Multimodal learning emerges as a promising solution for high-precision localization, a cornerstone of 6G integrated sensing and communications (ISAC), by integrating measurements from different data sources. Yet its real-world deployment remains challenging because(i)the quality and relevance of different modalities fluctuate with frequency, noise, and antenna heterogeneity and(ii)spatial and fingerprint ambiguities under non-line-of-sight (NLOS) propagation obscure the mapping between channel measurements and positions. To overcome these challenges, we propose a spatial-context-aware dynamicfusion architecture built on the mixture-of-experts (SCADF-MoE) backbone. We first construct a million-scale comprehensive ray-tracing dataset measuring synchronized angle, distance, gain, and channel across diverse carrier frequencies, antenna geometries, and noise levels. A three-stage pre-processing pipeline then clusters neighboring points into short trajectories, enriching data samples with spatial context information. The resulting sequences are fed into SCADF-MoE: first, multimodal soft MoE blocks with learnable routing matrices dynamically fuse heterogeneous inputs according to their modality relevance in different environmental contexts; second, a modality-task MoE formulates position estimation as a multi-objective problem, simultaneously predicting coordinates of neighboring points to leverage their shared spatial correlations. Additionally, we introduce a regularization loss that enforces expert diversity and mitigates gradient conflicts during multi-task optimization. Simulations across three environments (dense-urban, suburban, canyon) and three heterogeneity dimensions (frequency, noise, antenna) demonstrate that SCADF-MoE achieves consistent sub-meter accuracy in all conditions, reducing overall MSE by 63%, and cuts unseen-NLOS error by 55% compared to state-of-the-art methods. To the best of our knowledge, this is the first work that leverages large-scale multimodal MoEs for high-precision ISAC localization.

Current ophthalmology clinical workflows are plagued by over-referrals, long waits, and complex and heterogeneous medical records. Large language models (LLMs) present a promising solution to automate various procedures such as triaging, preliminary tests like visual acuity assessment, and report summaries. However, LLMs have demonstrated significantly varied performance across different languages in natural language question-answering tasks, potentially exacerbating healthcare disparities in Low and Middle-Income Countries (LMICs). This study introduces the first multilingual ophthalmological question-answering benchmark with manually curated questions parallel across languages, allowing for direct cross-lingual comparisons. Our evaluation of 6 popular LLMs across 7 different languages reveals substantial bias across different languages, highlighting risks for clinical deployment of LLMs in LMICs. Existing debiasing methods such as Translation Chain-of-Thought or Retrieval-augmented generation (RAG) by themselves fall short of closing this performance gap, often failing to improve performance across all languages and lacking specificity for the medical domain. To address this issue, We propose CLARA (Cross-Lingual Reflective Agentic system), a novel inference time de-biasing method leveraging retrieval augmented generation and self-verification. Our approach not only improves performance across all languages but also significantly reduces the multilingual bias gap, facilitating equitable LLM application across the globe.

Text-to-video (T2V) generation models have made significant progress in creating visually appealing videos. However, they struggle with generating coherent sequential narratives that require logical progression through multiple events. Existing T2V benchmarks primarily focus on visual quality metrics but fail to evaluate narrative coherence over extended sequences. To bridge this gap, we present SeqBench, a comprehensive benchmark for evaluating sequential narrative coherence in T2V generation. SeqBench includes a carefully designed dataset of 320 prompts spanning various narrative complexities, with 2,560 human-annotated videos generated from 8 state-of-the-art T2V models. Additionally, we design a Dynamic Temporal Graphs (DTG)-based automatic evaluation metric, which can efficiently capture long-range dependencies and temporal ordering while maintaining computational efficiency. Our DTG-based metric demonstrates a strong correlation with human annotations. Through systematic evaluation using SeqBench, we reveal critical limitations in current T2V models: failure to maintain consistent object states across multi-action sequences, physically implausible results in multi-object scenarios, and difficulties in preserving realistic timing and ordering relationships between sequential actions. SeqBench provides the first systematic framework for evaluating narrative coherence in T2V generation and offers concrete insights for improving sequential reasoning capabilities in future models. Please refer to https://videobench.github.io/SeqBench.github.io/ for more details.

Multiple Sclerosis (MS) disrupts nerve communication, potentially leading to permanent damage. Convolutional Neural Networks (CNNs) are commonly recommended to accelerate magnetic resonance imaging (MRI) analysis for MS. Traditional CNN-based methods often face challenges with feature selection, imbalanced classification, and hyperparameter optimization, which can reduce their effectiveness. To address these challenges, this paper introduces a reinforcement learning (RL) strategy for tackling feature selection and class imbalance, complemented by an advanced Differential Evolution (DE) algorithm for hyperparameter tuning. The RL algorithm dynamically adapts, enhancing feature selection and class balance, and consistently outperforms traditional static methods by continually optimizing its response to the complex patterns in MS detection. Furthermore, the DE is enhanced with a novel mutation mechanism that utilizes k-means clustering to identify significant clusters. Tested on the SNAC-MS and ISBI datasets, our model demonstrates robust accuracy with average F-measures of 92.873% and 93.783%, respectively, highlighting its effectiveness and adaptability in diagnosing MS.

Multi-modal multi-task learning holds significant promise in tackling complex diagnostic tasks and many significant medical imaging problems. It fulfills the needs in real-world diagnosis protocol to leverage information from different data sources and simultaneously perform mutually informative tasks. However, medical imaging domains introduce two key challenges: dynamic modality fusion and modality-task dependence. The quality and amount of task-related information from different modalities could vary significantly across patient samples, due to biological and demographic factors. Traditional fusion methods apply fixed combination strategies that fail to capture this dynamic relationship, potentially underutilizing modalities that carry stronger diagnostic signals for specific patients. Additionally, different clinical tasks may require dynamic feature selection and combination from various modalities, a phenomenon we term “modality-task dependence.” To address these issues, we propose M4oE, a novel Multi-modal Multi-task Mixture of Experts framework for precise Medical diagnosis. M4oE comprises Modality-Specific (MSoE) modules and a Modality-shared Modality-Task MoE (MToE) module. With collaboration from both modules, our model dynamically decomposes and learns distinct and shared information from different modalities and achieves dynamic fusion. MToE provides a joint probability model of modalities and tasks by using experts as a link and encourages experts to learn modality-task dependence via conditional mutual information loss. By doing so, M4oE offers sample and population-level interpretability of modality contributions. We evaluate M4oE on four public multi-modal medical benchmark datasets for solving two important medical diagnostic problems including breast cancer screening and retinal disease diagnosis. Results demonstrate our method's superiority over state-of-the-art methods under different metrics of classification and segmentation tasks like Accuracy, AUROC, AUPRC, and DICE.

Mental disorders are among the most widespread diseases globally. Analyzing functional brain networks through functional magnetic resonance imaging (fMRI) is crucial for understanding mental disorder behaviors. Although existing fMRI-based graph neural networks (GNNs) have demonstrated significant potential in brain network feature extraction, they often fail to characterize complex relationships between brain regions and demographic information in mental disorders. To overcome these limitations, we propose a learnable NeuroTree framework that integrates a $k$-hop AGE-GCN with neural ordinary differential equations (ODEs) and contrastive masked functional connectivity (CMFC) to enhance similarities and dissimilarities of brain region distance. Furthermore, NeuroTree effectively decodes fMRI network features into tree structures, which improves the capture of high-order brain regional pathway features and enables the identification of hierarchical neural behavioral patterns essential for understanding disease-related brain subnetworks. Our empirical evaluations demonstrate that NeuroTree achieves state-of-the-art performance across two distinct mental disorder datasets. It provides valuable insights into age-related deterioration patterns, elucidating their underlying neural mechanisms. The code and datasets are available at https://github.com/Ding1119/NeuroTree.

With the rapid development of the Visual Internet of Things (VIoT), person re-Identification (ReID) technology has made significant strides in research, particularly in the domains of urban security and intelligent surveillance. However, cloth-changing person re-identification (CC-ReID) poses significant challenges to traditional appearance-based methods due to frequent changes in clothing. To address this issue, this paper proposes a Person Parsing-Driven and Text-Guided for Cloth-Changing Person Re-identification (PT-ReID). This approach optimizes the image encoder of the CLIP model through a multi-branch design and leverages person parsing techniques to extract stable, clothing-invariant biological features. Additionally, the method incorporates Context Optimization (CoOp) technology, using learnable text descriptions as soft supervision to enhance cross-modal alignment. A saliency region loss mechanism is also introduced to minimize discrepancies between different feature representations, further improving the model’s ability to learn discriminative pedestrian features. Experimental results on three public datasets—PRCC, LTCC, and VC-Clothes—demonstrate that PT-ReID significantly outperforms state-of-the-art methods in Rank-1 accuracy and mean Average Precision (mAP) under cloth-changing scenarios, proving its effectiveness and robustness in handling complex clothing variations.

AI-based DR screening is promising in low- and middle-income countries (LMICs), where limited human resources constrain access to specialist-led programs. However, current systems often degrade under real-world image-quality variations, especially with portable devices that are vital for low- and middle-income countries. This study aims to develop Retsyn, a synthetic-data augmentation framework that improves screening robustness across devices and imaging conditions. RetSyn leverages advanced diffusion models to generate synthetic retinal images with diverse device and imaging quality characteristics. To address the challenges of (1) portable device data scarcity, (2) disease and quality distribution imbalance, and (3) varying image quality, RetSyn uses class and quality-conditioned diffusion for controllable synthesis, a group-balanced loss to increase coverage of minority (quality, disease) pairs, and a Direct Preference Optimization alignment step with a small paired smartphone–tabletop set. The synthesized images are then used to augment classifier training. The effectiveness of RetSyn-generated images was evaluated by training retinal diagnosis models on a combination of real and synthetic data. RetSyn yields consistent gains in-domain and out-of-domain. On low-quality tabletop images, F1 improves from 0.781 to 0.874 (binary) and 0.607 to 0.703 (three-class), while AUROC reaches 0.982 and 0.951, respectively. On out-of-domain portable images, RetSyn attains AUROC 0.813/F1 0.703 (binary) and AUROC 0.804/F1 0.609 (three-class), exceeding group-robustness baselines such as GroupDRO (binary: AUROC 0.786/F1 0.626; three-class: AUROC 0.789/F1 0.544). RetSyn presents an effective and scalable synthetic data framework that significantly enhances the robustness and generalizability of AI-based DR screening models in LMICs. By addressing the critical challenges posed by varying image quality and device characteristics, RetSyn facilitates more reliable deployment of AI diagnostics in underserved regions. Additionally, the release of the first publicly available paired smartphone-tabletop retinal image dataset will support further research into cross-device DR screening solutions. • RetSyn generates synthetic retinal images to enhance AI robustness across varying image qualities and devices. • Group-balanced training and preference optimization enable diverse medical image synthesis with minimal cross-device paired data. • Releases the first publicly available paired smartphone-tabletop retinal dataset. • Performance improved significantly on low-quality and portable device photographs. • Achieves first clinically acceptable results for smartphone-based DR screening in LMICs.

Electronic health records (EHRs) are multimodal by nature, consisting of structured tabular features like lab tests and unstructured clinical notes. In real-life clinical practice, doctors use complementary multimodal EHR data sources to get a clearer picture of patients' health and support clinical decision-making. However, most EHR predictive models do not reflect these procedures, as they either focus on a single modality or overlook the inter-modality interactions/redundancy. In this work, we propose MEDFuse, a Multimodal EHR Data Fusion framework that incorporates masked lab-test modeling and large language models (LLMs) to effectively integrate structured and unstructured medical data. MEDFuse leverages multimodal embeddings extracted from two sources: LLMs fine-tuned on free clinical text and masked tabular transformers trained on structured lab test results. We design a disentangled transformer module, optimized by a mutual information loss to 1) decouple modality-specific and modality-shared information and 2) extract useful joint representation from the noise and redundancy present in clinical notes. Through comprehensive validation on the public MIMIC-III dataset and the in-house FEMH dataset, MEDFuse demonstrates great potential in advancing clinical predictions, achieving over 90% F1 score in the 10-disease multi-label classification task.