Pith. sign in

REVIEW 2 major objections 3 minor 145 references

Trust in AI for digital health depends on treating robustness and explainability as one design problem, this review argues, and it organizes methods, application challenges, and evaluation metrics into a single framework.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 10:52 UTC pith:CPUKRHEN

load-bearing objection A mostly accurate, practitioner-friendly review of robustness and explainability in digital health, weakened mainly by an overclaimed 'unique gap' and heavy reliance on the authors' own prior systems. the 2 major comments →

arxiv 2608.02238 v1 pith:CPUKRHEN submitted 2026-08-03 cs.AI cs.LG

Trustworthy AI in Digital Health: A Comprehensive Review of Robustness and Explainability

Abdullah Mamun , Shovito Barua Soumma , Hassan Ghasemzadeh This is my paper
classification cs.AI cs.LG
keywords trustworthy AIdigital healthrobustnessexplainabilitycounterfactual explanationsevaluation metricsmachine learning lifecycleXAI
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Trustworthy AI for digital health is often split into two separate research streams: work on robustness (surviving missing data, sensor failure, noisy labels, and distribution shifts) and work on explainability (attribution methods, saliency maps, and counterfactual explanations). This review's central move is to combine them, contending that a health AI cannot earn clinical trust unless it is both reliable in the wild and able to explain its decisions in actionable terms. It does so by mapping challenges across the machine learning lifecycle, surveying application-specific trust concerns in areas such as intensive care, neonatal health, and metabolic monitoring, and assembling an evaluation toolkit—validity, fidelity, proximity, sparsity, diversity, and a trust coefficient—that makes trust measurable. If its synthesis holds, developers and regulators gain a shared frame for judging whether a system is worth deploying: check its robustness under realistic stress and check whether its explanations survive the same scrutiny.

Core claim

Central claim: the digital-health AI literature has developed robustness and explainability on separate tracks, and this split blocks real deployment. The authors argue the two are interdependent—a model that fails when a sensor is lost is untrustworthy however clear its explanations, and an accurate model whose reasoning cannot be inspected offers little decision support. The review's contribution is a unified frame: lifecycle challenges, application-specific trust demands, and an evaluation-metric taxonomy that make robustness and explainability commensurable and auditable on the same system.

What carries the argument

Two organizing devices carry the argument. The first is a lifecycle-and-domain map: it situates trust challenges—noise, missing data, class imbalance, distribution shift, opacity—at stages from problem definition through data collection, training, evaluation, and inference, and then links those challenges to both robustness and explainability methods and to clinical domains such as intensive care, neonatal care, and metabolic health. The second is the evaluation metric set: validity, fidelity, proximity, sparsity, diversity, and a mutual-information-based trust coefficient, which converts 'trust' from a slogan into measurable, comparable quantities. These devices allow the review to argue th

Load-bearing premise

The claim that this review uniquely bridges the robustness-explainability gap rests on the assumption that the cited works fairly represent the field and that the authors' own health-AI systems, featured prominently in the application showcases, are typical rather than favorable examples; no systematic search protocol is documented to ensure representativeness.

What would settle it

A protocol-driven systematic search of the digital-health AI literature combining robustness and explainability terms would falsify the uniqueness claim if it surfaces earlier reviews that already treat both dimensions together; alternatively, a blind clinical study showing that systems scoring high on the proposed metrics (validity, fidelity, proximity, sparsity, diversity) are not trusted more by clinicians than low-scoring systems would challenge the framework's practical value.

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

If this is right

  • Health AI products would come with a paired report: performance under sensor failure, missing channels, and distribution shift alongside validity, fidelity, and diversity scores for their explanations, making trust claims checkable.
  • The metric taxonomy gives researchers and clinicians a common vocabulary, so a counterfactual method validated in one domain can be compared directly with a saliency-map method in another.
  • Application-specific concerns—imbalance in neonatal risk data, missing wearables in metabolic tracking, label noise in free-living activity—map onto a menu of existing robust-learning techniques, pointing practitioners straight to solutions.
  • For LLM-based health assistants, the same standard applies: they need both stress-testing for hallucinated or outdated information and verifiable reasoning traces that clinicians can audit.

Where Pith is reading between the lines

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

  • A natural but untested consequence of the framework is that robustness techniques (augmentation, imputation, balancing) should also stabilize explanations; researchers could measure whether explanations change less when inputs are perturbed, unifying the two literatures empirically.
  • The metric set could be extended with uncertainty-aware criteria—for example, confidence intervals on attribution scores or counterfactual validity under distribution shift—making trust evaluations more honest in high-stakes settings.
  • The lifecycle map implies a regulatory audit template: a device submission would document robustness stress tests at each stage and explanation-quality metrics at inference, turning the review's frame into a checklist for approval processes.
  • A next synthesis might connect this framework to causal reasoning, testing whether counterfactual explanations that are causally faithful (not just prediction-changing) produce better clinician decisions than purely associative ones.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 3 minor

Summary. The paper is a narrative review of robustness and explainability as pillars of trustworthy AI in digital health. It opens with a survey of existing trustworthy-AI reviews and situates its contribution as jointly synthesizing robustness and explainability for healthcare. It then discusses application-specific trust concerns (radiology, cardiology, metabolic health, neonatal care, mental health, brain health, ICU, public health), robustness methods under data scarcity and sensor failure, explainable AI methods (LIME, SHAP, LRP, GradCAM, Integrated Gradients, NICE, DiCE, CFNOW), LLM-era trust issues, and evaluation metrics with formal definitions for validity, fidelity, proximity, sparsity, diversity, and trust. Many illustrative systems are drawn from the authors' own prior work (AIMEN, GlucoLens, CUDLE, sensor-failure reconstruction, MetaBoost), and the paper is framed as 'uniquely bridging' a gap in the literature.

Significance. If accepted as a synthesis, the review provides a useful entry point for researchers and practitioners: the descriptions of standard XAI methods are accurate; the taxonomy in Tables 1–4 is clearly organized; and the metric equations in Section 8 are a practical resource. The paper also explicitly engages with prior reviews, which strengthens its reliability as a survey. The main contribution is organizational rather than technical: it collects and links robustness and explainability concepts in a healthcare context. This is valuable despite the overstatement of uniqueness, and the paper's strengths lie in its breadth and clarity rather than in novel methodology.

major comments (2)
  1. [Section 2.1] The claim that 'This review uniquely bridges that gap' is contradicted by the paper's own Table 1, which lists Albahri et al. (2023) and Band et al. (2023) as reviews that jointly address trustworthiness and explainability in healthcare. Since this novelty claim is the primary positioning of the paper, it should be revised to a more defensible statement, e.g., that the review focuses specifically on digital-health applications and integrates robustness and explainability with a metric-oriented perspective, rather than claiming uniqueness.
  2. [Section 4.4 and Tables 2–3] The selection of illustrative systems is heavily weighted toward the authors' own prior work (AIMEN, GlucoLens, CUDLE, sensor-failure reconstruction, MetaBoost). Because the review does not document a systematic search strategy or inclusion criteria, the representativeness of this selection is unclear. This directly affects the 'comprehensive' claim. The authors should either add a methodology/limitations section disclosing the selection process and the self-citation concentration, or include more independent systems with comparable functionality.
minor comments (3)
  1. [Section 5.3] The heading 'Regularization and Novel Frameworks for Model Robustness' appears inside the Explainability section. The content (orthogonality constraints and hallucination detection) is about model robustness more than explanation, so the heading may mislead readers; consider renaming to reflect the joint robustness/explainability focus.
  2. [Section 8.2.4] The proximity equations use negative distances, but the text does not explicitly define the sign convention. State that Proximity_cont is a negated average distance so that higher values mean closer (better) counterfactuals.
  3. [References] Reference [94] lists the publisher as 'Lulu. com' with a space; correct to 'Lulu.com'. Also, some arXiv preprints (e.g., [12], [83], [88]) are cited without a version or access date; consider adding them for reproducibility.

Circularity Check

0 steps flagged

No significant circularity: the review is descriptive, externally grounded, and has no fitted-input/prediction chain.

full rationale

This is a narrative review rather than a derivation, so the standard circularity failure modes (fit-vs-prediction, definitional equivalence, ansatz smuggled in via citation) do not arise. The technical content—LIME, SHAP, Integrated Gradients, GradCAM, NICE, DiCE, and the evaluation metrics in Section 8—is described as standard external methods and is not used to justify any new predictive claim. Although several illustrative applications (AIMEN [85], GlucoLens [83], CUDLE [15], sensor-failure reconstruction [89]) are the authors' own work, they are presented as examples of robustness/explainability, not as evidence that forces the review's central claims; Table 1 explicitly acknowledges overlapping reviews (Albahri et al. [6], Band et al. [18], Ojha et al. [102]), so the 'uniquely bridges that gap' statement is an overclaim about novelty rather than a self-referential derivation. No equation in the paper reduces to its own input, and no fitted parameter is relabeled as a prediction. Accordingly, there is no significant circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

A review has no fitted parameters or invented entities. The burdens are conceptual: accepting a particular definitional framework (NIST/EU), assuming the literature selection is unbiased, and treating the authors' own prior systems as representative examples.

axioms (3)
  • domain assumption The NIST and EU definitions of trustworthy AI (robustness, explainability, fairness, privacy, accountability) are accepted as the organizing framework.
    Section 1 adopts these definitions as the backbone of the review without comparing them with alternative trust frameworks; the paper's structure depends on this choice.
  • domain assumption The 145 cited references constitute a representative sample of the relevant literature.
    No systematic search strategy, databases, or inclusion/exclusion criteria are reported; the review's comprehensiveness claim presumes representativeness.
  • ad hoc to paper The authors' own prior systems (CUDLE, AIMEN, GlucoLens, sensor-failure reconstruction, MetaBoost) are valid and representative exemplars of robust/explainable digital health AI.
    Sections 4 and 5.4 present these systems as the primary evidence that the proposed framework is practical; if those works are flawed or unrepresentative, the review's conclusions weaken.

pith-pipeline@v1.3.0-daily-deepseek · 22971 in / 12231 out tokens · 93418 ms · 2026-08-04T10:52:30.695849+00:00 · methodology

0 comments
read the original abstract

Ensuring trust in AI systems is essential for the safe and ethical integration of machine learning systems into high-stakes domains such as digital health. Key dimensions, including robustness, explainability, fairness, accountability, and privacy, need to be addressed throughout the AI lifecycle, from problem formulation and data collection to model deployment and human interaction. While various contributions address different aspects of trustworthy AI, a focused synthesis on robustness and explainability, especially tailored to the healthcare context, remains limited. This review addresses that need by organizing recent advancements into an accessible framework, highlighting both technical and practical considerations. We present a structured overview of methods, challenges, and solutions, aiming to support researchers and practitioners in developing reliable and explainable AI solutions for digital health. This review article is organized into three main parts. First, we introduce the pillars of trustworthy AI and discuss the technical and ethical challenges, particularly in the context of digital health. Second, we explore application-specific trust considerations across domains such as intensive care, neonatal health, and metabolic health, highlighting how robustness and explainability support trust. Lastly, we present recent advancements in techniques aimed at improving robustness under data scarcity and distributional shifts, as well as explainable AI methods ranging from feature attribution to gradient-based interpretations and counterfactual explanations. This paper is further enriched with detailed discussions of the contributions toward robustness and explainability in digital health, the development of trustworthy AI systems in the era of LLMs, and various evaluation metrics for measuring trust and related parameters such as validity, fidelity, and diversity.

Figures

Figures reproduced from arXiv: 2608.02238 by Abdullah Mamun, Hassan Ghasemzadeh, Shovito Barua Soumma.

Figure 1
Figure 1. Figure 1: Examples implicating the importance of robustness and explainability in AI systems. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Challenges in machine learning systems in different phases of the development lifecycle. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Example of a LIME plot for breast cancer detection. In this example, we applied LIME on [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example of a SHAP BeeSwarm plot for breast cancer detection. We applied SHAP on the [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Example of GradCAM and Integrated Gradients plots for detecting pneumonia from [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

145 extracted references · 24 linked inside Pith

  1. [1]

    [Accessed July 2025]

    CE Marking trade.gov.https://www.trade.gov/ce-marking. [Accessed July 2025]

    2025

  2. [2]

    https://digital-strategy.ec.europa.eu/en/library/ ethics-guidelines-trustworthy-ai

    Ethics guidelines for trustworthy AI — digital-strategy.ec.europa.eu. https://digital-strategy.ec.europa.eu/en/library/ ethics-guidelines-trustworthy-ai. [Accessed July 2025]

    2025

  3. [3]

    https://scikit-learn.org/stable/modules/ generated/sklearn.datasets.load_breast_cancer.html

    load_breast_cancer — scikit-learn.org. https://scikit-learn.org/stable/modules/ generated/sklearn.datasets.load_breast_cancer.html. [Accessed July 2025]

    2025

  4. [4]

    https://www.nist.gov/ trustworthy-and-responsible-ai

    Trustworthy and Responsible AI — nist.gov. https://www.nist.gov/ trustworthy-and-responsible-ai. [Accessed July 2025]

    2025

  5. [5]

    From attribution maps to human-understandable explanations through concept relevance propagation.Nature Machine Intelligence, 5(9):1006– 1019, 2023

    Reduan Achtibat, Maximilian Dreyer, Ilona Eisenbraun, Sebastian Bosse, Thomas Wiegand, Wojciech Samek, and Sebastian Lapuschkin. From attribution maps to human-understandable explanations through concept relevance propagation.Nature Machine Intelligence, 5(9):1006– 1019, 2023

    2023

  6. [6]

    A systematic review of trustworthy and explainable artificial intelligence in healthcare: Assessment of quality, bias risk, and data fusion.Information Fusion, 96:156–191, 2023

    Ahmed Shihab Albahri, Ali M Duhaim, Mohammed A Fadhel, Alhamzah Alnoor, Noor S Baqer, Laith Alzubaidi, Osamah Shihab Albahri, Abdullah Hussein Alamoodi, Jinshuai Bai, Asma Salhi, et al. A systematic review of trustworthy and explainable artificial intelligence in healthcare: Assessment of quality, bias risk, and data fusion.Information Fusion, 96:156–191, 2023

    2023

  7. [7]

    Actilabel: A combinatorial transfer learning framework for activity recognition.arXiv preprint arXiv:2003.07415, 2020

    Parastoo Alinia, Iman Mirzadeh, and Hassan Ghasemzadeh. Actilabel: A combinatorial transfer learning framework for activity recognition.arXiv preprint arXiv:2003.07415, 2020

    Pith/arXiv arXiv 2003

  8. [8]

    Evaluating the faithfulness of saliency maps in explaining deep learning models using realistic perturbations

    José P Amorim, Pedro H Abreu, João Santos, Marc Cortes, and Victor Vila. Evaluating the faithfulness of saliency maps in explaining deep learning models using realistic perturbations. Information Processing & Management, 60(2):103225, 2023

    2023

  9. [9]

    Deep learning for ecg arrhythmia detection and classification: an overview of progress for period 2017–2023

    Yaqoob Ansari, Omar Mourad, Khalid Qaraqe, and Erchin Serpedin. Deep learning for ecg arrhythmia detection and classification: an overview of progress for period 2017–2023. Frontiers in Physiology, 14:1246746, 2023

    2017

  10. [10]

    Designing user-centric behavioral interventions to prevent dysglycemia with novel counterfactual explanations.arXiv preprint arXiv:2310.01684, 2023

    Asiful Arefeen and Hassan Ghasemzadeh. Designing user-centric behavioral interventions to prevent dysglycemia with novel counterfactual explanations.arXiv preprint arXiv:2310.01684, 2023

    Pith/arXiv arXiv 2023

  11. [11]

    Glyman: Glycemic management using patient-centric counterfactuals

    Asiful Arefeen, Saman Khamesian, Maria Adela Grando, Bithika Thompson, and Hassan Ghasemzadeh. Glyman: Glycemic management using patient-centric counterfactuals. In2024 IEEE EMBS International Conference on Biomedical and Health Informatics (BHI), pages 1–5. IEEE, 2024

    2024

  12. [12]

    Glytwin: Digital twin for glucose control in type 1 diabetes through optimal be- havioral modifications using patient-centric counterfactuals.arXiv preprint arXiv:2504.09846, 2025

    Asiful Arefeen, Saman Khamesian, Maria Adela Grando, Bithika Thompson, and Hassan Ghasemzadeh. Glytwin: Digital twin for glucose control in type 1 diabetes through optimal be- havioral modifications using patient-centric counterfactuals.arXiv preprint arXiv:2504.09846, 2025

    Pith/arXiv arXiv 2025

  13. [13]

    Semi-factual explanations in ai

    Saugat Aryal. Semi-factual explanations in ai. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 23379–23380, 2024

    2024

  14. [14]

    A hybrid method for imputation of missing values using optimized fuzzy c-means with support vector regression and a genetic algorithm

    Ibrahim Berkan Aydilek and Ahmet Arslan. A hybrid method for imputation of missing values using optimized fuzzy c-means with support vector regression and a genetic algorithm. Information Sciences, 233:25–35, 2013

    2013

  15. [15]

    Cudle: Learning under label scarcity to detect cannabis use in uncontrolled environments using wearables.IEEE Sensors Journal, 2025

    Reza Rahimi Azghan, Nicholas C Glodosky, Ramesh Kumar Sah, Carrie Cuttler, Ryan McLaughlin, Michael J Cleveland, and Hassan Ghasemzadeh. Cudle: Learning under label scarcity to detect cannabis use in uncontrolled environments using wearables.IEEE Sensors Journal, 2025. 18

    2025

  16. [16]

    Robust counterfactual explanations on graph neural networks.Advances in Neural Information Processing Systems, 34:5644–5655, 2021

    Mohit Bajaj, Lingyang Chu, Zi Yu Xue, Jian Pei, Lanjun Wang, Peter Cho-Ho Lam, and Yong Zhang. Robust counterfactual explanations on graph neural networks.Advances in Neural Information Processing Systems, 34:5644–5655, 2021

    2021

  17. [17]

    Explainable ai is responsible ai: How explainability creates trustworthy and socially responsible artificial intelligence.arXiv preprint arXiv:2312.01555, 2023

    Stephanie Baker and Wei Xiang. Explainable ai is responsible ai: How explainability creates trustworthy and socially responsible artificial intelligence.arXiv preprint arXiv:2312.01555, 2023

    Pith/arXiv arXiv 2023

  18. [18]

    Application of explainable artificial intelligence in medical health: A systematic review of interpretability methods.Informatics in Medicine Unlocked, 40:101286, 2023

    Shahab S Band, Atefeh Yarahmadi, Chung-Chian Hsu, Meghdad Biyari, Mehdi Sookhak, Rasoul Ameri, Iman Dehzangi, Anthony Theodore Chronopoulos, and Huey-Wen Liang. Application of explainable artificial intelligence in medical health: A systematic review of interpretability methods.Informatics in Medicine Unlocked, 40:101286, 2023

    2023

  19. [19]

    Mhealth dataset.UCI machine learning repository, 2014

    Oresti Banos, Rafael Garcia, and Alejandro Saez. Mhealth dataset.UCI machine learning repository, 2014

    2014

  20. [20]

    Llama-nemotron: Efficient reasoning models.arXiv preprint arXiv:2505.00949, 2025

    Akhiad Bercovich, Itay Levy, Izik Golan, Mohammad Dabbah, Ran El-Yaniv, Omri Puny, Ido Galil, Zach Moshe, Tomer Ronen, Najeeb Nabwani, et al. Llama-nemotron: Efficient reasoning models.arXiv preprint arXiv:2505.00949, 2025

    arXiv 2025

  21. [21]

    Responsible development of clinical speech ai: Bridging the gap between clinical research and technology.NPJ Digital Medicine, 7(1):208, 2024

    Visar Berisha and Julie M Liss. Responsible development of clinical speech ai: Bridging the gap between clinical research and technology.NPJ Digital Medicine, 7(1):208, 2024

    2024

  22. [22]

    Jorge Bernal, Nima Tajkbaksh, Francisco Javier Sanchez, Bogdan J Matuszewski, Hao Chen, Lequan Yu, Quentin Angermann, Olivier Romain, Bjørn Rustad, Ilangko Balasingham, et al. Comparative validation of polyp detection methods in video colonoscopy: results from the miccai 2015 endoscopic vision challenge.IEEE transactions on medical imaging, 36(6):1231– 1249, 2017

    2015

  23. [23]

    Fine-tuning a llm using reinforcement learning from human feedback for a therapy chatbot application, 2023

    Desirée Bill and Theodor Eriksson. Fine-tuning a llm using reinforcement learning from human feedback for a therapy chatbot application, 2023

    2023

  24. [24]

    Layer-wise relevance propagation for neural networks with local renor- malization layers

    Alexander Binder, Grégoire Montavon, Sebastian Lapuschkin, Klaus-Robert Müller, and Wojciech Samek. Layer-wise relevance propagation for neural networks with local renor- malization layers. InArtificial Neural Networks and Machine Learning–ICANN 2016: 25th International Conference on Artificial Neural Networks, Barcelona, Spain, September 6-9, 2016, Proce...

    2016

  25. [25]

    Ethical issues in ai- enabled disease surveillance: perspectives from global health.Applied Sciences, 12(8):3890, 2022

    Ann Borda, Andreea Molnar, Cristina Neesham, and Patty Kostkova. Ethical issues in ai- enabled disease surveillance: perspectives from global health.Applied Sciences, 12(8):3890, 2022

    2022

  26. [26]

    Nice: an algorithm for nearest instance counterfactual explanations.Data mining and knowledge discovery, 38(5):2665–2703, 2024

    Dieter Brughmans, Pieter Leyman, and David Martens. Nice: an algorithm for nearest instance counterfactual explanations.Data mining and knowledge discovery, 38(5):2665–2703, 2024

    2024

  27. [27]

    Artificial intelligence revolution in turkish health consultancy: Development of llm-based virtual doctor assistants

    Muhammed Kayra Bulut and Banu Diri. Artificial intelligence revolution in turkish health consultancy: Development of llm-based virtual doctor assistants. In2024 8th International Artificial Intelligence and Data Processing Symposium (IDAP), pages 1–6. IEEE, 2024

    2024

  28. [28]

    e-snli: Nat- ural language inference with natural language explanations.Advances in Neural Information Processing Systems, 31, 2018

    Oana-Maria Camburu, Tim Rocktäschel, Thomas Lukasiewicz, and Phil Blunsom. e-snli: Nat- ural language inference with natural language explanations.Advances in Neural Information Processing Systems, 31, 2018

    2018

  29. [29]

    Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks

    Aditya Chattopadhay, Anirban Sarkar, Prantik Howlader, and Vineeth N Balasubramanian. Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks. In2018 IEEE winter conference on applications of computer vision (WACV), pages 839–847. IEEE, 2018

    2018

  30. [30]

    Rlhf deciphered: A critical analysis of reinforcement learning from human feedback for llms.ACM Computing Surveys, 2024

    Shreyas Chaudhari, Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, Ameet Deshpande, and Bruno Castro da Silva. Rlhf deciphered: A critical analysis of reinforcement learning from human feedback for llms.ACM Computing Surveys, 2024

    2024

  31. [31]

    Smote: synthetic minority over-sampling technique.Journal of artificial intelligence research, 16:321– 357, 2002

    Nitesh V Chawla, Kevin W Bowyer, Lawrence O Hall, and W Philip Kegelmeyer. Smote: synthetic minority over-sampling technique.Journal of artificial intelligence research, 16:321– 357, 2002

    2002

  32. [32]

    Diabetes: Non-invasive blood glucose monitoring using federated learning with biosensor signals.Biosensors, 15(4):255, 2025

    Narmatha Chellamani, Saleh Ali Albelwi, Manimurugan Shanmuganathan, Palanisamy Amirthalingam, and Anand Paul. Diabetes: Non-invasive blood glucose monitoring using federated learning with biosensor signals.Biosensors, 15(4):255, 2025. 19

    2025

  33. [33]

    Encoder-decoder with atrous separable convolution for semantic image segmentation

    Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. Encoder-decoder with atrous separable convolution for semantic image segmentation. In Proceedings of the European conference on computer vision (ECCV), pages 801–818, 2018

    2018

  34. [34]

    Do models explain themselves? counterfactual simulatability of natural language explanations

    Yanda Chen, Ruiqi Zhong, Narutatsu Ri, Chen Zhao, He He, Jacob Steinhardt, Zhou Yu, and Kathleen McKeown. Do models explain themselves? counterfactual simulatability of natural language explanations. InProceedings of the 41st International Conference on Machine Learning, pages 7880–7904, 2024

    2024

  35. [35]

    How machine learning is used to study addiction in digital healthcare: A systematic review.International Journal of Information Management Data Insights, 3(2):100175, 2023

    Bijoy Chhetri, Lalit Mohan Goyal, and Mamta Mittal. How machine learning is used to study addiction in digital healthcare: A systematic review.International Journal of Information Management Data Insights, 3(2):100175, 2023

    2023

  36. [36]

    Role of orthogonality constraints in improving properties of deep networks for image classification.arXiv preprint arXiv:2009.10762, 2020

    Hongjun Choi, Anirudh Som, and Pavan Turaga. Role of orthogonality constraints in improving properties of deep networks for image classification.arXiv preprint arXiv:2009.10762, 2020

    Pith/arXiv arXiv 2009

  37. [37]

    Evaluating visual explanations of attention maps for transformer-based medical imaging

    Minjae Chung, Jong Bum Won, Ganghyun Kim, Yujin Kim, and Utku Ozbulak. Evaluating visual explanations of attention maps for transformer-based medical imaging. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 110–120. Springer, 2024

    2024

  38. [38]

    The future landscape of large language models in medicine.Communications medicine, 3(1):141, 2023

    Jan Clusmann, Fiona R Kolbinger, Hannah Sophie Muti, Zunamys I Carrero, Jan-Niklas Eckardt, Narmin Ghaffari Laleh, Chiara Maria Lavinia Löffler, Sophie-Caroline Schwarzkopf, Michaela Unger, Gregory P Veldhuizen, et al. The future landscape of large language models in medicine.Communications medicine, 3(1):141, 2023

    2023

  39. [39]

    Evaluation of individual and ensemble probabilistic forecasts of covid-19 mortality in the united states.Proceedings of the National Academy of Sciences, 119(15):e2113561119, 2022

    Estee Y Cramer, Evan L Ray, Velma K Lopez, Johannes Bracher, Andrea Brennen, Alvaro J Castro Rivadeneira, Aaron Gerding, Tilmann Gneiting, Katie H House, Yuxin Huang, et al. Evaluation of individual and ensemble probabilistic forecasts of covid-19 mortality in the united states.Proceedings of the National Academy of Sciences, 119(15):e2113561119, 2022

    2022

  40. [40]

    Bias in medical ai: Implications for clinical decision-making.PLOS Digital Health, 3(11):e0000651, 2024

    James L Cross, Michael A Choma, and John A Onofrey. Bias in medical ai: Implications for clinical decision-making.PLOS Digital Health, 3(11):e0000651, 2024

    2024

  41. [41]

    Multi-objective coun- terfactual explanations

    Susanne Dandl, Christoph Molnar, Martin Binder, and Bernd Bischl. Multi-objective coun- terfactual explanations. InInternational conference on parallel problem solving from nature, pages 448–469. Springer, 2020

    2020

  42. [42]

    A multitask learning approach to per- sonalized blood glucose prediction.IEEE Journal of Biomedical and Health Informatics, 26(1):436–445, 2021

    John Daniels, Pau Herrero, and Pantelis Georgiou. A multitask learning approach to per- sonalized blood glucose prediction.IEEE Journal of Biomedical and Health Informatics, 26(1):436–445, 2021

    2021

  43. [43]

    Raphael Mazzine Barbosa de Oliveira, Kenneth Sörensen, and David Martens. A model- agnostic and data-independent tabu search algorithm to generate counterfactuals for tabular, image, and text data.European Journal of Operational Research, 317(2):286–302, 2024

    2024

  44. [44]

    Towards a rigorous science of interpretable machine learning.arXiv preprint arXiv:1702.08608, 2017

    Finale Doshi-Velez and Been Kim. Towards a rigorous science of interpretable machine learning.arXiv preprint arXiv:1702.08608, 2017

    Pith/arXiv arXiv 2017

  45. [45]

    Revealing hidden context bias in segmentation and object detection through concept-specific explanations

    Maximilian Dreyer, Reduan Achtibat, Thomas Wiegand, Wojciech Samek, and Sebastian Lapuschkin. Revealing hidden context bias in segmentation and object detection through concept-specific explanations. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3828–3838, 2023

    2023

  46. [46]

    Haloscope: Harnessing unlabeled llm generations for hallucination detection.Advances in Neural Information Processing Systems, 37:102948– 102972, 2025

    Xuefeng Du, Chaowei Xiao, and Sharon Li. Haloscope: Harnessing unlabeled llm generations for hallucination detection.Advances in Neural Information Processing Systems, 37:102948– 102972, 2025

    2025

  47. [47]

    The llama 3 herd of models.arXiv preprint arXiv:2407.21783, 2024

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. The llama 3 herd of models.arXiv preprint arXiv:2407.21783, 2024

    Pith/arXiv arXiv 2024

  48. [48]

    Attengluco: Multimodal transformer-based blood glucose forecasting on ai-readi dataset

    Ebrahim Farahmand, Reza Rahimi Azghan, Nooshin Taheri Chatrudi, Eric Kim, Gautham Kr- ishna Gudur, Edison Thomaz, Giulia Pedrielli, Pavan Turaga, and Hassan Ghasemzadeh. Attengluco: Multimodal transformer-based blood glucose forecasting on ai-readi dataset. arXiv preprint arXiv:2502.09919, 2025

    Pith/arXiv arXiv 2025

  49. [49]

    A trustworthy ai reality-check: the lack of transparency of artificial intelligence products in healthcare.Frontiers in Digital Health, 6:1267290, 2024

    Jana Fehr, Brian Citro, Rohit Malpani, Christoph Lippert, and Vince I Madai. A trustworthy ai reality-check: the lack of transparency of artificial intelligence products in healthcare.Frontiers in Digital Health, 6:1267290, 2024. 20

    2024

  50. [50]

    Towards an ai co-scientist.arXiv preprint arXiv:2502.18864, 2025

    Juraj Gottweis, Wei-Hung Weng, Alexander Daryin, Tao Tu, Anil Palepu, Petar Sirkovic, Artiom Myaskovsky, Felix Weissenberger, Keran Rong, Ryutaro Tanno, et al. Towards an ai co-scientist.arXiv preprint arXiv:2502.18864, 2025

    Pith/arXiv arXiv 2025

  51. [51]

    Interpretable machine learning model for new-onset atrial fibrillation prediction in critically ill patients: a multi-center study.Critical Care, 28(1):349, 2024

    Chengjian Guan, Angwei Gong, Yan Zhao, Chen Yin, Lu Geng, Linli Liu, Xiuchun Yang, Jingchao Lu, and Bing Xiao. Interpretable machine learning model for new-onset atrial fibrillation prediction in critically ill patients: a multi-center study.Critical Care, 28(1):349, 2024

    2024

  52. [52]

    Domain adaptation for medical image analysis: a survey.IEEE Transactions on Biomedical Engineering, 69(3):1173–1185, 2021

    Hao Guan and Mingxia Liu. Domain adaptation for medical image analysis: a survey.IEEE Transactions on Biomedical Engineering, 69(3):1173–1185, 2021

    2021

  53. [53]

    Counterfactual explanations and how to find them: literature review and benchmarking.Data Mining and Knowledge Discovery, 38(5):2770–2824, 2024

    Riccardo Guidotti. Counterfactual explanations and how to find them: literature review and benchmarking.Data Mining and Knowledge Discovery, 38(5):2770–2824, 2024

    2024

  54. [54]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

    Pith/arXiv arXiv 2025

  55. [55]

    Trustworthy artificial intelligence in medical imaging.PET clinics, 17(1):1, 2022

    Navid Hasani, Michael A Morris, Arman Rhamim, Ronald M Summers, Elizabeth Jones, Eliot Siegel, and Babak Saboury. Trustworthy artificial intelligence in medical imaging.PET clinics, 17(1):1, 2022

    2022

  56. [56]

    Adasyn: Adaptive synthetic sampling approach for imbalanced learning

    Haibo He, Yang Bai, Edwardo A Garcia, and Shutao Li. Adasyn: Adaptive synthetic sampling approach for imbalanced learning. In2008 IEEE international joint conference on neural networks (IEEE world congress on computational intelligence), pages 1322–1328. Ieee, 2008

    2008

  57. [57]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. Masked autoencoders are scalable vision learners. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16000–16009, 2022

    2022

  58. [58]

    Exploring inherent sensor redundancy for automotive anomaly detection

    Tianjia He, Lin Zhang, Fanxin Kong, and Asif Salekin. Exploring inherent sensor redundancy for automotive anomaly detection. In2020 57th ACM/IEEE Design Automation Conference (DAC), pages 1–6. IEEE, 2020

    2020

  59. [59]

    A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions.ACM Transactions on Information Systems, 43(2):1–55, 2025

    Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qiang- long Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, et al. A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions.ACM Transactions on Information Systems, 43(2):1–55, 2025

    2025

  60. [60]

    Self-supervised learning for medical image classification: a systematic review and implementation guidelines.NPJ Digital Medicine, 6(1):74, 2023

    Shih-Cheng Huang, Anuj Pareek, Malte Jensen, Matthew P Lungren, Serena Yeung, and Akshay S Chaudhari. Self-supervised learning for medical image classification: a systematic review and implementation guidelines.NPJ Digital Medicine, 6(1):74, 2023

    2023

  61. [61]

    Energy-efficient missing data recovery in wearable devices: A novel search-based approach

    Dina Hussein, Taha Belkhouja, Ganapati Bhat, and Janardhan Rao Doppa. Energy-efficient missing data recovery in wearable devices: A novel search-based approach. In2023 IEEE/ACM International Symposium on Low Power Electronics and Design (ISLPED), pages 1–6. IEEE, 2023

    2023

  62. [62]

    Cim: A novel clustering-based energy-efficient data impu- tation method for human activity recognition.ACM Transactions on Embedded Computing Systems, 22(5s):1–26, 2023

    Dina Hussein and Ganapati Bhat. Cim: A novel clustering-based energy-efficient data impu- tation method for human activity recognition.ACM Transactions on Embedded Computing Systems, 22(5s):1–26, 2023

    2023

  63. [63]

    Openai o1 system card

    Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, et al. Openai o1 system card. arXiv preprint arXiv:2412.16720, 2024

    Pith/arXiv arXiv 2024

  64. [64]

    Requirements for trustworthy artificial intelligence–a review

    Davinder Kaur, Suleyman Uslu, and Arjan Durresi. Requirements for trustworthy artificial intelligence–a review. InAdvances in Networked-Based Information Systems: The 23rd International Conference on Network-Based Information Systems (NBiS-2020) 23, pages 105–115. Springer, 2021

    2020

  65. [65]

    Trustworthy artificial intelligence: a review.ACM computing surveys (CSUR), 55(2):1–38, 2022

    Davinder Kaur, Suleyman Uslu, Kaley J Rittichier, and Arjan Durresi. Trustworthy artificial intelligence: a review.ACM computing surveys (CSUR), 55(2):1–38, 2022

    2022

  66. [66]

    Identifying medical diagnoses and treatable diseases by image-based deep learning.cell, 172(5):1122– 1131, 2018

    Daniel S Kermany, Michael Goldbaum, Wenjia Cai, Carolina CS Valentim, Huiying Liang, Sally L Baxter, Alex McKeown, Ge Yang, Xiaokang Wu, Fangbing Yan, et al. Identifying medical diagnoses and treatable diseases by image-based deep learning.cell, 172(5):1122– 1131, 2018. 21

    2018

  67. [67]

    The promise and perils of artificial intelligence in advancing participatory science and health equity in public health.JMIR Public Health and Surveillance, 11(1):e65699, 2025

    Abby C King, Zakaria N Doueiri, Ankita Kaulberg, and Lisa Goldman Rosas. The promise and perils of artificial intelligence in advancing participatory science and health equity in public health.JMIR Public Health and Surveillance, 11(1):e65699, 2025

    2025

  68. [68]

    Large language models are zero-shot reasoners.Advances in neural information processing systems, 35:22199–22213, 2022

    Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. Large language models are zero-shot reasoners.Advances in neural information processing systems, 35:22199–22213, 2022

    2022

  69. [69]

    Establishing and evaluating trustworthy ai: overview and research challenges.Frontiers in Big Data, 7:1467222, 2024

    Dominik Kowald, Sebastian Scher, Viktoria Pammer-Schindler, Peter Müllner, Kerstin Waxneg- ger, Lea Demelius, Angela Fessl, Maximilian Toller, Inti Gabriel Mendoza Estrada, Ilija Šimi´c, et al. Establishing and evaluating trustworthy ai: overview and research challenges.Frontiers in Big Data, 7:1467222, 2024

    2024

  70. [70]

    Self-supervised learning in medicine and healthcare.Nature Biomedical Engineering, 6(12):1346–1352, 2022

    Rayan Krishnan, Pranav Rajpurkar, and Eric J Topol. Self-supervised learning in medicine and healthcare.Nature Biomedical Engineering, 6(12):1346–1352, 2022

    2022

  71. [71]

    Trustworthy ai in the age of pervasive computing and big data

    Abhishek Kumar, Tristan Braud, Sasu Tarkoma, and Pan Hui. Trustworthy ai in the age of pervasive computing and big data. In2020 IEEE International Conference on Pervasive Computing and Communications Workshops (PerCom Workshops), pages 1–6. IEEE, 2020

    2020

  72. [72]

    dr llm, what do i have?

    Wojciech Kusa, Edoardo Mosca, and Aldo Lipani. “dr llm, what do i have?”: The impact of user beliefs and prompt formulation on health diagnoses. InProceedings of the Third Workshop on NLP for Medical Conversations, pages 13–19, 2023

    2023

  73. [73]

    Retrieval- augmented generation for knowledge-intensive nlp tasks.Advances in Neural Information Processing Systems, 33:9459–9474, 2020

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Na- man Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, et al. Retrieval- augmented generation for knowledge-intensive nlp tasks.Advances in Neural Information Processing Systems, 33:9459–9474, 2020

    2020

  74. [74]

    Trustworthy ai: From principles to practices.ACM Computing Surveys, 55(9):1–46, 2023

    Bo Li, Peng Qi, Bo Liu, Shuai Di, Jingen Liu, Jiquan Pei, Jinfeng Yi, and Bowen Zhou. Trustworthy ai: From principles to practices.ACM Computing Surveys, 55(9):1–46, 2023

    2023

  75. [75]

    Yolov6: A single-stage object detection framework for industrial applications.arXiv preprint arXiv:2209.02976, 2022

    Chuyi Li, Lulu Li, Hongliang Jiang, Kaiheng Weng, Yifei Geng, Liang Li, Zaidan Ke, Qingyuan Li, Meng Cheng, Weiqiang Nie, et al. Yolov6: A single-stage object detection framework for industrial applications.arXiv preprint arXiv:2209.02976, 2022

    Pith/arXiv arXiv 2022

  76. [76]

    Early prediction of alzheimer’s disease and related dementias using real-world electronic health records.Alzheimer’s & Dementia, 19(8):3506–3518, 2023

    Qian Li, Xi Yang, Jie Xu, Yi Guo, Xing He, Hui Hu, Tianchen Lyu, David Marra, Amber Miller, Glenn Smith, et al. Early prediction of alzheimer’s disease and related dementias using real-world electronic health records.Alzheimer’s & Dementia, 19(8):3506–3518, 2023

    2023

  77. [77]

    Llms for relational reasoning: How far are we? InProceedings of the 1st International Workshop on Large Language Models for Code, pages 119–126, 2024

    Zhiming Li, Yushi Cao, Xiufeng Xu, Junzhe Jiang, Xu Liu, Yon Shin Teo, Shang-Wei Lin, and Yang Liu. Llms for relational reasoning: How far are we? InProceedings of the 1st International Workshop on Large Language Models for Code, pages 119–126, 2024

    2024

  78. [78]

    Improving llm reasoning through scaling inference computation with collaborative verification

    Zhenwen Liang, Ye Liu, Tong Niu, Xiangliang Zhang, Yingbo Zhou, and Semih Yavuz. Improving llm reasoning through scaling inference computation with collaborative verification. arXiv preprint arXiv:2410.05318, 2024

    Pith/arXiv arXiv 2024

  79. [79]

    Trustworthy llms: a survey and guideline for evaluating large language models’ alignment.arXiv preprint arXiv:2308.05374, 2023

    Yang Liu, Yuanshun Yao, Jean-Francois Ton, Xiaoying Zhang, Ruocheng Guo, Hao Cheng, Yegor Klochkov, Muhammad Faaiz Taufiq, and Hang Li. Trustworthy llms: a survey and guideline for evaluating large language models’ alignment.arXiv preprint arXiv:2308.05374, 2023

    Pith/arXiv arXiv 2023

  80. [80]

    Shap: A unified approach to interpreting model predictions

    Scott Lundberg and Su-in Lee. Shap: A unified approach to interpreting model predictions. Advances in neural information processing systems, pages 1–10, 2017

    2017

Showing first 80 references.