Parameter-efficient fine-tuning (PEFT) methods have emerged as a practical solution for adapting large foundation models to downstream tasks, reducing computational and memory costs by updating only a small subset of parameters. Among them, approaches like LoRA aim to strike a balance between efficiency and expressiveness, but often suffer from slow convergence and limited adaptation capacity due to their inherent low-rank constraints. This trade-off hampers the ability of PEFT methods to capture complex patterns needed for diverse tasks. To address these challenges, we propose FRoD, a novel fine-tuning method that combines hierarchical joint decomposition with rotational degrees of freedom. By extracting a globally shared basis across layers and injecting sparse, learnable perturbations into scaling factors for flexible full-rank updates, FRoD enhances expressiveness and efficiency, leading to faster and more robust convergence. On 20 benchmarks spanning vision, reasoning, and language understanding, FRoD matches full model fine-tuning in accuracy, while using only 1.72% of trainable parameters under identical training budgets.

In this paper, we propose the IFresher framework to improve the timeliness of multi-agent mobile augmented reality (MAR) systems. Existing works have made strides in accuracy-latency trade-offs, but fail to directly address realtime task responsiveness and multi-agent contention challenges. To bridge this gap, we introduce the concept of the age of analytics information (AoAI), which quantifies the combined impact of video analytics (VA) accuracy, transmission delay, and computational efficiency. By deriving a closed-form expression for AoAI, IFresher establishes a central control mechanism that jointly optimizes bandwidth allocation and video configuration to minimize AoAI while ensuring accuracy. Due to the mixed-integer nonlinear characteristics of the problem and the fact that each agent only has local observations, the problem is reformulated into a decentralized partially observable Markov decision process (Dec-POMDP). We propose a multi-agent reinforcement learning (MARL) algorithm, named convex-embedded transformer QMIX (CTQMIX), using the centralized training and decentralized execution (CTDE) framework for agent collaboration. Specifically, the convex optimization ensures optimal bandwidth distribution, and the transformer captures temporal dependencies between observations and actions across time steps to improve decisionmaking in dynamic environments. Evaluations with real-world experiments show that the CTQMIX outperforms state-of-theart (SOTA) algorithms.

This paper presents an edge-assisted massive mobile live streaming (MMLS) framework named LOCA, integrating chunk-level analysis and long-term optimization to design resource allocation, bitrate adaptation, and source selection strategies. The proposed method ensures sustained real-time video delivery while minimizing latency and communication costs. Firstly, a chunk-level analysis of the entire process of video streaming is introduced, aiming at modeling fetch queue waiting time and rebuffering duration in each time slot. By embedding this mathamatical model into consideration, a long-term optimization is formulated to minimize rebuffering and communication overhead while maintaining high video qualities for massive users. Leveraging Lyapunov optimization, we transform this problem into a computationally tractable form. Further simplification via linearization achieves near-optimal solutions by adopting the mixed-integer linear programming method with enhanced computational efficiency. Simulation results demonstrate superior stability and long-term performance compared to the state-of-theart and baseline methods, validating the framework's efficacy in MMLS scenarios

Satellite networks are crucial for future 6G communications, offering global coverage. However, traditional Transmission Control Protocol (TCP) struggles with poor stability in dynamic satellite networks. In contrast, QUIC is capable of providing stability in such environments thanks to features like seamless connection migration. Nevertheless, the loss detection and recovery strategy of QUIC needs improvement in satellite networks with relatively high transmission errors. To address this, we propose Polar-QUIC, an adaptive packet-level polar coding-assisted QUIC scheme. By leveraging the robust error correction capabilities of polar codes, we design a packet-level encoder/decoder, loss detection tuning module, and adaptive redundancy module that combines packet loss rate and congestion window size. Experimental results demonstrate that Polar- QUIC effectively reduces the average Download Completion Time (DCT) of QUIC in satellite networks. Specifically, Polar-QUIC minimizes the average DCT by up to 35.8% and 74.5%, respectively, compared to general forward error correction schemes such as Reed-Solomon Codes and Random Linear Codes.

In this paper, we propose an edge selective sharing architecture (ESSA) for massive mobile video streaming (MMVS) and develop cross-layer optimizations for joint user scheduling and power allocation (JUSPA) in ESSA. This work aims to ensure users' high quality of experience (QoE) by efficiently using network resources over mobile edge computing (MEC)-assisted massive multiple-input and multiple-output (MIMO) networks. Initially, by taking advantage of the analytical and empirical characteristics of video streaming, ESSA is able to maintain MMVS by activating only a portion of facilities in wireless access networks (WANs) without sacrificing video quality. Following this, a cross-layer optimization problem is formulated for JUSPA. We simplify the problem into a generalized assignment (GA) problem by discussing the peculiarities of MMSV, whose approximate solution can be obtained in polynomial time. Moreover, load balancing based on user density is integrated to reformulate a capacitated facility location (CFL) problem solvable with low overall time complexity, effectively mitigating the increased overheads of MEC units and enhancing the overall performance of MMVS. Numerical results indicate that our ESSA with JUSPA schemes for MMVS achieves better streaming fluency and transmission efficiency than alternative methods.