Xichen Zhang, Ziyi He, Yinghao Zhu, Sitong Wu, Shaozuo Yu, Meng Chu, Wenhu Zhang, Haoru Tan, Jiaya Jia. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Urban Smart Sensing Devices (SSDs) require timely and secure code updates to maintain reliable operation. However, trust verification in vehicle-assisted code dissemination remains insufficiently explored, making systems vulnerable to malicious vehicles that distribute corrupted or outdated code. This paper proposes a Trust-driven Reliable Code Dissemination (TRCD) scheme, where mobile vehicles opportunistically act as “code mules” for efficient and secure dissemination. TRCD employs a two-layer trust evaluation mechanism: (i) selected anchor SSDs evaluate local vehicle trust in real time, and (ii) the code center constructs global trust by aggregating UAV-verified and Local High-Trust Vehicle (LHTV) reported reputation information to identify trustworthy vehicles and exclude malicious ones. To improve dissemination efficiency, we further design (i) an anchor SSD selection algorithm for cost-efficient trust collection, (ii) a token-based mechanism for prioritizing high-value tokens under storage constraints, and (iii) a genetic algorithm-based UAV trajectory optimization method for trust-aware dissemination. We also provide theoretical analysis on convergence behavior, computational complexity, and performance bounds. Extensive simulations using real-world taxi trajectory datasets show that TRCD achieves 98.2% trusted-vehicle identification accuracy, 99.4% malicious-vehicle detection accuracy, and reduces UAV flight distance by 64.5% compared with baseline methods, while significantly improving dissemination reliability. These results demonstrate that TRCD provides an efficient, scalable, and secure solution for urban SSD code dissemination.

Recruiting trust workers to achieve high data quality at low cost has become a promising approach in Mobile Crowdsensing (MCS). However, most existing trust evaluation methods only adopt a single-dimensional trust model, neglecting the fact that a worker's trustworthiness can vary across different task types, which leads to suboptimal task–worker matching, poor data quality, and inefficient cost utilization. To this end, we propose a Multidimensional Trust Evaluation and Task Matching (MTE-TM) based workers recruitment scheme to improve data quality while reducing costs for MCS. First, we represent worker trustworthiness by a composite of expected trust values and variance, enabling task-specific trust assessment that extends traditional single-dimensional trust to a multidimensional domain. A novel Expectation-Maximization (EM)-based trust evaluation mechanism is also introduced to improve accuracy. Second, we design a new worker selection method that combines a worker's trust level and the width of the confidence interval to compute their Upper Confidence Bound (UCB) index, which effectively guides worker selection toward optimal outcomes. Third, we propose an Optimized Data Quality Matching (ODQM) algorithm that assigns tasks to workers with high priority and low bid prices under budget constraints, thereby further improving data quality. The experimental results demonstrate the significant performance improvements of our scheme, achieving 45.54$\sim$95.55% optimization in trust evaluation, 65.01$\sim$72.95% improvement in data quality, and a notable reduction in regret.

Cervical spondylosis, a complex and prevalent condition, demands precise and efficient diagnostic techniques for accurate assessment. While MRI offers detailed visualization of cervical spine anatomy, manual interpretation remains labor-intensive and prone to error. To address this, we developed an innovative AI-assisted Expert-based Diagnosis System that automates both segmentation and diagnosis of cervical spondylosis using MRI. Leveraging multi-center datasets of cervical MRI images from patients with cervical spondylosis, our system features a pathology-guided segmentation model capable of accurately segmenting key cervical anatomical structures. The segmentation is followed by an expert-based diagnostic framework that automates the calculation of critical clinical indicators. Our segmentation model achieved an impressive average Dice coefficient exceeding 0.90 across four cervical spinal anatomies and demonstrated enhanced accuracy in herniation areas. Diagnostic evaluation further showcased the system's precision, with the lowest mean average errors (MAE) for the C2-C7 Cobb angle and the Maximum Spinal Cord Compression (MSCC) coefficient. In addition, our method delivered high accuracy, precision, recall, and F1 scores in herniation localization, K-line status assessment, T2 hyperintensity detection, and Kang grading. Comparative analysis and external validation demonstrate that our system outperforms existing methods, establishing a new benchmark for segmentation and diagnostic tasks for cervical spondylosis.

Many studies on trust relationship establishment in Social Networks (SNs) have assumed that the trustworthiness of partners can be identified by participants through interaction. However, in practice, participants not only struggle to discern the trustworthiness of their counterparts but also find it difficult to effectively determine whether the messages they spread are Useful Messages (UMs) or Malicious Messages (MMs). Therefore, designing an efficient information propagation scheme that promotes UMs dissemination while blocking MMs remains a challenging issue in real-world SNs. In this paper, we propose an efficient Trust-based Distributed Incentive (TDI) scheme that aligns with actual SN practices. First, an effective Bidirectional Trust Identification (BTI) approach is proposed to verify the trustworthiness of messages and participants without assuming that interacting participants can evaluate each other's trustworthiness. In BTI, the trustworthiness of participants is evaluated based on their evaluations of trusted participants and reliable messages, while the trust of messages is verified through feedback from trusted participants, laying a foundation for trust information propagation. Then, a Trust-based Message Forwarding (TMF) mechanism is proposed to facilitate the dissemination of trusted messages while blocking the forwarding of low-trust messages. Finally, a Proactive Trust Evaluation (PTE) mechanism is introduced to accelerate and effectively obtain participants' reliable evaluations. Specifically, some UMs are disseminated as Probing Messages (PMs) to accurately evaluate the trustworthiness of participants based on whether they evaluate them truthfully. Extensive simulations demonstrate that the TDI scheme outperforms the existing main schemes in terms of accurately identifying message trust, increasing UMs dissemination, blocking the spread of MMs, and purifying SNs.

Orchestration of Data as Services (ODS) at the Edge Layer (EL) in a Cloud-Edge-Seamless System (CESS) facilitates user access and avoids the long-distance transmission of massive raw data to Cloud Servers (CSs), thereby reducing network load. Building on this foundation, we argue that deploying both services and Data/Services Orchestration Programs (DSOPs), together with performing resource allocation at Edge Servers (ESs), further minimizing data processing time, service response time, and placement costs. To this end, we propose a novel service network architecture that integrates Service/DSOP placement and edge resource allocation to enhance overall system performance. First, a new service network architecture is proposed to jointly optimize Service/DSOP placement and resource allocation. Then, we design a Data-Driven Service/DSOP Placement (DDSDP) scheme that employs a Parameterized Deep$Q$-Network (P-DQN) to effectively tackle the hybrid action space optimization of such joint problem. Moreover, we develop a Demand-Calibrated Service/DSOP Placement (DCSDP) approach, which first leverages Long Short-Term Memory (LSTM) networks at the CS to capture spatio-temporal patterns of data and service demands, enabling ESs to dynamically train a demand-aware service/DSOP deployment model. Extensive simulations demonstrate that DDSDP and DCSDP significantly reduce service response time and adapt more effectively to network dynamics, achieving reductions of 47.89% and 53.88% compared to the baseline IFSP and P-DQN methods, respectively.

Shusheng Li, Jiale Li, Yifei Qu, Xinwei Shi, Yanliang Guo, Ziyi He, Yubo Wang, Wenjun Tan. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Groupwise registration has been an essential technique for addressing the growing demands for batch processing of medical images collected from various sites or time points. It is vital to select or construct an atlas for a given group when performing groupwise registration. This paper proposes a novel atlas generation framework with a latent diffusion model that generates an atlas from noise by leveraging the latent vectors of all subjects in the group. By incorporating the attention mechanism, the model accommodates groups of arbitrary sizes and enhances scalability during the atlas generation. A minimum group construction module is introduced to create minimal group units and their corresponding atlases for training the diffusion model. Experimental results on brain MRI slices demonstrate that the proposed method can generate comparable or higher quality atlases compared to baselines while significantly reducing runtime. To conclude, the method effectively balances image quality, registration actuary, atlas unbiasedness, and computational efficiency.

The rapid advancement of Large Language Models (LLMs) has stimulated interest in multi-agent collaboration for addressing complex medical tasks. However, the practical advantages of multi-agent collaboration approaches remain insufficiently understood. Existing evaluations often lack generalizability, failing to cover diverse tasks reflective of real-world clinical practice, and frequently omit rigorous comparisons against both single-LLM-based and established conventional methods. To address this critical gap, we introduce MedAgentBoard, a comprehensive benchmark for the systematic evaluation of multi-agent collaboration, single-LLM, and conventional approaches. MedAgentBoard encompasses four diverse medical task categories: (1) medical (visual) question answering, (2) lay summary generation, (3) structured Electronic Health Record (EHR) predictive modeling, and (4) clinical workflow automation, across text, medical images, and structured EHR data. Our extensive experiments reveal a nuanced landscape: while multi-agent collaboration demonstrates benefits in specific scenarios, such as enhancing task completeness in clinical workflow automation, it does not consistently outperform advanced single LLMs (e.g., in textual medical QA) or, critically, specialized conventional methods that generally maintain better performance in tasks like medical VQA and EHR-based prediction. MedAgentBoard offers a vital resource and actionable insights, emphasizing the necessity of a task-specific, evidence-based approach to selecting and developing AI solutions in medicine. It underscores that the inherent complexity and overhead of multi-agent collaboration must be carefully weighed against tangible performance gains. All code, datasets, detailed prompts, and experimental results are open-sourced at this link.

Assigning tasks to reliable workers to obtain reliable data is a critical issue in Mobile CrowdSensing (MCS). The challenge is compounded by the problem of Information Elicitation Without Verification (IEWV), which renders traditional data quality evaluation methods ineffective. While some studies attempt to address this, they often struggle to assess workers’ dynamic trustworthiness, resulting in unreliable data. To overcome these challenges, we propose the Trust and Time-sharing Task Allocation based Truth Discovery (TTTA-TD) scheme, designed to ensure reliable data collection in MCS. This scheme includes three components: (a) Classification-based Trust Evaluation (CTE) that classifies workers based on behavior and applies tailored penalties—lenient for honest workers and stricter for malicious ones, (b) Trust-based Truth Data Discovery (TTDD), which improves truth data accuracy by integrating trust scores, and (c) Trust and Time-sharing Task Allocation (TTTA) which allocates tasks to ensure data reliability and minimize time-sharing disparities. Experimental results show that the TTTA algorithm reduces average time-sharing by 93.95%. The TTDD algorithm improves truth estimates across all dataset qualities, and the TTTA-TD scheme enhances data reliability by 0.35%, 2.06%, and 7.41% in high, medium, and low-quality datasets respectively.

Template generation is a critical step in groupwise image registration, which involves aligning a group of subjects into a common space. While existing methods can generate high-quality template images, they often incur substantial time costs or are limited by fixed group scales. In this paper, we present InstantGroup, an efficient groupwise template generation framework based on variational autoencoder (VAE) models that leverage latent representations' arithmetic properties, enabling scalability to groups of any size. InstantGroup features a Dual VAE backbone with shared-weight twin networks to handle pairs of inputs and incorporates a Displacement Inversion Module (DIM) to maintain template unbiasedness and a Subject-Template Alignment Module (STAM) to improve template quality and registration accuracy. Experiments on 3D brain MRI scans from the OASIS and ADNI datasets reveal that InstantGroup dramatically reduces runtime, generating templates within seconds for various group sizes while maintaining superior performance compared to state-of-the-art baselines on quantitative metrics, including unbiasedness and registration accuracy.

BACKGROUND: Clinical studies have shown that miRNAs are closely related to human health. The study of potential associations between miRNAs and diseases will contribute to a profound understanding of the mechanism of disease development, as well as human disease prevention and treatment. MiRNA-disease associations predicted by computational methods are the best complement to biological experiments. RESULTS: In this research, a federated computational model KATZNCP was proposed on the basis of the KATZ algorithm and network consistency projection to infer the potential miRNA-disease associations. In KATZNCP, a heterogeneous network was initially constructed by integrating the known miRNA-disease association, integrated miRNA similarities, and integrated disease similarities; then, the KATZ algorithm was implemented in the heterogeneous network to obtain the estimated miRNA-disease prediction scores. Finally, the precise scores were obtained by the network consistency projection method as the final prediction results. KATZNCP achieved the reliable predictive performance in leave-one-out cross-validation (LOOCV) with an AUC value of 0.9325, which was better than the state-of-the-art comparable algorithms. Furthermore, case studies of lung neoplasms and esophageal neoplasms demonstrated the excellent predictive performance of KATZNCP. CONCLUSION: A new computational model KATZNCP was proposed for predicting potential miRNA-drug associations based on KATZ and network consistency projections, which can effectively predict the potential miRNA-disease interactions. Therefore, KATZNCP can be used to provide guidance for future experiments.

Groupwise registration is an important and challenging task for medical image processing and analysis. Traditional methods focus on generating a template and performing pairwise registration, which can be time-consuming to converge. In this paper, we propose an unsupervised end-to-end group-wise registration framework with multi-step mechanisms to progressively refine outputs. The framework can generate the displacement field for each subject directly without templates. Customized loss functions are designed to optimize the model and reduce the bias of generated common space. We experiment on 2D brain MRI coronal slices from OASIS and compare the results with two baseline methods using Dice score criterion. Results show that our framework achieves state-of-the-art performance with a much lower time cost.