pith. sign in

arxiv: 2606.24424 · v1 · pith:E26VYJ2Ynew · submitted 2026-06-23 · 📡 eess.SP

Explainable AI for Next-Generation Wireless Physical Layer: Basics, State-of-the-Art, and Open Challenges

Bingnan Xiao , Shuyan Hu , Xiaojing Chen , Zhiyuan Zhai , Bingcong Li , Wei Ni , Xin Wang , Ekram Hossain This is my paper

Pith reviewed 2026-06-25 22:32 UTC · model grok-4.3

classification 📡 eess.SP
keywords explainable AIwireless physical layerneural networkstaxonomyopen challengesAI-native systemsPHY layer protocol
0
0 comments X

The pith

Wireless physical layer systems require tailored explainable AI to address opacity risks in AI-native networks.

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

The paper formalizes responsibility-oriented goals for explainable AI specifically in wireless physical layers. It develops a taxonomy of explanation approaches and practical criteria for using them in communication settings. The survey reviews how these connect to different learning methods across the PHY protocol stack. It also maps out open challenges such as performance tradeoffs and the needs of emerging LLM-driven layers. A reader would care because next-generation wireless systems intend to embed neural networks throughout the PHY, where lack of explanations could affect reliability and safety.

Core claim

By formalizing responsibility-oriented goals for wireless XAI and building a systematic taxonomy of explainability approaches, the paper shows how to connect representative learning paradigms to suitable explanation techniques, evaluation metrics, and deployment considerations throughout the PHY layer.

What carries the argument

Responsibility-oriented goals combined with a systematic taxonomy of explainability approaches that maps learning paradigms to explanation techniques and deployment criteria.

If this is right

  • Explanations can be applied at multiple points in the PHY layer to mitigate reliability and privacy risks.
  • Practical criteria guide selection of explanation methods based on communication constraints like latency and spectrum use.
  • Explainability-performance tradeoffs must be managed when embedding XAI in time-varying channels.
  • Cross-layer consistency of explanations becomes required as systems integrate multiple PHY functions.
  • New explanation techniques will be needed for LLM- and agentic-AI components in future PHY layers.

Where Pith is reading between the lines

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

  • Early explainability-aware data processing at the PHY could lower the cost of later explanations in the stack.
  • Real-world channel measurements with and without explanations would test the distilled deployment criteria.
  • Similar taxonomy structures might extend to related areas such as integrated sensing and communication.

Load-bearing premise

The opacity of deep learning models creates increasing concerns about reliability, safety, and privacy in complex time-varying wireless networks.

What would settle it

A controlled wireless testbed deployment showing that black-box neural networks achieve the same measured reliability, safety metrics, and user trust levels as explained versions would challenge the need for wireless-specific XAI.

Figures

Figures reproduced from arXiv: 2606.24424 by Bingcong Li, Bingnan Xiao, Ekram Hossain, Shuyan Hu, Wei Ni, Xiaojing Chen, Xin Wang, Zhiyuan Zhai.

Figure 1
Figure 1. Figure 1: The organization of this survey. most all rule extraction techniques, e.g., [57], [58], substantiate their approach to the search for a simpler understanding of what the model internally does, stating that the knowledge (information) can be expressed in these simpler proxies that they consider explaining the antecedent [30]. In wireless sys￾tems, informativeness helps align AI models with operational needs… view at source ↗
Figure 2
Figure 2. Figure 2 [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Illustration of CAM [72], which enhances explainability by visualizing a CNN’s implicit attention for specific predictions. By computing a weighted sum of the last convolutional layer’s feature maps using class-specific weights from the final fully connected layer, it generates a localization heatmap. CAM enables local interpretation for a specific instance, visualized in the final output where red, high-a… view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of Deep SHAP [13], which approximates Shap￾ley values to interpret the predictions of DNNs. By employing a backpropagation-based technique, it attributes the model’s output to individual input features, such as pixels in an image. Deep SHAP enables both local interpretation, visualized as a heatmap where red areas contribute positively and blue areas negatively to the prediction for a specific… view at source ↗
Figure 5
Figure 5. Figure 5: The concepts of layers, units, and blocks. The concept of layers is defined as the processing modules of a network; the concept of units is defined as the neurons of a network; blocks are an aggregation of a bunch of units. spotlighting information crucial for other parts of the network. In [110], object-level attention models use selective search to generate candidate patches. The selected patches are fed… view at source ↗
Figure 6
Figure 6. Figure 6: Illustration of DSP [88]. The “Neural Network Controller” handles the complex mapping from “State S” to search probabilities θ. The crucial mechanism for explainability is that the final resulting agent is the “Symbolic Policy (Expression Tree),” instead of the neural network. Since “Action (a)” is derived from a structured, human-readable tree composed of explicit mathematical and logical operators (e.g.,… view at source ↗
Figure 7
Figure 7. Figure 7: The block diagram of XAI-CHEST in [129]. Based on the perturbation approach, the interpretability Model learns to generate a noise weight mask for the input features. With the premise that injecting noise into “irrelevant” features does not degrade the Estimator’s performance, the model identifies ”relevant” subcarriers by maximizing noise weights while maintaining estimation accuracy, visualizing the deci… view at source ↗
Figure 8
Figure 8. Figure 8: The concept bottleneck model of [150]. In this model, the Regressor first maps raw IQ signals into a set of pre-defined, human-understandable physical concepts (c1 . . . cn). Subsequently, the Classifier predicts the final modulation class based on these concepts. This structure forces the model to reason using explicit physical attributes, making the decision-making process transparent and verifiable. to … view at source ↗
Figure 9
Figure 9. Figure 9: The network of [169] with Grad-CAM. For the upper part, two CNNs extract the features of global time-frequency morphology and local fine-grained textures, respectively. The lower part (in grey) details the explainability process: gradients of the predicted jamming class are backpropagated to compute channel-wise importance weights. The resulting heatmaps are upsampled and overlaid onto the original input t… view at source ↗
Figure 10
Figure 10. Figure 10: The network mode of [233], where SHAP is utilized to visualize the feature contribution analysis. The plot ranks state features by their global importance, revealing how the model weighs both current and historical observations. The horizontal dispersion of the points indicates each feature’s impact on the Q-value output. By highlighting the significant negative contribution (red points with negative SHAP… view at source ↗
Figure 11
Figure 11. Figure 11: System model of [252]. The upper part depicts the DRQN structure, where the inputs are processed by LSTM to make channel access decisions under partial observability. The lower part (in grey) details the Reward Processing explainability mechanism, which decomposes the reward signal into semantically meaningful components such as “Success” and traces their propagation back through the recurrent hidden stat… view at source ↗
read the original abstract

Next-generation wireless systems are expected to be ``AI-native," with neural networks (NNs) embedded throughout the physical (PHY) layer protocol stack to improve spectral efficiency, latency, and network autonomy. However, the opacity of deep learning (DL) models raises increasing concerns about system reliability, safety, and privacy, especially under complex and time-varying network environments. This survey studies explainable AI (XAI) in wireless PHY layers from the explainability perspective. We first formalize a series of responsibility-oriented goals for wireless XAI. Then, we develop a systematic taxonomy of explainability approaches and distill practical criteria for deploying explanations in communication scenarios. We provide a comprehensive review of where and how XAI can be applied throughout the PHY layer, connecting representative learning paradigms to appropriate explanation techniques, evaluation metrics, and deployment considerations. Open challenges and future directions are discussed, including explainability-performance tradeoffs, explainability-aware data processing, customized XAI for communication-specific structures, cross-layer explanation consistency, and emerging needs for explaining LLM- and Agentic-AI-driven PHY layers.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 3 minor

Summary. The manuscript is a survey on explainable AI (XAI) for the physical (PHY) layer of next-generation wireless systems. It formalizes a series of responsibility-oriented goals for wireless XAI, develops a systematic taxonomy of explainability approaches, distills practical criteria for deploying explanations in communication scenarios, provides a comprehensive review connecting learning paradigms to explanation techniques and metrics throughout the PHY layer, and discusses open challenges including explainability-performance tradeoffs, explainability-aware data processing, customized XAI for communication structures, cross-layer consistency, and needs for LLM- and Agentic-AI-driven PHY layers.

Significance. If the taxonomy and deployment criteria prove accurate and complete upon verification, the survey could serve as a foundational reference for the emerging area of AI-native wireless systems. It organizes existing work around wireless-specific concerns such as time-varying environments and reliability, while highlighting forward-looking topics like LLM-driven PHY layers. The organizational framing of responsibility-oriented goals adds conceptual value beyond a simple literature aggregation.

minor comments (3)
  1. [Abstract] Abstract: the phrase 'responsibility-oriented goals' is introduced without a concise definition or example; adding one sentence in the introduction would improve accessibility for readers outside the XAI community.
  2. [Introduction] The claim of a 'comprehensive review' would be strengthened by an explicit statement in the introduction comparing coverage against the most recent prior surveys on wireless AI or XAI.
  3. [Open Challenges] Open challenges section: the discussion of 'explainability-aware data processing' would benefit from one concrete wireless example (e.g., channel estimation dataset curation) to make the point actionable.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the constructive review and the recommendation of minor revision. We appreciate the recognition that the survey organizes existing work around wireless-specific concerns and highlights forward-looking topics such as LLM-driven PHY layers. Since no specific major comments were raised, we have no points requiring direct rebuttal or revision at this stage.

Circularity Check

0 steps flagged

No significant circularity; survey is organizational and self-contained

full rationale

This is a survey paper whose contributions consist of formalizing responsibility-oriented goals, constructing a taxonomy of explainability approaches, reviewing applications across the PHY layer, and discussing open challenges. These are conceptual and organizational tasks with no equations, fitted parameters, predictions, or derivations that reduce to self-referential inputs, self-citations, or ansatzes. The motivating premise (DL opacity in wireless environments) is a standard field observation and does not create a load-bearing circular step. No patterns from the enumerated list apply, and the paper aggregates external work without reducing its own claims by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a survey paper synthesizing existing research on XAI for wireless PHY layers. No new free parameters, axioms, or invented entities are introduced; the contribution lies in organization and identification of challenges.

pith-pipeline@v0.9.1-grok · 5749 in / 1159 out tokens · 28862 ms · 2026-06-25T22:32:55.135346+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

275 extracted references · 7 linked inside Pith

  1. [1]

    Through-the-Earth magnetic induction communication and networking: A comprehensive survey,

    H. Ma, E. Liu, W. Niet al., “Through-the-Earth magnetic induction communication and networking: A comprehensive survey,”IEEE Com- mun. Surv. Tutorials, vol. 28, pp. 2263–2305, 2026

    2026

  2. [2]

    Toward quantum-native communica- tion systems: State-of-the-art, trends, and challenges,

    X. Zhou, A. Shen, S. Huet al., “Toward quantum-native communica- tion systems: State-of-the-art, trends, and challenges,”IEEE Commun. Surv. Tutorials, vol. 28, pp. 1436–1482, 2026

    2026

  3. [3]

    Privacy- preserving data-driven learning models for emerging communication networks: A comprehensive survey,

    M. M. Fouda, Z. M. Fadlullah, M. I. Ibrahem, and N. Kato, “Privacy- preserving data-driven learning models for emerging communication networks: A comprehensive survey,”IEEE Commun. Surveys Tuts., vol. 27, no. 4, pp. 2505–2542, 2025

    2025

  4. [4]

    Advancing 6G: Survey for explainable AI on communications and network slicing,

    H. Sun, Y . Liu, A. Al-Tahmeesschiet al., “Advancing 6G: Survey for explainable AI on communications and network slicing,”IEEE Open J. Commun. Soc., vol. 6, pp. 1372–1412, 2025

    2025

  5. [5]

    Explainability methods for identifying root-cause of SLA violation prediction in 5G network,

    A. Terra, R. Inam, S. Baskaranet al., “Explainability methods for identifying root-cause of SLA violation prediction in 5G network,” in Proc. IEEE Global Commun. Conf. (GLOBECOM), 2020, pp. 1–7

    2020

  6. [6]

    Toward trustworthy artificial intelligence (TAI) in the context of explainability and robustness,

    B. Chander, C. John, L. Warrier, and K. Gopalakrishnan, “Toward trustworthy artificial intelligence (TAI) in the context of explainability and robustness,”ACM Comput. Surv., vol. 57, no. 6, pp. 1–49, 2025

    2025

  7. [7]

    Zero-trust foundation models: A new paradigm for secure and collaborative artificial intelligence for Internet of Things,

    K. Li, C. Li, X. Yuanet al., “Zero-trust foundation models: A new paradigm for secure and collaborative artificial intelligence for Internet of Things,”IEEE Internet Things J., vol. 12, no. 22, pp. 46 269–46 293, 2025

    2025

  8. [8]

    Peeking inside the black-box: A survey on explainable artificial intelligence (XAI),

    A. Adadi and M. Berrada, “Peeking inside the black-box: A survey on explainable artificial intelligence (XAI),”IEEE Access, vol. 6, pp. 52 138–52 160, 2018

    2018

  9. [9]

    Explainable AI: From black box to glass box,

    A. Rai, “Explainable AI: From black box to glass box,”J. Acad. Market. Sci., vol. 48, no. 1, pp. 137–141, 2020

    2020

  10. [10]

    Discriminative feature attri- butions: Bridging post hoc explainability and inherent interpretability,

    U. Bhalla, S. Srinivas, and H. Lakkaraju, “Discriminative feature attri- butions: Bridging post hoc explainability and inherent interpretability,” Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 36, pp. 44 105– 44 122, 2023

    2023

  11. [11]

    Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead,

    C. Rudin, “Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead,”Nature Mach. Intell., vol. 1, no. 5, pp. 206–215, 2019

    2019

  12. [12]

    Why should I trust you? Explaining the predictions of any classifier,

    M. T. Ribeiro, S. Singh, and C. Guestrin, “Why should I trust you? Explaining the predictions of any classifier,” inProc. 22nd ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining (KDD), 2016, pp. 1135–1144

    2016

  13. [13]

    A unified approach to interpreting model predictions,

    S. M. Lundberg and S.-I. Lee, “A unified approach to interpreting model predictions,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 30, 2017

    2017

  14. [14]

    Explainable reinforce- ment learning: A survey and comparative review,

    S. Milani, N. Topin, M. Veloso, and F. Fang, “Explainable reinforce- ment learning: A survey and comparative review,”ACM Comput. Surv., vol. 56, no. 7, pp. 1–36, 2024

    2024

  15. [15]

    Toward learning model-agnostic explanations for deep learning-based signal modulation classifiers,

    Y . Tian, D. Xu, E. Tonget al., “Toward learning model-agnostic explanations for deep learning-based signal modulation classifiers,” IEEE Trans. Rel., vol. 73, no. 3, pp. 1529–1543, 2024

    2024

  16. [16]

    A review of taxonomies of explainable artificial intelligence (XAI) methods,

    T. Speith, “A review of taxonomies of explainable artificial intelligence (XAI) methods,” inProc. ACM Conf. Fairness, Accountability, Trans- parency (FAccT), 2022, pp. 2239–2250

    2022

  17. [17]

    Interpreting black-box models: A review on explainable artificial intelligence,

    V . Hassija, V . Chamola, A. Mahapatraet al., “Interpreting black-box models: A review on explainable artificial intelligence,”Cogn. Comput., vol. 16, no. 1, pp. 45–74, 2024

    2024

  18. [18]

    On the (in) fidelity and sensitivity of explanations,

    C.-K. Yeh, C.-Y . Hsieh, A. Suggala, D. I. Inouye, and P. K. Ravikumar, “On the (in) fidelity and sensitivity of explanations,”Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 32, 2019

    2019

  19. [19]

    Openxai: Towards a transparent evaluation of model explanations,

    C. Agarwal, S. Krishna, E. Saxena, M. Pawelczyk, N. Johnson, I. Puri, M. Zitnik, and H. Lakkaraju, “Openxai: Towards a transparent evaluation of model explanations,”Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 35, pp. 15 784–15 799, 2022

    2022

  20. [20]

    On the robustness of inter- pretability methods,

    D. Alvarez-Melis and T. S. Jaakkola, “On the robustness of inter- pretability methods,”arXiv preprint arXiv:1806.08049, 2018

    Pith/arXiv arXiv 2018

  21. [21]

    Onboard opti- mization and learning: A survey,

    M. I. Pavel, S. Hu, M. Pratama, and R. Kowalczyk, “Onboard opti- mization and learning: A survey,”arXiv preprint arXiv:2505.08793, 2025

    arXiv 2025

  22. [22]

    Edge ai deploying artificial intelligence models on edge devices for real-time analytics,

    S. Choudhary, S. Vijitha, D. D. Bhavani, N. Bhuvaneswari, M. Tiwari, and S. Subburam, “Edge ai deploying artificial intelligence models on edge devices for real-time analytics,” inProc. ITM Web Conf., vol. 76. EDP Sciences, 2025, p. 01009

    2025

  23. [23]

    Defensive adversarial CAPTCHA: A semantics-driven framework for natural adversarial example genera- tion,

    X. Du, X. Liu, J. Zhouet al., “Defensive adversarial CAPTCHA: A semantics-driven framework for natural adversarial example genera- tion,”IEEE Trans. Dependable Secure Comput., pp. 1–13, 2026, early access

    2026

  24. [24]

    Explainable AI (XAI): A systematic meta- survey of current challenges and future opportunities,

    W. Saeed and C. Omlin, “Explainable AI (XAI): A systematic meta- survey of current challenges and future opportunities,”Knowledge- Based Syst., vol. 263, p. 110273, 2023

    2023

  25. [25]

    Survey on explainable AI: From ap- proaches, limitations and applications aspects,

    W. Yang, Y . Wei, H. Weiet al., “Survey on explainable AI: From ap- proaches, limitations and applications aspects,”Human-Centric Intell. Syst., vol. 3, no. 3, pp. 161–188, 2023

    2023

  26. [27]

    A survey on XAI for 5G and beyond security: Technical aspects, challenges and research directions,

    T. Senevirathna, V . H. La, S. Marchaet al., “A survey on XAI for 5G and beyond security: Technical aspects, challenges and research directions,”IEEE Commun. Surveys Tuts., vol. 27, no. 2, pp. 941–973, 2025

    2025

  27. [28]

    Explainable AI for 6G use cases: Technical aspects and research challenges,

    S. Wang, M. A. Qureshi, L. Miralles-Pechu ´anet al., “Explainable AI for 6G use cases: Technical aspects and research challenges,”IEEE Open J. Commun. Soc., vol. 5, pp. 2490–2540, 2024

    2024

  28. [29]

    Explainable artificial intelligence (XAI) for internet of things: A survey,

    ˙I. K¨ok, F. Y . Okay,¨O. Muyanlı, and S. ¨Ozdemir, “Explainable artificial intelligence (XAI) for internet of things: A survey,”IEEE Internet of Things Journal, vol. 10, no. 16, pp. 14 764–14 779, 2023

    2023

  29. [30]

    Explainable Artificial Intelligence (XAI) for 6G: Improving Trust between Human and Machine,

    W. Guo, “Explainable Artificial Intelligence (XAI) for 6G: Improving Trust between Human and Machine,”IEEE Commun. Mag., vol. 58, no. 6, pp. 39–45, 2020

    2020

  30. [31]

    Explainable AI (XAI) for trustworthy and transparent decision-making: A theoretical framework for ai interpretability,

    A. Chinnaraju, “Explainable AI (XAI) for trustworthy and transparent decision-making: A theoretical framework for ai interpretability,”World J. Adv. Eng. Technol. Sci., vol. 14, no. 3, pp. 170–207, 2025

    2025

  31. [32]

    Exploiting deep learning for secure transmission in an underlay cognitive radio network,

    M. Zhang, K. Cumanan, J. Thiyagalingam, Y . Tang, W. Wang, Z. Ding, and O. A. Dobre, “Exploiting deep learning for secure transmission in an underlay cognitive radio network,”IEEE Trans. Veh. Technol., vol. 70, no. 1, pp. 726–741, 2021

    2021

  32. [33]

    Robust transmit power control with imperfect csi using a deep neural network,

    W. Lee and K. Lee, “Robust transmit power control with imperfect csi using a deep neural network,”IEEE Trans. Veh. Technol., vol. 70, no. 11, pp. 12 266–12 271, 2021

    2021

  33. [34]

    Adversarial attacks on deep-learning based radio signal classification,

    M. Sadeghi and E. G. Larsson, “Adversarial attacks on deep-learning based radio signal classification,”IEEE Wireless Commun. Lett., vol. 8, no. 1, pp. 213–216, 2019

    2019

  34. [35]

    DSIL: an effective spectrum prediction framework against spectrum concept drift,

    L. Guo, J. Lu, J. An, and K. Yang, “DSIL: an effective spectrum prediction framework against spectrum concept drift,”IEEE Trans. Cognit. Commun. Networking, vol. 10, no. 3, pp. 794–806, 2024

    2024

  35. [36]

    Building trust in iq-based deep learning models for wireless applications,

    S. Trudeau and K. Chowdhury, “Building trust in iq-based deep learning models for wireless applications,” inProc. Int. Symp. The- ory Algorithmic Found. Protoc. Des. Mob. Networks Mob. Comput. (MobiHoc), October 2025

    2025

  36. [37]

    Towards risk-free trustwor- thy artificial intelligence: Significance and requirements,

    L. Alzubaidi, A. Al-Sabaawi, J. Baiet al., “Towards risk-free trustwor- thy artificial intelligence: Significance and requirements,”Int. J. Intell. Syst., vol. 2023, no. 1, p. 4459198, 2023

    2023

  37. [38]

    Causality for trustworthy ar- tificial intelligence: status, challenges and perspectives,

    A. Rawal, A. Raglin, D. B. Rawatet al., “Causality for trustworthy ar- tificial intelligence: status, challenges and perspectives,”ACM Comput. Surv., vol. 57, no. 6, pp. 1–30, 2025

    2025

  38. [39]

    Can we open the black box of AI?

    D. Castelvecchi, “Can we open the black box of AI?”Nature, vol. 538, no. 7623, pp. 20–23, 2016

    2016

  39. [40]

    Optimal online control strategy for differentially private federated learning,

    X. Yuan, A. V . Savkin, W. Niet al., “Optimal online control strategy for differentially private federated learning,”IEEE Trans. Dependable Secure Comput., pp. 1–13, 2026, early access

    2026

  40. [41]

    Ethics guidelines for trustworthy AI,

    A. HLEG, “Ethics guidelines for trustworthy AI,”European Commis- sion, 2019. 26

    2019

  41. [42]

    A right to reasonable inferences: re- thinking data protection law in the age of big data and AI,

    S. Wachter and B. Mittelstadt, “A right to reasonable inferences: re- thinking data protection law in the age of big data and AI,”Colum. Bus. L. Rev., p. 494, 2019

    2019

  42. [43]

    From what to how: an initial review of publicly available AI ethics tools, methods and research to translate principles into practices,

    J. Morley, L. Floridi, L. Kinsey, and A. Elhalal, “From what to how: an initial review of publicly available AI ethics tools, methods and research to translate principles into practices,” inEthics, Governance, and Policies in Artificial Intelligence, 2021, pp. 153–183

    2021

  43. [44]

    Free privacy protection for wireless federated learning: Enjoy it or suffer from it?

    W. Li, T. Lv, X. Zhaoet al., “Free privacy protection for wireless federated learning: Enjoy it or suffer from it?”IEEE Transactions on Information Forensics and Security, vol. 20, pp. 6263–6278, 2025

    2025

  44. [45]

    From 5G to 6G: A survey on security, privacy, and standardization pathways,

    M. Yang, Y . Qu, T. Ranbadugeet al., “From 5G to 6G: A survey on security, privacy, and standardization pathways,”ACM Comput. Surv., Dec. 2025

    2025

  45. [46]

    Princi- pled artificial intelligence: Mapping consensus in ethical and rights- based approaches to principles for AI,

    J. Fjeld, N. Achten, H. Hilligoss, A. Nagy, and M. Srikumar, “Princi- pled artificial intelligence: Mapping consensus in ethical and rights- based approaches to principles for AI,”Berkman Klein Center for Internet & Society, 2020

    2020

  46. [47]

    Conscientious classifi- cation: A data scientist’s guide to discrimination-aware classification,

    B. D’Alessandro, C. O’Neil, and T. Lagatta, “Conscientious classifi- cation: A data scientist’s guide to discrimination-aware classification,” Big Data, vol. 5, no. 2, pp. 120–134, 2017

    2017

  47. [48]

    Enhancing convergence, privacy and fairness for wireless personalized federated learning: Quantization- assisted min-max fair scheduling,

    X. Zhao, Q. Cui, Z. Duet al., “Enhancing convergence, privacy and fairness for wireless personalized federated learning: Quantization- assisted min-max fair scheduling,”IEEE Transactions on Mobile Computing, vol. 24, no. 10, pp. 9902–9918, 2025

    2025

  48. [49]

    Data preprocessing techniques for classi- fication without discrimination,

    F. Kamiran and T. Calders, “Data preprocessing techniques for classi- fication without discrimination,”Knowl. Inf. Syst., vol. 33, no. 1, pp. 1–33, 2012

    2012

  49. [50]

    Learning fair representations,

    R. Zemel, Y . Wu, K. Swersky, T. Pitassi, and C. Dwork, “Learning fair representations,” inProc. 30th Int. Conf. Mach. Learn. (ICML), 2013, pp. 325–333

    2013

  50. [51]

    Mitigating unwanted biases with adversarial learning,

    B. H. Zhang, B. Lemoine, and M. Mitchell, “Mitigating unwanted biases with adversarial learning,” inProc. AAAI/ACM Conf. AI, Ethics, Soc. (AIES), 2018, pp. 335–340

    2018

  51. [52]

    Classification with fairness constraints: A meta-algorithm with provable guarantees,

    L. E. Celis, L. Huang, V . Keswani, and N. K. Vishnoi, “Classification with fairness constraints: A meta-algorithm with provable guarantees,” inProc. Conf. Fairness Accountability Transparency (FAT*), 2019, pp. 319–328

    2019

  52. [53]

    Equality of opportunity in super- vised learning,

    M. Hardt, E. Price, and N. Srebro, “Equality of opportunity in super- vised learning,”Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 29, 2016

    2016

  53. [54]

    Non-stationarity characterization and geometry-cluster-based stochastic model for high-speed train radio channels,

    Y . Zhang, K. Zhang, A. Ghazalet al., “Non-stationarity characterization and geometry-cluster-based stochastic model for high-speed train radio channels,”IEEE Trans. Intell. Transp. Syst., vol. 24, no. 7, pp. 7122– 7137, 2023

    2023

  54. [55]

    Model-driven deep learning for physical layer communications,

    H. He, S. Jin, C.-K. Wen, F. Gao, G. Y . Li, and Z. Xu, “Model-driven deep learning for physical layer communications,”IEEE Wireless Commun., vol. 26, no. 5, pp. 77–83, 2019

    2019

  55. [56]

    Digital twin-assisted explainable ai for robust beam prediction in mmwave mimo systems,

    N. Khan, A. Abdallah, A. Celik, A. M. Eltawil, and S. Coleri, “Digital twin-assisted explainable ai for robust beam prediction in mmwave mimo systems,”IEEE Trans. Wireless Commun., vol. 25, pp. 2435– 2451, 2026

    2026

  56. [57]

    Extracting rules from trained neural networks,

    H. Tsukimoto, “Extracting rules from trained neural networks,”IEEE Trans. Neural Netw., vol. 11, no. 2, pp. 377–389, 2000

    2000

  57. [58]

    Medical diagnostic expert system based on PDP model,

    K. Saito and R. Nakano, “Medical diagnostic expert system based on PDP model,” inProc. IEEE Int. Conf. Neural Netw., 2002

    2002

  58. [59]

    Explaining model confidence using counterfactuals,

    T. Le, T. Miller, R. Singh, and L. Sonenberg, “Explaining model confidence using counterfactuals,” inProc. AAAI Conf. Artif. Intell., vol. 37, no. 10, 2023, pp. 11 856–11 864

    2023

  59. [60]

    Enhancing the confidence of deep learning classifiers via interpretable saliency maps,

    E. Tjoa, H. J. Khok, T. Chouhan, and C. Guan, “Enhancing the confidence of deep learning classifiers via interpretable saliency maps,” Neurocomputing, vol. 562, p. 126825, 2023

    2023

  60. [61]

    Is your explanation reliable: Confidence-aware explanation on graph neural networks,

    J. Zhang, X. Liu, D. Luo, and H. Wei, “Is your explanation reliable: Confidence-aware explanation on graph neural networks,” inProc. 31st ACM SIGKDD Conf. Knowl. Discovery Data Mining (KDD), 2025, pp. 3740–3751

    2025

  61. [62]

    Consistent indi- vidualized feature attribution for tree ensembles,

    S. M. Lundberg, G. G. Erion, and S.-I. Lee, “Consistent indi- vidualized feature attribution for tree ensembles,”arXiv preprint arXiv:1802.03888, 2018

    Pith/arXiv arXiv 2018

  62. [63]

    DeepRED – rule extrac- tion from deep neural networks,

    J. R. Zilke, E. Loza Menc ´ıa, and F. Janssen, “DeepRED – rule extrac- tion from deep neural networks,”Knowledge Discovery, Knowledge Engineering and Knowledge Management, pp. 457–473, 2016

    2016

  63. [64]

    CERTIFAI: Counterfactual explanations for robustness, transparency, interpretability, and fairness of artificial intelligence models,

    S. Sharma, J. Henderson, and J. Ghosh, “CERTIFAI: Counterfactual explanations for robustness, transparency, interpretability, and fairness of artificial intelligence models,”arXiv preprint arXiv:1905.07857, 2019

    arXiv 1905

  64. [65]

    Learning structured sparsity in deep neural networks,

    W. Wen, C. Wu, Y . Wanget al., “Learning structured sparsity in deep neural networks,”Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 29, pp. 2074–2082, 2016

    2074

  65. [66]

    Distilling the knowledge in a neural network,

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,”arXiv preprint arXiv:1503.02531, 2015

    Pith/arXiv arXiv 2015

  66. [67]

    Towards black-box explainability with Gaussian discriminant knowledge dis- tillation,

    A. Haselhoff, J. Kronenberger, F. Kuppers, and J. Schneider, “Towards black-box explainability with Gaussian discriminant knowledge dis- tillation,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR) Workshops, 2021, pp. 21–28

    2021

  67. [68]

    EGNN: Constructing explainable graph neural networks via knowledge distillation,

    Y . Li, L. Liu, G. Wanget al., “EGNN: Constructing explainable graph neural networks via knowledge distillation,”Knowledge-Based Syst., vol. 241, p. 108345, 2022

    2022

  68. [69]

    Rule learning by searching on adapted nets,

    L. M. Fu, “Rule learning by searching on adapted nets,” inProc. 9th Nat. Conf. Artif. Intell. (AAAI), 1991

    1991

  69. [70]

    Extracting rules from artificial neural networks with dis- tributed representations,

    S. Thrun, “Extracting rules from artificial neural networks with dis- tributed representations,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), 1995, pp. 505–512

    1995

  70. [71]

    How transferable are features in deep neural networks?

    J. Yosinski, J. Clune, Y . Bengio, and H. Lipson, “How transferable are features in deep neural networks?” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 27, 2014, pp. 3320–3328

    2014

  71. [72]

    Learning deep features for discriminative localization,

    B. Zhou, A. Khosla, A. Lapedrizaet al., “Learning deep features for discriminative localization,” inProc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016

    2016

  72. [73]

    Grad-CAM: Visual explanations from deep networks via gradient-based localization,

    R. R. Selvaraju, M. Cogswell, A. Daset al., “Grad-CAM: Visual explanations from deep networks via gradient-based localization,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), 2017, pp. 618–626

    2017

  73. [74]

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

    A. Chattopadhay, A. Sarkar, P. Howlader, and V . N. Balasubrama- nian, “Grad-CAM++: Generalized gradient-based visual explanations for deep convolutional networks,” inProc. IEEE Winter Conf. Appl. Comput. Vis. (WACV), 2018, pp. 839–847

    2018

  74. [75]

    Deep inside convolu- tional networks: Visualising image classification models and saliency maps,

    K. Simonyan, A. Vedaldi, and A. Zisserman, “Deep inside convolu- tional networks: Visualising image classification models and saliency maps,”arXiv preprint arXiv:1312.6034, 2013

    Pith/arXiv arXiv 2013

  75. [76]

    Deep Taylor decomposition of neural networks,

    G. Montavon, S. Bach, A. Binder, W. Samek, and K.-R. M ¨uller, “Deep Taylor decomposition of neural networks,” inProc. Int. Conf. Mach. Learn. Wksp., 2016

    2016

  76. [77]

    Network dissection: Quantifying interpretability of deep visual representations,

    D. Bau, B. Zhou, A. Khoslaet al., “Network dissection: Quantifying interpretability of deep visual representations,” inProc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2017, pp. 6541–6549

    2017

  77. [78]

    Learning important features through propagating activation differences,

    A. Shrikumar, P. Greenside, and A. Kundaje, “Learning important features through propagating activation differences,” inProc. 34th Int. Conf. Mach. Learn. (ICML), 2017, pp. 3145–3153

    2017

  78. [79]

    Transformer interpretability beyond attention visualization,

    H. Chefer, S. Gur, and L. Wolf, “Transformer interpretability beyond attention visualization,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2021, pp. 782–791

    2021

  79. [80]

    White-box transformers via sparse rate reduction,

    Y . Yu, S. Buchanan, D. Paiet al., “White-box transformers via sparse rate reduction,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 36, 2023, pp. 9422–9457

    2023

  80. [81]

    Scaling white-box transformers for vision,

    J. Yang, X. Li, D. Paiet al., “Scaling white-box transformers for vision,”arXiv preprint arXiv:2405.20299, 2024

    arXiv 2024

Showing first 80 references.