arxiv: 2605.11516 · v1 · submitted 2026-05-12 · 💻 cs.NI

Recognition: no theorem link

Agents Should Replace Narrow Predictive AI as the Orchestrator in 6G AI-RAN

Pranshav Gajjar , Vijay K Shah

Authors on Pith no claims yet

Pith reviewed 2026-05-13 02:08 UTC · model grok-4.3

classification 💻 cs.NI

keywords 6G networksAI-RANLarge Language Modelsnetwork orchestrationRAN Intelligent Controllerintent-based networkingautonomous networks

0 comments

The pith

Large language models should serve as the central orchestrator for 6G radio access networks instead of narrow predictive models.

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

The paper contends that Level 5 autonomous 6G networks cannot emerge from collections of isolated deep neural networks and reinforcement learning agents because those models lack shared knowledge and cannot convert high-level operator instructions into workable network settings. It proposes placing multimodal LLMs or adapted Large Telecom Models inside the RAN Intelligent Controller so they act as the main reasoning layer, calling on narrow models only as needed subroutines while using retrieval methods to spot and resolve multi-vendor problems. This architecture would let networks handle unstructured human directives and unexpected conditions without constant manual reprogramming. The authors outline specific research directions to adapt such models to the strict speed, safety, and security demands of telecommunications.

Core claim

The central claim is that multimodal LLMs or domain-adapted Large Telecom Models should replace narrow DNN and DRL agents as the primary orchestrator inside the RAN Intelligent Controller, where they translate operator intent into policies, invoke narrow models as executable tools, and apply retrieval-augmented generation to diagnose complex network issues.

What carries the argument

The LLM or LTM positioned as cognitive operating system in the RIC that dynamically directs narrow predictive models as subroutines and employs retrieval to handle anomalies.

If this is right

Narrow DNN and DRL models would function only as subordinate tools invoked by the central LLM rather than as independent decision makers.
Networks would accept unstructured operator directives and convert them into concrete configurations without separate intent-mapping layers.
Anomaly detection and resolution would incorporate retrieval from network data sources to address issues across multiple vendors.
Development focus would shift to continuous network-driven alignment, extreme quantization, hallucination checks, and protection against prompt attacks.

Where Pith is reading between the lines

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

The approach could allow incremental upgrades to existing RIC deployments by layering the LLM layer on top of current narrow-model pipelines.
It might reduce the frequency of retraining narrow models by letting the central agent handle context shifts that would otherwise break them.
Success would require new benchmarks that measure end-to-end intent fulfillment rather than isolated prediction accuracy.

Load-bearing premise

Narrow predictive models are inherently limited by siloed knowledge and cannot bridge high-level human intent to network actions, while LLMs can be adapted to do so reliably under telecom constraints on speed and safety.

What would settle it

A controlled test in which an LLM orchestrator in a simulated 6G RAN either violates latency bounds, hallucinates incorrect policies, or fails to maintain reliability when given ambiguous operator directives or adversarial inputs.

Figures

Figures reproduced from arXiv: 2605.11516 by Pranshav Gajjar, Vijay K Shah.

**Figure 2.** Figure 2: Token consumption during AI5GTest automated validation across [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Figure 3: Adversarial prompt injection in a TeleMCP-enabled LLM reasoning agent (Non-RT RIC). Malicious payloads embedded in syslog fields are ingested [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

read the original abstract

This position paper argues that to achieve Level 5 autonomous 6G networks, the next generation of Artificial Intelligence in Radio Access Networks (AI-RAN) should transition away from fragmented, narrow predictive models and instead adopt multimodal Large Language Models (LLMs) as central reasoning agents. Current AI-RAN architectures rely on disjointed Deep Neural Networks (DNNs) and Deep Reinforcement Learning (DRL) agents that operate in isolated domains. These narrow models suffer from siloed knowledge, severe brittleness to out-of-distribution dynamics, and a fundamental inability to bridge the intent gap the semantic disconnect between high-level, unstructured operator directives and rigid numerical network configurations. We propose elevating LLMs, or domain-adapted Large Telecom Models (LTMs), to act as the cognitive operating system situated within the RAN Intelligent Controller (RIC), the control and orchestration layer of AI-RAN. In this architecture, LLMs do not replace narrow models but orchestrate them as executable subroutines, dynamically translating human intent into concrete policies and utilizing Retrieval-Augmented Generation (RAG) to autonomously diagnose complex, multi-vendor network anomalies. To make this architectural shift a reality, we call upon the machine learning community to prioritize critical foundational research tailored to the strict constraints of telecommunications, specifically focusing on continuous alignment via network-driven feedback (RLNF), extreme sub-8-bit edge quantization, neuro-symbolic verification to curb hallucinations, and securing orchestration frameworks against adversarial prompt injections.

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.

Desk Editor's Note private letter to a colleague

This is a position paper arguing for LLMs as orchestrators in 6G AI-RAN but offers only qualitative claims with no data or tests.

read the letter

This paper pushes for a shift in 6G AI-RAN where large language models or domain-adapted telecom versions sit in the RIC as the main reasoning layer. Narrow DNN and DRL models would become subroutines that the LLM directs, with RAG handling anomaly diagnosis across vendors. The core pitch is that this setup closes the intent gap and reduces brittleness better than today's fragmented approach. It also flags four concrete research needs: network feedback loops for alignment, extreme quantization, neuro-symbolic checks, and prompt-injection defenses. That list gives the community specific problems to tackle rather than a generic wish for better AI. The paper does a clear job naming real pain points in current AI-RAN, such as models that stay locked in their own domains and cannot translate loose operator goals into precise configurations. Those observations line up with known limits in the field. The soft spots are the absence of any experiments, simulations, or even back-of-the-envelope checks. Claims that LLMs can meet telecom latency, reliability, and security rules rest on general capabilities rather than targeted evidence. We do not see how the proposed architecture would handle real-time constraints or avoid hallucinations in a live network. Readers already working on autonomous 6G architectures or AI control planes will get the most value as a way to frame future work. Someone looking for new algorithms, measurements, or validated designs will find little to use directly. The paper deserves peer review as a position piece. Referees can test whether the vision holds up and suggest concrete next steps for validation.

Referee Report

2 major / 2 minor

Summary. This position paper argues that achieving Level 5 autonomous 6G networks requires replacing fragmented narrow predictive AI (DNNs and DRL agents) with multimodal LLMs or domain-adapted Large Telecom Models (LTMs) as the central orchestrator inside the RAN Intelligent Controller (RIC). Current narrow models are described as suffering from siloed knowledge, brittleness to out-of-distribution inputs, and an inability to translate high-level operator intent into network configurations. The proposed architecture positions LLMs to orchestrate narrow models as subroutines, employ RAG for multi-vendor anomaly diagnosis, and dynamically generate policies, while the paper calls for targeted research on RLNF, sub-8-bit quantization, neuro-symbolic verification, and prompt-injection defenses.

Significance. If realized under telecom constraints, the proposal could enable more integrated, intent-aware control planes for 6G AI-RAN and help close the gap between high-level directives and low-level configurations. The explicit research agenda (RLNF, extreme quantization, neuro-symbolic methods, and security) usefully identifies concrete open problems that must be solved before LLM-based orchestration becomes viable in latency- and reliability-critical environments.

major comments (2)

[Abstract] Abstract: the central motivation—that narrow DNN/DRL models inherently possess 'siloed knowledge' and a 'fundamental inability to bridge the intent gap'—is asserted qualitatively without citations to empirical studies, failure cases, or quantitative characterizations of these limitations in deployed AI-RAN systems. This justification is load-bearing for the proposed architectural shift.
[Architecture proposal] Architecture proposal (throughout): no concrete interface, latency budget, or reliability mechanism is sketched for how an LLM orchestrator would invoke narrow models as subroutines inside the RIC while meeting 6G sub-millisecond control-loop requirements and safety constraints. This absence leaves the feasibility of the core claim unaddressed.

minor comments (2)

[Abstract] The acronym 'RLNF' is introduced without expansion on first use; a brief parenthetical definition would improve readability.
[Introduction] The manuscript would benefit from a short related-work paragraph contrasting the proposal with existing RIC xApp/rApp frameworks and prior LLM-for-networking efforts.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and positive recommendation for minor revision. We address each major comment below, clarifying the scope of this position paper while indicating targeted revisions to strengthen the manuscript.

read point-by-point responses

Referee: [Abstract] Abstract: the central motivation—that narrow DNN/DRL models inherently possess 'siloed knowledge' and a 'fundamental inability to bridge the intent gap'—is asserted qualitatively without citations to empirical studies, failure cases, or quantitative characterizations of these limitations in deployed AI-RAN systems. This justification is load-bearing for the proposed architectural shift.

Authors: We agree that the abstract would benefit from explicit citations to ground these claims. As a position paper, the core contribution is the proposed architectural vision rather than a new empirical study; however, the limitations of narrow models (siloed operation, OOD brittleness, and intent-to-configuration gaps) are supported by references in the introduction and related-work sections to prior AI-RAN literature on multi-vendor orchestration failures and intent-based networking challenges. In revision we will insert 2–3 targeted citations (e.g., to studies on DRL policy transfer failures and RIC xApp isolation issues) directly into the abstract and expand the motivation paragraph with a short quantitative characterization drawn from the cited works. revision: yes
Referee: [Architecture proposal] Architecture proposal (throughout): no concrete interface, latency budget, or reliability mechanism is sketched for how an LLM orchestrator would invoke narrow models as subroutines inside the RIC while meeting 6G sub-millisecond control-loop requirements and safety constraints. This absence leaves the feasibility of the core claim unaddressed.

Authors: We acknowledge that the manuscript does not provide numerical latency budgets or detailed interface pseudocode, as the paper’s purpose is to define the high-level cognitive-orchestrator role and enumerate the open research problems required to realize it under telecom constraints. The architecture section already states that the LLM operates at a supervisory timescale while narrow models retain sub-millisecond loops; we will add a short clarifying paragraph describing a candidate interface (asynchronous API calls with hard timeouts and fallback to last-known-good policies) and explicitly note that concrete sub-millisecond verification and safety mechanisms are among the open problems listed in the research agenda (RLNF, neuro-symbolic verification). This keeps the position-paper scope intact while directly addressing the referee’s concern. revision: partial

Circularity Check

0 steps flagged

No significant circularity: high-level position paper with no derivations or fitted quantities

full rationale

The manuscript is a position paper that frames its contribution as an architectural proposal for using LLMs/LTMs as orchestrators in the RIC of AI-RAN, accompanied by a call for foundational research on topics such as RLNF, quantization, neuro-symbolic verification, and prompt-injection security. It contains no equations, no parameter fitting, no predictions of numerical quantities, and no self-citations that serve as load-bearing justifications for technical claims. The arguments about limitations of narrow DNN/DRL models and the potential role of LLMs are presented as directions to explore rather than as results derived from prior outputs or definitions within the paper itself. The derivation chain is therefore self-contained and does not reduce any claim to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central proposal rests on untested assumptions about LLM capabilities in constrained telecom environments rather than new evidence or derivations.

axioms (2)

domain assumption Narrow predictive models cannot bridge the semantic disconnect between high-level operator directives and network configurations
Invoked to justify elevating LLMs as the orchestrator.
domain assumption Domain-adapted LLMs can be made reliable enough for real-time network orchestration under latency and reliability constraints
Required for the proposed architecture to function without major new failures.

pith-pipeline@v0.9.0 · 5563 in / 1345 out tokens · 59732 ms · 2026-05-13T02:08:18.222136+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages · 1 internal anchor

[1]

Empowering next-gen networks: Ai-driven autonomy in o-ran and son architectures.IEEE Access, 13:186903–186936, 2025

Dalia Abouelmaati, Alireza Esfahani, Shidrokh Goudarzi, and Shahid Mumtaz. Empowering next-gen networks: Ai-driven autonomy in o-ran and son architectures.IEEE Access, 13:186903–186936, 2025

work page 2025
[2]

O-ran empowering vertical industry: Scenarios, solutions and best practice.White Paper, Dec, 2023

ORAN Alliance. O-ran empowering vertical industry: Scenarios, solutions and best practice.White Paper, Dec, 2023

work page 2023
[3]

Assessment of deep learning method- ology for self-organizing 5g networks.Applied Sciences, 9(15):2975, 2019

Muhammad Zeeshan Asghar, Mudassar Abbas, Khaula Zeeshan, Pyry Kotilainen, and Timo Hämäläinen. Assessment of deep learning method- ology for self-organizing 5g networks.Applied Sciences, 9(15):2975, 2019

work page 2019
[4]

Massive mimo channel estimation with an untrained deep neural network.IEEE Transactions on Wireless Communications, 19(3):2079–2090, 2020

Eren Balevi, Akash Doshi, and Jeffrey G Andrews. Massive mimo channel estimation with an untrained deep neural network.IEEE Transactions on Wireless Communications, 19(3):2079–2090, 2020

work page 2079
[5]

Llm-hric: Llm-empowered hierarchical ran intelligent control for o-ran.arXiv preprint arXiv:2504.18062, 2025

Lingyan Bao, Sinwoong Yun, Jemin Lee, and Tony QS Quek. Llm-hric: Llm-empowered hierarchical ran intelligent control for o-ran.arXiv preprint arXiv:2504.18062, 2025

work page arXiv 2025
[6]

Tele- knowledge pre-training for fault analysis

Zhuo Chen, Wen Zhang, Yufeng Huang, Mingyang Chen, Yuxia Geng, Hongtao Yu, Zhen Bi, Yichi Zhang, Zhen Yao, Wenting Song, et al. Tele- knowledge pre-training for fault analysis. In2023 IEEE 39th International Conference on Data Engineering (ICDE), pages 3453–3466. IEEE, 2023

work page 2023
[7]

Neural architecture search survey: A hardware perspective.ACM Computing Surveys, 55(4):1– 36, 2022

Krishna Teja Chitty-Venkata and Arun K Somani. Neural architecture search survey: A hardware perspective.ACM Computing Surveys, 55(4):1– 36, 2022

work page 2022
[8]

Zhiyi Cui, Faizan Qamar, Syed Hussain Ali Kazmi, Khairul Akram Zainol Ariffin, Ghazanfar Ali Safdar, and Muhammad Habib ur Rehman. A review of multi-agent deep reinforcement learning for resource allocation in beyond 5g network slicing: solutions, challenges and future research directions.PeerJ Computer Science, 12:e3728, 2026

work page 2026
[9]

Pacifista: Conflict evaluation and management in open ran.IEEE Transactions on Mobile Computing, 2025

Pietro Brach del Prever, Salvatore D’Oro, Leonardo Bonati, Michele Polese, Maria Tsampazi, Heiko Lehmann, and Tommaso Melodia. Pacifista: Conflict evaluation and management in open ran.IEEE Transactions on Mobile Computing, 2025

work page 2025
[10]

Dnn-based localization from channel estimates: Feature design and experimental results

Paul Ferrand, Alexis Decurninge, and Maxime Guillaud. Dnn-based localization from channel estimates: Feature design and experimental results. InGLOBECOM 2020-2020 IEEE Global Communications Conference, pages 1–6. IEEE, 2020

work page 2020
[11]

Oran-bench-13k: An open source benchmark for assessing llms in open radio access networks

Pranshav Gajjar and Vijay K Shah. Oran-bench-13k: An open source benchmark for assessing llms in open radio access networks. In 2025 IEEE 22nd Consumer Communications & Networking Conference (CCNC), pages 1–4. IEEE, 2025

work page 2025
[12]

Oransight-2.0: Foundational llms for o-ran.IEEE Transactions on Machine Learning in Communications and Networking, 2025

Pranshav Gajjar and Vijay K Shah. Oransight-2.0: Foundational llms for o-ran.IEEE Transactions on Machine Learning in Communications and Networking, 2025

work page 2025
[13]

Tele-llm-hub: Building context-aware multi-agent llm systems for telecom networks.arXiv preprint arXiv:2511.09087, 2025

Pranshav Gajjar, Cong Shen, and Vijay K Shah. Tele-llm-hub: Building context-aware multi-agent llm systems for telecom networks.arXiv preprint arXiv:2511.09087, 2025

work page arXiv 2025
[14]

Neuro-symbolic verification for preventing llm hallucinations in process control.Processes, 14(2), 2026

Boris Galitsky and Alexander Rybalov. Neuro-symbolic verification for preventing llm hallucinations in process control.Processes, 14(2), 2026

work page 2026
[15]

Ai5gtest: Ai- driven specification-aware automated testing and validation of 5g o-ran components

Abiodun Ganiyu, Pranshav Gajjar, and Vijay K Shah. Ai5gtest: Ai- driven specification-aware automated testing and validation of 5g o-ran components. In18th ACM Conference on Security and Privacy in Wireless and Mobile Networks, pages 53–64, 2025. 8

work page 2025
[16]

Sharing minds during marl training for enhanced cooperative llm agents

Jiaxuan Gao, Yule Wen, Chao Yu, and Yi Wu. Sharing minds during marl training for enhanced cooperative llm agents. InLanguage Gamification- NeurIPS 2024 Workshop

work page 2024
[17]

Comix: Generalized conflict management in o-ran xapps-architecture, workflow, and a power control case.IEEE Access, 2025

Anastasios Giannopoulos, Sotirios Spantideas, George Levis, Alexandros Kalafatelis, and Panagiotis Trakadas. Comix: Generalized conflict management in o-ran xapps-architecture, workflow, and a power control case.IEEE Access, 2025

work page 2025
[18]

Mamba: Linear-time sequence modeling with selective state spaces, 2024

Albert Gu and Tri Dao. Mamba: Linear-time sequence modeling with selective state spaces, 2024

work page 2024
[19]

Ai-driven intent-based networking approach for self-configuration of next generation networks.arXiv preprint arXiv:2603.23772, 2026

Md Kamrul Hossain and Walid Aljoby. Ai-driven intent-based networking approach for self-configuration of next generation networks.arXiv preprint arXiv:2603.23772, 2026

work page arXiv 2026
[20]

Model context protocol (mcp): Landscape, security threats, and future research directions

Xinyi Hou, Yanjie Zhao, Shenao Wang, and Haoyu Wang. Model context protocol (mcp): Landscape, security threats, and future research directions. ACM Transactions on Software Engineering and Methodology, 2025

work page 2025
[21]

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, Qianglong 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

work page 2025
[22]

Qutaiba Ibrahim and Mustafa Qassab. Theory, concepts and future of self organizing networks (son).Recent Advances in Computer Science and Communications (Formerly: Recent Patents on Computer Science), 15(7):904–928, 2022

work page 2022
[23]

Agen- tic ai empowered intent-based networking for 6g.arXiv preprint arXiv:2601.06640, 2026

Genze Jiang, Kezhi Wang, Xiaomin Chen, and Yizhou Huang. Agen- tic ai empowered intent-based networking for 6g.arXiv preprint arXiv:2601.06640, 2026

work page arXiv 2026
[24]

Ai-ran in 6g networks: State- of-the-art and challenges.IEEE Open Journal of the Communications Society, 5:294–311, 2023

Naveed Ali Khan and Stefan Schmid. Ai-ran in 6g networks: State- of-the-art and challenges.IEEE Open Journal of the Communications Society, 5:294–311, 2023

work page 2023
[25]

Open RAN Conflict Agents: Detecting and Mitigating xApp Conflicts with Generative Agents

Dae Cheol Kwon and Xinyu Zhang. Open RAN Conflict Agents: Detecting and Mitigating xApp Conflicts with Generative Agents. In Proceedings of the IEEE Conference on Computer Communications (INFOCOM), 2026

work page 2026
[26]

Architecture of neural processing unit for deep neural networks

Kyuho J Lee. Architecture of neural processing unit for deep neural networks. InAdvances in computers, volume 122, pages 217–245. Elsevier, 2021

work page 2021
[27]

Awq: Activation-aware weight quantization for on-device llm compression and acceleration.Proceedings of machine learning and systems, 6:87–100, 2024

Ji Lin, Jiaming Tang, Haotian Tang, Shang Yang, Wei-Ming Chen, Wei- Chen Wang, Guangxuan Xiao, Xingyu Dang, Chuang Gan, and Song Han. Awq: Activation-aware weight quantization for on-device llm compression and acceleration.Proceedings of machine learning and systems, 6:87–100, 2024

work page 2024
[28]

Leaf: Navigating concept drift in cellular networks.Proceedings of the ACM on Networking, 1(CoNEXT2):1–24, 2023

Shinan Liu, Francesco Bronzino, Paul Schmitt, Arjun Nitin Bhagoji, Nick Feamster, Hector Garcia Crespo, Timothy Coyle, and Brian Ward. Leaf: Navigating concept drift in cellular networks.Proceedings of the ACM on Networking, 1(CoNEXT2):1–24, 2023

work page 2023
[29]

Small language models: Survey, measurements, and insights.arXiv preprint arXiv:2409.15790, 2024

Zhenyan Lu, Xiang Li, Dongqi Cai, Rongjie Yi, Fangming Liu, Xiwen Zhang, Nicholas D Lane, and Mengwei Xu. Small language models: Survey, measurements, and insights.arXiv preprint arXiv:2409.15790, 2024

work page arXiv 2024
[30]

Llm-pruner: On the structural pruning of large language models.Advances in neural information processing systems, 36:21702–21720, 2023

Xinyin Ma, Gongfan Fang, and Xinchao Wang. Llm-pruner: On the structural pruning of large language models.Advances in neural information processing systems, 36:21702–21720, 2023

work page 2023
[31]

Model drift in dynamic networks.IEEE Communications Magazine, 61(10):78– 84, 2023

Dimitrios Michael Manias, Ali Chouman, and Abdallah Shami. Model drift in dynamic networks.IEEE Communications Magazine, 61(10):78– 84, 2023

work page 2023
[32]

Implementation of transformer-based llms with large-scale optoelectronic neurons on a cmos compatible platform.APL Machine Learning, 4(1), 2026

Neil Na, Chih-Hao Cheng, Shou-Chen Hsu, Che-Fu Liang, Chung-Chih Lin, Nathaniel Y Na, Andrew I Shieh, Erik Chen, Haisheng Rong, and Richard A Soref. Implementation of transformer-based llms with large-scale optoelectronic neurons on a cmos compatible platform.APL Machine Learning, 4(1), 2026

work page 2026
[33]

Toward autonomous o-ran: A multi-scale agentic ai framework for real- time network control and management.arXiv preprint arXiv:2602.14117, 2026

Hojjat Navidan, Mohammad Cheraghinia, Jaron Fontaine, Mohamed Seif, Eli De Poorter, H Vincent Poor, Ingrid Moerman, and Adnan Shahid. Toward autonomous o-ran: A multi-scale agentic ai framework for real- time network control and management.arXiv preprint arXiv:2602.14117, 2026

work page arXiv 2026
[34]

Llm-driven approach for safe and secure network management by design in iot-based systems.Symmetry, 18(2):337, 2026

Nenad Petrovic, Dragana Krstic, and Mariusz Gł ˛ abowski. Llm-driven approach for safe and secure network management by design in iot-based systems.Symmetry, 18(2):337, 2026

work page 2026
[35]

Understanding o-ran: Architecture, interfaces, algorithms, security, and research challenges.IEEE Communications Surveys & Tutorials, 25(2):1376–1411, 2023

Michele Polese, Leonardo Bonati, Salvatore D’oro, Stefano Basagni, and Tommaso Melodia. Understanding o-ran: Architecture, interfaces, algorithms, security, and research challenges.IEEE Communications Surveys & Tutorials, 25(2):1376–1411, 2023

work page 2023
[36]

Beyond connectivity: An open architecture for ai-ran convergence in 6g.IEEE Communications Magazine, 2026

Michele Polese, Niloofar Mohamadi, Salvatore D’Oro, Leonardo Bonati, and Tommaso Melodia. Beyond connectivity: An open architecture for ai-ran convergence in 6g.IEEE Communications Magazine, 2026

work page 2026
[37]

Direct preference optimization: Your language model is secretly a reward model.Advances in neural information processing systems, 36:53728–53741, 2023

Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. Direct preference optimization: Your language model is secretly a reward model.Advances in neural information processing systems, 36:53728–53741, 2023

work page 2023
[38]

Ai-native o-ran architectures for 6g: Towards real-time adapta- tion, conflict resolution, and efficient resource management.Authorea Preprints, 2025

Sif Eddine Salmi, Messaoud Ahmed Ouameur, Miloud Bagaa, George C Alexandropoulos, Abdellah Tahenni, Daniel Massicotte, and Adlen Ksentini. Ai-native o-ran architectures for 6g: Towards real-time adapta- tion, conflict resolution, and efficient resource management.Authorea Preprints, 2025

work page 2025
[39]

A mathematical theory of communication.The Bell system technical journal, 27(3):379–423, 1948

Claude Elwood Shannon. A mathematical theory of communication.The Bell system technical journal, 27(3):379–423, 1948

work page 1948
[40]

arXiv preprint arXiv:2407.18003

Luohe Shi, Hongyi Zhang, Yao Yao, Zuchao Li, and Hai Zhao. Keep the cost down: A review on methods to optimize llm’s kv-cache consumption. arXiv preprint arXiv:2407.18003, 2024

work page arXiv 2024
[41]

Addison-Wesley Boston, 1997

Michael John Sebastian Smith.Application-specific integrated circuits, volume 7. Addison-Wesley Boston, 1997

work page 1997
[42]

Spotlight: Accurate, explainable and efficient anomaly detection for open ran

Chuanhao Sun, Ujjwal Pawar, Molham Khoja, Xenofon Foukas, Ma- hesh K Marina, and Bozidar Radunovic. Spotlight: Accurate, explainable and efficient anomaly detection for open ran. InProceedings of the 30th Annual International Conference on Mobile Computing and Networking, pages 923–937, 2024

work page 2024
[43]

John Wiley & Sons, 2025

Nishith D Tripathi and Vijay K Shah.Fundamentals of O-RAN. John Wiley & Sons, 2025

work page 2025
[44]

Drl-based dynamic mac scheduler reconfiguration in o-ran

N Villegas, JL Herrera, L Diez, D Scotece, L Foschini, and R Agüero. Drl-based dynamic mac scheduler reconfiguration in o-ran. InICC 2025- IEEE International Conference on Communications, pages 5023–5028. IEEE, 2025

work page 2025
[45]

Reinforcement Learning for LLM Post-Training: A Survey

Zhichao Wang, Bin Bi, Shiva Kumar Pentyala, Kiran Ramnath, Sougata Chaudhuri, Shubham Mehrotra, Xiang-Bo Mao, Sitaram Asur, et al. A comprehensive survey of llm alignment techniques: Rlhf, rlaif, ppo, dpo and more.arXiv preprint arXiv:2407.16216, 2024

work page internal anchor Pith review Pith/arXiv arXiv 2024
[46]

Large language models for next-generation wireless network management: A survey and tutorial.arXiv preprint arXiv:2509.05946, 2025

Bisheng Wei, Ruihong Jiang, Ruichen Zhang, Yinqiu Liu, Dusit Niyato, Yaohua Sun, Yang Lu, Yonghui Li, Shiwen Mao, Chau Yuen, et al. Large language models for next-generation wireless network management: A survey and tutorial.arXiv preprint arXiv:2509.05946, 2025

work page arXiv 2025
[47]

Chain-of-thought prompting elicits reasoning in large language models.Advances in neural information processing systems, 35:24824–24837, 2022

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models.Advances in neural information processing systems, 35:24824–24837, 2022

work page 2022
[48]

Big data driven mobile traffic understanding and forecasting: A time series approach.IEEE transactions on services computing, 9(5):796–805, 2016

Fengli Xu, Yuyun Lin, Jiaxin Huang, Di Wu, Hongzhi Shi, Jeungeun Song, and Yong Li. Big data driven mobile traffic understanding and forecasting: A time series approach.IEEE transactions on services computing, 9(5):796–805, 2016

work page 2016
[49]

React: Synergizing reasoning and acting in language models

Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik R Narasimhan, and Yuan Cao. React: Synergizing reasoning and acting in language models. InThe eleventh international conference on learning representations, 2022

work page 2022
[50]

Mm-llms: Recent advances in multimodal large language models.Findings of the Association for Computational Linguistics: ACL 2024, pages 12401–12430, 2024

Duzhen Zhang, Yahan Yu, Jiahua Dong, Chenxing Li, Dan Su, Chenhui Chu, and Dong Yu. Mm-llms: Recent advances in multimodal large language models.Findings of the Association for Computational Linguistics: ACL 2024, pages 12401–12430, 2024

work page 2024
[51]

A survey of aiops in the era of large language models.ACM Computing Surveys, 58(2):1–35, 2025

Lingzhe Zhang, Tong Jia, Mengxi Jia, Yifan Wu, Aiwei Liu, Yong Yang, Zhonghai Wu, Xuming Hu, Philip Yu, and Ying Li. A survey of aiops in the era of large language models.ACM Computing Surveys, 58(2):1–35, 2025

work page 2025
[52]

When mcp servers attack: Taxonomy, feasibility, and mitigation.arXiv preprint arXiv:2509.24272, 2025

Weibo Zhao, Jiahao Liu, Bonan Ruan, Shaofei Li, and Zhenkai Liang. When mcp servers attack: Taxonomy, feasibility, and mitigation.arXiv preprint arXiv:2509.24272, 2025

work page arXiv 2025
[53]

Hao Zhou, Chengming Hu, Ye Yuan, Yufei Cui, Yili Jin, Can Chen, Haolun Wu, Dun Yuan, Li Jiang, Di Wu, et al. Large language model (llm) for telecommunications: A comprehensive survey on principles, key techniques, and opportunities.IEEE Communications Surveys & Tutorials, 27(3):1955–2005, 2024

work page 1955