Scalable Pre-Trained Large Channel Models for Wireless Communication Systems
IEEE Communications Magazine · 2026-05 · 64 (5) · 60-67
DOI: 10.1109/mcom.001.2500406 ↗At a glance
Journal article in IEEE Communications Magazine, 2026-05. Written by Jianhua Guo and 4 co-authors. 15 references. Funded by: National Key R&D Program of China, National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities.
Abstract (original language)
In the sixth generation (6G) era, large artificial intelligence models (LAMs), inspired by the success of large language models (LLMs), have emerged as promising solutions to overcome task/scenario-specific limitations of deep learning (DL)-based wireless communication models (WCMs). However, integrating LAMs into wireless communications faces several key challenges, including the absence of unified multitask pretraining paradigms, limited model/dataset sizes, and a lack of domain-specific scaling laws. This article presents a scalable pretrained masked channel model (MCM), specifically designed for channel-related tasks in wireless communications. Through a multidimensional comparative analysis of LLMs and WCMs, we highlight the necessity and challenges for building LAMs tailored to wireless communications based on their specific characteristics and requirements. A case study deploying MCM in an uplink multiple-input-multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) system demonstrates superior performance and cross-task/scenario generalization. Furthermore, we derive domain-specific scaling laws, providing empirical guidance for designing and deploying LAMs in wireless communications. Finally, we outline promising research directions to advance the integration of LAMs in future wireless networks.
Funding
- National Key R&D Program of China
- National Natural Science Foundation of China
- Fundamental Research Funds for the Central Universities
Cite
APA
Guo, J., Xue, J., Zhang, J., Qiao, Z., & Xu, Z. (2026). Scalable Pre-Trained Large Channel Models for Wireless Communication Systems. IEEE Communications Magazine, 64(5), 60-67. https://doi.org/10.1109/mcom.001.2500406
Vancouver
Guo J, Xue J, Zhang J, Qiao Z, Xu Z. Scalable Pre-Trained Large Channel Models for Wireless Communication Systems. IEEE Communications Magazine. 2026;64(5):60-67. doi:10.1109/mcom.001.2500406
BibTeX
@article{Guo2026,
author = {Guo, Jianhua and Xue, Jiang and Zhang, Junkai and Qiao, Zhen and Xu, Zongben},
title = {Scalable Pre-Trained Large Channel Models for Wireless Communication Systems},
journal = {IEEE Communications Magazine},
year = {2026},
volume = {64},
number = {5},
pages = {60--67},
doi = {10.1109/mcom.001.2500406}
}
EUVIMED