Integrating Movable Antennas and Intelligent Reflecting Surfaces (MA-IRS): Fundamentals, Practical Solutions, and Opportunities

Movable antennas (MAs) and intelligent reflecting surfaces (IRSs) enable active antenna repositioning and passive phase-shift tuning for channel reconfiguration, respectively. Integrating MAs and IRSs boosts spatial degrees of freedom, significantly enhancing wireless network capacity, coverage, and reliability. In this article, we first present the fundamentals of MA-IRS integration, involving clarifying the key design issues, revealing performance gain, and identifying the conditions where MA-IRS synergy persists. Then, we examine practical challenges and propose pragmatic design solutions, including optimization schemes, hardware architectures, deployment strategies, and robust designs for hardware impairments and mobility management. In addition, we highlight how MA-IRS architectures uniquely support advanced integrated sensing and communication, enhancing sensing performance and dual-functional flexibility. Overall, MA-IRS integration emerges as a compelling approach toward next-generation reconfigurable wireless systems.

Teochew-Wild: The First In-the-wild Teochew Dataset with Orthographic Annotations

This paper reports the construction of the Teochew-Wild, a speech corpus of the Teochew dialect. The corpus includes 18.9 hours of in-the-wild Teochew speech data from multiple speakers, covering both formal and colloquial expressions, with precise orthographic and pinyin annotations. Additionally, we provide supplementary text processing tools and resources to propel research and applications in speech tasks for this low-resource language, such as automatic speech recognition (ASR) and text-to-speech (TTS). To the best of our knowledge, this is the first publicly available Teochew dataset with accurate orthographic annotations. We conduct experiments on the corpus, and the results validate its effectiveness in ASR and TTS tasks.

Throughput Maximization for Movable Antenna Systems with Movement Delay Consideration

In this paper, we model the minimum achievable throughput within a transmission block of restricted duration and aim to maximize it in movable antenna (MA)-enabled multiuser downlink communications. Particularly, we account for the antenna moving delay caused by mechanical movement, which has not been fully considered in previous studies, and reveal the trade-off between the delay and signal-to-interference-plus-noise ratio at users. To this end, we first consider a single-user setup to analyze the necessity of antenna movement. By quantizing the virtual angles of arrival, we derive the requisite region size for antenna moving, design the initial MA position, and elucidate the relationship between quantization resolution and moving region size. Furthermore, an efficient algorithm is developed to optimize MA position via successive convex approximation, which is subsequently extended to the general multiuser setup. Numerical results demonstrate that the proposed algorithms outperform fixed-position antenna schemes and existing ones without consideration of movement delay. Additionally, our algorithms exhibit excellent adaptability and stability across various transmission block durations and moving region sizes, and are robust to different antenna moving speeds. This allows the hardware cost of MA-aided systems to be reduced by employing low rotational speed motors.

Movable Antennas Enabled Wireless-Powered NOMA: Continuous and Discrete Positioning Designs

This paper investigates a movable antenna (MA)-enabled wireless-powered communication network (WPCN), where multiple wireless devices (WDs) first harvest energy from the downlink (DL) signal broadcast by a hybrid access point (HAP) and then transmit information in the uplink (UL) using non-orthogonal multiple access. Unlike conventional WPCNs with fixed-position antennas (FPAs), this MA-enabled WPCN allows the MAs at the HAP and the WDs to adjust their positions twice: once before DL wireless power transfer and once before DL wireless information transmission. Our goal is to maximize the system sum throughput by jointly optimizing the MA positions, the time allocation, and the UL power allocation. Considering the characteristics of antenna movement, we explore both continuous and discrete positioning designs, which, after formulation, are found to be non-convex optimization problems. Before tackling these problems, we rigorously prove that using identical MA positions for both DL and UL is the optimal strategy in both scenarios, thereby greatly simplifying the problems and enabling easier practical implementation of the system. We then propose alternating optimization-based algorithms for the resulting simplified problems. Simulation results show that: 1) the proposed continuous MA scheme can enhance the sum throughput by up to 395.71% compared to the benchmark with FPAs, even when additional compensation transmission time is provided to the latter; 2) a step size of one-quarter wavelength for the MA motion driver is generally sufficient for the proposed discrete MA scheme to achieve over 80% of the sum throughput performance of the continuous MA scheme; 3) when each moving region is large enough to include multiple optimal positions for the continuous MA scheme, the discrete MA scheme can achieve comparable sum throughput without requiring an excessively small step size.

Human-like object concept representations emerge naturally in multimodal large language models

The conceptualization and categorization of natural objects in the human mind have long intrigued cognitive scientists and neuroscientists, offering crucial insights into human perception and cognition. Recently, the rapid development of Large Language Models (LLMs) has raised the attractive question of whether these models can also develop human-like object representations through exposure to vast amounts of linguistic and multimodal data. In this study, we combined behavioral and neuroimaging analysis methods to uncover how the object concept representations in LLMs correlate with those of humans. By collecting large-scale datasets of 4.7 million triplet judgments from LLM and Multimodal LLM (MLLM), we were able to derive low-dimensional embeddings that capture the underlying similarity structure of 1,854 natural objects. The resulting 66-dimensional embeddings were found to be highly stable and predictive, and exhibited semantic clustering akin to human mental representations. Interestingly, the interpretability of the dimensions underlying these embeddings suggests that LLM and MLLM have developed human-like conceptual representations of natural objects. Further analysis demonstrated strong alignment between the identified model embeddings and neural activity patterns in many functionally defined brain ROIs (e.g., EBA, PPA, RSC and FFA). This provides compelling evidence that the object representations in LLMs, while not identical to those in the human, share fundamental commonalities that reflect key schemas of human conceptual knowledge. This study advances our understanding of machine intelligence and informs the development of more human-like artificial cognitive systems.

Joint Transmitter and Receiver Design for Movable Antenna Enhanced Multicast Communications

Movable antenna (MA) is an emerging technology that utilizes localized antenna movement to pursue better channel conditions for enhancing communication performance. In this paper, we study the MA-enhanced multicast transmission from a base station equipped with multiple MAs to multiple groups of single-MA users. Our goal is to maximize the minimum weighted signal-to-interference-plus-noise ratio (SINR) among all the users by jointly optimizing the position of each transmit/receive MA and the transmit beamforming. To tackle this challenging problem, we first consider the single-group scenario and propose an efficient algorithm based on the techniques of alternating optimization and successive convex approximation. Particularly, when optimizing transmit or receive MA positions, we construct a concave lower bound for the signal-to-noise ratio (SNR) of each user by applying only the second-order Taylor expansion, which is more effective than existing works utilizing two-step approximations. The proposed design is then extended to the general multi-group scenario. Simulation results demonstrate that significant performance gains in terms of achievable max-min SNR/SINR can be obtained by our proposed algorithm over benchmark schemes. Additionally, the proposed algorithm can notably reduce the required amount of transmit power or antennas for achieving a target level of max-min SNR/SINR performance compared to benchmark schemes.

Beyond Finite Data: Towards Data-free Out-of-distribution Generalization via Extrapolation

Out-of-distribution (OOD) generalization is a favorable yet challenging property for deep neural networks. The core challenges lie in the limited availability of source domains that help models learn an invariant representation from the spurious features. Various domain augmentation have been proposed but largely rely on interpolating existing domains and frequently face difficulties in creating truly "novel" domains. Humans, on the other hand, can easily extrapolate novel domains, thus, an intriguing question arises: How can neural networks extrapolate like humans and achieve OOD generalization? We introduce a novel approach to domain extrapolation that leverages reasoning ability and the extensive knowledge encapsulated within large language models (LLMs) to synthesize entirely new domains. Starting with the class of interest, we query the LLMs to extract relevant knowledge for these novel domains. We then bridge the gap between the text-centric knowledge derived from LLMs and the pixel input space of the model using text-to-image generation techniques. By augmenting the training set of domain generalization datasets with high-fidelity, photo-realistic images of these new domains, we achieve significant improvements over all existing methods, as demonstrated in both single and multi-domain generalization across various benchmarks. With the ability to extrapolate any domains for any class, our method has the potential to learn a generalized model for any task without any data. To illustrate, we put forth a much more difficult setting termed, data-free domain generalization, that aims to learn a generalized model in the absence of any collected data. Our empirical findings support the above argument and our methods exhibit commendable performance in this setting, even surpassing the supervised setting by approximately 1-2\% on datasets such as VLCS.

IRS-Aided Multi-Antenna Wireless Powered Communications in Interference Channels

This paper investigates intelligent reflecting surface (IRS)-aided multi-antenna wireless powered communications in a multi-link interference channel, where multiple IRSs are deployed to enhance the downlink/uplink communications between each pair of hybrid access point (HAP) and wireless device. Our objective is to maximize the system sum throughput by optimizing the allocation of communication resources. To attain this objective and meanwhile balance the performance-cost tradeoff, we propose three transmission schemes: the IRS-aided asynchronous (Asy) scheme, the IRS-aided time-division multiple access (TDMA) scheme, and the IRS-aided synchronous (Syn) scheme. For the resulting three non-convex design problems, we propose a general algorithmic framework capable of addressing all of them. Numerical results show that our proposed IRS-aided schemes noticeably surpass their counterparts without IRSs in both system sum throughput and total transmission energy consumption at the HAPs. Moreover, although the IRS-aided Asy scheme consistently achieves the highest sum throughput, the IRS-aided TDMA scheme is more appealing in scenarios with substantial cross-link interference and limited IRS elements, while the IRS-aided Syn scheme is preferable in low cross-link interference scenarios.

Understanding Adversarial Transferability in Federated Learning

We investigate the robustness and security issues from a novel and practical setting: a group of malicious clients has impacted the model during training by disguising their identities and acting as benign clients, and only revealing their adversary position after the training to conduct transferable adversarial attacks with their data, which is usually a subset of the data that FL system is trained with. Our aim is to offer a full understanding of the challenges the FL system faces in this practical setting across a spectrum of configurations. We notice that such an attack is possible, but the federated model is more robust compared with its centralized counterpart when the accuracy on clean images is comparable. Through our study, we hypothesized the robustness is from two factors: the decentralized training on distributed data and the averaging operation. We provide evidence from both the perspective of empirical experiments and theoretical analysis. Our work has implications for understanding the robustness of federated learning systems and poses a practical question for federated learning applications.

Diverse Cotraining Makes Strong Semi-Supervised Segmentor

Deep co-training has been introduced to semi-supervised segmentation and achieves impressive results, yet few studies have explored the working mechanism behind it. In this work, we revisit the core assumption that supports co-training: multiple compatible and conditionally independent views. By theoretically deriving the generalization upper bound, we prove the prediction similarity between two models negatively impacts the model's generalization ability. However, most current co-training models are tightly coupled together and violate this assumption. Such coupling leads to the homogenization of networks and confirmation bias which consequently limits the performance. To this end, we explore different dimensions of co-training and systematically increase the diversity from the aspects of input domains, different augmentations and model architectures to counteract homogenization. Our Diverse Co-training outperforms the state-of-the-art (SOTA) methods by a large margin across different evaluation protocols on the Pascal and Cityscapes. For example. we achieve the best mIoU of 76.2%, 77.7% and 80.2% on Pascal with only 92, 183 and 366 labeled images, surpassing the previous best results by more than 5%.