Federated Recommender Systems (FedRecs) leverage federated learning to protect user privacy by retaining data locally. However, user embeddings in FedRecs often encode sensitive attribute information, rendering them vulnerable to attribute inference attacks. Attribute unlearning has emerged as a promising approach to mitigate this issue. In this paper, we focus on user-level FedRecs, which is a more practical yet challenging setting compared to group-level FedRecs. Adversarial training emerges as the most feasible approach within this context. We identify two key challenges in implementing adversarial training-based attribute unlearning for user-level FedRecs: i) mitigating training instability caused by user data heterogeneity, and ii) preventing attribute information leakage through gradients. To address these challenges, we propose FedAU2, an attribute unlearning method for user-level FedRecs. For CH1, we propose a adaptive adversarial training strategy, where the training dynamics are adjusted in response to local optimization behavior. For CH2, we propose a dual-stochastic variational autoencoder to perturb the adversarial model, effectively preventing gradient-based information leakage. Extensive experiments on three real-world datasets demonstrate that our proposed FedAU2 achieves superior performance in unlearning effectiveness and recommendation performance compared to existing baselines.

While stacked intelligent metasurfaces (SIMs) have demonstrated significant technical and cost advantages in multiuser scenarios, existing literature universally assumes ideal transceiver hardware. Addressing this gap, this paper investigates the design and optimization of a SIM-assisted multiuser downlink multiple-input single-output (MISO) system under practical transceiver hardware impairments (HWIs). To accurately capture distortion effects at both the base station (BS) and user equipment, we adopt an aggregate HWI model based on improper Gaussian statistics. The considered impairments include finite-resolution digital-to-analog converters (DACs), power amplifier (PA) nonlinearities, in-phase/quadrature (I/Q) imbalance, and other radio-frequency (RF) front-end non-idealities. We formulate a sum-rate (SR) maximization problem that jointly optimizes digital beamforming at the BS and multi-layer analog beamforming at the SIM. To tackle this highly non-convex optimization challenge, we propose a closed-form-based iterative algorithm that alternately updates BS and SIM beamforming with guaranteed convergence. Extensive simulations validate the effectiveness of the proposed algorithm, quantify the impact of different HWI sources, and demonstrate that SIM deployment significantly improves system robustness, mitigates HWI-induced performance degradation, and reduces DAC resolution requirements without substantial performance loss.

Stacked intelligent metasurface (SIM) extends the concept of single-layer reconfigurable holographic surfaces (RHS) by incorporating a multi-layered structure, thereby providing enhanced control over electromagnetic wave propagation and improved signal processing capabilities. This study investigates the potential of SIM in enhancing the rate fairness in multiuser downlink systems by addressing two key optimization problems: maximizing the minimum rate (MR) and maximizing the geometric mean of rates (GMR). The former strives to enhance the minimum user rate, thereby ensuring fairness among users, while the latter relaxes fairness requirements to strike a better trade-off between user fairness and system sum-rate (SR). For the MR maximization, we adopt a consensus alternating direction method of multipliers (ADMM)-based approach, which decomposes the approximated problem into sub-problems with closed-form solutions. For GMR maximization, we develop an alternating optimization (AO)-based algorithm that also yields closed-form solutions and can be seamlessly adapted for SR maximization. Numerical results validate the effectiveness and convergence of the proposed algorithms. Comparative evaluations show that MR maximization ensures near-perfect fairness, while GMR maximization balances fairness and system SR. Furthermore, the two proposed algorithms respectively outperform existing related works in terms of MR and SR performance. Lastly, SIM with lower power consumption achieves performance comparable to that of multi-antenna digital beamforming.

To meet the emerging demands for rapid data aggregation and reliable information transmission in future wireless applications, a novel system architecture integrating Over-the-Air Computation (AirComp) and downlink multi-user communication via a simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) is proposed in this paper. In the considered cellular scenario, an unmanned aerial vehicle (UAV) carries a STAR-RIS beneath its fuselage, creating a programmable aerial platform that concurrently serves Internet-of-Things (IoT) devices and conventional mobile users. The STAR-RIS operates in transmission mode to enable efficient wireless data aggregation of IoT devices, while its reflection mode establishes high-quality downlink channels from the base station (BS) to multiple users. Capturing the true electromagnetic behavior of the STAR-RIS, we explicitly model the practical coupling between the reflection and transmission phase shifts. Two optimization problems are then formulated: one minimizes AirComp distortion and the other maximizes the minimum user rate in the downlink. Both non-convex problems are tackled by efficient iterative algorithms derived from the penalty dual decomposition (PDD) framework. Extensive simulations confirm that the proposed design markedly outperforms baseline approaches and its performance can approach that of ideal phase-shift control by enhancing the key system parameters. Additionally, the trade-off between computation and communication performance is demonstrated.

Utilizing Graphic User Interfaces (GUIs) for human-computer interaction is essential for accessing various digital tools. Recent advancements in Vision Language Models (VLMs) reveal significant potential for developing versatile agents that assist humans in navigating GUIs. However, current VLMs face challenges related to fundamental abilities, such as OCR and grounding, as well as a lack of knowledge about GUI elements functionalities and control methods. These limitations hinder their effectiveness as practical GUI agents. To address these challenges, we introduce GUICourse, a series of datasets for training visual-based GUI agents using general VLMs. First, we enhance the OCR and grounding capabilities of VLMs using the GUIEnv dataset. Next, we enrich the GUI knowledge of VLMs using the GUIAct and GUIChat datasets. Our experiments demonstrate that even a small-sized GUI agent (with 3.1 billion parameters) performs effectively on both single-step and multi-step GUI tasks. We further finetune our GUI agents on other GUI tasks with different action spaces (AITW and Mind2Web), and the results show that our agents are better than their baseline VLMs. Additionally, we analyze the impact of OCR and grounding capabilities through an ablation study, revealing a positive correlation with GUI navigation ability.

Zhenyu Li, Yike Zhang, Tengyu Pan, Yutao Sun, Zhichao Duan, Junjie Fang, Rong Han, Zixuan Wang, Jianyong Wang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

With the growing demand for safeguarding sensitive user information in recommender systems, recommendation attribute unlearning is receiving increasing attention. Existing studies predominantly focus on single-attribute unlearning. However, privacy protection requirements in the real world often involve multiple sensitive attributes and are dynamic. Existing single-attribute unlearning methods cannot meet these real-world requirements due to CH1: the inability to handle multiple unlearning requests simultaneously, and CH2: the lack of efficient adaptability to dynamic unlearning needs. To address these challenges, we propose LEGO, a lightweight and efficient multiple-attribute unlearning framework. Specifically, we divide the multiple-attribute unlearning process into two steps: i) Embedding Calibration removes information related to a specific attribute from user embedding, and ii) Flexible Combination combines these embeddings into a single embedding, protecting all sensitive attributes. We frame the unlearning process as a mutual information minimization problem, providing LEGO a theoretical guarantee of simultaneous unlearning, thereby addressing CH1. With the two-step framework, where Embedding Calibration can be performed in parallel and Flexible Combination is flexible and efficient, we address CH2. Extensive experiments on three real-world datasets across three representative recommendation models demonstrate the effectiveness and efficiency of our proposed framework.

Riccati matrix equation (RME), a critical nonlinear matrix equation in autonomous driving and deep learning. However, memory-compute separation in traditional solving systems leads to latency and inefficiency when solving nonlinear equations, particularly under real-time requirements. Existing hardware lacks dedicated accelerators for RME, no specialized solvers quick addressing its nonlinear complexity. To address this issue, we propose a novel RME solver based on a memristive array, which leverages the parallel and fast computing advantages of analog circuits to quickly solve any order RME. Inspired by Neurodynamics for non-linear matrix equations, we first introduce an innovative Neurodynamics-based RME solving algorithm specifically designed from an analog circuit perspective. Based on this algorithm, we constructed a pioneering closed-loop analog circuit solver, overcoming bottlenecks in using circuits for such nonlinear matrix equations. Our evaluation demonstrates that the proposed solver achieves over 90% accuracy for a 128th-order parameter Riccati matrix equation. Compared to traditional digital processors, the solver offers significant energy efficiency advantages and is three orders of magnitude faster than CPU. Additionally, the solver successfully accelerates our proposed dung beetle optimizer-linear quadratic control algorithm for vehicle suspension control, achieving high precision while significantly reducing time and energy consumption compared to CPU and GPU.

In this paper, we consider an intelligent reflecting surface (IRS) assisted downlink multiuser communication system with improper Gaussian signaling (IGS) that serves as generalized Gaussian signaling and can effectively combat multiuser interference. We focus on the max-min achievable rate optimization problem by jointly optimizing the transmit beamforming vectors and reflecting phase shifts, subject to the transmit power budget constraint at the access point (AP). We propose a low-complexity iterative algorithm based on a two-layer iterative procedure, which differs from these existing algorithms that rely on inefficient alternating optimization framework and high computational complexity convex optimization tools. Specifically, in the outer layer procedure, we employ a tractable lower bound of user communication rate to reformulate the original problem and repeatedly update the lower bound in each iteration. In the inner layer procedure, based on the alternating direction method of multipliers (ADMM), we decompose the reformulated problem into several convex sub-problems, which can be alternately solved by closed-form solutions. Furthermore, we study the initialization, convergence, and computational complexity of the proposed algorithm. Additionally, we simplify the algorithm to make it applicable for the cases of conventional proper Gaussian signaling (PGS) and without IRS. Finally, numerical results validate the advantages of the proposed algorithm over benchmarking algorithm in terms of rate performance and average execution time.

In this paper, we focus on simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) assisted multiuser multiple-input single-output (MISO) interference channels (ICs), where the RIS can serve the users in both forward and backward half-spaces at the same time by transmitting and reflecting incident signals. To suppress the inevitable inter-user interference, we adopt improper Gaussian signaling (IGS) and propose to jointly optimize the beamforming vectors of APs and the passive transmission and reflection coefficients of RIS to maximize the minimum achievable information rate among these users. In order to address the non-convex optimization problem, we provide an efficient iterative optimization algorithm to attain high-quality solutions. Numerical results demonstrate the superiority of IGS over conventional proper Gaussian signaling in STAR-RIS assisted MU-MISO ICs.