QwenMoE-SC: A Mixture-of-Expert Semantic Communication Model with GNN-Based Unequal Error Protection, NEFTune Technique and Direct Preference Optimization

Zhang, Runwei; Zhu, Yibo; Yang, Chia Chen; Tian, Zhen; Chen, Shiyong

doi:10.3390/math14111894

This is an early access version, the complete PDF, HTML, and XML versions will be available soon.

Open AccessArticle

QwenMoE-SC: A Mixture-of-Expert Semantic Communication Model with GNN-Based Unequal Error Protection, NEFTune Technique and Direct Preference Optimization

by

Runwei Zhang

^1,†,

Yibo Zhu

^2,†,

Chia Chen Yang

³,

Zhen Tian

⁴

and

Shiyong Chen

^5,*

¹

Beijing-Dublin International College, Beijing University of Technology, Beijing 100124, China

²

School of Engineering, University of Glasgow, Glasgow G12 8QQ, UK

³

College of Management, National Pingtung University, Pingtung 91201, Taiwan

⁴

James Watt School of Engineering, University of Glasgow, Glasgow G12 8QQ, UK

⁵

School of Electronic and Information Engineering, Beihang University, Beijing 100191, China

^*

Author to whom correspondence should be addressed.

^†

These authors contributed equally to this work.

Mathematics 2026, 14(11), 1894; https://doi.org/10.3390/math14111894

Submission received: 6 April 2026 / Revised: 1 May 2026 / Accepted: 9 May 2026 / Published: 29 May 2026

(This article belongs to the Special Issue New Advances in Graph Neural Networks (GNNs) and Applications)

Download Versions Notes

Abstract

We propose QwenMoE-SC, a semantic communication framework that integrates a Mixture-of-Experts (MoE) Large Language Model with three complementary modules: (1) a Graph Neural Network (GNN)-based Unequal Error Protection (UEP) plug-in that assigns semantic importance scores via syntactic dependency graph message passing for adaptive bit allocation, without modifying the pre-trained LLM; (2) NEFTune noise injection during fine-tuning for channel robustness; and (3) a Communication-aware Direct Preference Optimization (C-DPO) strategy that favors semantically faithful yet token-efficient transmissions. Comprehensive ablation studies on AWGN and Rayleigh fading channels show that each component contributes distinct gains, and their combination consistently outperforms traditional separation-based methods and neural baselines in sentence similarity, BLEU score, and semantic-level BER, with the largest improvements at low-to-mid SNR regimes. QwenMoE-SC can also serve as a semantic interface layer within expert and decision-support systems, enabling robust, bandwidth-efficient communication between data sources, inference engines, and human users.

Keywords: semantic communication; large language models; mixture-of-experts; graph neural networks; unequal error protection; Qwen1.5; NEFTune; direct preference optimization; wireless communication

Share and Cite

MDPI and ACS Style

Zhang, R.; Zhu, Y.; Yang, C.C.; Tian, Z.; Chen, S. QwenMoE-SC: A Mixture-of-Expert Semantic Communication Model with GNN-Based Unequal Error Protection, NEFTune Technique and Direct Preference Optimization. Mathematics 2026, 14, 1894. https://doi.org/10.3390/math14111894

AMA Style

Zhang R, Zhu Y, Yang CC, Tian Z, Chen S. QwenMoE-SC: A Mixture-of-Expert Semantic Communication Model with GNN-Based Unequal Error Protection, NEFTune Technique and Direct Preference Optimization. Mathematics. 2026; 14(11):1894. https://doi.org/10.3390/math14111894

Chicago/Turabian Style

Zhang, Runwei, Yibo Zhu, Chia Chen Yang, Zhen Tian, and Shiyong Chen. 2026. "QwenMoE-SC: A Mixture-of-Expert Semantic Communication Model with GNN-Based Unequal Error Protection, NEFTune Technique and Direct Preference Optimization" Mathematics 14, no. 11: 1894. https://doi.org/10.3390/math14111894

APA Style

Zhang, R., Zhu, Y., Yang, C. C., Tian, Z., & Chen, S. (2026). QwenMoE-SC: A Mixture-of-Expert Semantic Communication Model with GNN-Based Unequal Error Protection, NEFTune Technique and Direct Preference Optimization. Mathematics, 14(11), 1894. https://doi.org/10.3390/math14111894

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

QwenMoE-SC: A Mixture-of-Expert Semantic Communication Model with GNN-Based Unequal Error Protection, NEFTune Technique and Direct Preference Optimization

Abstract

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI