Siming Huang, Tianhao Cheng, Jason Klein Liu, Weidi Xu, Jiaran Hao, Liuyihan Song, Yang Xu, Jian Yang, Jiaheng Liu, Chenchen Zhang, Linzheng Chai, Ruifeng Yuan, Xianzhen Luo, Qiufeng Wang, YuanTao Fan, Qingfu Zhu, Zhaoxiang Zhang, Yang Gao, Jie Fu, Qian Liu, Houyi Li, Ge Zhang, Yuan Qi, Xu Yinghui, Wei Chu, Zili Wang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

This paper investigates a novel anti-eavesdropping strategy based on unmanned aerial vehicle (UAV)-mounted simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS). In particular, a UAV equipped with a STAR-RIS acts as a passive relay to reflect desired signals and simultaneously acts as a friendly jammer to transmit artificial noise (AN) against eavesdroppers. Based on the phase shift coupling characteristics of STAR-RIS, three phase shift configuration strategies are proposed, namely reliability-priority (RP), security-priority (SP), and element-partitioning (EP) schemes. Analytical closed-form expressions of connection outage probability (COP), secrecy outage probability (SOP), effective secrecy throughput (EST) and secrecy energy efficiency (SEE) are derived to evaluate the reliable and secure performance achieved by the proposed schemes, respectively. The asymptotic analysis is also performed for further insights. Analysis and simulation results demonstrate that the proposed three schemes outperform traditional benchmark schemes. From the perspective of reliability, the RP scheme can achieve the best COP. In terms of security, as the number of STAR-RIS elements increases, the SOPs of the SP and EP exponentially decrease, whereas the SOP of the RP scheme increases. The EP scheme achieves the optimal EST, and the asymptotic EST is independent of phase estimation errors. Additionally, it is recommended that the UAV be deployed near the eavesdropper for the SP and EP schemes to enhance SEE.

Pediatric pneumonia is the leading cause of death among children under five years worldwide, imposing a substantial burden on affected families. Currently, there are three significant hurdles in diagnosing and treating pediatric pneumonia. Firstly, pediatric pneumonia shares similar symptoms with other respiratory diseases, making rapid and accurate differential diagnosis challenging. Secondly, primary hospitals often lack sufficient medical resources and experienced doctors. Lastly, providing personalized diagnostic reports and treatment recommendations is labor-intensive and time-consuming. To tackle these challenges, we proposed a Medical Multimodal Large Language Model for Pediatric Pneumonia (P2Med-MLLM). It was capable of handling diverse clinical tasks-such as generating free-text medical records and radiology reports-within a unified framework. Specifically, P2Med-MLLM was trained on a large-scale dataset, including real clinical information from 163,999 outpatient and 8,684 inpatient cases. It can process both plain text data (e.g., outpatient and inpatient records) and interleaved image-text pairs (e.g., 2D chest X-ray images, 3D chest Computed Tomography images, and corresponding radiology reports). We designed a three-stage training strategy to enable P2Med-MLLM to comprehend medical knowledge and follow instructions for various clinical decision-support tasks. To rigorously evaluate P2Med-MLLM's performance, we conducted automatic scoring by the large language model and manual scoring by the specialist on the test set of 642 samples, meticulously verified by pediatric pulmonology specialists. The results demonstrated the reliability of automated scoring and the superiority of P2Med-MLLM. This work plays a crucial role in assisting doctors with prompt diagnosis and treatment planning, reducing severe symptom mortality rates, and optimizing the allocation of medical resources.

Although conventional false data injection attacks can circumvent the detection of bad data detection (BDD) in sustainable power grid cyber physical systems, they are easily detected by well-trained deep learning-based detectors. Still, state estimation models with deep leaning-based detectors are not secure due to the vulnerabilities and fragility of deep learning models. Using the related laws of conventional false data injection attacks and adversarial sample attacks, this paper proposes the targEted adVersarial fAlse Data injEction (EVADE) strategy to explore targeted adversarial false data injection attacks for state estimation in Smart Grid. The proposed EVADE attack strategy selects key state variables based on adversarial saliency maps to improve the attack efficiency and perturbs as few state variables as possible to reduce the attack cost. In this way, the EVADE attack strategy can bypass the detection of BDD and neural attack detection (NAD) methods (that is, maintaining deep stealthy) with a high success rate and achieve the attack target simultaneously. Experimental results demonstrate the effectiveness of the proposed strategy, posing serious and pressing concerns for sustainable cyber physical power system security.

Multimodal Question Answering (MMQA) is crucial as it enables comprehensive understanding and accurate responses by integrating insights from diverse data representations such as tables, charts, and text. Most existing researches in MMQA only focus on two modalities such as image-text QA, table-text QA and chart-text QA, and there remains a notable scarcity in studies that investigate the joint analysis of text, tables, and charts. In this paper, we present CT2C-QA, a pioneering Chinese reasoning-based QA dataset that includes an extensive collection of text, tables, and charts, meticulously compiled from 200 selectively sourced webpages. Our dataset simulates real webpages and serves as a great test for the capability of the model to analyze and reason with multimodal data, because the answer to a question could appear in various modalities, or even potentially not exist at all. Additionally, we present AED (Allocating, Expert and Decision), a multi-agent system implemented through collaborative deployment, information interaction, and collective decision-making among different agents. Specifically, the Assignment Agent is in charge of selecting and activating expert agents, including those proficient in text, tables, and charts. The Decision Agent bears the responsibility of delivering the final verdict, drawing upon the analytical insights provided by these expert agents. We execute a comprehensive analysis, comparing AED with various state-of-the-art models in MMQA, including GPT-4. The experimental outcomes demonstrate that current methodologies, including GPT-4, are yet to meet the benchmarks set by our dataset.

As a promising technique for the next generation communication network, integrated sensing and communication (ISAC) has attracted incremental research attentions due to its capabilities in spectrum sharing, cost saving, and data collecting. In this paper we utilize unmanned aerial vehicle (UAV) swarm to perform downlink ISAC transmission to serve multiple terrestrial legitimate users and sensing targets. To accommodate more practical application scenarios, we assume that there are also multiple malicious eavesdroppers in the network attempting to eavesdrop on the confidential signal. In order to enhance the security of the downlink transmission while maintaining sufficient sensing performance, we propose a joint optimization of the centralized trajectory of UAV swarm, the transmit beamforming on each UAV, and the ISAC schedule, which is eventually formulated as an average secrecy rate (ASR) maximization problem. A deep reinforcement learning (DRL) based algorithm is developed to solve the considered optimization problem and its effectiveness is validated via experimental simulations, which also proves its superiority over benchmark methods.

We investigate the secrecy transmission of uplink non-orthogonal multiple access (NOMA) with the aid of energy harvesting (EH) jammers. During each time frame, communication is divided into two phases. At the first phase, the base station (BS) transfers wireless power to EH receivers (EHRs). At the second phase, users perform uplink NOMA transmission to BS, while one of EHRs is selected as a friendly jammer that uses the energy harvested from the previous phase to emit the artificial noise for confusing the eavesdropper. In terms of the requirement of channel state information (CSI), we propose three friendly EH jammer selection schemes, namely random EH jammer selection (REJS) scheme without the requirement of any CSI, maximal EH jammer selection (MEJS) scheme with the CSI between BS and each EHR, and optimal EH jammer selection (OEJS) scheme where both the CSIs from BS to EHRs and from EHRs to the eavesdropper need to be known. Analytical closed-form expressions for the connection outage probability (COP), secrecy outage probability (SOP) and effective secrecy throughput (EST) are derived to evaluate the system performance achieved by the proposed schemes, respectively. Also, the asymptotic analysis is provided to gain further insights. The analytical and numerical results indicate that the proposed schemes can realize better secrecy performance than conventional scheme without an EH jammer. Both the secrecy diversity orders of the REJS and MEJS schemes are one while the OEJS scheme can achieve a full secrecy diversity order. Furthermore, owing to the impact of connection outage, the three schemes converge to the same EST floor with the increase of signal-to-noise ratio (SNR).