Pith. sign in

REVIEW 3 major objections 5 minor 59 references

On the Convergence of Large Language Model Optimizer for Black-Box Network Management

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read The paper proves that the LLM optimizer framework converges to a globally optimal action whenever it uses elitist sampling, and derives how multiple parallel LLMs accelerate that convergence.

desk verdict First serious Markov-chain model of LLM-based black-box optimization, but the central convergence proof depends on a P2-positivity claim that Lemma 2 does not prove. read the letter →

arxiv 2507.02689 v1 pith:TDBJL2LV submitted 2025-07-03 cs.IT eess.SPmath.IT

classification cs.ITeess.SPmath.IT MSC 60J1090C2668T50
keywords largelanguagemodelsblack-boxoptimizationfinite-stateMarkovchainelitistsamplingconvergenceanalysiswirelessnetworkmanagementmulti-LLM
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that the large language model optimizer (LLMO) framework is not just a heuristic that works in simulations: it is a convergent stochastic search algorithm. The authors model the LLMO loop as a finite-state Markov chain and prove that with the elitist sampler, which feeds back the best actions seen so far, the probability of being at an optimal state tends to $1$ as iterations grow, regardless of the reward function's convexity or the initial actions. They further prove that the multi-LLM variant converges faster, with the optimality gap reduced by a factor of $\lambda^L$ per iteration when $L$ identical LLMs share a memory. If these results hold, prompt-based LLMs can be used as universal black-box solvers for network management tasks that lack closed-form models and gradients.

What carries the argument

The central object is the finite state space $\mathcal{S}$ of byte-pair-encoding (BPE) token vectors for every possible action population, together with the optimal-state subset $\mathcal{S}^\star$. The argument runs through the transition matrix $P_{\mathrm{LLM}}$, split into blocks $P_1,\dots,P_4$ according to moves inside, into, and out of the optimal set. Lemma 2 is the load-bearing mechanism: with elitist sampling, $P_1$ and $P_4$ are upper triangular, $P_2$ has at least one positive entry in every column, and $P_3=0$; the Perron-Frobenius theorem then forces $P_4^{\infty}=0$, which is exactly the claim that non-optimal states are left forever. The convergence-rate analysis uses the average convergence rate $\gamma(t)$ and its limit $q_{\max}=\max_{s\in\mathcal{S}'}q_{ss}$, which becomes $\lambda^L$ for identical LLMs.

What would settle it

Run the LLMO with elitist sampling on a reward where the global optimum is isolated and many nearby states are better than their predecessors but never optimal; if the LLM's token sampler from some non-optimal state assigns zero probability to the token vector of the global optimum, the process will never enter $\mathcal{S}^\star$ and the empirical probability of reaching the optimum stays below $1$. That can be checked directly by enumerating the support of the LLM's output distribution from that state.

Watch

Extended reading notes

Core claim

The core claim is that every LLMO run is a finite-state Markov chain, because tokenization turns candidate solutions into a finite language space. In that chain, elitist sampling imposes a special block structure on the transition matrix: rewards never decrease, transitions leaving the optimal set are impossible, and from every non-optimal state there is a positive-probability path into the optimal set. Theorem 2 then concludes $\lim_{t\to\infty} \Pr\{s(t)\in\mathcal{S}^\star\}=1$ for any initial distribution, and states that the LIFO sampler, which feeds back only the most recent proposals, cannot guarantee this. For the multi-LLM architecture, the paper characterizes the average convergence rate and shows the optimality gap shrinks by a factor $q_{\max}<1$ per iteration, reducing to $\lambda^L$ when all $L$ LLMs are identical. The paper also acknowledges that many local optima can still make practical convergence difficult, since the theorem is asymptotic.

Load-bearing premise

The proof assumes that from every non-optimal state the LLM can eventually produce a globally optimal proposal with positive probability; showing that the LLM can improve is not the same as showing it can reach the true optimum, and the latter is not proven.

Editorial extensions

If this is right

  • A single LLM with elitist sampling is guaranteed to converge to a globally optimal action for any bounded black-box reward, convex or not, so the LLMO cannot be dismissed as a lucky heuristic.
  • The choice of sampler is decisive: LIFO sampling, which keeps only recent proposals, fails the guarantee and can stay stuck in non-optimal states, while elitist sampling retains the best actions and forces monotone improvement.
  • In the multi-LLM architecture, the optimality gap contracts by a fixed factor each iteration, and with $L$ identical models the factor is $\lambda^L$; thus increasing $L$ exponentially accelerates convergence.
  • The theoretical slopes from Theorem 3 match measured convergence curves on power control and massive MIMO tasks, giving practitioners a quantitative way to predict how many iterations and LLMs a deployment needs.

Reading between the lines

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

  • Editorial inference: The guarantee is asymptotic and inherits the unproven reachability premise that the LLM can always propose the global optimum with positive probability; this is why the paper's own caveat about many local optima is the practical failure mode to watch.
  • Editorial inference: Nothing in the Markov-chain argument is specific to language models as such; any tokenizable stochastic proposal generator with positive support on the optimum would inherit the same convergence theorem, so the result acts as a general template for elitist in-context search.
  • Editorial inference: A direct way to test the theoretical rate in practice is to estimate $q_{\max}$ from the LLM's empirical token-level transition distribution and compare predicted versus observed optimality-gap slopes across base models, prompt templates, and population sizes.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper develops a Markov chain model for the LLM optimizer (LLMO) framework applied to black-box network management. The authors formalize the tokenization of action populations, define a finite state space, and prove (Theorem 2) that with elitist sampling the LLMO converges almost surely to a state containing a globally optimal action, regardless of the convexity of the reward function or the initial distribution. They then extend the analysis to a multi-LLM architecture (Section IV), deriving an average convergence rate that decreases exponentially with the number of identical LLMs (Theorem 3). The theoretical claims are supported by extensive simulations on interference channels, broadcast channels, and massive MIMO systems using several LLMs.

Significance. If the theoretical results were correct, this would be a valuable first rigorous foundation for LLM-based black-box optimizers, justifying the widely observed empirical success of elitist sampling and multi-LLM designs. The paper also contributes a clean tokenizer-level state space construction and a broad experimental study. However, as the major comments detail, the central convergence proof and the multi-LLM transition analysis contain load-bearing gaps. The significance is therefore conditional on a successful revision; the current manuscript does not establish its advertised claims.

major comments (3)
  1. [Appendix C, Lemma 2] The proof that P2 has at least one positive entry per column is a non-sequitur. Upper-triangularity of P4 only implies that from a non-optimal state there is a positive transition to some higher-ranked state; it does not imply that this state lies in the optimal set S*. For example, order states as o ≻ a ≻ b with o optimal and a,b non-optimal, and set p(a|b)=ε, p(b|b)=1-ε, p(a|a)=1, with no transitions from a or b to o. Then P4 is upper triangular and P3=0, yet P2 is identically zero. The proof of Theorem 2 in Appendix D relies precisely on the column sums of P4 being less than 1, which requires the unproven P2 positivity. Thus the almost-sure convergence claim is unsupported.
  2. [Section IV-A, Lemma 3] The identity s(t) = max{s(t)_1, ..., s(t)_L, s(t-1)} in (20) is incorrect for P>1. The elitist sampler selects the P best actions from the union of the L generated populations and the previous in-context examples, which is not the lexicographic maximum of states as defined in Definition 4. For instance, with P=2, let state A have rewards (10,1) and state B have rewards (9,8). The lexicographic maximum is A, but the top two actions from the union are (10,9), a state different from both A and B. Consequently, the transition probability formula (19) and the subsequent convergence-rate conclusions in Theorem 3 are not established.
  3. [Section III-C, Theorem 2] The theorem claims convergence regardless of the convexity of the reward function or the initial distribution, but the proof assumes, through Lemma 2, that from every non-optimal state there is a positive transition to an optimal state. This reachability is not implied by Assumptions 1-2, which only ensure token availability and finite precision. The softmax positivity in (9)-(10) could provide such a guarantee if the restricted vocabulary set T contains all tokens in (13), but this argument is not made, and the notation T is ambiguous because it is reused for the restricted sampling vocabulary in (9) and the token set in (13). The proof therefore effectively assumes the very reachability that it sets out to establish.
minor comments (5)
  1. [Section III-A, equations (9) and (13)] The symbol T is used both for the restricted sampling vocabulary in (9) and for the set of relevant tokens in (13). Please disambiguate these two sets, for example by using T_samp and T_tok.
  2. [Appendix A] The sentence 'We need three extra tokens for the decimal point, minus sign, comma, and newline character' lists four items; clarify that the comma and newline are alternative separators for the CSV format, so that the total number of extra tokens is three.
  3. [Definition 4] The state ordering should be stated explicitly as a lexicographic order on the sorted reward sequences of the P actions; the current formulation is informal and could lead to ambiguity.
  4. [Proof of Lemma 3] The proof uses conditional independence of s(t)_l for l=1,...,L given s(t-1) without explicitly stating that the L LLMs generate independently given the same prompt; this assumption should be stated.
  5. [Lemma 1] The state space S is defined as all token vectors of length P D N_token, which includes token strings that do not correspond to valid CSV action matrices; this over-approximation is harmless for the Markov chain construction but should be acknowledged.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the LLMO convergence proof has a proof gap in Lemma 2, but the missing P2-positivity property is not assumed as an input and is derivable from the softmax positivity in (9)-(10).

full rationale

The paper's derivation chain is not circular. Theorem 1 constructs the finite-state Markov chain from the tokenizer analysis, the LLM conditional distribution in (10), the memory update in (5), and the sampling operator in (2); none of these pieces is defined in terms of the convergence conclusion. Theorem 2 relies on Lemma 2, and Appendix C's proof of Lemma 2 contains a genuine non-sequitur: upper triangularity of P4 only yields positive transitions to better non-optimal states, and does not by itself place mass in S*, so the claim that P2 has a positive entry in each column is not established as written. This is an unproven step (and Theorem 3 inherits it), but it is a correctness gap rather than circularity: the P2 positivity is not an input assumption or a renamed version of the theorem's conclusion, and it can be derived from (9)-(10), where every token in the restricted vocabulary has positive conditional probability, so from any prompt every token-vector state in S, including states containing a globally optimal action, has positive probability. The authors' self-citations [25]-[26] are used to motivate the multi-LLM architecture, not to supply the convergence argument, and the convergence-rate results are cited from the external reference [49]. The numerical fit of log10 lambda at L=1 is used to predict the L=3 and L=5 slopes on independent runs, which is a legitimate cross-validation rather than a fitted parameter renamed as a prediction. No self-definitional, fitted-input, imported-uniqueness, or ansatz-smuggling circularity was found.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The only genuinely new mathematical content is the Markov chain interpretation. The convergence guarantee requires either an explicit assumption that the LLM can reach the global optimum from any state, or a proof of that property from LLM inference; neither is provided. The finite-precision assumptions are reasonable but are not enough to establish reachability.

free parameters (2)
  • Number of digits Ndigit = 3 (in simulations)
    The finite state space requires finite-precision actions; the value is chosen by the user and the proof holds for any finite Ndigit.
  • Population size P = 5 (in simulations)
    The number of in-context examples and generated actions per iteration; chosen by the user, not fitted.
assumptions (3)
  • domain assumption Assumption 1: The restricted vocabulary set T contains tokens relevant for generating the action population in CSV format.
    Needed for the finite state space to include all required tokens; stated in Section III.B. It may fail for some LLM sampling configurations (e.g., top-K excluding numeric tokens).
  • domain assumption Assumption 2: The LLM generates actions as finite-precision floating-point numbers with Ndigit digits.
    Needed for finiteness of the action space; stated in Section III.B. Reasonable for current LLMs but not guaranteed for all decoding settings.
  • ad hoc to paper Reachability of the optimal set: from every non-optimal state, the transition block P2 has at least one positive entry to S*.
    This is asserted in Lemma 2 (Appendix C) but not proven; the given argument from the upper-triangular structure of P4 is insufficient. It is effectively an unstated assumption that the LLM can generate the global optimum with positive probability from any state.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On the Convergence of Large Language Model Optimizer for Black-Box Network Management." pith.science (2026). https://pith.science/paper/TDBJL2LV

@misc{pith2026250702689,
  author       = {Pith},
  title        = {Pith review of: On the Convergence of Large Language Model Optimizer for Black-Box Network Management},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TDBJL2LV}},
  note         = {Machine review of arXiv:2507.02689}
}
read the original abstract

Future wireless networks are expected to incorporate diverse services that often lack general mathematical models. To address such black-box network management tasks, the large language model (LLM) optimizer framework, which leverages pretrained LLMs as optimization agents, has recently been promoted as a promising solution. This framework utilizes natural language prompts describing the given optimization problems along with past solutions generated by LLMs themselves. As a result, LLMs can obtain efficient solutions autonomously without knowing the mathematical models of the objective functions. Although the viability of the LLM optimizer (LLMO) framework has been studied in various black-box scenarios, it has so far been limited to numerical simulations. For the first time, this paper establishes a theoretical foundation for the LLMO framework. With careful investigations of LLM inference steps, we can interpret the LLMO procedure as a finite-state Markov chain, and prove the convergence of the framework. Our results are extended to a more advanced multiple LLM architecture, where the impact of multiple LLMs is rigorously verified in terms of the convergence rate. Comprehensive numerical simulations validate our theoretical results and provide a deeper understanding of the underlying mechanisms of the LLMO framework.

Figures

Figures reproduced from arXiv: 2507.02689 by the authors.

Figure 1
Figure 1. LLMO framework [18]. Algorithm 1 LLMO Framework [18] Initialize M(0) , x (0) best, and r (0) best. for iteration t = 1, · · · , T do Sample [X (t−1) ex , r (t−1) ex ] using S(·) in (2). Generate pmpt(t−1) using P(·) in [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. An example prompt for LLMO. with new tasks that are not involved in the training corpus [16], [43]. Thus, the LLM can extract patterns of actions with high reward values, thereby enhancing the likelihood of generating optimal actions. In the LLMO, a selection of good sampling operators is crucial. Popular choices include an elitist sampler which extracts the top P actions from the memory M(t−1) [20], [23], [25], and… view at source ↗
Figure 3
Figure 3. Implementation of LLMO in wireless networks. [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (14 more)
Figure 4
Figure 4. Figure 4: Outline of convergence analysis. Finally, we can identify natural language outputs using the decoding process T −1 (·). Conse￾quently, the LLM inference (4) of the LLMO can be rewritten by X(t) = T −1  σ [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Interpretation of transition matrices. Since the states are sorted in the descending order, PLLM can be expressed as PLLM =   P1 P2 P3 P4   , (17) where P1 ≜ {pss˜:∀s, s˜ ∈ S ⋆} ∈ R |S ⋆|×|S ⋆| and P4 ≜ {pss˜:∀s, s˜ ∈ S ′} ∈ R |S ′ |×|S ′ | indicate the transition …
Figure 6
Figure 6. Figure 6: Multi-LLMO framework with L LLMs [25]. IV. MULTI-LLM OPTIMIZER FRAMEWORK Section III has proved that the LLMO attains the optimum as t → ∞. Thus, even with elitist sampling, the LLMO might suffer from slow convergence, as reported in various application scenarios [18]–…
Figure 7
Figure 7. Figure 7: Convergence behavior of various schemes for [PITH_FULL_IMAGE:figures/full_fig_p026_7.png]
Figure 8
Figure 8. Figure 8: Average EE with respect to L with t = 20. functions, the LLMO-E outperforms other BBO schemes, e.g., the GA and the BO. Since these methods require a careful optimization of various hyperparameters, they need to be fine-tuned for each given network scenario. In contras…
Figure 9
Figure 9. Figure 9: Average EE performance for various LLMO configurations with [PITH_FULL_IMAGE:figures/full_fig_p028_9.png]
Figure 10
Figure 10. Figure 10: Convergence behavior of LLMO with heterogeneous LLMs. [PITH_FULL_IMAGE:figures/full_fig_p028_10.png]
Figure 11
Figure 11. Figure 11: Convergence behavior of LLMO-E with various [PITH_FULL_IMAGE:figures/full_fig_p029_11.png]
Figure 12
Figure 12. Figure 12: Average convergence rate with respect to [PITH_FULL_IMAGE:figures/full_fig_p030_12.png]
Figure 13
Figure 13. Figure 13: Convergence behavior of various schemes with [PITH_FULL_IMAGE:figures/full_fig_p031_13.png]
Figure 14
Figure 14. Figure 14: Constraint violation of LLMO-E with L = 3. 0 5 10 15 20 SNR [dB] 0 1 2 3 4 5 6 Average SE [nats/Hz] Local optimal LLMO-E LLMO-L GA Brute-force [PITH_FULL_IMAGE:figures/full_fig_p032_14.png]
Figure 15
Figure 15. Figure 15: Average SE with respect to SNR with L = 3 and t = 100. penalty method improves more slowly than that of the language constraint method, and it eventually approaches the local optimum. This demonstrates that the penalty reward successfully guides the LLMO-E toward a fe…
Figure 16
Figure 16. Figure 16: Convergence behavior for cell-average EE reward with [PITH_FULL_IMAGE:figures/full_fig_p034_16.png]
Figure 17
Figure 17. Figure 17: An example of BPE tokenizer is presented in [PITH_FULL_IMAGE:figures/full_fig_p036_17.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

59 extracted references · 49 canonical work pages

  1. [1]

    Joint user selection, power allocation, and precoding design with imperfect CSIT for multi-cell MU-MIMO downlink systems,

    J. Choi, N. Lee, S.-N. Hong, and G. Caire, “Joint user selection, power allocation, and precoding design with imperfect CSIT for multi-cell MU-MIMO downlink systems,” IEEE Trans. Wireless Commun. , vol. 19, pp. 162–176, Jan. 2020

  2. [2]

    Robust precoding designs for multiuser MIMO systems with limited feedback,

    W. Zhou, D. Zhang, M. Debbah, and I. Lee, “Robust precoding designs for multiuser MIMO systems with limited feedback,” IEEE Trans. Wireless Commun. , vol. 23, pp. 9583–9595, Aug. 2024

  3. [3]

    Optimal design of energy-efficient multi-user MIMO systems: Is massive MIMO the answer?,

    E. Bj ¨ornson, L. Sanguinetti, J. Hoydis, and M. Debbah, “Optimal design of energy-efficient multi-user MIMO systems: Is massive MIMO the answer?,” IEEE Trans. Wireless Commun. , vol. 14, pp. 3059–3075, Jun. 2015

  4. [4]

    P. M. Pardalos, V . Rasskazova, and M. N. Vrahatis, Black Box Optimization, Machine Learning, and No-Free Lunch Theorems. Springer, 2021

  5. [5]

    Boyd and L

    S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge University Press, 2004

  6. [6]

    A unified algorithmic framework for block-structured optimization involving big data: With applications in machine learning and signal processing,

    M. Hong, M. Razaviyayn, Z.-Q. Luo, and J.-S. Pang, “A unified algorithmic framework for block-structured optimization involving big data: With applications in machine learning and signal processing,” IEEE Signal Process. Mag. , vol. 33, pp. 57–77, Jan. 2016

  7. [7]

    Deep learning for distributed optimization: Applications to wireless resource management,

    H. Lee, S. H. Lee, and T. Q. S. Quek, “Deep learning for distributed optimization: Applications to wireless resource management,” IEEE J. Sel. Areas Commun. , vol. 37, pp. 2251–2266, Oct. 2019

  8. [8]

    Artificial intelligence meets autonomy in wireless networks: A distributed learning approach,

    H. Lee, S. H. Lee, and T. Q. S. Quek, “Artificial intelligence meets autonomy in wireless networks: A distributed learning approach,” IEEE Netw., vol. 36, pp. 100–107, Nov. 2022

Show all 59 references
  1. [9]

    MOSAIC: Multiobjective optimization strategy for AI-aided internet of things communications,

    H. Lee, S. H. Lee, and T. Q. S. Quek, “MOSAIC: Multiobjective optimization strategy for AI-aided internet of things communications,” IEEE Internet Things J. , vol. 9, pp. 15657–15673, Sep. 2022

  2. [10]

    Learning optimal fronthauling and decentralized edge computation in fog radio access networks,

    H. Lee, J. Kim, and S.-H. Park, “Learning optimal fronthauling and decentralized edge computation in fog radio access networks,” IEEE Trans. Wireless Commun. , vol. 20, pp. 5599–5612, Sep. 2021

  3. [12]

    J. H. Holland, Adaptation in Natural and Artificial Systems . Ann Arbor: The University of Michigan Press, 1975

  4. [13]

    Taking the human out of the loop: A review of bayesian optimization,

    B. Shahriari, K. Swersky, Z. Wang, R. P. Adams, and N. de Freitas, “Taking the human out of the loop: A review of bayesian optimization,” Proc. IEEE, vol. 104, pp. 148–175, Jan. 2016

  5. [14]

    R. S. Sutton and A. G. Barto, Reinforcement Learning: An Introduction . MIT Press, 2nd ed., 2018

  6. [15]

    Language models are unsupervised multitask learners,

    A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, and I. Sutskever, “Language models are unsupervised multitask learners,” 2019. [Online] Available: https://cdn.openai.com/better-language-models/language models are unsupervised multitask learners.pdf

  7. [16]

    Language models are few-shot learners,

    T. B. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, et al., “Language models are few-shot learners,” Jul. 2020. [Online] Available: https://arxiv.org/abs/2005.14165

  8. [17]

    LLaMA: Open and efficient foundation language models,

    H. Touvron et al. , “LLaMA: Open and efficient foundation language models,” Feb. 2023. [Online] Available: https://arxiv.org/abs/2302.13971

  9. [18]

    Large language models as optimizers,

    C. Yang, X. Wang, Y . Lu, H. Liu, Q. V . Le, D. Zhou, and X. Chen, “Large language models as optimizers,” in Proc. Int. Conf. Learn. Represent. (ICLR) , Apr. 2024

  10. [19]

    Exploring the true potential: Evaluating the black-box optimization capability of large language models,

    B. Huang, X. Wu, Y . Zhou, J. Wu, L. Feng, R. Cheng, and K. C. Tan, “Exploring the true potential: Evaluating the black-box optimization capability of large language models,” Jul. 2024. [Online] Available: https://arxiv.org/abs/2404.06290. 41

  11. [20]

    Large language models as evolutionary optimizers,

    S. Liu, C. Chen, X. Qu, K. Tang, and Y .-S. Ong, “Large language models as evolutionary optimizers,” 2024. [Online] Available: https://arxiv.org/abs/2310.19046

  12. [21]

    Towards optimizing with large language model,

    P.-F. Guo, Y .-H. Chen, Y .-D. Tsai, and S.-D. Lin, “Towards optimizing with large language model,” 2023. [Online] Available: https://arxiv.org/abs/2310.05204

  13. [22]

    Importance of directional feedback for LLM-based optimizers,

    A. Nie, C.-A. Cheng, A. Kolobov, and A. Swaminathan, “Importance of directional feedback for LLM-based optimizers,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS) , pp. 1–12, Dec. 2023

  14. [23]

    Large language model-based evolutionary optimizer: Reasoning with elitism,

    S. Brahmachary et al. , “Large language model-based evolutionary optimizer: Reasoning with elitism,” Neurocomput., vol. 622, p. 129272, 2025

  15. [24]

    Large language model for multi-objective evolutionary optimization,

    F. Liu, X. Lin, Z. Wang, S. Yao, X. Tong, M. Yuan, and Q. Zhang, “Large language model for multi-objective evolutionary optimization,” Mar. 2024. [Online] Available: https://arxiv.org/abs/2310.12541

  16. [25]

    Large language models for knowledge-free network management: Feasibility study and opportunities,

    H. Lee, M. Kim, S. Baek, N. Lee, M. Debbah, and I. Lee, “Large language models for knowledge-free network management: Feasibility study and opportunities,” Sep. 2024. [Online] Available: https://arxiv.org/abs/2410.17259

  17. [26]

    AI-driven decentralized network management: Leveraging multi-agent large language models for scalable optimization,

    H. Lee, M. Kim, S. Baek, W. Zhu, M. Debbah, and I. Lee, “AI-driven decentralized network management: Leveraging multi-agent large language models for scalable optimization,” IEEE Commun. Mag. , vol. 63, pp. 50–56, Jun. 2025

  18. [27]

    Wireless multi-agent generative AI: From connected intelligence to collective intelligence,

    H. Zou, Q. Zhao, L. Baria, M. Bennis, and M. Debbah, “Wireless multi-agent generative AI: From connected intelligence to collective intelligence,” submitted to IEEE Commun. Mag. , 2023. [Online] Available: https://arxiv.org/abs/2307.02757

  19. [28]

    GenAINet: Enabling wireless collective intelligence via knowledge transfer and reasoning,

    H. Zou, Q. Zhao, L. Baria, Y . Tian, M. Bennis, S. Lasaulce, M. Debbah, and F. Bader, “GenAINet: Enabling wireless collective intelligence via knowledge transfer and reasoning,” submitted to IEEE Commun. Mag., 2024. [Online] Available: https://arxiv.org/abs/2402.16631

  20. [29]

    Large language model based multi-objective optimization for integrated sensing and communications in UA V networks,

    H. Li, M. Xiao, K. Wang, D. I. Kim, and M. Debbah, “Large language model based multi-objective optimization for integrated sensing and communications in UA V networks,” Oct. 2024. [Online] Available: https://arxiv.org/abs/2410.05062

  21. [30]

    Large language model-based wireless network design,

    K. Qiu, S. Bakirtzis, I. Wassell, H. Song, J. Zhang, and K. Wang, “Large language model-based wireless network design,” IEEE Wireless Commun. Lett. , vol. 13, pp. 3340–3344, Dec. 2024

  22. [31]

    WirelessAgent: Large language model agents for intelligent wireless networks,

    J. Tong, J. Shao, Q. Wu, W. Guo, Z. Li, Z. Lin, and J. Zhang, “WirelessAgent: Large language model agents for intelligent wireless networks,” Sep. 2024. [Online] Available: https://arxiv.org/abs/2409.07964

  23. [32]

    Chain-of-thought prompting elicits reasoning in large language models,

    J. Wei, X. Wang, D. Schuurmans, M. Bosma, B. Ichter, F. Xia, E. H. Chi, Q. V . Le, and D. Zhou, “Chain-of-thought prompting elicits reasoning in large language models,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), pp. 24824–24837, 2022

  24. [33]

    Self-consistency improves chain of thought reasoning in language models,

    X. Wang, J. Wei, D. Schuurmans, Q. Le, E. H. Chi, S. Narang, A. Chowdhery, and D. Zhou, “Self-consistency improves chain of thought reasoning in language models,” in Proc. Int. Conf. Learn. Represent. (ICLR) , Apr. 2023

  25. [34]

    Large language models as general pattern machines,

    S. Mirchandani, F. Xia, P. Florence, B. Ichter, D. Driess, M. G. Arenas, K. Rao, D. Sadigh, and A. Zeng, “Large language models as general pattern machines,” in Proc. Annu. Conf. Robot Learn. (CoRL) , pp. 1–21, Nov. 2023

  26. [35]

    LLMSense: Harnessing LLMs for high-level reasoning over spatiotemporal sensor traces,

    X. Ouyang and M. Srivastava, “LLMSense: Harnessing LLMs for high-level reasoning over spatiotemporal sensor traces,” Mar. 2024. [Online] Available: https://arxiv.org/abs/2403.19857

  27. [36]

    ReAct: Synergizing reasoning and acting in language models,

    S. Yao, D. Yu, J. Zhao, I. Shafran, T. L. Griffiths, Y . Cao, and K. Narasimhan, “ReAct: Synergizing reasoning and acting in language models,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS) , pp. 1–19, Dec. 2022

  28. [37]

    Reflextion: Language agents with verbal reinforcement learning,

    N. Shinn, F. Cassano, A. Gopinath, K. Narasimhan, and S. Yao, “Reflextion: Language agents with verbal reinforcement learning,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS) , pp. 1–19, Dec. 2023

  29. [38]

    OptiMUS: Scalable optimization modeling with (MI)LP solvers and large language models,

    A. AhmadiTeshnizi, W. Gao, and M. Udell, “OptiMUS: Scalable optimization modeling with (MI)LP solvers and large language models,” in Proc. Int. Conf. Machine Learn. (ICML) , pp. 577 – 596, Jul. 2024

  30. [39]

    Optimizing the focusing performance 42 of non-ideal cell-free mMIMO using genetic algorithm for indoor scenario,

    K. Shen, S. Safapourhajari, T. De Pessemier, L. Martens, W. Joseph, and Y . Miao, “Optimizing the focusing performance 42 of non-ideal cell-free mMIMO using genetic algorithm for indoor scenario,” IEEE Trans. Wireless Commun. , vol. 21, pp. 8832–8845, Oct. 2022

  31. [40]

    A block quantum genetic interference mitigation algorithm for dynamic metasurface antennas and field trials,

    T. Yang, H. Yin, R. Song, and L. Zhang, “A block quantum genetic interference mitigation algorithm for dynamic metasurface antennas and field trials,” IEEE Wireless Commun. Lett. , vol. 13, pp. 3678–3682, Dec. 2024

  32. [41]

    Decentralized computation offloading with cooperative UA Vs: A multi-agent deep reinforcement learning perspective,

    S. Hwang, H. Lee, J. Park, and I. Lee, “Decentralized computation offloading with cooperative UA Vs: A multi-agent deep reinforcement learning perspective,” IEEE Wireless Commun., vol. 29, pp. 24–31, Aug. 2022

  33. [42]

    Cooperative multi-agent deep reinforcement learning methods for UA V-aided mobile edge computing networks,

    M. Kim, H. Lee, S. Hwang, M. Debbah, and I. Lee, “Cooperative multi-agent deep reinforcement learning methods for UA V-aided mobile edge computing networks,” IEEE Internet Things J. , vol. 11, pp. 38040–38053, Dec. 2024

  34. [43]

    An explanation of in-context learning as implicit Bayesian inference,

    S. M. Xie, A. Raghunathan, P. Liang, and T. Ma, “An explanation of in-context learning as implicit Bayesian inference,” in Proc. Int. Conf. Learn. Represent. (ICLR) , Apr. 2022

  35. [44]

    Large language models are zero-shot reasoners,

    T. Kojima, S. S. Gu, M. Reid, Y . Matsuo, and Y . Iwasawa, “Large language models are zero-shot reasoners,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS) , pp. 22199–22213, Dec. 2022

  36. [45]

    Survey of hallucination in natural language generation,

    Z. Ji, N. Lee, R. Frieske, T. Yu, D. Su, Y . Xu, E. Ishii, Y . J. Bang, A. Madotto, and P. Fung, “Survey of hallucination in natural language generation,” ACM Comput. Surv., vol. 55, pp. 1–38, Mar. 2023

  37. [46]

    Towards mitigating hallucination in large language models via self- reflection,

    Z. Ji, T. Yu, Y . Xu, N. Lee, E. Ishii, and P. Fung, “Towards mitigating hallucination in large language models via self- reflection,” in Proc. Conf. Empirical Methods Natural Lang. Process. (EMNLP) , 2023

  38. [47]

    Convergence analysis of canonical genetic algorithms,

    G. Rudolph, “Convergence analysis of canonical genetic algorithms,” IEEE Trans. Neural Netw. , vol. 5, pp. 96–101, Jan. 1994

  39. [48]

    A Markov chain analysis on simple genetic algorithms,

    J. Suzuki, “A Markov chain analysis on simple genetic algorithms,” IEEE Trans. Syst. Man, Cybern., vol. 25, pp. 655–659, Apr. 1995

  40. [49]

    Average convergence rate of evolutionary algorithms,

    J. He and G. Lin, “Average convergence rate of evolutionary algorithms,” IEEE Trans. Evol. Comput., vol. 20, pp. 316–321, Apr. 2016

  41. [50]

    R. S. Varga, Matrix Iterative Analysis . Springer, 2009

  42. [51]

    On the convergence rates of genetic algorithms,

    J. He and L. Kang, “On the convergence rates of genetic algorithms,” Theore. Comput. Sci. , vol. 229, pp. 23–39, Nov. 1999

  43. [52]

    Average convergence rate of evolutionary algorithms in continuous optimization,

    Y . Chen and J. He, “Average convergence rate of evolutionary algorithms in continuous optimization,” Inf. Sci., vol. 562, pp. 200–219, Jul. 2021

  44. [53]

    An iteratively weighted MMSE approach to distributed sum-utility maximization for a MIMO interfering broadcast channel,

    Q. Shi, M. Razaviyayn, Z.-Q. Luo, and C. He, “An iteratively weighted MMSE approach to distributed sum-utility maximization for a MIMO interfering broadcast channel,” IEEE Trans. Signal Process. , vol. 59, pp. 4331–4340, Sep. 2011

  45. [54]

    Learning autonomy in management of wireless random networks,

    H. Lee, S. H. Lee, and T. Q. S. Quek, “Learning autonomy in management of wireless random networks,” IEEE Trans. Wireless Commun., vol. 20, pp. 8039–8053, Dec. 2021

  46. [55]

    Fractional programming for communication systems—Part I: Power control and beamforming,

    K. Shen and W. Yu, “Fractional programming for communication systems—Part I: Power control and beamforming,” IEEE Trans. Signal Process., vol. 66, pp. 2616–2630, May 2018

  47. [56]

    Horn and C

    R. Horn and C. Johnson, Matrix Analysis. Cambridge University Press, 1985

  48. [57]

    Genetic algorithms: What fitness scaling is optimal?,

    V . Kreinovich, C. Quintana, and O. Fuentes, “Genetic algorithms: What fitness scaling is optimal?,” Cybern. Syst., vol. 24, no. 1, pp. 9–26, 1993

  49. [58]

    DeepMIMO: A generic deep learning dataset for millimeter wave and massive MIMO applications,

    A. Alkhateeb, “DeepMIMO: A generic deep learning dataset for millimeter wave and massive MIMO applications,” in Proc. Inf. Theory Appl. Workshop (ITA) , pp. 1–8, Feb. 2019

  50. [59]

    TelecomGPT: A framework to build telecom- specfic large language models,

    H. Zou, Q. Zhao, Y . Tian, L. Bariah, F. Bader, T. Lestable, and M. Debbah, “TelecomGPT: A framework to build telecom- specfic large language models,” Jul. 2024. [Online] Available: https://arxiv.org/abs/2407.09424

  51. [60]

    Isoifescu, Finite Markov Processes and Their Applications

    M. Isoifescu, Finite Markov Processes and Their Applications . Chichester: Wiley, 1980

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.