. Janavs, K. Sheehan, R. Shytle, Y . Bannon, J. Rogers, K. Milo, S. Stock, and B. Wilkinson, “Reliability and validity of the mini international neuropsychiatric inter- view for children and adolescents (mini-kid),”The Journal of Clinical Psychiatry, vol. 71, no. ...

    2010

  18. [26]

    Verbal fluency: Language or executive function measure?,

    D. Whiteside, T. Kealey, M. Semla, H. Luu, L. Rice, M. Basso, and B. Roper, “Verbal fluency: Language or executive function measure?,”Applied Neuropsychology: Adult, vol. 23, pp. 29– 34, 2016

    2016

  19. [27]

    Blowing in the wind: Using ‘North Wind and the Sun’ texts to sample phoneme in- ventories,

    L. Baird, N. Evans, and S. Greenhill, “Blowing in the wind: Using ‘North Wind and the Sun’ texts to sample phoneme in- ventories,”Journal of the International Phonetic Association, vol. 52, pp. 453–494, 2022

    2022

  20. [28]

    Construction and analysis of emotional corpus,

    L. Xu, H. Lin, and J. Zhao, “Construction and analysis of emotional corpus,”Journal of Chinese Information Process- ing, vol. 22, pp. 116–122, 2008

    2008

  21. [29]

    Study on Chinese word sentiment po- larity automatic estimation,

    J. Zhang and H. Jin, “Study on Chinese word sentiment po- larity automatic estimation,”Jisuanji Gongcheng/ Computer Engineering, vol. 36, 2010

    2010

  22. [30]

    The racially diverse affective expression (RADIATE) face stimulus set,

    M. Conley, D. Dellarco, E. Rubien-Thomas, A. Cohen, A. Cervera, N. Tottenham, and B. Casey, “The racially diverse affective expression (RADIATE) face stimulus set,”Psychiatry research, vol. 270, pp. 1059–1067, 2018

    2018

  23. [31]

    Divergent thinking as an indicator of creative potential,

    M. Runco and S. Acar, “Divergent thinking as an indicator of creative potential,”Creativity research journal, vol. 24, pp. 66–75, 2012

    2012

  24. [32]

    LoRA: Low-rank adaptation of large language models,

    E. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-rank adaptation of large language models,” inProc. ICLR, 2022

    2022

  25. [33]

    On calibration of modern neural networks,

    C. Guo, G. Pleiss, Y . Sun, and K. Weinberger, “On calibration of modern neural networks,” inProc. ICML, Sydney, 2017

    2017

  26. [34]

    Qwen2.5-omni technical report,

    J. Xu, Z. Guo, J. He, H. Hu, T. He, S. Bai, K. Chen, J. Wang, Y . Fan, K. Dang, et al., “Qwen2.5-omni technical report,” arXiv preprint arXiv:2503.20215, 2025

    2025 arXiv

  27. [35]

    Robust speech recognition via large-scale weak supervision,

    A. Radford, J. Kim, T. Xu, G. Brockman, C. McLeavey, and I. Sutskever, “Robust speech recognition via large-scale weak supervision,” inProc. ICML, Honolulu, 2023

    2023

  28. [36]

    Obtaining well calibrated probabilities using bayesian binning,

    M. P. Naeini, G. Cooper, and M. Hauskrecht, “Obtaining well calibrated probabilities using bayesian binning,” inProc. AAAI, Austin, 2015

    2015

  29. [37]

    Improved estimation, evaluation and applications of confidence measures for speech recognition.,

    M. h. Siu, H. Gish, and F. Richardson, “Improved estimation, evaluation and applications of confidence measures for speech recognition.,” inProc. Eurospeech, Rhodes Island, 1997. 5

    1997

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.