The rise of vehicle-infrastructure (V2I) collaboration enables safer and broader perception. To process large-scale V2I video streams, vision-language models (VLMs) are promising as they unify multi-view vision into end-to-end task grounding, reducing handcrafted design. We use Vision Mixture-of-Experts (V-MoE) as the distributed visual backbone of VLMs, leveraging sparse expert routing to enable conditional computation across diverse viewpoints under resource constraints. Yet, V-MoEs face a critical challenge: large-scale data shifts over minutes to hours in V2I systems, amplified by agnostic participants and biased features propagating through experts. To maintain accuracy efficiently, we find it beneficial to co-adapt multiple V-MoEs on edge servers, avoiding the latency and privacy risks of cloud offloading and the accuracy sacrifices of on-device methods. However, the resource-constrained edge poses challenges for efficient co-adaptation: i) DRAM fragmentation and imbalance limit expert parallelism, ii) memory-I/O bottlenecks restrict computation reuse, and iii) asynchronous adaptation increases task-switch overhead. Also, prior work rarely explores the upper bound of concurrent tasks under limited edge resources, a critical factor for practical V2I deployment. To address these, we present AdaSprite. By combining cooperative elastic scaling with multi-level multiplexing, AdaSprite optimizes expert lifespans to reduce DRAM fragmentation, exploits predictable activation patterns for efficient I/O reuse, and employs twin-buffer scheduling to leverage sparsity. On a weak edge, AdaSprite supports up to 17 concurrent V2I tasks (vs. up to 6 for baselines), improving SLO attainment by 1.6x and throughput by 2.1x. Also, it allows users to trade accuracy and concurrency for second-level adaptation.

We propose a Cross-Modal Contrastive Learning (CMCL) framework that integrates multimodal MRI features extracted from resting-state functional magnetic resonance imaging (Rs-fMRI) and voxel-based morphometry (VBM), aiming to identify drug addiction circuits in the brain. To enhance structural information modeling, we introduce, for the first time, a structural covariance graph constructed based on the cosine similarity of gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) volumes. This graph is combined with a functional connectivity graph derived from traditional Pearson correlation, enabling unified modeling of functional and structural information. CMCL aligns the latent representations of functional and structural modalities through a contrastive learning mechanism, effectively capturing multiscale brain network abnormalities associated with addiction and revealing functional-structural coupling changes in reward and executive control circuits. To address the inadequate interaction modeling in traditional multimodal approaches, we design a unified feature embedding module, an improved InfoNCE loss function, and a complementary loss to mitigate cross-modal heterogeneity and enhance cross-modal complementarity. Experimental results on a drug addiction dataset demonstrate that CMCL significantly outperforms existing methods in classification performance. Moreover, the learned subgraph structures closely match clinical findings, accurately identifying key brain regions and pathways related to drug addiction. These results highlight the functional-structural co-alteration patterns induced by addiction and provide a novel methodological support for addiction mechanism research and clinical auxiliary diagnosis.

Mobile video applications today have attracted significant attention. Deep learning model (e.g., deep neural network, DNN) compression is widely used to enable on-device inference for facilitating robust and private mobile video applications. The compressed DNN, however, is vulnerable to the agnostic data drift of the live video captured from the dynamically changing mobile scenarios. To combat the data drift, mobile ends rely on edge servers to continuously evolve and re-compress the DNN with freshly collected data. We design a framework, AdaEvo, that efficiently supports the resource-limited edge server handling mobile DNN evolution tasks from multiple mobile ends. The key goal of AdaEvo is to maximize the average quality of experience (QoE), i.e., the proportion of high-quality DNN service time to the entire life cycle, for all mobile ends. Specifically, it estimates the DNN accuracy drops at the mobile end without labels and performs a dedicated video frame sampling strategy to control the size of retraining data. In addition, it balances the limited computing and memory resources on the edge server and the competition between asynchronous tasks initiated by different mobile users. With an extensive evaluation of real-world videos from mobile scenarios and across four diverse mobile tasks, experimental results show that AdaEvo enables up to 34% accuracy improvement and 32% average QoE improvement.

Mobile and IoT vision applications increasingly utilize multitask deep learning (DL) models for real-time inference. The integration of Mixture of Experts (MoE) and Vision Transformers (ViTs) is particularly effective due to the task-agnostic backbone and scalable task-specific heads. However, data drift in open-world environments can lead to accuracy drops and safety risks, as observed in systems like Google Waymo One and NVIDIA NoTraffic. We present AdaSprite to integrate ViT-based MoE co-adaptation into resource-limited IoT systems, first addressing the expert dynamic sparsity during retraining as an opportunity in multi-task settings through cross-task collaboration in computation, I/O, and resource scheduling. Implemented as microservices, AdaSprite improves adaptation accuracy by 34.5% and latency by 77.1%, outperforming baselines across four practical scenarios.

The rise of mobile devices equipped with numerous sensors, such as LiDAR and cameras, has spurred the adoption of multi-modal deep intelligence for distributed sensing tasks, such as smart cabins and driving assistance. However, the arrival times of mobile sensory data vary due to modality size and network dynamics, which can lead to delays (if waiting for slower data) or accuracy decline (if inference proceeds without waiting). Moreover, the diversity and dynamic nature of mobile systems exacerbate this challenge. In response, we present a shift to opportunistic inference for asynchronous distributed multi-modal data, enabling inference as soon as partial data arrives. While existing methods focus on optimizing modality consistency and complementarity, known as modal affinity, they lack a computational approach to control this affinity in open-world mobile environments. AdaFlow pioneers the formulation of structured cross-modality affinity in mobile contexts using a hierarchical analysis-based normalized matrix. This approach accommodates the diversity and dynamics of modalities, generalizing across different types and numbers of inputs. Employing an affinity attention-based conditional GAN (ACGAN), AdaFlow facilitates flexible data imputation, adapting to various modalities and downstream tasks without retraining. Experiments show that AdaFlow significantly reduces inference latency by up to 79.9% and enhances accuracy by up to 61.9%, outperforming status quo approaches. Also, this method can enhance LLM performance to preprocess asynchronous data.