Pith. sign in

REVIEW 5 major objections 6 minor 9 cited by

Large Language Model Enabled Multi-Task Physical Layer Network

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a single frozen LLM shared across task-specific adapters can perform multi-user precoding, signal detection, and channel prediction with performance comparable to dedicated single-task networks.

desk verdict The multi-task LLM architecture is genuinely new and the math is mostly sound, but the headline empirical claim rests on unreleased, seed-free simulations and a confounded ablation; worth a serious referee, not a desk reject. read the letter →

arxiv 2412.20772 v2 pith:CYPHHLJ6 submitted 2024-12-30 cs.IT eess.SPmath.IT

Tianyue Zheng , Linglong Dai This is my paper · ORCID
classification cs.ITeess.SPmath.IT
keywords largelanguagemodelsmulti-tasklearningphysicallayercommunicationsmulti-userprecodingsignaldetectionchannelpredictionLoRAquantization
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

The paper sets out to show that one large language model can serve as a shared brain for several physical-layer jobs at once. The three test tasks are downlink multi-user precoding, uplink signal detection, and channel prediction for moving users. The proposed network pairs a frozen LLAMA2-7B backbone with LoRA adapters and small task-specific encoders and decoders, and uses a natural-language instruction template to tell the model which task it is solving. The paper reports that this single multi-task network reaches performance comparable to dedicated single-task networks on all three tasks, and that quantization-aware LoRA initialization cuts fine-tuning storage by 75 percent with almost no performance loss. If that holds, a base station could deploy one pre-trained model across many PHY functions instead of retraining a separate network for each.

What carries the argument

The load-bearing mechanism is a shared frozen transformer decoder (LLAMA2-7B) modified by LoRA low-rank matrices, surrounded by per-task encoders and decoders joined by a prompt template of the form [Task Identifier] Task description <Instruction>. For precoding, the decisive object is the structural beamforming identity $\mathbf{w}_k^* = \sqrt{p_k}\left(\mathbf{I}_{N_T} + \sum_{k'=1}^K \frac{\lambda_{k'}}{\sigma^2} \mathbf{h}_{k'} \mathbf{h}_{k'}^H\right)^{-1} \mathbf{h}_k$, normalized, taken from [22]; because the optimal precoder is asserted to have this form, the network only has to produce the $2K$ scalars $\lambda$ and $p$ instead of the full beamforming matrix. For quantization, the decisive procedure alternates between quantizing the residual $\mathbf{W} - \mathbf{A}\mathbf{B}^T$ and taking its rank-$r$ SVD so that the 4-bit backbone plus adapters approximates the full-precision weights.

What would settle it

Train the multi-task network on a channel set whose optimal precoders are known to lie outside the family in Eq. (4) (for example, frequency-selective per-subcarrier precoding or measured near-field channels), and compare its sum rate against WMMSE with many iterations; if the LLM-based sum rate trails by a growing margin, the reported near-optimal precoding comes from the structural assumption rather than from the learned model.

Watch

Extended reading notes

Core claim

The authors claim that a single frozen LLM, augmented with LoRA and task-specific modules, can simultaneously learn multi-user precoding, signal detection, and channel prediction, and that this multi-task network is statistically indistinguishable in performance from separately fine-tuned single-task LLMs while costing much less to store during fine-tuning. The evidence is the QuaDRiGa UMa simulations: the shared network's channel-prediction NMSE tracks the single-task model across speeds and SNRs, its precoding sum rate is near WMMSE across transmit powers and user numbers, and its detection NMSE and SER are competitive, best at low SNR and slightly behind the noise-statistics-aware OAMP-Net at high SNR. The authors also claim the quantization-aware LoRA initialization keeps these numbers nearly unchanged while storing the backbone in 4-bit NF4 format, a 75% storage cut, and that the instruction prompt is responsible for faster convergence and improved task performance.

Load-bearing premise

The precoding claim's load-bearing premise is that the true optimal beamformer always has the structural form in Eq. (4), so a network that only outputs the 2K scalars $\lambda$ and $p$ cannot be better than that structure allows, no matter how well it learns.

Editorial extensions

If this is right

  • A single pretrained backbone can be reused for any new PHY task by attaching a small encoder, a decoder, and a task instruction, so the marginal cost of adding a task is millions of parameters instead of billions.
  • Multi-user precoding becomes a lightweight regression of 2K scalars followed by a closed-form beamformer, replacing iterative WMMSE at runtime with one forward pass.
  • LoRA fine-tuning-aware quantization makes it feasible to fine-tune the 7B backbone on a resource-limited device: the 4-bit backbone uses 75% less storage during fine-tuning with roughly equal NMSE and sum rate.
  • Because the LLM processes variable-length token sequences, the same trained network serves 4 to 8 users without re-architecting input tensors, unlike fixed-size deep-learning baselines.
  • The multi-task instruction prompt accelerates convergence and improves the network's adaptability, so prompt design is part of the training efficiency gain, not just bookkeeping.

Reading between the lines

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

  • A clean test the paper does not run is replacing the pretrained LLAMA2-7B with a randomly initialized transformer of the same size; if the sum rate and NMSE hold, the benefit comes from scale and architecture rather than from language pretraining, and the 'LLM' framing becomes incidental.
  • The shown results indicate multi-task training matches, but does not beat, the single-task LLM on each task; the likely economic win is memory and deployment cost rather than positive transfer, and low-data or out-of-distribution regimes could reveal either interference or synergy.
  • For real base stations the comparison to dedicated networks should include inference latency and energy: a 7B-parameter decoder adds per-symbol cost that the paper's accuracy figures do not capture.
  • The natural-language instruction could be ablated against a one-hot task token to test whether semantic task descriptions, rather than mere task identity, drive the faster convergence reported in the paper.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes a multi-task physical-layer network built on a single frozen LLAMA2-7B backbone, augmented with task-specific input encoders, output decoders, and LoRA adapters, to perform multi-user precoding, signal detection, and channel prediction simultaneously. The design uses natural-language instructions as prompts, a structural optimality result (Eq. (4)) to reduce the precoding output space to 2K parameters (λ, p), and a LoRA fine-tuning-aware quantization method adopted from LoftQ to reduce storage. The authors report simulations against dedicated baselines (WMMSE, DetNet, OAMP-Net, LSTM/GRU/Transformer, etc.) and claim comparable or better performance, with the quantization yielding 75% storage reduction during fine-tuning.

Significance. If the empirical claims are reliable, the paper would provide a useful demonstration that a large frozen language-model backbone, shared across several physical-layer tasks, can approach or beat task-specific networks with only small trainable adapters. The use of the exact structural parameterization for precoding (Eq. (4)) is principled, and the LoRA fine-tuning-aware quantization is a sensible adoption with clear practical benefit. The paper also gives a clean multi-task framework with prompts and encoders/decoders that can be extended to other tasks. However, the significance is substantially tempered by the fact that all headline conclusions rest on single-run simulations with no released code and several unreported hyperparameters, and by a confounded backbone ablation.

major comments (5)
  1. [§V, Figs. 5–10 and Tables I–II] All reported performance numbers are point estimates from a single simulation run; the paper never reports error bars, confidence intervals, or repeated-seed statistics. Since the central claim is that the proposed method achieves 'comparable' or 'near-optimal' performance, run-to-run variance could change the comparisons, particularly for the small differences in Table I (e.g., 29.3658 vs. 29.3632 bit/s/Hz). Please provide results over at least several random seeds with mean±std, or justify why the reported numbers are deterministic.
  2. [§V-G, Table II] The ablation removing the LLM backbone is confounded: removing the backbone also removes prompt processing, changes the number of trainable parameters (19.777M vs. 23.9713M), and changes the effective model depth. Therefore the comparison cannot isolate the contribution of the frozen LLM. To support the claim that a single shared LLM 'enables' multi-task PHY performance, the authors should train a matched non-LLM multi-task backbone (e.g., a transformer with a similar number of trainable parameters and the same prompts, or at least a shared multi-task network without the LLM) under identical data, loss, and training schedules.
  3. [§I and §V-D, Fig. 10] The abstract and introduction claim 'comparable performance with the dedicated designed network for each task,' but Fig. 10 shows OAMP-Net outperforming the proposed method at high SNR in signal detection. The body acknowledges this in Section V-D, but the high-level claim is therefore not uniformly supported. The abstract and conclusions should be qualified to state that the method is competitive at low SNR and slightly worse than OAMP-Net at high SNR in detection, or the framing should be revised.
  4. [§V-A and §IV-C] Several hyperparameters that are load-bearing for reproducibility are omitted: LoRA rank r, learning rate, batch size, optimizer, patch size N in the channel-prediction encoder, the number of CSI attention modules L in Eq. (20), the number of transformer blocks L in the precoding encoder, and the precise schedule for switching from supervised to unsupervised training in precoding (Eqs. (36)-(37)). Without these, the numerical results cannot be reproduced, and the claimed near-optimal precoding performance depends crucially on the two-stage training schedule.
  5. [footnote 1, §I] The paper states that 'Simulation codes will be provided to reproduce the results' but the link points to a general publications page, not to a specific repository or release. Given that the paper's evidence is entirely empirical, the absence of a direct code archive (e.g., a DOI or persistent repository) is a significant reproducibility gap. Please provide an accessible, permanent link to the code and data-generation scripts.
minor comments (6)
  1. [§II-A, Eq. (4)] The denominator in Eq. (4) appears garbled (the norm expression is not typeset correctly). Please fix the equation and ensure the normalization is clear.
  2. [§II-A, P1 (5)] Eq. (5) is stated as a minimization of the sum-rate objective, but the problem is a maximization; the sign is corrected only in the unsupervised loss (37). Please fix the formulation or explain the sign convention.
  3. [§IV-B, Algorithm 1] Line 4 of Algorithm 1 refers to '[36]-[37]' for the SVD step; these reference numbers do not match the LoftQ method cited in the text and the bibliography. Please correct the citation.
  4. [Table II] The 'Network parameters' column (23.9713 M for LLAMA2-7B) is confusing: the text says the backbone has 7B parameters while all other modules are about 19M. The table likely reports trainable parameters including LoRA, but this is not stated. Please clarify whether the numbers are trainable or total, and define the units consistently.
  5. [§III-A and §V-B] The 'CSI attention module' is attributed to [25], which is Squeeze-and-Excitation Networks; the connection is not explained. Either describe how the SE module is used for CSI attention or cite the appropriate channel-prediction attention reference.
  6. [References [8]] Reference [8] has an obviously wrong author list ('R. K. Alec'); it should be Radford et al. Please correct.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's central claims are empirical benchmark comparisons against external baselines, and its key structural ingredients (the Eq. (4) precoding parameterization and the LoRA fine-tuning-aware quantization) are imported from independent prior work rather than derived from the paper's own fitted inputs.

full rationale

The derivation chain in this paper does not reduce to its own inputs. The multi-user precoding problem is reformulated using the beamforming structure in Eq. (4), which is cited to Bjornson et al. [22], an external optimality result; the network is then asked to learn only the 2K scalar parameters lambda and p, with the WMMSE algorithm independently generating supervised labels and an unsupervised sum-rate loss directly optimizing the original objective. No fitted parameter is renamed as a prediction: the reported sum rate, NMSE, and SER results are computed from held-out test data against external baselines such as WMMSE, ZF, CNN, DetNet, OAMP-Net, LSTM, GRU, and LLM4CP. The quantization portion explicitly follows LoftQ [21] and QLoRA's NF4 format [27], and the claimed 75% storage reduction is the direct arithmetic consequence of moving from 16-bit to 4-bit weights, not a retrofitted number. The few self-citations, notably [5] and [17], are used as a baseline method and as general context, respectively, and neither carries the load of the paper's central claim. The paper's limitations are empirical and reproducibility-related - single-run simulations, no reported error bars, a confounded backbone ablation in Table II - but those are correctness and evidence-quality concerns, not circularity. Therefore the appropriate circularity score is 0.

Assumptions & free parameters 7 free parameters · 4 assumptions · 0 invented entities

No new physical entities, particles, forces, or dimensions are introduced. The architectural components (instruction prompts, encoders, decoders, LoRA adapters) are engineering constructs, not ontological entities. The ledger lists the hand-chosen hyperparameters and the domain assumptions that the central empirical claims rest on.

free parameters (7)
  • LoRA rank r
    Controls the capacity of the low-rank adapters in the LLM backbone; the paper never reports its value, so all performance results depend on an unstated hyperparameter.
  • Fine-tuning epochs = 200
    Hand-chosen training duration in Section V-A; the no-prompt baseline requires 300 epochs, so this choice affects the convergence claims.
  • Quantization iterations Iter = 5
    Set in Section V-E for the LoRA fine-tuning-aware quantization; larger values would improve the approximation at higher cost.
  • Transformer depth L in precoding encoder = 3
    Hand-chosen number of blocks in Section III-B1; no sensitivity analysis is provided.
  • Received time slots L0 in detection encoder = 8
    Selected in Section III-B2 based on training improvement; this is a data-dependent architectural choice.
  • Patch size N in channel prediction
    Introduced in Section III-B3 but never specified, making the channel prediction encoder impossible to reproduce exactly.
  • Precoding supervised-to-unsupervised transition
    The two-stage training in Section IV-C has no stated switch epoch, mixing ratio, or stopping criterion.
assumptions (4)
  • domain assumption The optimal downlink beamforming structure in Eq. (4) is exact, and learning only scalar parameters λ and p is sufficient for near-optimal sum rate.
    Invoked in Section II-A and problem P1; the network never predicts the full beamforming matrix W, so this structural result bounds the achievable performance.
  • domain assumption Pretrained LLAMA2-7B and GPT2 weights transfer to numerical wireless tasks after linear projection and LoRA fine-tuning.
    The framework in Sections III-C and IV depends on this transfer; the paper offers only simulation evidence, not a mechanism or broader validation.
  • domain assumption QuaDRiGa with 3GPP UMa NLOS parameters is representative of the claimed multi-task PHY performance.
    All experiments in Section V use this simulated channel model; no measured data or alternative standard model is tested.
  • domain assumption The alternating quantization and SVD initialization in Algorithm 1, adopted from LoftQ [21], provides a good low-rank approximation of the full-precision weights.
    The method is taken from [21] and not re-derived; the paper relies on its correctness to claim 75% storage reduction with negligible degradation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Large Language Model Enabled Multi-Task Physical Layer Network." pith.science (2026). https://pith.science/paper/CYPHHLJ6

@misc{pith2026241220772,
  author       = {Pith},
  title        = {Pith review of: Large Language Model Enabled Multi-Task Physical Layer Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CYPHHLJ6}},
  note         = {Machine review of arXiv:2412.20772}
}
read the original abstract

The advance of Artificial Intelligence (AI) is continuously reshaping the future 6G wireless communications. Particularly, the development of Large Language Models (LLMs) offers a promising approach to effectively improve the performance and generalization of AI in different physical-layer (PHY) tasks. However, most existing works finetune dedicated LLM networks for a single wireless communication task separately. Thus performing diverse PHY tasks requires extremely high training resources, memory usage, and deployment costs. To solve the problem, we propose a LLM-enabled multi-task PHY network to unify multiple tasks with a single LLM, by exploiting the excellent semantic understanding and generation capabilities of LLMs. Specifically, we first propose a multi-task LLM framework, which finetunes LLM to perform multi-user precoding, signal detection and channel prediction simultaneously. Besides, multi-task instruction module, input encoders, as well as output decoders, are elaborately designed to distinguish different tasks. The proposed design allows different wireless data types to be well aligned with the LLM input format. Moreover, low-rank adaptation (LoRA) is utilized for LLM fine-tuning. To reduce the memory requirement during LLM fine-tuning, a LoRA fine-tuning-aware quantization method is introduced. Extensive numerical simulations are also displayed to verify the effectiveness of the proposed method.

Figures

Figures reproduced from arXiv: 2412.20772 by the authors.

Figure 1
Figure 1. The model framework of our method. In this figure, “CP” denotes [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Detailed illustration of encoders and decoders. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The illustration of a transformer block in LLAMA2. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: The comparison of LoRA fine-tuning-aware quantization with tradi [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: The NMSE performance of proposed method and other baselines [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: The NMSE performance of proposed method and other baselines [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 9
Figure 9. Figure 9: The NMSE performance of the proposed method and other baselines [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: The SER performance of the proposed method and other baselines [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 9 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Modular PE-Structured Learning for Cross-Task Wireless Communications

    eess.SP 2025-09 conditional novelty 6.0 of 10

    By exploiting permutation equivariance, the authors assemble a compact modular Transformer that learns several wireless tasks with 100 samples per task and 9.71k parameters.

  2. LVM4CSI: Enabling Direct Application of Pre-Trained Large Vision Models for Wireless Channel Tasks

    cs.IT 2025-07 conditional novelty 6.0 of 10

    A frozen pre-trained vision model can extract wireless channel paths and features, beating conventional estimators in channel estimation and matching specialized networks in sensing with far fewer trainable parameters.

  3. Prompt-Enabled Large AI Models for CSI Feedback

    cs.IT 2025-01 conditional novelty 6.0 of 10

    A transformer-based large model, conditioned on the average channel magnitude of a scenario, improves CSI feedback accuracy and generalization across many simulated wireless environments.

  4. BERT4beam: Large AI Model Enabled Generalized Beamforming Optimization

    eess.SY 2025-09 conditional novelty 5.0 of 10

    A BERT-based transformer, BERT4beam, learns to output beamforming vectors from CSI and achieves near-SCA performance across multiple MU-MISO tasks and system scales.

  5. A Wireless Foundation Model for Multi-Task Prediction

    cs.AI 2025-07 conditional novelty 5.0 of 10

    A single pretrained wireless time-series model, using univariate decomposition and a causal Transformer, predicts channel, angle, and traffic data, and zero-shot predicts a new delay task better than traditional full-...

  6. Large Language Models for Next-Generation Wireless Network Management: A Survey and Tutorial

    cs.NI 2025-09 conditional novelty 4.0 of 10

    A survey and tutorial that organizes LLM-enabled wireless network optimization into formulation, solution, and verification stages, with case studies drawn from the authors' own prior papers.

  7. Empowering Near-Field Communications in Low-Altitude Economy with LLM: Fundamentals, Potentials, Solutions, and Future Directions

    eess.SP 2025-06 conditional novelty 4.0 of 10

    A proposed GPT-2-based framework for jointly classifying far versus near-field users and designing multiuser precoding in low-altitude drone networks, supported by a brief simulation and many open questions.

  8. Foundation Model Empowered Synesthesia of Machines (SoM): AI-native Intelligent Multi-Modal Sensing-Communication Integration

    eess.SP 2025-06 conditional novelty 4.0 of 10

    The paper proposes a systematic classification and two roadmaps for using foundation models (LLMs and wireless foundation models) to design Synesthesia of Machines systems for 6G, with preliminary case-study evidence ...

  9. A Comprehensive Survey of Large AI Models for Future Communications: Foundations, Applications and Challenges

    cs.IT 2025-05 conditional novelty 1.0 of 10

    A survey organizing the growing literature on large AI models for 6G communications, with a classification of model types, training and evaluation methods, and a list of challenges.

Reference graph

Works this paper leans on

37 extracted references · 24 canonical work pages · cited by 9 Pith papers

  1. [1]

    Toward a 6G AI-native air interface,

    J. Hoydis, F. A. Aoudia, A. Valcarce, and H. Viswanathan, “Toward a 6G AI-native air interface,” IEEE Commun. Mag. , vol. 59, no. 5, pp. 76–81, May 2021

    2021

  2. [2]

    An introduction to deep learning for the physical layer,

    T. O’Shea and J. Hoydis, “An introduction to deep learning for the physical layer,” IEEE Trans. Cog. Commun. Netw. , vol. 3, no. 4, pp. 563–575, Dec. 2017

    work page 2017

  3. [3]

    Deep learning for massive MIMO CSI feedback,

    C.-K. Wen, W.-T. Shih, and S. Jin, “Deep learning for massive MIMO CSI feedback,” IEEE Wireless Commun. Lett., vol. 7, no. 5, pp. 748–751, Oct. 2018. 13

    work page 2018

  4. [4]

    Deep learning-based channel estimation for beamspace mmwave massive MIMO systems,

    H. He, C.-K. Wen, S. Jin, and G. Y . Li, “Deep learning-based channel estimation for beamspace mmwave massive MIMO systems,” IEEE Wireless Commun. Lett., vol. 7, no. 5, pp. 852–855, Oct. 2018

    work page 2018

  5. [5]

    Accurate channel prediction based on transformer: Making mobility negligible,

    H. Jiang, M. Cui, D. W. K. Ng, and L. Dai, “Accurate channel prediction based on transformer: Making mobility negligible,” IEEE J. Sel. Areas Commun., vol. 40, no. 9, pp. 2717–2732, Sep. 2022

    work page 2022

  6. [6]

    Model-driven deep learning for MIMO detection,

    H. He, C.-K. Wen, S. Jin, and G. Y . Li, “Model-driven deep learning for MIMO detection,” IEEE Trans. Sigal Process. , vol. 68, pp. 1702–1715, Feb. 2020

    work page 2020

  7. [7]

    Artificial intelligence for 6G networks: Technology advancement and standardization,

    M. K. Shehzad, L. Rose, M. M. Butt, I. Z. Kov ´acs, M. Assaad, and M. Guizani, “Artificial intelligence for 6G networks: Technology advancement and standardization,” IEEE Veh. Tech. Mag., vol. 17, no. 3, pp. 16–25, May 2022

    work page 2022

  8. [8]

    Language models are unsupervised multitask learners,

    R. K. Alec, J. Wu, R. Child, D. Luan, D. Amodei, and I. Sutskever, “Language models are unsupervised multitask learners,” OpenAI blog , vol. 1, no. 8, p. 9, 2019

    work page 2019

Show all 37 references
  1. [9]

    Llama 2: Open foundation and fine-tuned chat models,

    H. Touvron, L. Martin, K. R. Stone, P. Albert, A. Almahairi, Y . Babaei, N. Bashlykov, S. Batra, P. Bhargava, and et al. . Shruti Bhosale, “Llama 2: Open foundation and fine-tuned chat models,” arXiv preprint arXiv:2307.09288, 2023

    2023 arXiv

  2. [10]

    Language models are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, and et al. . Amanda Askell, “Language models are few-shot learners,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS) , Dec. 2020, pp. 1–5

    2020

  3. [11]

    LLM-empowered resource allocation in wireless communications systems,

    W. Lee and J. Park, “LLM-empowered resource allocation in wireless communications systems,” arXiv preprint arXiv:2408.02944 , 2024

    2024 arXiv

  4. [12]

    Prompt-enabled large AI models for CSI feedback,

    J. Guo, Y . Cui, C.-K. Wen, and S. Jin, “Prompt-enabled large AI models for CSI feedback,” arXiv preprint arXiv:2501.10629 , 2025

    2025 arXiv

  5. [13]

    LLM4CP: Adapting large language models for channel prediction,

    B. Liu, X. Liu, S. Gao, X. Cheng, and L. Yang, “LLM4CP: Adapting large language models for channel prediction,” J. Commun. Inf. Netw. , vol. 9, no. 2, pp. 113–125, Jun. 2024

    2024

  6. [14]

    Beam predic- tion based on large language models,

    Y . Sheng, K. Huang, L. Liang, P. Liu, S. Jin, and G. Y . Li, “Beam predic- tion based on large language models,” arXiv preprint arXiv:2408.08707, 2024

    2024 arXiv

  7. [15]

    WirelessLLM: Empowering large language models towards wireless intelligence,

    J. Shao, J. Tong, Q. Wu, W. Guo, Z. Li, Z. Lin, and J. Zhang, “WirelessLLM: Empowering large language models towards wireless intelligence,” J. Commun. Info. Netw. , vol. 9, no. 2, pp. 99–112, Jun. 2024

    2024

  8. [16]

    Large language model (LLM) for telecom- munications: A comprehensive survey on principles, key techniques, and opportunities,

    H. Zhou, C. Hu, Y . Yuan, Y . Cui, Y . Jin, C. Chen, H. Wu, D. Yuan, L. Jiang, and et al.. Wu, Di, “Large language model (LLM) for telecom- munications: A comprehensive survey on principles, key techniques, and opportunities,” IEEE Commun. Surv. Tutor., Sep. 2024

    2024

  9. [17]

    AI and deep learning for THz ultra-massive MIMO: From model-driven approaches to foundation models,

    W. Yu, H. He, S. Song, J. Zhang, L. Dai, L. Zheng, and K. B. Letaief, “AI and deep learning for THz ultra-massive MIMO: From model-driven approaches to foundation models,” arXiv preprint arXiv:2412.09839 , 2024

    2024 arXiv

  10. [18]

    Large language model enhanced multi-agent systems for 6G communications,

    F. Jiang, Y . Peng, L. Dong, K. Wang, K. Yang, C. Pan, D. Niyato, and O. A. Dobre, “Large language model enhanced multi-agent systems for 6G communications,” IEEE Wireless Communications , vol. 31, no. 6, pp. 48–55, Dec. 2024

    2024

  11. [19]

    Large generative AI models for telecom: The next big thing?

    L. Bariah, Q. Zhao, H. Zou, Y . Tian, F. Bader, and M. Debbah, “Large generative AI models for telecom: The next big thing?” arXiv preprint arXiv:2306.10249, 2023

    2023 arXiv

  12. [20]

    LoRA: Low-rank adaptation of large language models,

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-rank adaptation of large language models,” arXiv preprint arXiv:2106.09685 , 2021

    2021 arXiv

  13. [21]

    LoftQ: LoRA-Fine-Tuning-Aware Quantization for large lan- guage models,

    Y . Li, Y . Yu, C. Liang, P. He, N. Karampatziakis, W. Chen, and T. Zhao, “LoftQ: LoRA-Fine-Tuning-Aware Quantization for large lan- guage models,” arXiv preprint arXiv:2310.08659 , 2023

    2023 arXiv

  14. [22]

    Optimal multiuser trans- mit beamforming: A difficult problem with a simple solution structure,

    E. Bj ¨ornson, M. Bengtsson, and B. Ottersten, “Optimal multiuser trans- mit beamforming: A difficult problem with a simple solution structure,” IEEE Signal Process. Mag. , vol. 31, no. 4, pp. 142–148, Jul. 2014

    2014

  15. [23]

    Effects of channel aging in massive MIMO systems,

    K. T. Truong and R. W. Heath, “Effects of channel aging in massive MIMO systems,” J. Commun. Netw., vol. 15, no. 4, pp. 338–351, Aug. 2013

    2013

  16. [24]

    A time series is worth 64 words: Long-term forecasting with transformers,

    Y . Nie, N. H. Nguyen, P. Sinthong, and J. Kalagnanam, “A time series is worth 64 words: Long-term forecasting with transformers,” arXiv preprint arXiv:2211.14730, 2022

    2022 arXiv

  17. [25]

    Squeeze-and-excitation networks,

    J. Hu, L. Shen, and G. Sun, “Squeeze-and-excitation networks,” in 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 7132–7141

    2018

  18. [26]

    Qwen2.5 technical report,

    A. Yang, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, and C. L. et al., “Qwen2.5 technical report,” arXiv preprint arXiv:2412.15115 , 2024

    2024 arXiv

  19. [27]

    QLoRA: Efficient finetuning of quantized LLMs,

    T. Dettmers, A. Pagnoni, A. Holtzman, and L. Zettlemoyer, “QLoRA: Efficient finetuning of quantized LLMs,” arXiv preprint arXiv:2305.14314, 2023

    2023 arXiv

  20. [28]

    QuaDRiGa: A 3-D multi-cell channel model with time evolution for enabling virtual field trials,

    S. Jaeckel, L. Raschkowski, K. B ¨orner, and L. Thiele, “QuaDRiGa: A 3-D multi-cell channel model with time evolution for enabling virtual field trials,” IEEE Trans. Antennas and Propag. , vol. 62, no. 6, pp. 3242–3256, Jun. 2014

    2014

  21. [29]

    Study on channel model for frequencies from 0.5 to 100 GHz(Release 15)[R]

    3GPP RADIO ACCESS NETWORK WORKING GROUP, “Study on channel model for frequencies from 0.5 to 100 GHz(Release 15)[R].” 3GPP TR 38.901, 2018

    2018

  22. [30]

    Neural network-based fading channel pre- diction: A comprehensive overview,

    W. Jiang and H. D. Schotten, “Neural network-based fading channel pre- diction: A comprehensive overview,” IEEE Access, vol. 7, pp. 118 112– 118 124, Aug. 2019

    2019

  23. [31]

    Deep learning for fading channel prediction,

    ——, “Deep learning for fading channel prediction,” IEEE Open J. Commun. Soc., vol. 1, pp. 320–332, Mar. 2020

    2020

  24. [32]

    LSTM-GRU model-based channel prediction for one-bit massive MIMO system,

    I. Helmy, P. Tarafder, and W. Choi, “LSTM-GRU model-based channel prediction for one-bit massive MIMO system,” IEEE Trans. Veh. Tech., vol. 72, no. 8, pp. 11 053–11 057, Mar. 2023

    2023

  25. [33]

    Performance analysis for massive MIMO downlink with low complexity approximate zero-forcing precoding,

    C. Zhang, Y . Jing, Y . Huang, and L. Yang, “Performance analysis for massive MIMO downlink with low complexity approximate zero-forcing precoding,” IEEE Trans. Commun., vol. 66, no. 9, pp. 3848–3864, Sep. 2018

    2018

  26. [34]

    Weighted sum-rate maximization using weighted MMSE for MIMO- BC beamforming design,

    S. S. Christensen, R. Agarwal, E. De Carvalho, and J. M. Cioffi, “Weighted sum-rate maximization using weighted MMSE for MIMO- BC beamforming design,” IEEE Trans. Wireless Commun., vol. 7, no. 12, pp. 4792–4799, Dec. 2008

    2008

  27. [35]

    A deep learning framework for optimization of MISO downlink beam- forming,

    W. Xia, G. Zheng, Y . Zhu, J. Zhang, J. Wang, and A. P. Petropulu, “A deep learning framework for optimization of MISO downlink beam- forming,” IEEE Trans. Commun. , vol. 68, no. 3, pp. 1866–1880, Dec. 2020

    2020

  28. [36]

    Power of deep learning for channel estimation and signal detection in OFDM systems,

    H. Ye, G. Y . Li, and B.-H. Juang, “Power of deep learning for channel estimation and signal detection in OFDM systems,” IEEE Wireless Commun. Lett., vol. 7, no. 1, pp. 114–117, Sep. 2018

    2018

  29. [37]

    Deep MIMO detection,

    N. Samuel, T. Diskin, and A. Wiesel, “Deep MIMO detection,” in Proc. 18th IEEE Int. Workshop Signal Process. Advances Wireless Commun (SPAWC), Hokkaido, Japan, Jul. 2017, pp. 1–5

    2017

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.