Singular Value Decomposition (SVD) has recently gained traction as an effective compression technique for large language models (LLMs), with many studies reporting 20-80% parameter reduction at minimal accuracy cost. However, despite reducing weight memory, existing SVD-based approaches still rely on standard dense CUDA kernels during inference, which incur substantial-and ultimately unnecessary-activation memory overhead. Our analysis reveals that this kernel-induced cost, which grows with sequence length and hidden size, in worst case prevents any real reduction in peak inference memory, limiting the practical impact of SVD compression for on-device deployment. To address this bottleneck, we propose FlashSVD, an end-to-end, rank-aware streaming inference framework for SVD-compressed LLMs. FlashSVD integrates seamlessly with any SVD-based model and directly fuses low-rank projection kernels into self-attention and feed-forward pipelines. This design avoids materializing large activation buffers by streaming small tiles of truncated factors through on-chip SRAM, performing on-the-fly multiplication and reduction, and immediately evicting results–thus preserving high GPU occupancy without introducing latency. On standard benchmarks (e.g., BERT-Base), FlashSVD reduces peak activation memory by up to 70.2% and transient memory by 75%, with zero accuracy loss against low-rank baselines, enabling truly memory-efficient deployment of low-rank LLMs.

In complex and variable terrains, affine formation control, with its flexible formation adjustment ability, can achieve various formation shapes to adapt well to the environment. Notably, the existing affine formation research based on stress matrices require the variation parameters for translation, rotation, scaling, and shearing of formations to be predesigned offline. To address this, we propose a novel method for online affine parameter adjustment that enables self-reconfiguration of formations in multiobstacle environments. By adopting artificial potential field environment excitation, the proposed motion planning algorithm can dynamically adjust the affine transformation parameters online, and realize the self-reconfiguration of formation shape to avoid collision. Then, a distributed model predictive controller is proposed for multiagent systems, which actively utilizes historical control input information to flexibly adjust controller performance while avoiding algebraic loops between neighboring agent controllers. The algorithm separates stability and performance optimization within the nonlinear model predictive control framework, ensuring both the feasibility and stability of the underlying optimization. Finally, the simulation results confirm the effectiveness of the proposed controller.

This article considers the formation control of multiple autonomous aerial vehicles (AAVs), where the AAVs operate in dynamic environments with multiple obstacles and narrow (constrained) areas. A new fuzzy model predictive interval output-constrained maneuver formation control method for AAVs is proposed. The interval output constraint algorithm utilizes the distance information to implement output constraints and formation changes, which are more practical but more challenging than the output constraint problem based on user-assigned settling time, especially in narrow areas. One of the distinctive advantages of the proposed output-constrained model predictive controller is that it can activate formation changes and output constraints for AAVs when passing through narrow space, and it can automatically restore the formation of AAVs and deactivate the output constraints when the AAVs are far away from the narrow space. Unlike most nonlinear model predictive control strategy, where predictive control relies on the accurate underlying dynamical system, the article introduces adaptive fuzzy updating law to receding horizon optimization algorithm to estimate and compensate unknown dynamics and external disturbances. Two potential field functions are designed to safely track in 3-D environments with obstacles. Finally, several examples are provided to illustrate the effectiveness of the proposed controller.

In federated learning systems, the unexpected quitting of participants is inevitable. Such quittings generally do not incur serious consequences in horizontal federated learning (HFL), but they do damage to vertical federated learning (VFL), which has been underexplored in previous research. In this paper, we show that there are two major vulnerabilities when passive parties unexpectedly quit in the deployment phase of VFL — severe performance degradation and intellectual property (IP) leakage of the active party’s labels. To solve these issues, we design PlugVFL to improve the VFL model’s robustness against the unexpected exit of passive parties and protect the active party’s IP in the deployment phase simultaneously. We evaluate our framework on multiple datasets against different inference attacks. The results show that PlugVFL effectively maintains model performance after the passive party quits and successfully disguises label information from the passive party’s feature extractor, thereby mitigating IP leakage.

Pre-trained language models (PLM) have revolutionized the NLP landscape, achieving stellar performances across diverse tasks. These models, while benefiting from vast training data, often require fine-tuning on specific data to cater to distinct downstream tasks. However, this data adaptation process has inherent security and privacy concerns, primarily when leveraging user-generated, device-residing data. Federated learning (FL) provides a solution, allowing collaborative model fine-tuning without centralized data collection. However, applying FL to finetune PLMs is hampered by challenges, including restricted model parameter access due to the high encapsulation, high computational requirements, and communication overheads. This paper introduces Federated Black-box Prompt Tuning (FedBPT), a framework designed to address these challenges. FedBPT allows the clients to treat the model as a black-box inference API. By focusing on training optimal prompts and utilizing gradient-free optimization methods, FedBPT reduces the number of exchanged variables, boosts communication efficiency, and minimizes computational and storage costs. Experiments highlight the framework’s ability to drastically cut communication and memory costs while maintaining competitive performance. Ultimately, FedBPT presents a promising solution for efficient, privacy-preserving fine-tuning of PLM in the age of large language models.

This article investigates the formation tracking-containment control problem for multiple unmanned aerial vehicles (UAVs), while considering collision avoidance in the three-dimensional (3-D) environment. A distributed model predictive formation control method is developed for UAVs with obstacle avoidance. Unlike most existing trajectory tracking control schemes, the proposed method is derived from a smooth shifting function and a distributed Lyapunov-based model predictive controller, together with two collision-free functions, to obtain an improved control algorithm with the following characteristics: 1) the swarm system of UAVs realizes the cooperative hunting process from tracking formation to containing formation, fencing a target to their convex hull; 2) the Lyapunov-based model predictive formation control method inherits the stability properties of backstepping technique and adopts the receding horizon optimization of model predictive control (MPC) technique; and 3) by exploiting the Lyapunov-based MPC algorithm and collision-free functions, the swarm system of UAVs can avoid collision with obstacles or collision with each other in 3-D scenarios with multiple obstacles. Finally, the simulation and comparison results confirm that the proposed controller outperforms the traditional backstepping controller in terms of safety and tracking performance.

Over the past decade, there has been a persistent trend in edge computing, driving the migration of intelligence closer to the edge. The increasing need to process data locally has fueled the deployment of highly efficient computing hardware and artificial intelligence (AI) models onto edge devices. The performance and robustness of edge computing systems are significantly influenced by the heterogeneity of computing systems and the diverse nature of data to be processed by each edge device. This paper aims to explore the principles of software/hardware co-design for edge computing systems in AI applications. We will delve into the robustness concerns faced by edge AI due to the inherent heterogeneity of systems and data. Furthermore, we will present various solutions that effectively mitigate these adverse effects and enhance the resilience of edge AI systems.

This article investigates the adaptive performance guaranteed tracking control problem for multiagent systems (MASs) with power integrators and measurement sensitivity. Different from the structural characteristics of existing results, the dynamic of each agent is a power exponential function. A method called adding a power integrator technique is introduced to guarantee that the consensus is achieved of the MASs with power integrators. Different from existing prescribed performance tracking control results for MASs, a new performance guaranteed control approach is proposed in this article, which can guarantee that the relative position error between neighboring agents can converge into the prescribed boundary within preassigned finite time. By utilizing the Nussbaum gain technique and neural networks, a novel control scheme is proposed to solve the unknown measurement sensitivity on the sensor, which successfully relaxes the restrictive condition that the unknown measurement sensitivity must be within a specific range. Based on the Lyapunov functional method, it is proven that the relative position error between neighboring agents can converge into the prescribed boundary within preassigned finite time. Finally, a simulation example is proposed to verify the availability of the control strategy.

Performance degradation due to domain shift remains a major challenge in medical image analysis. Unsupervised domain adaptation that transfers knowledge learned from the source domain with ground truth labels to the target domain without any annotation is the mainstream solution to resolve this issue. In this paper, we present a novel unsupervised domain adaptation framework for cross-modality cardiac segmentation, by explicitly capturing a common cardiac structure embedded across different modalities to guide cardiac segmentation. In particular, we first extract a set of 3D landmarks, in a self-supervised manner, to represent the cardiac structure of different modalities. The high-level structure information is then combined with another complementary feature, the Canny edges, to produce accurate cardiac segmentation results both in the source and target domains. We extensively evaluate our method on the MICCAI 2017 MM-WHS dataset for cardiac segmentation. The evaluation, comparison and comprehensive ablation studies demonstrate that our approach achieves satisfactory segmentation results and outperforms state-of-the-art unsupervised domain adaptation methods by a significant margin.