Pith. sign in

REVIEW 3 major objections 4 minor 180 references

LLM-powered agentic AI is positioned as a first-class architectural paradigm for 5G/6G networks, supported by the first unified survey of its concepts, protocols, evaluation, and standardization.

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

2026-08-01 21:26 UTC pith:XHGVCCKA

load-bearing objection Solid, genuinely useful survey of agentic AI for 5G/6G, but the 'first comprehensive' claim is softer than it looks and Table 4's latency numbers need sources. the 3 major comments →

arxiv 2607.16066 v1 pith:XHGVCCKA submitted 2026-07-17 cs.NI cs.AI

LLM-Powered Agentic AI for 5G/6G Networks: A Tutorial and Survey on Architectures, Protocols, and Standardization

Mazene Ameur , Abdelkader Mekrache , Bouziane Brik , Adlen Ksentini This is my paper
classification cs.NI cs.AI
keywords agentic AIlarge language models5G/6G networksintent-based networkingzero-touch network and service managementagentic protocolsnetwork standardizationtelecom benchmarks
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.

This paper tries to establish that existing research treats large language models and telecom networks in isolation, and that no previous work jointly addresses LLM agents, their protocols, evaluation, benchmarks, network integration, and standardization for next-generation networks. To close that gap, it offers a two-part tutorial-and-survey: the first part formalizes the control, management, and AI-native planes of 5G/6G and explains agentic foundations such as memory, planning, tool use, and multi-agent coordination; the second maps those capabilities onto network surfaces, reviews five agentic protocols, telecom benchmarks, and standardization efforts, and lists open challenges. A sympathetic reader would care because a unified cross-domain reference could align AI research with telecom constraints and give standardization bodies a shared vocabulary.

Core claim

On its own terms, the paper's central claim is that agentic AI is not an overlay on 6G but a first-class architectural component: a vertical AI-native plane intersecting the control and management planes. It argues that no prior work jointly and systematically covers LLMs, agentic concepts, agentic protocols, evaluation methods, benchmarks, telecom integration, and standardization, and it positions the paper as the first comprehensive tutorial-style study of that intersection. The paper substantiates the claim by synthesizing the three-plane abstraction, comparing five agentic protocols (MCP, A2A, ANP, ACP, AP2) against 6G integration points, cataloging telecom-oriented benchmarks, and mappi

What carries the argument

The central organizing device is the three-plane abstraction: the control plane (RAN, transport, core, edge/cloud), the management plane (Intent-Based Networking and Zero-touch Service Management), and a vertical AI-native plane (data operations, AI operations, explainability) that cuts across both. The paper also formalizes an agent as a tuple A=⟨S,O,Ac,M,T,πθ,G⟩—state space, observations, actions, memory, tools, LLM policy, and goal—and uses it to structure a taxonomy of agentic capabilities. These constructs do the work of mapping five agentic protocols to specific 6G integration points and of organizing the surveyed literature into a coherent framework.

Load-bearing premise

The load-bearing premise is that the literature selection behind 'no prior work' is comprehensive and unbiased; if a prior unified survey was missed, the central claim of being the first collapses.

What would settle it

A reproducible literature search with explicit inclusion and exclusion criteria that surfaces a comparable prior survey would falsify the 'first comprehensive survey' claim. Separately, re-measuring the protocol overheads in Table 4 on a defined testbed (e.g., an MCP tool chain on a loaded edge node) would show whether the reported 50–200 ms latencies hold.

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

If this is right

  • If the survey's framework is adopted, researchers get a shared vocabulary for agentic network intelligence, making results comparable across control, management, and AI-native planes.
  • Standardization bodies gain a concrete map of where agentic protocols fit: MCP at the non-RT RIC and management plane, A2A across non-RT and near-RT tiers, ANP in core/edge orchestration, ACP as a legacy bridge, AP2 in business-layer transactions.
  • The paper's synthesis formulates a central design constraint with concrete consequences: LLM reasoning cannot fit sub-10 ms control loops, so agentic 6G designs must be hierarchical, with LLM strategic planning on slower horizons and lightweight policies on fast loops.
  • The identified gaps—ephemeral agent identity, missing wire-level serialization between protocols, and absent cross-SDO benchmarks—become specific, testable work items for 3GPP, ETSI, TM Forum, and IETF.
  • Telecom-specific benchmarks (TeleQnA, TeleTables, TelAgentBench, MMTelCo, TeleMath, TSpec-LLM, TeleYAML, and α3-Bench) form a starting point for standardized evaluation that the paper argues is currently fragmented.

Where Pith is reading between the lines

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

  • My inference: The overhead figures in Table 4 (e.g., MCP at 50–200 ms per tool chain, AP2 signing at 1–2 ms per mandate) are presented without sources; any downstream engineering decision that relies on these numbers should treat them as provisional until reproduced in a defined testbed.
  • My inference: If the 'first comprehensive survey' claim is to hold, the literature selection process needs to be made reproducible; this suggests a natural testable extension: a formal systematic review with explicit inclusion/exclusion criteria and a documented search trail.
  • My inference: The strong latency numbers imply that LLM-based agents will, for the foreseeable future, be confined to non-real-time and management planes; RAN-tier agentic control will depend on either much smaller models or hardware that drastically cuts inference latency, a testable prediction.
  • My inference: The three-plane abstraction could generalize beyond telecom to other real-time cyber-physical systems that need to layer generative intelligence over existing control loops, such as power grids or industrial automation.

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

3 major / 4 minor

Summary. The paper is a two-part tutorial-plus-survey that positions LLM-powered agentic AI as a first-class architectural paradigm for 5G/6G networks. Part I covers NGN control, management, and AI-native planes and agentic AI foundations (LLM adaptation, memory, tools, planning, architecture patterns, protocols, evaluation). Part II surveys applications across RAN/transport/core/edge, the management plane (IBN/ZSM), and the AI-native plane, then maps standardization (3GPP, ETSI, TM Forum, O-RAN, IETF, GSMA) and 6G research projects. The central claim, stated in Section 1.2, is that no prior work jointly and systematically addresses LLMs, agentic concepts/protocols, evaluation/benchmarks, telecom integration, and standardization, making this 'the first comprehensive and tutorial-style study' of the intersection.

Significance. If the gap claim holds, this is a timely cross-disciplinary reference for both AI and telecommunications researchers and for standards bodies. The paper's value lies in its synthesis: the seven-dimension comparison in Table 1, the protocol-to-6G mapping in Table 4, the management-plane comparison in Table 6, and the SDO/project tables (Tables 8 and 9) provide structured entry points to a very recent literature. The tutorial sections (Sections 2–3) are accessible and cover relevant fundamentals. The paper does not introduce new algorithms or datasets, but as a survey that explicitly aims to be comprehensive, its utility depends on the reliability and verifiability of its selection process and of the quantitative protocol figures it reports.

major comments (3)
  1. [Section 1.2 (Selection Process)] The 'first comprehensive' claim and the Table 1 comparison rest on the literature-selection procedure, but the description is non-reproducible. The text reports only that Google Scholar searches and a 'rigorous pre-selection process' were used, without inclusion/exclusion criteria, exact search strings, date ranges, screening logs, or a list of excluded candidates. Table 1 lists only 15 comparator surveys, and the checkmark assignments have no coding rubric. A single missed survey that jointly covers the seven dimensions would falsify the headline contribution. Please provide a PRISMA-style flow diagram, a full search protocol (databases, queries, date), and a coding rule for the 'addressed/partially/not addressed' levels.
  2. [Table 4 / Section 3.5.6] Quantitative protocol metrics are presented without sources, measurement protocols, or uncertainty ranges: MCP '50–200 ms per tool chain on loaded edge', AP2 'Signing latency (1–2 ms/mandate)', ANP 'identity resolution latency', and 'JSON serialization overhead of 0.5–several Kilobyte'. These numbers are load-bearing for the discussion of 6G integration timescales. Either remove them, explicitly label them as author estimates, or cite empirical benchmarks/measurements. As written, downstream readers may propagate unsourced figures.
  3. [Section 4.1 (Synthesis and Takeaway Lessons)] The synthesis states as fact that 'LLM-based reasoning cannot satisfy sub-10 ms control-loop requirements' and that hierarchical decomposition introduces 'insufficiently quantified E2/A1/O1 signaling overhead' without citation to any measurement. Such statements are acceptable as author assessments in a tutorial-style survey only if explicitly flagged as such. Please qualify them as the authors' synthesis, or support them with measured or cited evidence, to avoid appearing as established quantitative results.
minor comments (4)
  1. [Section 3.5.3 (ANP)] 'Built on Decentralized Identifiers (DIDs), Verifiable and Credentials' — the phrase appears garbled; it should likely be 'Verifiable Credentials'.
  2. [Table 1 legend] The legend uses the symbol 'G #' to mean 'Partially Addressed', which looks like a rendering artifact. Replace it with a Unicode half-filled circle or a textual marker (e.g., '◐' or 'Partial').
  3. [Figure 1] The survey-structure diagram contains many small text labels; at print resolution some are difficult to read. Consider simplifying the figure or enlarging the font.
  4. [Section 6.4] The statistic '82.4% of evaluated models are compromised through inter-agent communication' is cited to [177]; please verify that the percentages exactly match the cited source, since the cited paper's title focuses on agent-based attacks rather than a comparative vulnerability study.

Circularity Check

0 steps flagged

No circularity: the survey's 'first comprehensive' claim is an empirical gap claim; selection-process and unsourced latency figures are evidence-quality concerns, not self-referential derivation.

full rationale

This paper is a tutorial-and-survey, not a derivation: it contains no fitted parameters, no predictions, and no first-principles result that could be equivalent to its own inputs. Its central claim (Section 1.2) is an explicitly hedged gap claim: 'To the best of our knowledge, no prior work jointly and systematically addresses LLMs, agentic concepts, agentic protocols, evaluation methodologies, benchmarks, telecom network integration, and standardization initiatives in a unified study.' That claim is supported by Table 1's comparison of 23 surveys, and the comparison depends on the Section 1.2 'Selection Process' (Google Scholar searches plus a 'pre-selection process' with no inclusion/exclusion criteria) and on the unsourced Table 4 latency figures (e.g., '50–200 ms per tool chain on loaded edge' for MCP). These are reproducibility and evidence-rigor weaknesses that could affect the 'first comprehensive survey' novelty claim, but they are not circularity: the checkmarks in Table 1 are not derived from the paper's own conclusions, and the gap claim is not equivalent to its inputs by construction. Self-citations (e.g., [8, 53, 73, 77, 121–123, 140]) appear throughout Sections 3 and 4, but they are descriptive citations to prior work by the same authors; none is invoked as a load-bearing authority defining or proving the survey's novelty, and no uniqueness theorem or ansatz is imported from the authors' prior work. No specific circular step can be quoted, so the honest finding is no significant circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

As a survey, there are no fitted parameters or invented entities. The load-bearing premises are the completeness/accuracy of the cited literature, the existence and content of the cited standards, and the reliability of unsourced quantitative protocol-overhead figures.

axioms (3)
  • domain assumption The related-work set (23 surveys in Table 1) is representative and complete enough to substantiate the 'no prior work' claim.
    Section 1.2's selection process uses Google Scholar with unstated inclusion/exclusion criteria; the central gap claim depends on this coverage.
  • domain assumption Cited agentic protocol and standards documents (MCP, A2A, ANP, ACP, AP2; ETSI GS 059, 3GPP TR 22.870, etc.) exist and contain what the survey describes.
    The survey's protocol and standards tables in Sections 3.5 and 5.1 are not independently verified; their accuracy is assumed from the cited references.
  • domain assumption The quantitative latency and overhead figures in Table 4 (e.g., MCP 50-200 ms per tool chain, AP2 1-2 ms/mandate) are accurate.
    These figures are presented as facts without citation, derivation, or measurement protocol, yet they support the protocol-to-6G integration mapping.

pith-pipeline@v1.3.0-alltime-deepseek · 34515 in / 11919 out tokens · 99028 ms · 2026-08-01T21:26:09.652665+00:00 · methodology

0 comments
read the original abstract

Agentic Artificial Intelligence (AI), enabled by Large Language Models, marks a shift from rule-based automation toward autonomous, goal-driven control of Next-Generation Networks (NGNs). Existing surveys treat the two domains in isolation, leaving protocol integration, evaluation, and standardization alignment underexplored. To address this gap, a two-part tutorial-and-survey is presented. Part I formalises the control, management, and AI-native planes of 5G and 6G. It then covers the foundations of agentic systems: reasoning, planning, tool use, multi-agent coordination, and evaluation. Part II maps agentic capabilities onto 5G/6G control surfaces, standardization, and major 6G initiatives. Finally, it identifies open challenges shaping autonomous telecommunications.

Figures

Figures reproduced from arXiv: 2607.16066 by Abdelkader Mekrache, Adlen Ksentini, Bouziane Brik, Mazene Ameur.

Figure 1
Figure 1. Figure 1: Survey Structure. 2.1 Control plane In next-generation networks, the control plane is decomposed across multiple technological domains, namely the Radio Access Network (RAN), Transport Network (TN), Core Network (CN), and Edge/Cloud infrastructure. Each of these domains adopts the principle of Control and User Plane Separation (CUPS), enabling independent scaling, modularization, and programmable control o… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of AI-native 6G technologies. The TN adopts Software-Defined Networking (SDN) to centralize routing decisions in a controller that maintains a global topology view, enabling dynamic path computation, slicing, and load balancing [30]. Mainstream controllers such as ONOS4 , OpenDaylight5 , and the 6G-oriented TeraFlowSDN6 expose northbound APIs through which AI agents can inject routing intents or t… view at source ↗
Figure 3
Figure 3. Figure 3: LLM-based Agentic AI paradigm Evolution. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Telecom-oriented Agentic AI Framework Overview. [PITH_FULL_IMAGE:figures/full_fig_p010_4.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

180 extracted references · 2 canonical work pages

  1. [1]

    Andrews, Stefano Buzzi, Wan Choi, Stephen V

    Jeffrey G. Andrews, Stefano Buzzi, Wan Choi, Stephen V. Hanly, Angel Lozano, Anthony C. K. Soong, and Jianzhong Charlie Zhang. What will 5g be?IEEE Journal on Selected Areas in Communications, 32(6):1065–1082, 2014. doi: 10.1109/JSAC.2014.2328098

    arXiv 2014

  2. [2]

    A vision of 6g wireless systems: Applications, trends, technologies, and open research problems.IEEE Network, 34(3):134–142, 2020

    Walid Saad, Mehdi Bennis, and Mingzhe Chen. A vision of 6g wireless systems: Applications, trends, technologies, and open research problems.IEEE Network, 34(3):134–142, 2020. doi: 10.1109/MNET.001.1900287

    doi:10.1109/mnet.001.1900287 2020

  3. [3]

    Dual self-attention is what you need for model drift detection in 6g networks

    Mazene Ameur, Bouziane Brik, and Adlen Ksentini. Dual self-attention is what you need for model drift detection in 6g networks. IEEE Transactions on Machine Learning in Communications and Networking, 3:690–709, 2025. doi: 10.1109/TMLCN.2025.3576727

    arXiv 2025

  4. [4]

    Ameur, B

    M. Ameur, B. Brik, and A. Ksentini. Adapt but do not forget: Towards enhancing drift handling in 6g networks. InProceedings of the 2025 IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN), pages 1–6, Barcelona, Spain,

    2025

  5. [5]

    A speculative study on 6g.IEEE Wireless Communications, 27(4):118–125, August 2020

    Faisal Tariq et al. A speculative study on 6g.IEEE Wireless Communications, 27(4):118–125, August 2020. doi: 10.1109/MWC.001.1900334

    work page doi:10.1109/mwc.001.1900334 2020

  6. [6]

    A survey on intent-based networking.IEEE Communications Surveys & Tutorials, 25(1):625–655,

    Aris Leivadeas and Matthias Falkner. A survey on intent-based networking.IEEE Communications Surveys & Tutorials, 25(1):625–655,

  7. [7]

    ETSI GR ZSM 009-3 V1.1.1: Closed-loop automation (part 3)

    ETSI. ETSI GR ZSM 009-3 V1.1.1: Closed-loop automation (part 3). Technical report, European Telecommunications Standards Institute (ETSI), August 2023

    2023

  8. [8]

    Mekrache, A

    A. Mekrache, A. Ksentini, and C. Verikoukis. Intent-based management of next-generation networks: An llm-centric approach.IEEE Network, 38(5):29–36, 2024

    2024

  9. [9]

    The rise of agentic AI: Implications, concerns, and the path forward.IEEE Intelligent Systems, 40(2):8–14, March 2025

    San Murugesan. The rise of agentic AI: Implications, concerns, and the path forward.IEEE Intelligent Systems, 40(2):8–14, March 2025. doi: 10.1109/MIS.2025.3544940

    arXiv 2025

  10. [10]

    A. K. Pati. Agentic AI: A comprehensive survey of technologies, applications, and societal implications.IEEE Access, 13:151824–151837, 2025. 30•Mazene et al

    2025

  11. [11]

    Z. Chen, Q. Sun, N. Li, X. Li, Y. Wang, and I. Chih-Lin. Enabling mobile AI agent in 6g era: Architecture and key technologies.IEEE Network, 38(5):66–75, 2024

    2024

  12. [12]

    Jiang et al

    F. Jiang et al. A comprehensive survey of large AI models for future communications: Foundations, applications and challenges.arXiv preprint arXiv:cs.IT, 2025

    2025

  13. [13]

    Kong et al

    D. Kong et al. A survey of LLM-driven AI agent communication: Protocols, security risks, and defense countermeasures.arXiv preprint arXiv:cs.CR, 2025

    2025

  14. [14]

    Long et al

    S. Long et al. A survey on intelligent network operations and performance optimization based on large language models.IEEE Communications Surveys & Tutorials, 27(6):3915–3949, 2025

    2025

  15. [15]

    Saad et al

    W. Saad et al. Artificial general intelligence (AGI)-native wireless systems: A journey beyond 6g.Proceedings of the IEEE, PP(99):1–39, 2025

    2025

  16. [16]

    Abdulkadir Celik and Ahmed M. Eltawil. At the dawn of generative AI era: A tutorial-cum-survey on new frontiers in 6g wireless intelligence.IEEE Open Journal of the Communications Society, 5:2433–2489, 2024

    2024

  17. [17]

    Zhao et al

    C. Zhao et al. Edge general intelligence through world models and agentic AI: Fundamentals, solutions, and challenges.arXiv preprint arXiv:cs.LG, 2025

    2025

  18. [18]

    Mohammadi, Y

    M. Mohammadi, Y. Li, J. Lo, and W. Yip. Evaluation and benchmarking of LLM agents: A survey. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, volume 2, pages 6129–6139, 2025

    2025

  19. [19]

    Zhu et al

    J. Zhu et al. Evolutionary perspectives on the evaluation of LLM-based AI agents: A comprehensive survey.arXiv preprint arXiv:cs.CL, 2025

    2025

  20. [20]

    Jiang, C

    F. Jiang, C. Pan, L. Dong, K. Wang, O. A. Dobre, and M. Debbah. From large AI models to agentic AI: A tutorial on future intelligent communications.arXiv preprint arXiv:cs.AI, 2025

    2025

  21. [21]

    Masterman, S

    T. Masterman, S. Besen, M. Sawtell, and A. Chao. The landscape of emerging AI agent architectures for reasoning, planning, and tool calling: A survey.arXiv preprint arXiv:cs.AI, 2024

    2024

  22. [22]

    Sapkota, K

    R. Sapkota, K. I. Roumeliotis, and M. Karkee. AI agents vs. agentic AI: A conceptual taxonomy, applications and challenges.Information Fusion, 126:103599, 2026

    2026

  23. [23]

    M. A. Ferrag, N. Tihanyi, and M. Debbah. From LLM reasoning to autonomous AI agents: A comprehensive review.arXiv preprint arXiv:cs.AI, 2025

    2025

  24. [24]

    Zhang et al

    R. Zhang et al. Toward edge general intelligence with agentic AI and agentification: Concepts, technologies, and future directions. IEEE Communications Surveys & Tutorials, 2026. doi: 10.1109/COMST.2026.3651702

    arXiv 2026

  25. [25]

    Toward 6g networks: Use cases and technologies.IEEE Communications Magazine, 58(3):55–61, 2020

    Marco Giordani, Michele Polese, Marco Mezzavilla, Sundeep Rangan, and Michele Zorzi. Toward 6g networks: Use cases and technologies.IEEE Communications Magazine, 58(3):55–61, 2020

    2020

  26. [26]

    Majid Butt, Istvan Z

    Muhammad Kamran Shehzad, Luc Rose, M. Majid Butt, Istvan Z. Kovacs, Mohamad Assaad, and Mohsen Guizani. Artificial intelligence for 6g networks: Technology advancement and standardization.IEEE Vehicular Technology Magazine, 17(3):16–25, 2022

    2022

  27. [27]

    RAN-CN converged control-plane for 6g cellular networks

    Junyoung Choi et al. RAN-CN converged control-plane for 6g cellular networks. InProceedings of GLOBECOM 2022 - 2022 IEEE Global Communications Conference, pages 1253–1258, 2022

    2022

  28. [28]

    A survey on open radio access networks: Challenges, research directions, and open source approaches.Sensors, 24 (3):1038, 2024

    Wuri Azariah et al. A survey on open radio access networks: Challenges, research directions, and open source approaches.Sensors, 24 (3):1038, 2024

    2024

  29. [29]

    A survey of 5g network: Architecture and emerging technologies.IEEE Access, 3:1206–1232, 2015

    Ankur Gupta and Rakesh Kumar Jha. A survey of 5g network: Architecture and emerging technologies.IEEE Access, 3:1206–1232, 2015

    2015

  30. [30]

    Software defined 5g and 6g networks: A survey.Mobile Networks and Applications, 27(5):1792–1812, 2022

    Qiang Long, Yunfei Chen, Haijun Zhang, and Xianfu Lei. Software defined 5g and 6g networks: A survey.Mobile Networks and Applications, 27(5):1792–1812, 2022

    2022

  31. [31]

    Prabha, A

    C. Prabha, A. Goel, and J. Singh. A survey on SDN controller evolution: A brief review. InProceedings of the 7th International Conference on Communication and Electronics Systems (ICCES), pages 569–575, 2022

    2022

  32. [32]

    3GPP 5g core network: An overview and future directions.Array, 2022

    Syed Husain, Andreas Kunz, and Jaeseung Song. 3GPP 5g core network: An overview and future directions.Array, 2022

    2022

  33. [33]

    An in-kernel solution based on XDP for 5g UPF: Design, prototype and performance evaluation

    Tiago Alfredo Nascimento do Amaral, Raphael Vicente Rosa, Daniel Fernandes Campos Moura, and Christian Esteve Rothenberg. An in-kernel solution based on XDP for 5g UPF: Design, prototype and performance evaluation. InProceedings of the 17th International Conference on Network and Service Management (CNSM), pages 146–152, 2021

    2021

  34. [34]

    Mekrache, K

    A. Mekrache, K. Boutiba, and A. Ksentini. Combining network data analytics function and machine learning for abnormal traffic detection in beyond 5g. InProceedings of GLOBECOM 2023 - 2023 IEEE Global Communications Conference, pages 1204–1209, 2023

    2023

  35. [35]

    When mlops meets nwdaf to enable autonomous next generation network analytics

    Mazene Ameur, Bouziane Brik, and Adlen Ksentini. When mlops meets nwdaf to enable autonomous next generation network analytics. InICMLCN 2026, IEEE International Conference on Machine Learning in Communications and Networking, 2026

    2026

  36. [36]

    NEFSim: An open experimentation framework utilizing 3GPP’s exposure services

    Dimitris Fragkos, George Makropoulos, Athanasios Gogos, Harilaos Koumaras, and Alexandros Kaloxylos. NEFSim: An open experimentation framework utilizing 3GPP’s exposure services. InProceedings of the 2022 Joint European Conference on Networks and Communications & 6G Summit (EuCNC/6G Summit), pages 303–308, 2022

    2022

  37. [37]

    A comprehensive analysis of cloud service models: IaaS, PaaS, and SaaS in the context of emerging technologies and trend

    Raed Younis et al. A comprehensive analysis of cloud service models: IaaS, PaaS, and SaaS in the context of emerging technologies and trend. InProceedings of the 2024 International Conference on Electrical, Communication and Computer Engineering (ICECCE), pages 1–6, 2024. LLM-Powered Agentic AI for 5G/6G Networks: A Tutorial and Survey on Architectures, P...

    2024

  38. [38]

    Overview of research trends and challenges in 6g mobile networks and the computing continuum.Scientific and Practical Cyber Security Journal (SPCSJ), 8(4):54–65, 2024

    Slavko Kitanov et al. Overview of research trends and challenges in 6g mobile networks and the computing continuum.Scientific and Practical Cyber Security Journal (SPCSJ), 8(4):54–65, 2024

    2024

  39. [39]

    Zero touch management: A survey of network automation solutions for 5g and 6g networks.IEEE Communications Surveys & Tutorials, 24(4):2535–2578, 2022

    Estefania Coronado et al. Zero touch management: A survey of network automation solutions for 5g and 6g networks.IEEE Communications Surveys & Tutorials, 24(4):2535–2578, 2022

    2022

  40. [40]

    Zero-touch network and service management (ZSM); intent-driven autonomous networks

    ETSI ISG ZSM. Zero-touch network and service management (ZSM); intent-driven autonomous networks. Technical Report GR ZSM 011 V2.1.1, European Telecommunications Standards Institute (ETSI), 2024

    2024

  41. [41]

    Explainable AI (XAI): Core ideas, techniques, and solutions.ACM Computing Surveys, 55(9):1–33, 2023

    Rudresh Dwivedi et al. Explainable AI (XAI): Core ideas, techniques, and solutions.ACM Computing Surveys, 55(9):1–33, 2023

    2023

  42. [42]

    Toward native artificial intelligence in 6g networks: System design, architectures, and paradigms.arXiv preprint arXiv:2103.02823, 2021

    Jing Wu et al. Toward native artificial intelligence in 6g networks: System design, architectures, and paradigms.arXiv preprint arXiv:2103.02823, 2021

    Pith/arXiv arXiv 2021

  43. [43]

    Chochliouros et al

    Ioannis P. Chochliouros et al. Developing a 6g data and ML operations automation via an end-to-end AI framework: The 6G-DALI context. InProceedings of the IFIP International Conference on Artificial Intelligence Applications and Innovations, pages 129–144, 2025

    2025

  44. [44]

    Santos et al

    João F. Santos et al. Managing O-RAN networks: xapp development from zero to hero.IEEE Communications Surveys & Tutorials, 2025

    2025

  45. [45]

    In-band network telemetry: A survey.Computer Networks, 186:107763, 2021

    Liang Tan et al. In-band network telemetry: A survey.Computer Networks, 186:107763, 2021

    2021

  46. [46]

    Ksentini et al

    A. Ksentini et al. Lightweight resource exposure framework for efficient service and resource orchestration in the cloud-edge continuum. InProceedings of the 2025 IEEE International Conference on Communications Workshops (ICC Workshops), pages 2081–2087, 2025

    2025

  47. [47]

    MLOps as a service for AI-native 6g networks

    Slawomir Kukliński, Robert Kołakowski, and Bartosz Mastej. MLOps as a service for AI-native 6g networks. InProceedings of the 2025 IEEE 11th International Conference on Network Softwarization (NetSoft), pages 61–66, 2025

    2025

  48. [48]

    Vincent Poor, and Shuguang Cui

    Zhaohui Yang, Mingzhe Chen, Kai-Kit Wong, H. Vincent Poor, and Shuguang Cui. Federated learning for 6g: Applications, challenges, and opportunities.Engineering, 8:33–41, 2022

    2022

  49. [49]

    Ameur, A

    M. Ameur, A. Bradai, and N. Lagraa. Exploring teacher-student learning with multi-agent DRL for QoS routing in SDN. InProceedings of ICC 2025 - IEEE International Conference on Communications, pages 5933–5938, Montreal, QC, Canada, 2025. doi: 10.1109/ICC52391. 2025.11161544

    arXiv 2025

  50. [50]

    Knowledge distillation and dataset distillation of large language models: Emerging trends, challenges, and future directions.Artificial Intelligence Review, 59(1):17, 2026

    Lei Fang et al. Knowledge distillation and dataset distillation of large language models: Emerging trends, challenges, and future directions.Artificial Intelligence Review, 59(1):17, 2026

    2026

  51. [51]

    Explaining anomalies detected by autoencoders using shapley additive explanations.Expert Systems with Applications, 186:115736, 2021

    Liat Antwarg, Ronald Meir Miller, Bracha Shapira, and Lior Rokach. Explaining anomalies detected by autoencoders using shapley additive explanations.Expert Systems with Applications, 186:115736, 2021

    2021

  52. [52]

    Deterministic local interpretable model-agnostic explanations for stable explainability

    Muhammad Rehman Zafar and Naimul Khan. Deterministic local interpretable model-agnostic explanations for stable explainability. Machine Learning and Knowledge Extraction, 3(3):525–541, 2021

    2021

  53. [53]

    Mekrache et al

    A. Mekrache et al. On combining xai and llms for trustworthy zero-touch network and service management in 6g.IEEE Communications Magazine, 2025

    2025

  54. [54]

    A survey on in-context learning

    Qingxiu Dong et al. A survey on in-context learning. InProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 1107–1128, Miami, Florida, USA, 2024. Association for Computational Linguistics

    2024

  55. [55]

    How abilities in large language models are affected by supervised fine-tuning data composition

    Guanting Dong et al. How abilities in large language models are affected by supervised fine-tuning data composition. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 177–198, 2024

    2024

  56. [56]

    Jane Qin, Xiaohui Tao, and Fu Lee Wang

    Liang Xu, Haoran Xie, S.-Z. Jane Qin, Xiaohui Tao, and Fu Lee Wang. Parameter-efficient fine-tuning methods for pretrained language models: A critical review and assessment.arXiv preprint arXiv:cs.CL, 2023

    2023

  57. [57]

    LoRA: Low-rank adaptation of large language models

    Edward Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. LoRA: Low-rank adaptation of large language models. InProceedings of the International Conference on Learning Representations (ICLR), 2022

    2022

  58. [58]

    DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning.arXiv preprint arXiv:cs.CL, 2026

    DeepSeek-AI et al. DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning.arXiv preprint arXiv:cs.CL, 2026

    2026

  59. [59]

    A survey on knowledge distillation of large language models.arXiv preprint arXiv:2402.13116, 2024

    Xiaohan Xu et al. A survey on knowledge distillation of large language models.arXiv preprint arXiv:2402.13116, 2024

    Pith/arXiv arXiv 2024

  60. [60]

    ReFT: Representation finetuning for language models.arXiv preprint arXiv:cs.CL, 2024

    Zhengxuan Wu et al. ReFT: Representation finetuning for language models.arXiv preprint arXiv:cs.CL, 2024

    2024

  61. [61]

    Large language models are zero-shot reasoners

    Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. Large language models are zero-shot reasoners. InAdvances in Neural Information Processing Systems, volume 35, pages 22199–22213, 2022

    2022

  62. [62]

    The unreliability of explanations in few-shot prompting for textual reasoning

    Xi Ye and Greg Durrett. The unreliability of explanations in few-shot prompting for textual reasoning. InAdvances in Neural Information Processing Systems, volume 35, pages 30378–30392, 2022

    2022

  63. [63]

    Reasoning for translation: Comparative analysis of chain-of-thought and tree-of-thought prompting for LLM translation

    Lan Nguyen and Yue Xu. Reasoning for translation: Comparative analysis of chain-of-thought and tree-of-thought prompting for LLM translation. InProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics, pages 259–275, 2025

    2025

  64. [64]

    Retrieval-augmented generation for knowledge-intensive NLP tasks

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. Retrieval-augmented generation for knowledge-intensive NLP tasks. InAdvances in Neural Information Processing Systems (NeurIPS), volume 33, pages 9459–9474, 2020

    2020

  65. [65]

    API-Bank: A comprehensive benchmark for tool-augmented LLMs.arXiv preprint arXiv:2304.08244, 2023

    Minghao Li et al. API-Bank: A comprehensive benchmark for tool-augmented LLMs.arXiv preprint arXiv:2304.08244, 2023

    Pith/arXiv arXiv 2023

  66. [66]

    Venkata Bharat Parthasarathy, Ahsan Zafar, Asif Khan, and Arslan Shahid. The ultimate guide to fine-tuning LLMs from basics to breakthroughs: An exhaustive review of technologies, research, best practices, applied research challenges and opportunities.arXiv preprint arXiv:cs.LG, 2024. 32•Mazene et al

    2024

  67. [67]

    A practical guide to building agents

    OpenAI. A practical guide to building agents. https://cdn.openai.com/business-guides-and-resources/a-practical-guide-to-building- agents.pdf, 2026. Accessed: 20 January 2026

    2026

  68. [68]

    Building effective AI agents

    Anthropic. Building effective AI agents. https://www.anthropic.com/engineering/building-effective-agents, 2026. Accessed: 20 January 2026

    2026

  69. [69]

    Tool learning with foundation models

    Libo Qin et al. Tool learning with foundation models. InAdvances in Neural Information Processing Systems (NeurIPS), volume 36, pages 6215–6230, 2023

    2023

  70. [70]

    The evolution of agentic AI: Architecture and workflows for autonomous systems.Journal of Multidisciplinary, 5:418–427, 2025

    Manish Jaggavarapu. The evolution of agentic AI: Architecture and workflows for autonomous systems.Journal of Multidisciplinary, 5:418–427, 2025

    2025

  71. [71]

    ReAct: Synergizing reasoning and acting in language models

    Shunyu Yao, Jeffrey Zhao, Dian Yu, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. ReAct: Synergizing reasoning and acting in language models. InProceedings of the International Conference on Machine Learning (ICML), 2023

    2023

  72. [72]

    AutoGPT: Autonomous GPT for complex task solving.arXiv preprint arXiv:2304.03442, 2023

    Siru Qian, Hao Wang, and Chuan Shi. AutoGPT: Autonomous GPT for complex task solving.arXiv preprint arXiv:2304.03442, 2023

    Pith/arXiv arXiv 2023

  73. [73]

    Ameur, B

    M. Ameur, B. Brik, and A. Ksentini. EUROCOMPLY: Enabling zero-touch AI compliance auditing via LLM-based agentic AI.IEEE Communications Magazine, 2025. doi: 10.1109/MCOM.001.2500366

    work page doi:10.1109/mcom.001.2500366 2025

  74. [74]

    Multi-agent collaboration: Harnessing the power of LLMs.arXiv preprint arXiv:2305.11598, 2023

    Ankit Gupta, Siddharth Karamcheti, Ranjay Krishna, and Christopher Manning. Multi-agent collaboration: Harnessing the power of LLMs.arXiv preprint arXiv:2305.11598, 2023

    Pith/arXiv arXiv 2023

  75. [75]

    A survey of agent interoperability protocols: Model context protocol (MCP), agent communication protocol (ACP), agent-to-agent protocol (A2A), and agent network protocol (ANP)

    Aamir Ehtesham, Abhishek Singh, Gaurav Kumar Gupta, and Saurabh Kumar. A survey of agent interoperability protocols: Model context protocol (MCP), agent communication protocol (ACP), agent-to-agent protocol (A2A), and agent network protocol (ANP). arXiv preprint arXiv:cs.AI, 2025

    2025

  76. [76]

    E. Li, H. Du, and K. Huang. Netmcp: Network-aware model context protocol platform for llm capability extension, 2025

    2025

  77. [77]

    Ameur, B

    M. Ameur, B. Brik, and A. Ksentini. Agentic-nwdaf: Enabling intent-driven agentic intelligence for autonomous 6g network analytics. InProceedings of the IEEE International Conference on Communications (ICC), Glasgow, Scotland, UK, 2026

    2026

  78. [78]

    Agent network protocol (ANP)

    Agent Network Protocol Contributors. Agent network protocol (ANP). https://github.com/agent-network-protocol/ AgentNetworkProtocol, 2024. Accessed: 30 April 2025

    2024

  79. [79]

    Introduction to agent communication protocol (ACP)

    IBM BeeAI. Introduction to agent communication protocol (ACP). https://docs.beeai.dev/acp/alpha/introduction, 2024. Accessed: April 2025

    2024

  80. [80]

    Announcing agent payments protocol (AP2)

    Siddharth Parikh and Rohan Surapaneni. Announcing agent payments protocol (AP2). Google Cloud Blog, September 2025

    2025

Showing first 80 references.