Reinforcement Learning-Based Heterogeneous Multi-Task Optimization in Semantic Broadcast Communications

Honggang Zhang; Rongpeng Li; Zhifeng Zhao; Zhilin Lu

REVIEW 3 major objections 5 minor 46 references

Reinforcement Learning-Based Heterogeneous Multi-Task Optimization in Semantic Broadcast Communications

T0 review · 3 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 SemanticBC-TriRL claims that a tri-level alternating framework—supervised decoder updates, a PPO-driven shared encoder, and an adaptive multi-gradient task-weighting module—resolves conflicts between image reconstruction and…

desk verdict Useful engineering recipe for heterogeneous semantic broadcast, but the convergence theorem is overclaimed and should be revised before publication. read the letter →

arxiv 2504.19806 v1 pith:RSKPFU5I submitted 2025-04-28 eess.SP

Zhilin Lu , Rongpeng Li , Zhifeng Zhao , Honggang Zhang This is my paper · ORCID

classification eess.SP

keywords semanticbroadcastcommunicationsmulti-objectiveoptimizationtri-levelself-criticalreinforcementlearningproximalpolicymulti-gradientaggregationjointsource-channelcodingimagereconstruction

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 proposes SemanticBC-TriRL, a training scheme for semantic broadcast systems in which one encoder at the transmitter serves multiple receivers with different tasks, specifically image reconstruction and image classification. The central claim is that a tri-level alternating optimization—task-specific decoders trained by supervised learning, a shared encoder trained by proximal policy optimization (PPO), and an adaptive task-weighting module built on multi-gradient aggregation—resolves conflicts among heterogeneous receiver objectives. If the claim holds, a single transmitted bitstream can simultaneously support several downstream tasks with higher SSIM and classification accuracy than end-to-end Deep JSCC, conventional BPG+LDPC coding, or an equal-weight variant, particularly at low signal-to-noise ratios. The paper also provides a non-asymptotic convergence guarantee for the alternating procedure.

What carries the argument

The load-bearing mechanism is the tri-level alternating learner with a multi-gradient aggregation weight module. The first level relies on the singleton condition: for a fixed encoder parameter $\theta_l$, each decoder has a unique optimum $\phi^*_{n,l}$ after $\kappa$ local updates. The second level reformulates the encoder as a reinforcement-learning agent: its policy is a Gaussian distribution over channel symbols, and PPO (a policy-gradient update that clips how far the policy moves in one step) maximizes the task-weighted reward $r=\sum_n w_n \Theta_n(m,\hat{m}_n)$ with an auxiliary supervised loss. The third level computes the gradient of the encoder loss with respect to the weights and the encoder, subject to the approximated feasibility constraint, and updates the weights so that tasks whose gradients point against the common descent direction receive lower weight. This closed-form quadratic-program update is what lets the scheme rebalance conflicting objectives during training.

What would settle it

Run the tri-level training on a batch and record the stationarity gap $\psi(v_l)=\|-\nabla F(v_l)-\lambda_l \nabla g(v_l)\|^2$ and the feasibility gap $g(v_l)$; if the strong-convexity assumption holds, these should decay roughly like $O(\Gamma(H/2)+\sqrt{\eta}+1/\sqrt{L\eta})$, and with $\eta=O(L^{-1/2})$ the gap should fall as $O(\Gamma(H/2)+L^{-1/4})$. Any observed divergence or much slower convergence would indicate that the theory is not describing the implemented system.

Watch

Extended reading notes

Core claim

The paper's discovery is that the conflict among heterogeneous receiver tasks in semantic broadcast can be converted into a constrained tri-level optimization problem with a bottom-up alternating solution. At the first level, each receiver's decoder performs $\kappa$ local gradient steps toward its own task loss, giving an approximate decoder optimum $\phi^*_{n,l}$ for the current encoder parameter $\theta_l$. At the second level, the encoder is updated by PPO, treating the encoder as an agent whose Gaussian sampling policy produces channel symbols and whose reward is the task-weighted receiver performance; the PPO loss is augmented with a decoder-side supervised loss to stabilize the update. At the third level, a weight-assignment module finds a descent direction for the joint variable $v_l=(\{w_{n,l}\},\theta_l)$ by minimizing the first-order Taylor term of the encoder loss subject to a constraint on an approximated feasibility function, solved with a Lagrange multiplier. The paper proves in Theorem 1 that, under strong-convexity and Lipschitz-gradient assumptions, the procedure reaches a weak stationary point with error $O(\Gamma(H/2)+\sqrt{\eta}+1/\sqrt{L\eta})$, and reports that this scheme outperforms the baselines on MNIST and CIFAR-10, with the largest gains at low SNR.

Load-bearing premise

The convergence theorem requires the encoder loss to be strongly convex in the encoder parameters with Lipschitz-continuous gradients, and each decoder to reach a unique optimum after its local updates; these conditions are not guaranteed by the deep-network PPO objective with binary quantization and channel noise, so the stated rate may not apply to the system actually simulated.

Editorial extensions

If this is right

A single transmitter can serve receivers with different tasks from one encoded bitstream, avoiding per-task dedicated transmissions and reducing the transmitter's computational load.
At low SNR, the adaptive weighting should yield graceful degradation for both reconstruction and classification, in contrast to the abrupt cliff effect of conventional BPG+LDPC transmission.
Adding a new receiver task requires training only its decoder and letting the weight module rebalance the encoder loss, which is the scalability claim validated with up to six receivers.
For fixed inner steps $H$, increasing the outer iteration count $L$ and choosing $\eta=O(L^{-1/2})$ drives the tri-level procedure to a weak stationary point at rate $O(\Gamma(H/2)+L^{-1/4})$.
The performance difference from the equal-weight variant supports the claim that learned task weights, not just the larger model, drive part of the improvement.

Reading between the lines

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

Inference, not stated in the paper: the multi-gradient weighting module could transfer to other multi-objective joint source-channel coding settings, for example detection plus reconstruction, whenever each objective supplies a differentiable reward, since the quadratic-program update does not depend on the specific task losses.
Inference, not stated in the paper: because the encoder is frozen during decoder updates, the scheme could be deployed incrementally, adding a new receiver's decoder and then fine-tuning the encoder with adjusted weights while old receivers keep working.
Inference, not stated in the paper: a natural stress test is to initialize the two task gradients in direct opposition and check whether the learned weights track the Pareto front computed by explicit multi-objective search; the paper reports weight trajectories but does not compare them to the true Pareto set.
Inference, not stated in the paper: if the strong-convexity assumption is the bottleneck, replacing it with a Polyak-Łojasiewicz condition or a smoothed relaxation would likely preserve a similar non-asymptotic rate and make the proof applicable to the actual deep-network objective.

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

Quick take: this is a useful engineering recipe, not a fundamental advance, and the convergence result is the weak link. The tri-level stack—supervised decoder updates, PPO-trained shared encoder, multi-gradient task weighting—is new as a combination for heterogeneous semantic broadcast. I don't see that exact assembly in the cited literature. The comparison work is also decent: Deep JSCC, BPG+LDPC, and an equal-weight ablation, under AWGN and Rayleigh, on MNIST and CIFAR-10. The adaptive weighting clearly does something: the curves are consistent, and the equal-weight ablation is the right control.

The main problem is exactly where the stress-test note points. Theorem 1 imports Lemmas 2 and 3 from [43]/[23], and those lemmas require c-strong convexity of the lower-level objective and a singleton decoder optimum. The implemented encoder loss in (14) is a clipped PPO surrogate over a stochastic Gaussian policy, with binary quantization and channel noise. That objective cannot be c-strongly convex. The assertion in Assumption 1 that bounded μ(t) and A(t) imply Lipschitz gradients is a non sequitur; bounded first moments do not give strong convexity, and the rest of the paragraph merely claims Hessian boundedness, which also does not follow. The decoder singleton condition is equally unsatisfied after κ=100 local updates on nonconvex nets. So the non-asymptotic rate in (19) is a correct conditional statement about an idealized problem, but it is not a convergence proof for the system that was simulated. That claim should be reframed or removed.

Secondary soft spots: no error bars, no code, and several implementation details about the quantized PPO pipeline—especially how gradients pass through QTX(·)—are left unspecified. The authors lean on their own earlier works [12], [40] for foundational pieces, which is legitimate, but the reliance is heavy enough that reproducibility suffers. Those are fixable.

Empirically the paper is plausible; I don't see the components being fraudulent or the curves fake. It is a reasonable systems contribution: it tells people how to combine alternating training with adaptive task weights and gives evidence that it helps. A serious referee should engage with it, but should send back the theory for revision. If the authors either condition the theorem on the actual objective or honestly reposition it as heuristic motivation, the paper becomes acceptable.

Referee Report

3 major / 5 minor

Summary. This paper proposes SemanticBC-TriRL, a tri-level alternating optimization framework for semantic broadcast communications with one shared transmitter-side encoder and multiple receiver-side task-specific decoders. The three levels are: (i) supervised local decoder updates, (ii) PPO-based shared encoder updates with a stochastic Gaussian policy and binary quantization, and (iii) a multi-gradient aggregation weight module that adaptively assigns task weights. The authors formulate the system as a constrained tri-level optimization problem, propose Algorithm 1, and state Theorem 1 with a non-asymptotic convergence rate of O(Γ(H/2) + √η + √(1/(Lη))). Simulations on MNIST and CIFAR-10 over AWGN, Rayleigh, and Rician channels compare against Deep JSCC, BPG+LDPC, and an equal-weight variant, reporting higher SSIM and classification accuracy at low SNR, together with scalability up to six receivers.

Significance. If the theoretical and empirical claims hold, the framework is a practically useful extension of semantic broadcast to heterogeneous receivers, and the adaptive weight module addresses a real gradient-conflict problem. The paper's strengths include the clear problem decomposition, the comparison with an equal-weight counterpart that isolates the effect of the weight assignment, and the breadth of experiments across datasets, channel models, CBR values, and numbers of receivers. The empirical curves consistently point in the claimed direction, particularly the low-SNR robustness that is relevant for semantic communications. However, the central theoretical contribution, Theorem 1, is not supported for the implemented system because Assumption 1 is implausible for the clipped PPO objective, and the proof imports key lemmas from strongly-convex bi-level literature without verifying their premises. The empirical study also lacks statistical error bars and code release, which limits confidence in the magnitude of the reported gains.

major comments (3)

[IV-D] Assumption 1 and Theorem 1 do not cover the system actually simulated in Section V. The encoder loss L_TX in Eq. (14) is a weighted combination of the clipped PPO surrogate (12), the value loss (13), and decoder losses, evaluated through the stochastic Gaussian policy (10), the binary quantizer Q_TX in (2), and channel noise. This objective is not c-strongly convex in theta_l and is not differentiable unless a straight-through estimator is silently introduced; the paper neither states nor analyzes such an estimator. Since Lemmas 2 and 3 are imported from the strongly-convex bi-level analyses of [43] and [23], their premises do not hold for the clipped, quantized, stochastic-policy objective, so Eq. (19) cannot be asserted for SemanticBC-TriRL as trained. Please either prove the required properties for the actual objective, replace the theorem with a convergence statement for an idealized smooth unclipped surrogate and clearly delimit its scope, or explicitly label the convergence analysis as heuristic.
[III-B] The decoder singleton condition in (8a) is not satisfied by the implemented training procedure. The manuscript assumes a unique optimal decoder phi*_{n,l} after sufficient local updates, but the decoder networks in Tables IV and V are non-convex, and Table III sets kappa=100, which does not guarantee convergence to a global optimum. Because the second-level gradients in (17b) and (18) use phi*_{n,l} while Algorithm 1 uses phi^kappa_{n,l}, the first-level exactness in (8a) is approximate at best. The paper should justify or relax this condition, for example by stating an inexact lower-level bound in the spirit of [28].
[IV-C] The gradient formulas for the weight module rely on an undocumented simplification after Eq. (16). The sentence 'it is assumed that the clipping of a single sample does not impact the overall gradient computation' asserts away the non-differentiability of the min and clip operations in Eq. (12). Without specifying a surrogate gradient, the expressions for nabla_theta F(v_l) and nabla_w F(v_l) in (17a)-(17b) are not derivatives of the actual loss (14). This directly affects the descent direction d_l in Eq. (15) and the weight updates in Algorithm 1. Please state the gradient estimator used and analyze its bias, or reformulate F(v_l) to use a smooth surrogate.

minor comments (5)

[Table II] The caption reads 'Notions used in this paper' but should read 'Notations used in this paper'.
[Eq. (13)] Equation (13) has a formatting error: V_chi(s(t)) is missing a closing parenthesis in the manuscript text, which should be corrected.
[V-A] In the comparison schemes list, the bullet 'JSCC: JSCC share the same encoder...' contains a subject-verb agreement error, and the word 'comparision' is misspelled as 'comparision'.
[V-B] The paper would benefit from releasing code and the full set of training hyperparameters, and from reporting error bars or multiple seed statistics for the performance curves in Figs. 5, 8, and 13, since the reported differences are sometimes small (e.g., the SSIM gap between SemanticBC-TriRL and SemanticBC-TriRL-EW).
[Eq. (17a)] In Eq. (17a), the reward term uses inconsistent indices: the subscript in Theta_{phi^kappa_{1,l}} appears to be an artifact, and the notation for the N receivers should be harmonized with (11) and (17b).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the convergence proof imports external bi-level optimization lemmas under explicit assumptions, and the empirical gains are benchmarked against independent baselines.

full rationale

The derivation chain is self-contained in the relevant sense. The tri-level problem (8) is a reformulation of (7); the decoder level uses supervised losses (5)-(6), the encoder level uses the PPO objective (14), and the weight module solves the quadratic program (15)-(18). The convergence theorem in Section IV-D is proved from Lemma 2 (credited to Theorem 1 of [43]) and Lemma 3 (credited to Lemma 10 of [43] and Lemma B.2 of [23]) under Assumptions 1 and 2. These are external results imported from non-overlapping groups, not from the authors' own prior work, and the assumptions are stated premises rather than conclusions of the theorem. The empirical claims are compared against Deep JSCC, BPG+LDPC, and the equal-weight variant SemanticBC-TriRL-EW, so the reported gains are not constructed from the training objective. Self-citations to [12] and [40] are contextual: they motivate alternating training, binary quantization, and the SSIM reward, but they do not carry the convergence claim, which rests on the external lemmas. The main rigor concern is that Assumption 1's c-strong convexity is implausible for the clipped PPO objective in Eq. (14), and the kappa=100 decoder updates do not ensure the singleton optimum; however, an unverified premise is a correctness risk, not a circularity, because the theorem explicitly assumes the property instead of having the conclusion define the premise. No step reduces a prediction to a fitted input or to a self-citation chain.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The load-bearing theoretical result is purchased with strong convexity and Lipschitz assumptions that a deep PPO objective does not satisfy; the empirical comparison adds several hand-chosen hyperparameters with no sensitivity analysis. No new physical entities are introduced.

free parameters (7)

PPO clipping ratio epsilon
Used in clip(mu_t, 1-epsilon, 1+epsilon) in Eq. (12); no value or sensitivity study is reported.
Barrier coefficient beta
Sets rho_l = beta ||nabla g~(v_l)||^2 in Eq. (16); no default value is reported.
Number of Swin Transformer stages = 2
Section III-A states two stages are set according to empirical results; this is a hand-chosen architecture capacity.
Local decoder iterations kappa = 100
Table III; the singleton-condition assumption treats kappa local steps as sufficient for decoder convergence.
Inner encoder steps H = 5
Table III; H determines the approximation theta~_l^H and appears in the convergence rate Gamma(H/2).
Learning rates gamma, eta, zeta = eta=1e-3; gamma and zeta not reported
Algorithm 1 uses three learning rates but Table III only lists eta.
Number of training epochs Ee = 71
Table III; the stopping point is chosen by hand rather than by a convergence criterion.

assumptions (6)

ad hoc to paper The encoder loss f(v_l)=L_TX(v_l;{phi*_n,l}) is c-strongly convex with respect to theta_l and has Lf-Lipschitz gradients.
Assumption 1 in Section IV-D. A clipped PPO surrogate over a stochastic Gaussian policy is not globally strongly convex; this assumption is needed for the convergence rate in Theorem 1 but is not checked.
standard math The third-level objective F(v_l) has LF-Lipschitz gradients and bounded norm.
Assumptions 1(2) and 2 in Section IV-D; standard regularity conditions in bi-level optimization, but no verification is offered for the learned weight and encoder parameters.
domain assumption Singleton condition: for fixed theta_l and w_n,l there exists a unique optimal decoder phi*_n,l, reached after kappa local updates.
Section III-B and Algorithm 1 lines 4-8. Nonconvex decoders have multiple optima and kappa=100 is finite, so uniqueness is unrealistic; this underpins the formulation of problem (8).
domain assumption The channel is non-differentiable and no training gradient is transmitted from TX to RXs.
Section III-A. This justifies decoupled decoder training, but means encoder supervision is limited to scalar task rewards and decoder losses.
ad hoc to paper PPO clipping of a single sample does not affect the overall gradient computation for the weighting module.
Section IV-C after Eq. (16). The closed-form update direction ignores per-sample clipping; no justification or empirical check is provided.
domain assumption One-step MDP with discount factor gamma=1.
Section IV-B. This simplifies the advantage function to r-V(s); it is reasonable for the broadcast setting but remains a modeling assumption.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Reinforcement Learning-Based Heterogeneous Multi-Task Optimization in Semantic Broadcast Communications." pith.science (2026). https://pith.science/paper/RSKPFU5I

@misc{pith2026250419806,
  author       = {Pith},
  title        = {Pith review of: Reinforcement Learning-Based Heterogeneous Multi-Task Optimization in Semantic Broadcast Communications},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RSKPFU5I}},
  note         = {Machine review of arXiv:2504.19806}
}

read the original abstract

Semantic broadcast communications (Semantic BC) for image transmission have achieved significant performance gains for single-task scenarios. Nevertheless, extending these methods to multi-task scenarios remains challenging, as different tasks typically require distinct objective functions, leading to potential conflicts within the shared encoder. In this paper, we propose a tri-level reinforcement learning (RL)-based multi-task Semantic BC framework, termed SemanticBC-TriRL, which effectively resolves such conflicts and enables the simultaneous support of multiple downstream tasks at the receiver side, including image classification and content reconstruction tasks. Specifically, the proposed framework employs a bottom-up tri-level alternating learning strategy, formulated as a constrained multi-objective optimization problem. At the first level, task-specific decoders are locally optimized using supervised learning. At the second level, the shared encoder is updated via proximal policy optimization (PPO), guided by task-oriented rewards. At the third level, a multi-gradient aggregation-based task weighting module adaptively adjusts task priorities and steers the encoder optimization. Through this hierarchical learning process, the encoder and decoders are alternately trained, and the three levels are cohesively integrated via constrained learning objective. Besides, the convergence of SemanticBC-TriRL is also theoretically established. Extensive simulation results demonstrate the superior performance of the proposed framework across diverse channel conditions, particularly in low SNR regimes, and confirm its scalability with increasing numbers of receivers.

Figures

Figures reproduced from arXiv: 2504.19806 by the authors.

**Figure 2.** The parameters relationship between the encoder and de [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Optimization process under the alternate learning mechanism. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (10 more)

**Figure 4.** Figure 4: The encoder learning process via weighted PPO algorithm. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Comparison of semantic BC schemes in terms of SSIM and classification accuracy for MNIST image transmission under AWGN [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]

**Figure 6.** Figure 6: Examples of visual comparison for MNIST image transmission through AWGN, Rayleigh, and Rician fading channels under [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]

**Figure 7.** Figure 7: Examples of visual comparison for MNIST image transmission through AWGN channel under [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

**Figure 8.** Figure 8: Comparison of semantic BC schemes in terms of SSIM, PSNR and classification accuracy for CIFAR-10 image transmission under [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]

**Figure 9.** Figure 9: Examples of visual comparison for CIFAR-10 image transmission through AWGN channel under [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]

**Figure 10.** Figure 10: Convergence of encoder of SemanticBC-TriRL for MNIST image transmission with two RXs under AWGN, Rayleigh fading channels, wherein RX1 and RX2 execute image reconstruction and image classification tasks respectively. the image reconstruction task. But, SemanticBC-TriR…

**Figure 11.** Figure 11: The training loss at decoder sides of SemanticBC-TriRL for MNIST image transmission through AWGN and Rayleigh fading channels. 0 10 20 30 40 50 60 70 Epoch 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 W eig h t w n Reconstruction task (CBR=0.02) Classification task (C…

**Figure 12.** Figure 12: Training weights evolution of our proposed [PITH_FULL_IMAGE:figures/full_fig_p014_12.png]

**Figure 13.** Figure 13: Impact of CBR on task performance of SemanticBC-TriRL for MNIST image transmission over AWGN channel optimization strategy. On the other hand, as the number of RXs increases from 2 to 6 in [PITH_FULL_IMAGE:figures/full_fig_p015_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 44 canonical work pages

[43]

Bome! bilevel optimization made easy: A simple first-order approach,

B. Liu, M. Ye, S. Wright, P. Stone, and Q. Liu, “Bome! bilevel optimization made easy: A simple first-order approach,” in Proc. Adv. Neural Inf. Proces. Syst. (NIPS) , New Orleans, LA, US, 2022, pp. 17 248–17 262

work page 2022
[23]

A first-order multi-gradient algorithm for multi-objective bi-level optimization,

F. Ye, B. Lin, X. Cao, Y . Zhang, and I. W. Tsang, “A first-order multi-gradient algorithm for multi-objective bi-level optimization,” Jul

work page
[12]

Self- critical alternate learning based semantic broadcast communication,

Z. Lu, R. Li, M. Lei, C. Wang, Z. Zhao, and H. Zhang, “Self- critical alternate learning based semantic broadcast communication,” IEEE Trans. Commun. early access, Oct. 2024

work page 2024
[40]

Alternate learning- based snr-adaptive sparse semantic visual transmission,

S. Tong, X. Yu, R. Li, K. Lu, Z. Zhao, and H. Zhang, “Alternate learning- based snr-adaptive sparse semantic visual transmission,” IEEE Trans. Wirel. Commun. early access, Dec. 2024

work page 2024
[28]

A generic first-order algorithmic framework for bi-level programming beyond lower-level singleton,

R. Liu, P. Mu, X. Yuan, S. Zeng, and J. Zhang, “A generic first-order algorithmic framework for bi-level programming beyond lower-level singleton,” in Proc. Int. Conf. Mach. Learn. (ICML) . Virtual, Online: PMLR, Jul. 2020, pp. 6305–6315

work page 2020
[1]

Task-oriented image semantic communication based on rate-distortion theory,

F. Liu, W. Tong, Y . Yang, Z. Sun, and C. Guo, “Task-oriented image semantic communication based on rate-distortion theory,” Dec. 2022. [Online]. Available: https://arxiv.org/abs/2201.10929

arXiv 2022
[2]

Semantics-empowered communication: A Tutorial-cum-Survey,

Z. Lu, R. Li, K. Lu, X. Chen, E. Hossain, Z. Zhao, and H. Zhang, “Semantics-empowered communication: A Tutorial-cum-Survey,” IEEE Communications Surveys & Tutorials , vol. 26, no. 1, pp. 41–79, Nov. 2023

work page 2023
[3]

Rethinking modern communication from semantic coding to semantic communication,

K. Lu, Q. Zhou, R. Li, Z. Zhao, X. Chen, J. Wu, and H. Zhang, “Rethinking modern communication from semantic coding to semantic communication,” IEEE Wireless Commun., vol. 30, no. 1, pp. 158–164, Feb. 2023. 16

work page 2023

Show all 46 references

[4]

Adaptive bit rate control in semantic communication with incremental knowledge-based HARQ,

Q. Zhou, R. Li, Z. Zhao, Y . Xiao, and H. Zhang, “Adaptive bit rate control in semantic communication with incremental knowledge-based HARQ,” IEEE Open J. Commun. Soc., vol. 3, pp. 1076–1089, Jul. 2022

2022
[5]

Topology data analysis-based error detection for semantic image transmission with incremental knowledge- based HARQ,

F. Ni, R. Li, Z. Zhao, and H. Zhang, “Topology data analysis-based error detection for semantic image transmission with incremental knowledge- based HARQ,” China Communications, vol. 22, no. 1, pp. 235–255, Jan. 2025

2025
[6]

Deep learning enabled semantic communication systems,

H. Xie, Z. Qin, G. Y . Li, and B.-H. Juang, “Deep learning enabled semantic communication systems,” IEEE Transactions on Signal Pro- cessing, vol. 69, pp. 2663–2675, Apr. 2021

2021
[7]

Wireless image transmission using deep source channel coding with attention modules,

J. Xu, B. Ai, W. Chen, A. Yang, P. Sun, and M. Rodrigues, “Wireless image transmission using deep source channel coding with attention modules,” IEEE Trans. Circ. Syst. Vid. , vol. 32, no. 4, pp. 2315–2328, May 2022

2022
[8]

Task-oriented multi- user semantic communications,

H. Xie, Z. Qin, X. Tao, and K. B. Letaief, “Task-oriented multi- user semantic communications,” IEEE Journal on Selected Areas in Communications, vol. 40, no. 9, pp. 2584–2597, Jul. 2022

2022
[9]

Semantic-preserving image compression,

N. Patwa, N. Ahuja, S. Somayazulu, O. Tickoo, S. Varadarajan, and S. Koolagudi, “Semantic-preserving image compression,” in Proc. IEEE Int. Conf. Image Process. (ICIP) . Abu Dhabi, United Arab Emirates: IEEE, Oct. 2020, pp. 1281–1285

2020
[10]

A multi-task semantic communi- cation system for natural language processing,

Y . Sheng, F. Li, L. Liang, and S. Jin, “A multi-task semantic communi- cation system for natural language processing,” in Proc. Veh. Technol. Conf. (VTC). London, United Kingdom: IEEE, Sep. 2022, pp. 1–5

2022
[11]

Features disentangled semantic broadcast communication networks,

S. Ma, Z. Zhang, Y . Wu, H. Li, G. Shi, D. Gao, Y . Shi, S. Li, and N. Al-Dhahir, “Features disentangled semantic broadcast communication networks,” IEEE Trans. Wirel. Commun., vol. 23, no. 6, pp. 6580 – 6594, Nov. 2023

2023
[13]

One-to-many semantic communication systems: Design, implementation, performance evaluation,

H. Hu, X. Zhu, F. Zhou, W. Wu, R. Q. Hu, and H. Zhu, “One-to-many semantic communication systems: Design, implementation, performance evaluation,” IEEE Commu. Lett. , vol. 26, no. 12, pp. 2959–2963, Sep. 2022

2022
[14]

Multi-task semantic communication with graph attention-based feature correlation extraction,

X. Yu, T. Lv, W. Li, W. Ni, D. Niyato, and E. Hossain, “Multi-task semantic communication with graph attention-based feature correlation extraction,” IEEE Trans. Mobile Comput. early access, Jan. 2025

2025
[15]

Semantic communications for image recovery and classification via deep joint source and channel coding,

Z. Lyu, G. Zhu, J. Xu, B. Ai, and S. Cui, “Semantic communications for image recovery and classification via deep joint source and channel coding,” IEEE Trans. Wirel. Commun. , vol. 23, no. 8, pp. 8388–8404, Jan. 2024

2024
[16]

Swin transformer-based dynamic semantic communication for multi-user with different computing capacity,

L. X. Nguyen, Y . L. Tun, Y . K. Tun, M. N. Nguyen, C. Zhang, Z. Han, and C. S. Hong, “Swin transformer-based dynamic semantic communication for multi-user with different computing capacity,” IEEE Transactions on Vehicular Technology , vol. 73, no. 6, pp. 8957–8972, Feb. 2024

2024
[17]

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

W. Xu, Y . Zhang, F. Wang, Z. Qin, C. Liu, and P. Zhang, “Semantic communication for the internet of vehicles: A multiuser cooperative approach,” IEEE Veh. Technol. Mag., vol. 18, no. 1, pp. 100–109, Jan. 2023

2023
[18]

Collaborative semantic communication for edge inference,

W. F. Lo, N. Mital, H. Wu, and D. Gündüz, “Collaborative semantic communication for edge inference,” IEEE Wirel. Commun. Le., vol. 12, no. 7, pp. 1125–1129, Mar. 2023

2023
[19]

A unified multi-task semantic communication system with domain adaptation,

G. Zhang, Q. Hu, Z. Qin, Y . Cai, and G. Yu, “A unified multi-task semantic communication system with domain adaptation,” in Proc. IEEE Glob. Commun. Conf. (GLOBECOM) . Rio de Janeiro, Brazil: IEEE, Dec. 2022, pp. 3971–3976

2022
[20]

Multi-user semantic fusion for semantic communications over degraded broadcast channels,

T. Wu, Z. Chen, M. Tao, B. Xia, and W. Zhang, “Multi-user semantic fusion for semantic communications over degraded broadcast channels,” Jun. 2024. [Online]. Available: arXivpreprintarXiv:2406.10556

2024 arXiv
[21]

Multi-task learning as multi-objective opti- mization,

O. Sener and V . Koltun, “Multi-task learning as multi-objective opti- mization,” in Proc. Adv. Neural Inf. Proces. Syst. (NIPS) . Montreal, QC, Canada: NIPS foundation, Dec. 2018, pp. 527–538

2018
[22]

A multi-objective/multi-task learning framework induced by pareto stationarity,

M. Momma, C. Dong, and J. Liu, “A multi-objective/multi-task learning framework induced by pareto stationarity,” in Proc. Int. Conf. Machin. Learn. (ICML) . Baltimore, MD, US.: PMLR, Jul. 2022, pp. 15 895– 15 907

2022
[24]

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. Commun., vol. 72, no. 7, pp. 4101–4116, Feb. 2024

2024
[25]

Multi-receiver task- oriented communications via multi-task deep learning,

Y . E. Sagduyu, T. Erpek, A. Yener, and S. Ulukus, “Multi-receiver task- oriented communications via multi-task deep learning,” in Proc. IEEE Future Netw. World Forum (FNWF) , Baltimore, MD, US, Nov. 2023, pp. 1–6

2023
[26]

Semantic commu- nication for cooperative multi-task processing over wireless networks,

A. H. Razlighi, C. Bockelmann, and A. Dekorsy, “Semantic commu- nication for cooperative multi-task processing over wireless networks,” IEEE Wireless Communications Letters, vol. 13, no. 10, pp. 2867–2871, Aug. 2024

2024
[27]

Multi-objective meta learning,

F. Ye, B. Lin, Z. Yue, P. Guo, Q. Xiao, and Y . Zhang, “Multi-objective meta learning,” in Proc. Adv. Neural Inf. Proces. Syst. (NIPS) . Virtual, Online: NIPS foundation, Dec. 2021, pp. 21 338–21 351

2021
[29]

Deep joint source- channel coding for wireless image transmission,

E. Bourtsoulatze, D. B. Kurka, and D. Gündüz, “Deep joint source- channel coding for wireless image transmission,” IEEE Transactions on Cognitive Communications and Networking , vol. 5, no. 3, pp. 567–579, May 2019

2019
[30]

Wireless image re- trieval at the edge,

M. Jankowski, D. Gündüz, and K. Mikolajczyk, “Wireless image re- trieval at the edge,” IEEE J. Sel. Area. Commun. , vol. 39, no. 1, pp. 89–100, Nov. 2020

2020
[31]

Robust semantic communications with masked vq-vae enabled codebook,

Q. Hu, G. Zhang, Z. Qin, Y . Cai, G. Yu, and G. Y . Li, “Robust semantic communications with masked vq-vae enabled codebook,” IEEE Trans. Wirel. Commun., vol. 22, no. 12, pp. 8707–8722, Apr. 2023

2023
[32]

Deep learning enabled semantic communications with speech recognition and synthesis,

Z. Weng, Z. Qin, X. Tao, C. Pan, G. Liu, and G. Y . Li, “Deep learning enabled semantic communications with speech recognition and synthesis,” IEEE Trans. Wirel. Commun., vol. 22, no. 9, pp. 6227–6240, Feb. 2023

2023
[33]

Task-oriented image transmission for scene classification in unmanned aerial systems,

X. Kang, B. Song, J. Guo, Z. Qin, and F. R. Yu, “Task-oriented image transmission for scene classification in unmanned aerial systems,” IEEE Trans.Commun., vol. 70, no. 8, pp. 5181–5192, Jun. 2022

2022
[34]

Task- oriented semantic communication for stereo-vision 3d object detection,

Z. Cao, H. Zhang, L. Liang, H. Wang, S. Jin, and G. Y . Li, “Task- oriented semantic communication for stereo-vision 3d object detection,” IEEE Trans.Commun. early access, Feb. 2025

2025
[35]

Feature importance-aware task-oriented semantic transmission and op- timization,

Y . Wang, S. Han, X. Xu, H. Liang, R. Meng, C. Dong, and P. Zhang, “Feature importance-aware task-oriented semantic transmission and op- timization,” IEEE Trans. Cogn. Commun. Netw. , vol. 10, no. 4, Mar. 2024

2024
[36]

Non-orthogonal multiple access enhanced multi-user semantic communication,

W. Li, H. Liang, C. Dong, X. Xu, P. Zhang, and K. Liu, “Non-orthogonal multiple access enhanced multi-user semantic communication,” IEEE Trans. Cogn. Commun., vol. 9, no. 6, pp. 1438–1453, Aug, 2023

2023
[37]

Multi-task oriented semantic communication model for deep images,

T. Gao, N. Li, P. Liu, and Z. Zhou, “Multi-task oriented semantic communication model for deep images,” in Proc. 4th Int. Conf. Adv. Manuf. Technol. Electron. Inf. (AMTEI) , Chongqing, China, Sep. 2025, pp. 454–461

2025
[38]

Witt: A wireless image transmission transformer for semantic communications,

K. Yang, S. Wang, J. Dai, K. Tan, K. Niu, and P. Zhang, “Witt: A wireless image transmission transformer for semantic communications,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process. (ICASSP) . Rhodes Island, Greece: IEEE, Jun. 2023, pp. 1–5

2023
[39]

Swin transformer: Hierarchical vision transformer using shifted windows,

Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” in Proc. IEEE Int. Conf. Comput. Vision. (ICCV) . Virtual, Online: IEEE, Oct. 2021, pp. 10 012–10 022

2021
[41]

Separate source channel coding is still what you need: An LLM-based rethinking,

T. Ren, R. Li, M.-m. Zhao, X. Chen, G. Liu, Y . Yang, Z. Zhao, and H. Zhang, “Separate source channel coding is still what you need: An LLM-based rethinking,” ZTE Commun., vol. 23, no. 1, pp. 30–44, Mar. 2025

2025
[42]

Bi-objective trade-off with dynamic barrier gradient descent,

C. Gong and X. Liu, “Bi-objective trade-off with dynamic barrier gradient descent,” in Proc. Adv. Neural Inf. Proces. Syst. (NIPS), Virtual, Online, Dec. 2021, pp. 29 630–29 642

2021
[44]

Design of low-density parity check codes for 5g new radio,

T. Richardson and S. Kudekar, “Design of low-density parity check codes for 5g new radio,” IEEE Commun. Mag., vol. 56, no. 3, pp. 28–34, Mar. 2018. 17 Zhilin Lu received the M.S. degree from North China Electric Power University, Beijing, China, in

2018
[2021]

degree with the College of Information Science and Elec- tronic Engineering, Zhejiang University, Hangzhou, China

She is currently pursuing the Ph.D. degree with the College of Information Science and Elec- tronic Engineering, Zhejiang University, Hangzhou, China. Her research interests include semantic com- munication and deep reinforcement learning. Rongpeng Li (Senior Member, IEEE) is ...

2015
[2024]

Available: https://arxiv.org/abs/2401.09257

[Online]. Available: https://arxiv.org/abs/2401.09257

arXiv

Pith tools

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

[1] [43]

Bome! bilevel optimization made easy: A simple first-order approach,

B. Liu, M. Ye, S. Wright, P. Stone, and Q. Liu, “Bome! bilevel optimization made easy: A simple first-order approach,” in Proc. Adv. Neural Inf. Proces. Syst. (NIPS) , New Orleans, LA, US, 2022, pp. 17 248–17 262

work page 2022

[2] [23]

A first-order multi-gradient algorithm for multi-objective bi-level optimization,

F. Ye, B. Lin, X. Cao, Y . Zhang, and I. W. Tsang, “A first-order multi-gradient algorithm for multi-objective bi-level optimization,” Jul

work page

[3] [12]

Self- critical alternate learning based semantic broadcast communication,

Z. Lu, R. Li, M. Lei, C. Wang, Z. Zhao, and H. Zhang, “Self- critical alternate learning based semantic broadcast communication,” IEEE Trans. Commun. early access, Oct. 2024

work page 2024

[4] [40]

Alternate learning- based snr-adaptive sparse semantic visual transmission,

S. Tong, X. Yu, R. Li, K. Lu, Z. Zhao, and H. Zhang, “Alternate learning- based snr-adaptive sparse semantic visual transmission,” IEEE Trans. Wirel. Commun. early access, Dec. 2024

work page 2024

[5] [28]

A generic first-order algorithmic framework for bi-level programming beyond lower-level singleton,

R. Liu, P. Mu, X. Yuan, S. Zeng, and J. Zhang, “A generic first-order algorithmic framework for bi-level programming beyond lower-level singleton,” in Proc. Int. Conf. Mach. Learn. (ICML) . Virtual, Online: PMLR, Jul. 2020, pp. 6305–6315

work page 2020

[6] [1]

Task-oriented image semantic communication based on rate-distortion theory,

F. Liu, W. Tong, Y . Yang, Z. Sun, and C. Guo, “Task-oriented image semantic communication based on rate-distortion theory,” Dec. 2022. [Online]. Available: https://arxiv.org/abs/2201.10929

arXiv 2022

[7] [2]

Semantics-empowered communication: A Tutorial-cum-Survey,

Z. Lu, R. Li, K. Lu, X. Chen, E. Hossain, Z. Zhao, and H. Zhang, “Semantics-empowered communication: A Tutorial-cum-Survey,” IEEE Communications Surveys & Tutorials , vol. 26, no. 1, pp. 41–79, Nov. 2023

work page 2023

[8] [3]

Rethinking modern communication from semantic coding to semantic communication,

K. Lu, Q. Zhou, R. Li, Z. Zhao, X. Chen, J. Wu, and H. Zhang, “Rethinking modern communication from semantic coding to semantic communication,” IEEE Wireless Commun., vol. 30, no. 1, pp. 158–164, Feb. 2023. 16

work page 2023