Time-Slotted Multi-Cluster UAV AirComp with Energy-Awareness: A Pointer Network-Assisted Soft Actor-Critic Learning Framework

Over-the-air computation (AirComp) has emerged as a promising approach for massive data aggregation, which is yet challenged by the channel variations, task distributions, and inherent energy limitation of the computation nodes. In this paper, we propose an unmanned aerial vehicle (UAV)-assisted Aircomp system to serve multi-cluster computation tasks over time, where the UAV mobility-facilitated spatial and time diversity is exploited for efficient and accurate data computation. Specifically, we aim for the minimization of AirComp aggregation error and the energy consumption by jointly optimizing the transceiver beamforming, normalizing factors, sensor scheduling, and UAV trajectory. To solve the formulated problem, we decompose it into two layers where the inner layer addresses the optimization-based AirComp transceiver design, and the outer layer focuses on the deep reinforcement learning (DRL)-based scheduling and trajectory design. In particular, a pointer network actor-critic learning is developed to tackle the binary scheduling problem, and a soft actor-critic DRL algorithm is employed to determine the UAV trajectory. Simulation results validate the convergence of the proposed hierarchical learning framework and demonstrate its significant performance gains in terms of AirComp aggregation error and energy consumption as compared with baseline schemes.

Bridging the Cognitive Gap: A Unified Memory Paradigm for 6G Agentic AI-RAN

As 6G evolves, the radio access network must transcend traditional automation to embrace agentic AI capable of perception, reasoning, and evolution. A fundamental cognitive gap persists in current disaggregated architectures, where interfaces force the physical layer to compress high-dimensional states into low-dimensional metrics, trapping reasoning agents behind a semantic bottleneck. This article envisions a shift from interface-bound to memory-centric architectures. We propose a unified memory paradigm that dissolves the boundaries between sensing and reasoning by mapping biological memory hierarchies onto heterogeneous computing fabrics. Enabled by emerging coherent interconnects, this approach creates a cognitive continuum where microsecond-level reflexes, millisecond-level reasoning, and long-term evolution share state across time scales. By replacing message passing with zero-copy observability, we empower AI agents to bridge the gap between real-time responsiveness and long-horizon context for truly autonomous 6G networks.

FedVSSAM: Mitigating Flatness Incompatibility in Sharpness-Aware Federated Learning

Sharpness-aware minimization (SAM) is an effective method for improving the generalization of federated learning (FL) by steering local training toward flat minima. Under data heterogeneity, however, device-side SAM searches for locally flat basins that are incompatible with the flat region preferred by the global objective. We identify this structural failure mode as flatness incompatibility, which explains why improving local flatness alone may provide limited training and generalization improvement for the global model. We reveal that flatness incompatibility arises from data heterogeneity and the friendly adversary phenomenon, and is further amplified by local updates and partial device participation. To mitigate this issue, we propose Federated Learning with variance-suppressed sharpness-aware minimization (FedVSSAM), which constructs a variance-suppressed adjusted direction and uses it consistently in local flatness search, local descent, and global update. FedVSSAM anchors both perturbation and update directions to a more stable global direction, instead of correcting only an isolated local perturbation. We establish non-convex convergence guarantees of FedVSSAM and prove that the mean-square deviation between the adjusted direction and the global gradient is effectively controlled. Experiments demonstrate that FedVSSAM mitigates flatness incompatibility and outperforms the baselines across diverse FL settings.

Rotatable Antenna Enabled Multi-Cell Mixed Near-Field and Far-Field Communications

Prior studies on mixed near-field and far-field communications have focused exclusively on single-cell scenarios, where both near-field and far-field users are served by the same base station (BS), leading to intra-cell mixed-field interference. In this paper, we consider a more general and practical multi-cell mixed-field scenario consisting of multiple cells, each serving multiple users, thus resulting in more complex inter-cell mixed-field interference. To address this new challenge, we propose leveraging rotatable antenna (RA) technology to enhance multi-cell mixed-field communication performance by exploiting the additional spatial degree-of-freedom introduced by RA rotation to mitigate interference in an efficient way. Specifically, we study an RA-enabled multi-cell mixed-field communication system in which each BS is equipped with an RA array to serve its associated users. We formulate a network-wide sum-rate maximization problem that jointly optimizes the transmit beamforming and the rotation angles of the RA arrays, subject to per-BS power constraints and admissible array rotation limits. To gain useful insights into the role of RAs in multi-cell mixed-field communications, we first analyze a special case with a single user per cell. For this case, we obtain a closed-form expression for the rotation-aware inter-cell mixed-field interference using the Fresnel integrals and analytically show that RA rotation can effectively mitigate such interference, thereby substantially improving system performance. For the general case with multiple users per cell, we develop an efficient double-layer algorithm: the inner layer optimizes the transmit beamforming at each BS via semidefinite relaxation and successive convex approximation; while the outer layer determines the rotation angles of the RA arrays using particle swarm optimization.

SLA-Aware Distributed LLM Inference Across Device-RAN-Cloud

Embodied AI requires sub-second inference near the Radio Access Network (RAN), but deployments span heterogeneous tiers (on-device, RAN-edge, cloud) and must not disrupt real-time baseband processing. We report measurements from a 5G Standalone (SA) AI-RAN testbed using a fixed baseline policy for repeatability. The setup includes an on-device tier, a three-node RAN-edge cluster co-hosting a containerized 5G RAN, and a cloud tier. We find that on-device execution remains multi-second and fails to meet sub-second budgets. At the RAN edge, SLA feasibility is primarily determined by model variant choice: quantized models concentrate below 0.5\,s, while unquantized and some larger quantized models incur deadline misses due to stalls and queuing. In the cloud tier, meeting a 0.5\,s deadline is challenging on the measured WAN path (up to 32.9\% of requests complete within 0.5\,s), but all evaluated variants meet a 1.0\,s deadline (100\% within 1.0\,s). Under saturated downlink traffic and up to $N{=}20$ concurrent inference clients, Multi-Instance GPU (MIG) isolation preserves baseband timing-health proxies, supporting safe co-location under fixed partitioning.

From Specialist to Large Models: A Paradigm Evolution Towards Semantic-Aware MIMO

The sixth generation (6G) network is expected to deploy larger multiple-input multiple-output (MIMO) arrays to support massive connectivity, which will increase overhead and latency at the physical layer. Meanwhile, emerging 6G demands such as immersive communications and environmental sensing pose challenges to traditional signal processing. To address these issues, we propose the ``semantic-aware MIMO'' paradigm, which leverages specialist models and large models to perceive, utilize, and fuse the inherent semantics of channels and sources for improved performance. Moreover, for representative MIMO physical-layer tasks, e.g., random access activity detection, channel feedback, and precoding, we design specialist models that exploit channel and source semantics for better performance. Additionally, in view of the more diversified functions of 6G MIMO, we further explore large models as a scalable solution for multi-task semantic-aware MIMO and review recent advances along with their advantages and limitations. Finally, we discuss the challenges, insights, and prospects of the evolution of specialist models and large models empowered semantic-aware MIMO paradigms.

Wireless Federated Multi-Task LLM Fine-Tuning via Sparse-and-Orthogonal LoRA

Decentralized federated learning (DFL) based on low-rank adaptation (LoRA) enables mobile devices with multi-task datasets to collaboratively fine-tune a large language model (LLM) by exchanging locally updated parameters with a subset of neighboring devices via wireless connections for knowledge integration.However, directly aggregating parameters fine-tuned on heterogeneous datasets induces three primary issues across the DFL life-cycle: (i) \textit{catastrophic knowledge forgetting during fine-tuning process}, arising from conflicting update directions caused by data heterogeneity; (ii) \textit{inefficient communication and convergence during model aggregation process}, due to bandwidth-intensive redundant model transmissions; and (iii) \textit{multi-task knowledge interference during inference process}, resulting from incompatible knowledge representations coexistence during inference. To address these issues in a fully decentralized scenario, we first propose a sparse-and-orthogonal LoRA that ensures orthogonality between model updates to eliminate direction conflicts during fine-tuning.Then, we analyze how device connection topology affects multi-task performance, prompting a cluster-based topology design during aggregation.Finally, we propose an implicit mixture of experts (MoE) mechanism to avoid the coexistence of incompatible knowledge during inference. Simulation results demonstrate that the proposed approach effectively reduces communication resource consumption by up to $73\%$ and enhances average performance by $5\%$ compared with the traditional LoRA method.

Duality-Guided Graph Learning for Real-Time Joint Connectivity and Routing in LEO Mega-Constellations

Laser inter-satellite links (LISLs) of low Earth orbit (LEO) mega-constellations enable high-capacity backbone connectivity in non-terrestrial networks, but their management is challenged by limited laser communication terminals, mechanical pointing constraints, and rapidly time-varying network topologies. This paper studies the joint problem of LISL connection establishment, traffic routing, and flow-rate allocation under heterogeneous global traffic demand and gateway availability. We formulate the problem as a mixed-integer optimization over large-scale, time-varying constellation graphs and develop a Lagrangian dual decomposition that interprets per-link dual variables as congestion prices coordinating connectivity and routing decisions. To overcome the prohibitive latency of iterative dual updates, we propose DeepLaDu, a Lagrangian duality-guided deep learning framework that trains a graph neural network (GNN) to directly infer per-link (edge-level) congestion prices from the constellation state in a single forward pass. We enable scalable and stable training using a subgradient-based edge-level loss in DeepLaDu. We analyze the convergence and computational complexity of the proposed approach and evaluate it using realistic Starlink-like constellations with optical and traffic constraints. Simulation results show that DeepLaDu achieves up to 20\% higher network throughput than non-joint or heuristic baselines, while matching the performance of iterative dual optimization with orders-of-magnitude lower computation time, suitable for real-time operation in dynamic LEO networks.

Vision-Language-Model-Guided Differentiable Ray Tracing for Fast and Accurate Multi-Material RF Parameter Estimation

Accurate radio-frequency (RF) material parameters are essential for electromagnetic digital twins in 6G systems, yet gradient-based inverse ray tracing (RT) remains sensitive to initialization and costly under limited measurements. This paper proposes a vision-language-model (VLM) guided framework that accelerates and stabilizes multi-material parameter estimation in a differentiable RT (DRT) engine. A VLM parses scene images to infer material categories and maps them to quantitative priors via an ITU-R material table, yielding informed conductivity initializations. The VLM further selects informative transmitter/receiver placements that promote diverse, material-discriminative paths. Starting from these priors, the DRT performs gradient-based refinement using measured received signal strengths. Experiments in NVIDIA Sionna on indoor scenes show 2-4$\times$ faster convergence and 10-100$\times$ lower final parameter error compared with uniform or random initialization and random placement baselines, achieving sub-0.1\% mean relative error with only a few receivers. Complexity analyses indicate per-iteration time scales near-linearly with the number of materials and measurement setups, while VLM-guided placement reduces the measurements required for accurate recovery. Ablations over RT depth and ray counts confirm further accuracy gains without significant per-iteration overhead. Results demonstrate that semantic priors from VLMs effectively guide physics-based optimization for fast and reliable RF material estimation.

CoCo-Fed: A Unified Framework for Memory- and Communication-Efficient Federated Learning at the Wireless Edge

The deployment of large-scale neural networks within the Open Radio Access Network (O-RAN) architecture is pivotal for enabling native edge intelligence. However, this paradigm faces two critical bottlenecks: the prohibitive memory footprint required for local training on resource-constrained gNBs, and the saturation of bandwidth-limited backhaul links during the global aggregation of high-dimensional model updates. To address these challenges, we propose CoCo-Fed, a novel Compression and Combination-based Federated learning framework that unifies local memory efficiency and global communication reduction. Locally, CoCo-Fed breaks the memory wall by performing a double-dimension down-projection of gradients, adapting the optimizer to operate on low-rank structures without introducing additional inference parameters/latency. Globally, we introduce a transmission protocol based on orthogonal subspace superposition, where layer-wise updates are projected and superimposed into a single consolidated matrix per gNB, drastically reducing the backhaul traffic. Beyond empirical designs, we establish a rigorous theoretical foundation, proving the convergence of CoCo-Fed even under unsupervised learning conditions suitable for wireless sensing tasks. Extensive simulations on an angle-of-arrival estimation task demonstrate that CoCo-Fed significantly outperforms state-of-the-art baselines in both memory and communication efficiency while maintaining robust convergence under non-IID settings.