Task-Agnostic Semantic Communications Relying on Information Bottleneck and Federated Meta-Learning

Dusit Niyato; Hao Wei; Ping Zhang; Wanli Ni; Wenjun Xu; Wen Wang; Yongming Huang

REVIEW 4 major objections 5 minor 34 references

Task-Agnostic Semantic Communications Relying on Information Bottleneck and Federated Meta-Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A single task-agnostic semantic communication receiver can serve image classification, visual question answering, and sentiment analysis with accuracy close to that of task-specific systems, while adapting to new users in a few gradient…

desk verdict A credible task-agnostic semantic communication system with a load-bearing convergence theorem that is not proved as stated; worth a serious referee, but only with substantial revision. read the letter →

arxiv 2504.21723 v2 pith:6YPA63E3 submitted 2025-04-30 eess.SP

Hao Wei , Wen Wang , Wanli Ni , Wenjun Xu , Yongming Huang , Dusit Niyato , Ping Zhang This is my paper · ORCID

classification eess.SP

keywords task-agnosticsemanticcommunicationinformationbottleneckdistributedmultimodalfederatedmeta-learningadaptivefeaturetransmissionresourceallocationconvergenceanalysiswirelessnetworks

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 aims to show that one semantic communication system can serve multiple tasks and input modalities without task-specific retraining. It proposes TASC, in which user-side semantic transmitters compress unimodal features while a shared base-station receiver fuses them into multimodal representations for classification, visual question answering, and sentiment analysis. Training combines a confluent distributed multimodal information-bottleneck loss (C-DMIB) with federated meta-learning, so the shared model can adapt to new users in one or a few gradient steps. The experiments report performance comparable to task-specific SemCom with a better rate-distortion tradeoff, and a roughly 59% reduction in test loss compared with federated learning after few adaptation steps.

What carries the argument

The central object is the confluent distributed multimodal information bottleneck (C-DMIB) loss, the sum of two variational information-bottleneck terms: a univocal term per modality that penalizes mutual information between input and transmitted feature while rewarding task-prediction accuracy, and a syncretic term applied to the fused multimodal representation that filters semantic noise. Variational upper bounds make the mutual-information terms tractable, and the same MLP channel encoder uses a dimension-importance function $\phi_j(\sigma^2)$ to adaptively transmit only the most useful semantic dimensions under dynamic channel conditions. The convergence argument rests on four smoothness and heterogeneity assumptions and produces a per-user loss divergence $\delta_k(t)$ that ranks semantic users for selection; a joint user-selection, power-control, and CPU-frequency allocation algorithm then solves the resulting non-convex mixed-integer problem.

What would settle it

Compute the empirical gradient norm and Hessian spectral norm of the CIFAR-10, CLEVR, and CMU-MOSI local losses at the pretrained initialization used in the experiments; if the gradients grow unboundedly or the Hessian is not Lipschitz, the FOSP bound of Theorem 1 and the $\delta_k(t)$ user ranking lose their theoretical backing. Alternatively, compare the global meta-loss after training with semantic users selected by $\delta_k(t)$ versus randomly selected users: if ranked selection does not consistently descend faster, Corollary 1's selection rule is contradicted.

Watch

Extended reading notes

Core claim

On its own terms, TASC claims that task-agnostic semantic communication is achievable: one receiver can handle multiple tasks and modalities, with accuracy close to that of a system separately trained for each task. The design principle is to apply the information bottleneck separately to each unimodal stream (univocal DMIB) and jointly to the fused multimodal representation (syncretic DMIB); combining both gives the confluent loss used in training. Over fading channels, an adaptive feature-transmission rule prunes or keeps semantic dimensions according to the channel noise level. Trained with federated meta-learning, the system reaches a model initialization from which new users adapt quickly; in the reported experiments, TASC cuts test loss by about 59% compared with FedAvg after few gradient-descent iterations, and its user-selection and resource-allocation rules follow from an explicit non-convex convergence analysis.

Load-bearing premise

The load-bearing premise is that the local loss functions of the pretrained Transformer encoders and MLP decoders satisfy the four smoothness, Lipschitz-Hessian, bounded-variance, and heterogeneity assumptions in Section IV-A; the paper states these without loss of generality but does not check them for the architectures used in the simulations.

Editorial extensions

If this is right

A single shared receiver trained once can serve image classification, VQA, and sentiment analysis with accuracy near that of task-specific models, reducing storage and switching overhead at the base station.
New users or new data distributions can be served with one or a few gradient steps rather than full retraining, which the paper supports with the reported 59% test-loss reduction versus federated learning.
The confluent DMIB loss yields a better rate-distortion tradeoff than either univocal or syncretic IB alone, because it prunes redundant unimodal dimensions and filters multimodal semantic noise.
The joint SU-selection, resource-block, power, and CPU-frequency allocation reduces per-round training latency and energy cost compared with greedy and random baselines.
The convergence and adaptation behavior transfers across the three evaluated tasks, suggesting the same transceiver can be deployed without retraining when the task set changes.

Reading between the lines

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

If the result holds at scale, task-agnostic semantic receivers could decouple model deployment from task specification, so operators could add a new task by adding a task-query embedding rather than retraining the radio access stack.
The dimension-importance function $\phi_j(\sigma^2)$ suggests a channel-aware variable-rate semantic code; a natural testable extension is to compare its rate-distortion curve against learned compression codes across many SNR values, not only the 8 dB operating point.
The same C-DMIB and federated-meta-learning combination could be applied to other modality triples, such as video plus sensor plus text for robotic control, although the paper does not test whether the convergence guarantees remain tight outside the three evaluated tasks.
The 59% rapid-adaptation result comes from the image-classification experiment; replicating it on VQA and sentiment analysis with brand-new users would directly test the generality of the claimed few-step adaptation.

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.

Desk Editor's Note

read the letter

Plainly: TASC is a sensible systems paper—shared task-agnostic receiver, three VIB-style losses, adaptive feature-dimension transmission, FML training, and a resource allocation scheme—but its theoretical contribution is not in the shape the paper claims. Theorem 1 is the load-bearing result for SU selection and the joint optimization, and as written it is unproved: Appendix A derives the bound only for τ0=1, and the printed term (K−1)/(K/K_t−1) blows up at K_t=K while the corresponding appendix term is its reciprocal. A referee needs to see this fixed before the convergence and resource-allocation claims can be used.

What is genuinely useful: the architecture is clearly described, the DMIB losses reduce to standard VIB with a fusion filter, and the simulation story is internally consistent. The adaptive transmission trick—ordering dimensions by importance with no mask overhead—is a nice practical detail. The rate-distortion comparison and the adaptation result (about 59% test-loss reduction over FL after a few steps) support the main qualitative claims. Credit is due for an honest, readable write-up.

The soft spots are in proportion: the experiments lack error bars and repeated runs, no code or data is released, and the closest prior multimodal unified SemCom baseline (reference [15]) does not get a head-to-head comparison. Several hyperparameters are left unspecified. The proof gap is not minor, because the SU-selection rule δ_k(t) and Algorithm 2 are presented as theoretically guaranteed, and Figs. 5–6 are interpreted through that lens. The self-citation to [33]—the same arXiv paper—means the proof isn't actually available elsewhere; that should be cleaned up. Nothing here is constructed by definition; circularity is not the issue.

Who is this for: researchers building task-oriented or multimodal SemCom systems, and anyone working on FML convergence in wireless settings. The system part deserves attention; the theory part needs serious rework. I would send it to a knowledgeable referee rather than desk-reject, with instructions to check Appendix A line by line and to require either a corrected theorem or an honest limitation statement.

Referee Report

4 major / 5 minor

Summary. The paper proposes TASC, a task-agnostic semantic communication framework in which multiple single-modality users transmit compressed semantic features to a base station that runs one shared, multimodal receiver. The training objective is a combination of a univocal distributed multimodal information bottleneck (U-DMIB), a syncretic variant (S-DMIB), and an adaptive selection of transmitted feature dimensions based on channel noise. The model is trained with federated meta-learning, and the authors provide a convergence analysis, a user-selection rule derived from per-user gradient divergence, and a joint user-selection/resource-allocation algorithm. The central claims are that TASC performs comparably to task-specific SemCom on image classification, VQA, and sentiment analysis, that the confluent DMIB gives a better rate-distortion tradeoff, and that new users adapt quickly, with about a 59% reduction in test loss relative to FL after a few gradient steps.

Significance. If the results hold, TASC would be a useful step toward a single shared receiver serving multiple tasks and modalities, with quantifiable benefits in convergence speed, latency, and energy consumption. The system design is ambitious, the DMIB formulation is a sensible variational IB extension, and the resource-allocation subproblems are given explicit closed-form solutions. The paper also deserves credit for reporting results on three distinct tasks and for making the full training protocol explicit in Algorithm 1. However, the theoretical convergence guarantee, which is the stated basis for the user-selection rule and for the resource-allocation objective, is not established as written; and the empirical evaluation lacks error bars, repeated runs, and a comparison against the closest prior unified multimodal SemCom baselines. The significance of the contribution is therefore conditional on a corrected proof and a strengthened experimental section.

major comments (4)

[IV-B, Appendix A] Theorem 1 is stated for a general number of meta-learning steps τ0, but the proof in Appendix A ends with "If we set τ0=1, we obtain the desired result." The displayed bounds in Appendix A contain factors τ0(τ0−1) or τ(τ−1), which vanish for τ0=1, so the argument as written establishes only the τ0=1 case. Corollary 1 and the user-selection criterion δ_k(t) in Eq. (62) inherit this gap. Please either provide the general-τ0 proof with explicit dependence of all constants on τ0, or restrict Theorem 1 and Corollary 1 to τ0=1 and correspondingly qualify the claims in Sections IV-C and V.
[Theorem 1, Eq. (58)] The factor (K−1)/(K/K_t−1) in Eq. (58) diverges when K_t=K, i.e., when every SU is selected, which is an admissible case in Algorithm 2 and in the optimization problem P. The corresponding Appendix A expression (1−K_t/K)/(K_t−K_t/K) vanishes at K_t=K and is the reciprocal of the Theorem 1 factor in general. At least one of the two expressions is algebraically wrong. Because the bound is the stated justification for the convergence improvement with increasing K_t, this error must be corrected and the dependence on the selected-user fraction re-derived.
[Eq. (106) and Theorem 1] The descent step from Eqs. (104) and (105) to Eq. (106) is not valid for the stated learning-rate range β∈(0,1/H_L]. Combining the cross-term lower bound (104) and the second-moment bound (105) gives a coefficient of E[||∇L(·)||²] equal to β²H_L − β/2, and requiring β²H_L − β/2 ≤ −β/4 amounts to β≤1/(4H_L), which is stronger than the range stated in Theorem 1. The proof therefore does not substantiate the contraction claimed in Eq. (106) under the theorem's hypotheses.
[VI, Figs. 3–7] The experimental section reports single curves without error bars, confidence intervals, or multiple seeds, and the baselines do not include the closest prior unified multimodal SemCom systems (e.g., refs. [13]–[15]). As a result, the headline claims that TASC is "comparable to task-specific SemCom" across tasks and achieves a "better rate-distortion tradeoff" are not yet quantified against the relevant state of the art. Please add repeated-run statistics and at least one task-agnostic/unified multimodal SemCom baseline to support the central empirical claims.

minor comments (5)

[References [33]] Reference [33] is the arXiv identifier of this same paper, yet the main text repeatedly cites it as an external "technical report" for proofs of Lemma 4, Lemma 5, Lemma 6, and the theorems; please cite the appendices in the manuscript directly or clearly identify the companion report as a distinct document.
[Theorem 1 statement] The statement of Theorem 1 refers to "the number of communication rounds K," but K elsewhere denotes the number of SUs and C denotes the number of communication rounds; please fix this notation for consistency.
[Fig. 4] The caption and surrounding text describe a "rate-distortion curve," while the axis labeling and the discussion refer to classification accuracy versus latency; please align the caption, axes, and discussion with the quantity actually plotted.
[VI-A] The reproducibility of the experiments would be improved by reporting the number of SUs per modality, the data partition for the support/query sets, the pretrained model versions, and the per-task hyperparameters used in the simulations.
[VI-A] The text states that TASC and Specific SC are trained at an SNR of 8 dB, while Fig. 3 evaluates a range of SNR values; please clarify whether the models are trained once at 8 dB and evaluated across SNRs, or retrained at each operating point.

Circularity Check

1 steps flagged · score 2.0 of 10

Central DMIB and FML derivations are self-contained; the only circularity-adjacent issue is that the convergence proofs are attributed to the paper's own identical arXiv preprint as a 'technical report.'

self citation load bearing [Section IV-B, Theorem 1 proof line; Section IV-C, Eq. (62); Section V-D, Algorithm 2]
"Proof: See Appendix A of the technical report [33]."

The referenced technical report [33] is the same arXiv:2504.21723 paper under review. Theorem 1 and Corollary 1 are the stated theoretical basis for the SU-selection divergence delta_k(t) in Eq. (62) and for the resource-allocation objective Psi(a) in Eq. (76). Attributing these proofs to the authors' own identical preprint makes the theoretical premise rest on a self-citation rather than on an independent, externally checkable source. Since the proof is reproduced in the manuscript's Appendix A, the derivation is not hidden, and the mathematical argument can be audited in the same document; this is therefore a minor self-citation issue rather than a full reduction of the result to its inputs.

full rationale

The derivation chain is largely self-contained and not circular by construction. The DMIB objective is obtained from the standard variational information bottleneck: Eq. (9) upper-bounds the rate term with a KL divergence to a Gaussian prior, and Eqs. (12), (14), and (16) give Monte Carlo training losses; no fitted quantity is later renamed as a prediction. The FML objective in Eq. (33) and the inner/outer updates in Eqs. (34)-(37) follow the standard model-agnostic meta-learning template, and the convergence analysis uses conventional smoothness, variance, and heterogeneity assumptions. The SU-selection weight delta_k(t) in Eq. (62) is exactly the per-user term of the Corollary 1 upper bound, so the resource-allocation objective Psi(a) in Eq. (76) is a derived surrogate for the bound rather than a retrofitted prediction. Experimental claims are compared against external baselines such as Specific SC, FedAvg, vanilla FML, and CL, so the central empirical claims are not constructed from the model's own outputs. The one circularity-adjacent flaw is that the proofs of Theorem 1, Corollary 1, and Lemmas 4-6 are cited to '[33]', which is the same arXiv paper; however, the proofs are included in the manuscript's appendices, so this is a minor self-citation rather than a hidden imported theorem. The reviewer concern that Appendix A only finalizes the bound for tau0=1, and that the SU-selection factor in Eq. (58) degenerates when K_t=K, is a proof-completeness and correctness risk, not a definitional circularity, and therefore does not raise the circularity score further.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central results rest on standard variational IB approximations and on smoothness and heterogeneity assumptions imported from FL convergence theory. The main free choices are the IB weight, the transmission threshold, the resource-allocation weights, and the uninstantiated constants in the SU divergence heuristic. No new physical entities are postulated.

free parameters (5)

zeta (IB rate-distortion weight) = not reported
Appears in every DMIB objective, Eqs. (5), (11), (14), balancing compression against task accuracy; no value or sensitivity analysis is given in Section VI.
phi_th (adaptive transmission threshold) = not reported
Decides how many semantic feature dimensions are kept in Section II-C; the entire overhead-reduction claim depends on it, but no value or schedule is reported.
eta_1, eta_2 (SU divergence constants) = not specified numerically
Used in delta_k(t), Eq. (62), and in Algorithm 2; only lower bounds are given in Eqs. (60)-(61), so the implemented selection rule effectively treats them as hand-set weights.
rho_1, rho_2 (resource allocation weights) = not reported (simulation notes rho_1 = rho_2)
Weight the convergence contribution, energy, and latency terms in problem P, Eq. (77a); the reported plots depend on these weights yet their values are not listed.
FML learning rates alpha, beta and inner rates alpha_B, alpha_S, beta_B, beta_S = alpha = beta = 0.001 in Section VI-D; Adam lr 2e-5 used for training
Hyperparameters controlling inner and outer updates; convergence guarantees require alpha <= 1/H and beta <= 1/H_L, but the corresponding H and H_L are not computed for the actual networks.

assumptions (5)

domain assumption Markov chain Y_k <-> X^m_k <-> Z^m_k <-> Z_k <-> S_k in Eq. (4) screens the task label from raw data through the latent features.
This is the information-theoretic setup of the whole DMIB derivation; if the chain fails, the IB objectives do not characterize the intended task information.
standard math Variational bounds: D_KL(p(z|x)||q(z)) upper-bounds I(X;Z), and cross-entropy or MAE losses approximate H(Y|Z).
These are standard VIB results from Alemi et al. [29] and are applied in Eqs. (9)-(12).
domain assumption The encoder posterior p_theta(z|x) and the marginal q(z) are Gaussian, namely N(mu, Sigma) and N(0, I).
This variational family is chosen for tractability; there is no proof that the true IB-optimal distribution lies in this family.
domain assumption Assumptions 1-4: smoothness, bounded gradients, Lipschitz Hessians, bounded per-sample variance, and bounded heterogeneity for every local loss.
These are imported from non-convex FL analysis and are not verified for the Transformer-based transceivers; Theorem 1 and Corollary 1 depend on them.
ad hoc to paper The feature-importance network psi(sigma^2) can be constrained to non-negative increasing outputs, and sequential thresholding of phi_j(sigma^2) preserves task-relevant information.
Section II-C asserts monotonicity from the MLP parameter range, but it does not show that the pruned dimensions do not contain information needed for the task.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Task-Agnostic Semantic Communications Relying on Information Bottleneck and Federated Meta-Learning." pith.science (2026). https://pith.science/paper/6YPA63E3

@misc{pith2026250421723,
  author       = {Pith},
  title        = {Pith review of: Task-Agnostic Semantic Communications Relying on Information Bottleneck and Federated Meta-Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6YPA63E3}},
  note         = {Machine review of arXiv:2504.21723}
}

read the original abstract

As a paradigm shift towards pervasive intelligence, semantic communication (SemCom) has shown great potentials to improve communication efficiency and provide user-centric services by delivering task-oriented semantic meanings. However, the exponential growth in connected devices, data volumes, and communication demands presents significant challenges for practical SemCom design, particularly in resource-constrained wireless networks. In this work, we first propose a task-agnostic SemCom (TASC) framework that can handle diverse tasks with multiple modalities. Aiming to explore the interplay between communications and intelligent tasks from the information-theoretical perspective, we leverage information bottleneck (IB) theory and propose a distributed multimodal IB (DMIB) principle to learn minimal and sufficient unimodal and multimodal information effectively by discarding redundancy while preserving task-related information. To further reduce the communication overhead, we develop an adaptive semantic feature transmission method under dynamic channel conditions. Then, TASC is trained based on federated meta-learning (FML) for rapid adaptation and generalization in wireless networks. To gain deep insights, we rigorously conduct theoretical analysis and devise resource management to accelerate convergence while minimizing the training latency and energy consumption. Moreover, we develop a joint user selection and resource allocation algorithm to address the non-convex problem with theoretical guarantees. Extensive simulation results validate the effectiveness and superiority of the proposed TASC compared to baselines.

Figures

Figures reproduced from arXiv: 2504.21723 by the authors.

**Figure 2.** The procedure of TASC in one communication round. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗

**Figure 3.** Accuracy performance versus SNR for various tasks under Rayleigh channels. [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Rate distortion curve for classification task with [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]

**Figure 5.** Figure 5: Comparison of convergence, training time, and energy cost under image classification task. [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]

**Figure 6.** Figure 6: Test accuracy versus number of communication rounds. [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 30 canonical work pages

[15]

A unified multi- task semantic communication system for multimodal data,

G. Zhang, Q. Hu, Z. Qin, Y . Cai, G. Yu, and X. Tao, “A unified multi- task semantic communication system for multimodal data,” IEEE Trans. Wireless Commun., vol. 72, no. 7, pp. 4101-4116, Jul. 2024

work page 2024
[33]

Task-Agnostic Semantic Communications Relying on Information Bottleneck and Federated Meta-Learning

H. Wei et al. , “Task-agnostic semantic communications relying on information bottleneck and federated meta-learning,” arxiv: 2504.21723

work page Pith review arXiv
[13]

Task-oriented multi-user semantic communications for VQA,

H. Xie, Z. Qin, and G. Y . Li, “Task-oriented multi-user semantic communications for VQA,” IEEE Wireless Commun. Lett. , vol. 11, no. 3, pp. 553-557, Mar. 2022

work page 2022
[1]

Beyond transmitting bits: Context, semantics, and task-oriented communications,

D. G ¨und¨uz et al. , “Beyond transmitting bits: Context, semantics, and task-oriented communications,” IEEE J. Sel. Areas Commun. , vol. 41, no. 1, pp. 5-41, Jan. 2023

work page 2023
[2]

Toward wisdom-evolutionary and primitive-concise 6G: A new paradigm of semantic communication networks,

P. Zhang et al., “Toward wisdom-evolutionary and primitive-concise 6G: A new paradigm of semantic communication networks,” Engineering, vol. 8, pp. 60-73, Jan. 2022

work page 2022
[3]

When AI meets sustainable 6G,

X. You et al. , “When AI meets sustainable 6G,” Sci. China Inf. Sci. , vol. 68, no. 1, p. 110301, 2025

work page 2025
[4]

Intellicise wireless networks from semantic commu- nications: A survey, research issues, and challenges,

P. Zhang et al. , “Intellicise wireless networks from semantic commu- nications: A survey, research issues, and challenges,” IEEE Commun. Surveys Tuts., 2024, early access, doi: 10.1109/COMST.2024.3443193

work page arXiv 2024
[5]

Task-oriented explainable semantic communications,

S. Ma et al. , “Task-oriented explainable semantic communications,” IEEE Trans. Wireless Commun. , vol. 22, no. 12, pp. 9248-9262, Dec. 2023

work page 2023

Show all 34 references

[6]

Federated learning and meta learning: Approaches, applications, and directions,

X. Liu, Y . Deng, A. Nallanathan, and M. Bennis, “Federated learning and meta learning: Approaches, applications, and directions,” IEEE Commun. Surveys Tuts., vol. 26, no. 1, pp. 571-618, 1st Quart. 2024

2024
[7]

On forecasting-oriented time series transmission: A fed- erated semantic communication system,

B. Zhao et al., “On forecasting-oriented time series transmission: A fed- erated semantic communication system,” IEEE Trans. Mobile Comput. , vol. 23, no. 12, pp. 13728-13744, Dec. 2024

2024
[8]

An efficient federated learning framework for training semantic communication systems,

L. X. Nguyen et al. , “An efficient federated learning framework for training semantic communication systems,” IEEE Trans. Veh. Technol., vol. 73, no. 10, pp. 15872-15877, Oct. 2024

2024
[9]

Distributed image transmission using deep joint source- channel coding,

S. Wang et al., “Distributed image transmission using deep joint source- channel coding,” Proc. IEEE Int. Conf. Acoust., Speech, Signal Process., Singapore, Singapore, May, 2022, pp. 5208-5212

2022
[10]

Semantic communication for the internet of vehicles: A multiuser cooperative approach,

W. Xu et al., “Semantic communication for the internet of vehicles: A multiuser cooperative approach,” IEEE Veh. Technol. Mag., vol. 18, no. 1, pp. 100-109, Mar. 2023

2023
[11]

Semantic MIMO systems for speech-to-text transmission,

Z. Weng, Z. Qin, H. Xie, X. Tao, and K. B. Letaief, “Semantic MIMO systems for speech-to-text transmission,” IEEE Trans. Wireless Commun., vol. 23, no. 12, pp. 18697-18710, Dec. 2024

2024
[12]

Deep joint source- channel coding for multi-task network,

M. Wang, Z. Zhang, J. Li, M. Ma, and X. Fan, “Deep joint source- channel coding for multi-task network,” IEEE Signal Process. Lett., vol. 28, pp. 1973-1977, Sep. 2021

1973
[14]

Task-oriented multi-user semantic communications,

H. Xie, Z. Qin, X. Tao, and K. B. Letaief, “Task-oriented multi-user semantic communications,” IEEE J. Sel. Areas Commun. , vol. 40, no. 9, pp. 2584-2597, Sept. 2022

2022
[16]

Transitioning from federated learning to quantum federated learning in internet of things: A comprehensive survey,

C. Qiao et al. , “Transitioning from federated learning to quantum federated learning in internet of things: A comprehensive survey,” IEEE Commun. Surveys Tuts., vol. 27, no. 1, pp. 509-545, Feb. 2025

2025
[17]

A joint learning and communications framework for federated learning over wireless networks,

M. Chen et al. , “A joint learning and communications framework for federated learning over wireless networks,” IEEE Trans. Wireless Commun., vol. 20, no. 1, pp. 269-283, Jan. 2021

2021
[18]

Federated learning based audio semantic communication over wireless networks,

H. Tong, Z. Yang, S. Wang, Y . Hu, W. Saad, and C. Yin, “Federated learning based audio semantic communication over wireless networks,” in Proc. IEEE Global Comm. Conf. , Madrid, Spain, Dec. 2021, pp. 1-6

2021
[19]

Federated semantic learning driven by information bottleneck for task-oriented communications,

H. Wei et al. , “Federated semantic learning driven by information bottleneck for task-oriented communications,” IEEE Commun. Lett., vol. 27, no. 10, pp. 2652-2656, Oct. 2023

2023
[20]

Federated contrastive learning for personalized semantic communication,

Y . Wang, W. Ni, W. Yi, X. Xu, P. Zhang, and A. Nallanathan, “Federated contrastive learning for personalized semantic communication,” IEEE Commun. Lett., vol. 28, no. 8, pp. 1875-1879, Aug. 2024

2024
[21]

Federated knowledge distillation enabled image semantic communication,

X. Xu et al., “Federated knowledge distillation enabled image semantic communication,” in Proc. IEEE Global Comm. Conf., Cape Town, South Africa, Dec. 2024, pp. 3152-3157

2024
[22]

Federated meta-learning with fast convergence and efficient communication,

F. Chen et al. , “Federated meta-learning with fast convergence and efficient communication,” arxiv:1802.07876

arXiv
[23]

Model-agnostic meta-learning for fast adaptation of deep networks,

C. Finn, P. Abbeel, and S. Levine, “Model-agnostic meta-learning for fast adaptation of deep networks,” in Proc. Int. Conf. Learn. Represent., 2017, pp. 1126-1135

2017
[24]

Personalized federated learning with theoretical guarantees: A model-agnostic meta-learning approach,

A. Fallah, A. Mokhtari, and A. Ozdaglar, “Personalized federated learning with theoretical guarantees: A model-agnostic meta-learning approach,” in Proc. Adv. Neural Inf. Process. Syst. , 2020, pp. 1-12

2020
[25]

The information bottleneck problem and its applications in machine learning,

Z. Goldfeld and Y . Polyanskiy, “The information bottleneck problem and its applications in machine learning,” IEEE J. Sel. Areas Inf. Theory, vol. 1, no. 1, pp. 19-38, May 2020

2020
[26]

Energy efficient federated learning over heterogeneous mobile devices via joint design of weight quantization and wireless Transmission,

R. Chen, L. Li, K. Xue, C. Zhang, M. Pan, and Y . Fang, “Energy efficient federated learning over heterogeneous mobile devices via joint design of weight quantization and wireless Transmission,” IEEE Trans. Mobile Comput., vol. 22, no. 12, pp. 7451-7465, Dec. 2023

2023
[27]

OFDM-guided deep joint source channel coding for wireless multipath fading channels,

M. Yang, C. Bian, and H. -S. Kim, “OFDM-guided deep joint source channel coding for wireless multipath fading channels,” IEEE Trans. Cogn. Commun. Netw., vol. 8, no. 2, pp. 584-599, Jun. 2022

2022
[28]

Task-oriented communication for mul- tidevice cooperative edge inference,

J. Shao, Y . Mao, and J. Zhang, “Task-oriented communication for mul- tidevice cooperative edge inference,” IEEE Trans. Wireless Commun. , vol. 22, no. 1, pp. 73-87, Jan. 2023

2023
[29]

Deep variational information bottleneck,

A. A. Alemi, I. Fischer, J. V . Dillon, and K. Murphy, “Deep variational information bottleneck,” in Proc. Int. Conf. Learn. Represent. , Toulon, France, Apr. 2017, pp. 1-19

2017
[30]

Semantic communication with memory,

H. Xie, Z. Qin, and G. Y . Li, “Semantic communication with memory,” IEEE J. Sel. Areas Commun. , vol. 41, no. 8, pp. 2658-2669, Aug. 2023

2023
[31]

Attention is all you need,

A. Vaswani et al., “Attention is all you need,” in Proc. Adv. Neural Inf. Process. Syst., Long Beach, CA, USA, Dec. 2017, pp. 6000-6010

2017
[32]

UniT: Multimodal multitask learning with a unified transformer,

R. Hu and A. Singh, “UniT: Multimodal multitask learning with a unified transformer,” in Proc. IEEE/CVF Int. Conf. Comput. Vis., Oct. 2021, pp. 1419-1429

2021
[34]

1 K KX k=1 θk t,τ−θk t,τ 2 # ≤ 35β2H2 Lτ0(τ0− 1) 2σ2 L +µ2 L . (102) E∥C3∥2≤ 1 Kt E

E. W. Weisstein. (2011). Hungarian Maximum Matching Algorithm. [Online]. Available: https://mathworld.wolfram.com/ 14 APPENDIX A PROOF OF THEOREM 1 For brevity, let θk t,τ =θk(t,τ ). We denote that θt,τ = 1/KPK k=1θk t,τ , and ¯θt,τ = 1/KtPKt k=1θk t,τ . Based on Lemma 1, L(·)...

2011

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [15]

A unified multi- task semantic communication system for multimodal data,

G. Zhang, Q. Hu, Z. Qin, Y . Cai, G. Yu, and X. Tao, “A unified multi- task semantic communication system for multimodal data,” IEEE Trans. Wireless Commun., vol. 72, no. 7, pp. 4101-4116, Jul. 2024

work page 2024

[2] [33]

Task-Agnostic Semantic Communications Relying on Information Bottleneck and Federated Meta-Learning

H. Wei et al. , “Task-agnostic semantic communications relying on information bottleneck and federated meta-learning,” arxiv: 2504.21723

work page Pith review arXiv

[3] [13]

Task-oriented multi-user semantic communications for VQA,

H. Xie, Z. Qin, and G. Y . Li, “Task-oriented multi-user semantic communications for VQA,” IEEE Wireless Commun. Lett. , vol. 11, no. 3, pp. 553-557, Mar. 2022

work page 2022

[4] [1]

Beyond transmitting bits: Context, semantics, and task-oriented communications,

D. G ¨und¨uz et al. , “Beyond transmitting bits: Context, semantics, and task-oriented communications,” IEEE J. Sel. Areas Commun. , vol. 41, no. 1, pp. 5-41, Jan. 2023

work page 2023

[5] [2]

Toward wisdom-evolutionary and primitive-concise 6G: A new paradigm of semantic communication networks,

P. Zhang et al., “Toward wisdom-evolutionary and primitive-concise 6G: A new paradigm of semantic communication networks,” Engineering, vol. 8, pp. 60-73, Jan. 2022

work page 2022

[6] [3]

When AI meets sustainable 6G,

X. You et al. , “When AI meets sustainable 6G,” Sci. China Inf. Sci. , vol. 68, no. 1, p. 110301, 2025

work page 2025

[7] [4]

Intellicise wireless networks from semantic commu- nications: A survey, research issues, and challenges,

P. Zhang et al. , “Intellicise wireless networks from semantic commu- nications: A survey, research issues, and challenges,” IEEE Commun. Surveys Tuts., 2024, early access, doi: 10.1109/COMST.2024.3443193

work page arXiv 2024

[8] [5]

Task-oriented explainable semantic communications,

S. Ma et al. , “Task-oriented explainable semantic communications,” IEEE Trans. Wireless Commun. , vol. 22, no. 12, pp. 9248-9262, Dec. 2023

work page 2023