22 November 2024 in IEEE Transactions on Wireless Communications
The emergence of large language models (LLMs) with multi-task generalization capabilities is expected to improve the performance of artificial intelligence (AI)-as-a-service provision in 6G networks. By fine-tuning LLMs, AI services can become more precise and tailored to the demands of different downstream tasks. However, centralized fine-tuning paradigms pose a potential risk to user privacy, and existing distributed fine-tuning methods incur significant wireless transmission burdens due to the large-scale parameter transmission of LLMs. To tackle these challenges, by leveraging the low rank feature in LLM fine-tuning, we propose a wireless over-the-air federated learning (AirFL) based low-rank adaptation (LoRA) framework that integrates LoRA and over-the-air computation (AirComp) to achieve efficient fine-tuning and aggregation. Based on multiple-input multiple-output (MIMO) and orthogonal frequency division multiplexing (OFDM), we design a multi-stream AirComp scheme to fulfill the aggregation requirement of AirFL-LoRA. Furthermore, by deriving an optimality gap, we gain theoretical insights into the joint impact of rank selection and gradient aggregation distortion on the fine-tuning performance of AirFL-LoRA. Next, we formulate a non-convex problem to minimize the optimality gap, which is solved by the proposed backtracking-based alternating algorithm and the manifold optimization algorithm iteratively. Through fine-tuning LLMs for different downstream tasks, experimental results reveal that the AirFL-LoRA framework outperforms the state-of-the-art baselines on both training loss and perplexity, closely approximating the performance of FL with ideal aggregation.
19 September 2025 in IEEE Communications Standards Magazine
Artificial intelligence (AI) promises to revolutionize the design, optimization and management of next-generation communication systems. In this article, we explore the integration of large AI models (LAMs) into semantic communications (SemCom) by leveraging their multi-modal data processing and generation capabilities. Although LAMs bring unprecedented abilities to extract semantics from raw data, this integration entails multifaceted challenges including high resource demands, model complexity, and the need for adaptability across diverse modalities and tasks. To overcome these challenges, we propose a LAM-based multi-task SemCom (MTSC) architecture, which includes an adaptive model compression strategy and a federated split fine-tuning approach to facilitate the efficient deployment of LAM-based semantic models in resource-limited networks. Furthermore, a retrieval-augmented generation scheme is implemented to synthesize the most recent local and global knowledge bases to enhance the accuracy of semantic extraction and content generation, thereby improving the inference performance. Finally, simulation results demonstrate the efficacy of the proposed LAM-based MTSC architecture, highlighting the performance enhancements across various downstream tasks under varying channel conditions.
21 March 2025 in IEEE Internet of Things Journal
In existing deep learning-based semantic communication systems, centralized training of semantic models brings a risk of privacy leakage, whereas distributed training imposes a huge computational burden on user equipments (UEs). To address these challenges, we propose a hybrid federated learning (Hybrid-FL) framework to alleviate the computational burden on UEs while protecting the user privacy. Specifically, each UE uploads local gradients and semantic symbols to the base station for the collaborative training of global and local semantic models. Furthermore, we propose a joint communication and computation scheme for supporting the model aggregation and semantics transmission. To gain deep insights, we expose the joint impact of communication and computation on the convergence behavior of Hybrid-FL by deriving an upper bound. Then, we formulate a mixed-integer nonlinear programming problem to improve the convergence performance of Hybrid-FL, which is then effectively solved by using our proposed algorithm that developed based on alternating and matching theory. Experimental results demonstrate that Hybrid-FL outperforms conventional FL by achieving a 20% accuracy gain and a 80% latency reduction.
6 June 2025 in IEEE Internet of Things Magazine
Large artificial intelligence (AI) models exhibit remarkable capabilities in various application scenarios, but deploying them at the network edge poses significant challenges due to issues such as data privacy, computational resources, and latency. In this paper, we explore federated fine-tuning and collaborative reasoning techniques to facilitate the implementation of large AI models in resource-constrained wireless networks. Firstly, promising applications of large AI models within specific domains are discussed. Subsequently, federated fine-tuning methods are proposed to adapt large AI models to specific tasks or environments at the network edge, effectively addressing the challenges associated with communication overhead and enhancing communication efficiency. These methodologies follow clustered, hierarchical, and asynchronous paradigms to effectively tackle privacy issues and eliminate data silos. Furthermore, to enhance operational efficiency and reduce latency, efficient frameworks for model collaborative reasoning are developed, which include decentralized horizontal collaboration, cloud-edge-end vertical collaboration, and multi-access collaboration. Next, simulation results demonstrate the effectiveness of our proposed methods in reducing the fine-tuning loss of large AI models across various downstream tasks. Finally, several open challenges and research opportunities are outlined.
6 December 2023 in IEEE Wireless Communications Letters
To reduce latency and improve the convergence of hierarchical federated learning (HFL), this letter investigates an HFL framework with partial aggregation in a resource-constrained multi-tier wireless network. Due to device heterogeneity and resource scarcity, device selection and resource allocation are jointly considered in the presence of inter-cell interference (ICI). To capture the impacts of device selection and resource allocation on the convergence behavior of HFL, an upper bound of the expected optimality gap is derived. Then, a mixed-integer non-linear programming (MINLP) problem is formulated to minimize the weighted sum of the optimality gap and overall latency. To solve the MINLP problem efficiently, a two-layer solution based on genetic algorithm and alternating optimization is designed to jointly optimize the resource block association, uplink transmission power, and central processing unit frequency allocation. Simulation results on two real-world datasets demonstrate that the proposed algorithm achieves better learning performance and ICI management than the federated learning with client selection (FedCS) and random resource block association schemes.
9 December 2025 in IEEE Transactions on Vehicular Technology
The overwhelming scale of large language models (LLMs) exhausts the on-device communication and computation resources in vehicular networks, limiting its application in performing inference for vehicle intelligence. To address this challenge, we resort to caching technique, specifically developing a novel collaborative LLM inference framework. In the framework, the road side unit selectively caches intermediate computational results of input tokens from vehicle users, whereby the redundant transmission and computation can be reduced. Next, we evaluate the communication and computation overheads requested for implementing the proposed collaborative LLM inference framework. Then, we formulate a mixed-integer non-linear programming problem to minimize the collaborative inference latency by jointly optimizing the allocation of communication, computation, and caching resources. Simulation results validate the superiority of the caching-empowered LLM collaborative inference framework over baselines in terms of latency and energy efficiency.