Vehicle re-identification (ReID) plays a key role in intelligent transportation systems. However, this task is complicated by intra-class variations due to viewpoint changes, occlusions, and inter-class differences among visually similar vehicles. The issues of feature coupling and limited fine-grained discrimination affect accurate vehicle matching. We propose DualDis, a novel vehicle re-identification framework that decouples identity-related features. DualDis consists of two key modules, Adaptive Component Disentangling (ACD) and Progressive Dimensional Attention (PDA). ACD uses multi-head attention to separate vehicle parts, while PDA deploys a region-aware sparse channel and symmetry-aware contextual attention to distinguish symmetrical and asymmetrical features. This dual-path structure enables the given model to concentrate on the most discriminative features while minimizing redundant information. Extensive experiments on the VeRi776 and VehicleID datasets reveal that DualDis outperforms state-of-the-art methods in multi-view retrieval, which obtains superior accuracy and demonstrates its generalization capabilities across different datasets. The source code is publicly available at https://github.com/711L/DualDis.
论文检索
输入标题、作者或关键词，从 954 篇学术成果中精准定位
In web-based AI services, providers typically host multiple large language models (LLMs) that exhibit diverse capabilities and incur different API costs. Meanwhile, LLM's performance depends not only on its inherent capacity but also on the reasoning strategy it employs, which together influence both answer quality and computational cost. A key challenge is therefore how to adaptively allocate models and strategies to achieve high-quality responses under constrained costs. To address this challenge, we propose Route-To-Reason (RTR), a unified routing framework that simultaneously selects suitable LLMs and reasoning strategies according to query complexity and user budget. Specifically, RTR learns dense vector representations of models and strategies that capture their behavioral characteristics in handling different queries. Leveraging these embeddings, RTR builds a routing table that estimates the cost and performance of different model–strategy pairs. During inference, RTR consults this routing table to dynamically assign the most appropriate pair, enabling adaptive and cost-efficient reasoning tailored to query difficulty and budget scenarios. Extensive experiments across multiple reasoning benchmarks show that RTR achieves comparable or higher accuracy than the best single LLM while substantially reducing both token usage and API cost (by up to 60%), achieving a superior trade-off between performance and efficiency. By lowering the overhead of large-scale LLM inference, RTR contributes to cost-aware and environmentally sustainable deployment of web-based AI services.
Low-latency live streaming (LLS) has emerged as a popular web application, with many platforms adopting real-time protocols such as WebRTC to minimize end-to-end latency. However, we observe a counter-intuitive phenomenon: even when the actual encoded bitrate does not fully utilize the available bandwidth, stalling events remain frequent. This insufficient bandwidth utilization arises from the intrinsic temporal variations of real-time video encoding, which cause conventional packet-level congestion control algorithms to misestimate available bandwidth. When a high-bitrate frame is suddenly produced, sending at the wrong rate can either trigger packet loss or increase queueing delay, resulting in playback stalls. To address these issues, we present Camel, a novel frame-level congestion control algorithm (CCA) tailored for LLS. Our insight is to use frame-level network feedback to capture the true network capacity, immune to the irregular sending pattern caused by encoding. Camel comprises three key modules: the Bandwidth and Delay Estimator and the Congestion Detector, which jointly determine the average sending rate, and the Bursting Length Controller, which governs the emission pattern to prevent packet loss. We evaluate Camel on both large-scale real-world deployments and controlled simulations. In the real-world platform with 250M users and 2B sessions across 150+ countries, Camel achieves up to a 70.8% increase in 1080P resolution ratio, a 14.4% increase in media bitrate, and up to a 14.1% reduction in stalling ratio. In simulations under undershooting, shallow buffers, and network jitter, Camel outperforms existing congestion control algorithms, with up to 19.8% higher bitrate, 93.0% lower stalling ratio, and 23.9% improvement in bandwidth estimation accuracy.
Account migration in sharded blockchains presents a critical trade-off between optimization effectiveness and system availability. While dynamically reallocating accounts across shards can significantly reduce cross-shard transaction overhead, existing migration mechanisms cause service disruptions that intensify as state data volumes grow. To address this challenge, we propose BIND, a batch-wise account migration protocol that eliminates service interruptions by enabling continuous transaction processing throughout migration. BIND introduces a dual transaction pool architecture that isolates transactions involving migrating accounts while allowing non-migrating accounts to operate uninterrupted. To optimize migration efficiency, we design a reverse greedy heuristic algorithm that partitions accounts into batches based on community cohesion, maximizing intra-batch connectivity to front-load cross-shard communication reduction. We evaluate BIND using real Ethereum transactions, demonstrating superior performance over existing mechanisms. BIND achieves 12% higher overall throughput, reduces migration time to 23.6%-39.3% of the one-shot baseline (across 1-10Gbps bandwidth), and lowers cross-shard transaction rates by 24.1% compared to random batching. These results confirm BIND as a practical solution for large-scale, non-disruptive account migration in production sharded blockchains.
With the rapid growth and rising complexity of web content, edge-deployed LLMs have become essential for enhancing users' online experiences. However, sparsely-activated Mixture-of-Experts (MoE) models, which are well-suited for edge scenarios, face significant memory bottleneck challenges. Offload-based methods have been proposed to mitigate the problem, but they face difficulties with expert prediction. To promote the application of MoE models in edge scenarios, we propose Fate, an offloading system designed for MoE models to enable efficient inference in resource-constrained environments. The key insight behind Fate is that gate inputs from adjacent layers can be effectively used for expert prefetching, achieving high prediction accuracy. Furthermore, Fate employs a shallow-favoring expert caching strategy that increases the expert hit rate to 99%. Additionally, Fate integrates tailored quantization strategies for cache optimization and I/O efficiency. Experimental results show that, compared to baselines, Fate achieves up to 1.34×-5.07× prefill speedup and 1.26×–4.41× decoding speedup, while maintaining inference quality.
Real-world multichannel time series prediction faces growing demands for efficiency across edge and cloud environments, making channel compression a timely and essential problem. Motivated by success of Multiple-Input Multiple-Output (MIMO) methods in signal processing, we propose a predictability-aware compression–decompression framework to reduce runtime, decrease communication cost, and maintain prediction accuracy across diverse predictors. The core idea involves using a circular seasonal key matrix with orthogonality to capture underlying time series predictability during compression and to mitigate reconstruction errors during decompression by introducing more realistic data assumptions. Theoretical analyses show that the proposed framework is both time-efficient and accuracy-preserving under a large number of channels. Extensive experiments on six datasets across various predictors demonstrate that the proposed method achieves superior overall performance by jointly considering prediction accuracy and runtime, while maintaining strong compatibility with diverse predictors.
Modern cloud applications, from interative web services to mobile and WoT workloads, generate highly dynamic multi-tenant network demands. Guaranteeing priority-aware bandwidth remains challenging: legacy shapers like Linux Traffic Control Hierarchical Token Bucket are static and unscalable, while cloud-native solutions such as Cilium offer only coarse-grained rate limiting. We present BeeQoS, a cloud-native QoS system that delivers low-latency, adaptive, and scalable multi-priority bandwidth guarantees. BeeQoS consists of an eBPF-powered data plane for high-performance, fine-grained per-packet shaping, a demand-aware control plane that senses real-time flow requirements and adaptively reallocates bandwidth, and seamless Kubernetes integration for expressive policy specification and cluster-wide scalable deployment. Evaluation shows that BeeQoS scales to 1K+ flows with stable performance, boosts high-/medium-priority throughput by 14.6%/36.4%, cuts median latency by 72.4%, reduces deployment overhead, and improves video QoE by 27.3% over state-of-practice baselines.
Split Learning (SL) offers a framework for collaborative model training that respects data privacy by allowing participants to share the same dataset while maintaining distinct feature sets. However, SL is susceptible to backdoor attacks, in which malicious clients subtly alter their embeddings to insert hidden triggers that compromise the final trained model. To address this vulnerability, we introduce SecureSplit, a defense mechanism tailored to SL. SecureSplit applies a dimensionality transformation strategy to accentuate subtle differences between benign and poisoned embeddings, facilitating their separation. With this enhanced distinction, we develop an adaptive filtering approach that uses a majority-based voting scheme to remove contaminated embeddings while preserving clean ones. Rigorous experiments across four datasets (CIFAR-10, MNIST, CINIC-10, and ImageNette), five backdoor attack scenarios, and seven alternative defenses confirm the effectiveness of SecureSplit under various challenging conditions.
Sharding is a key approach to scale blockchain, where transaction scheduling critically impacts performance. Traditional sharding uses non-perception scheduling, where each shard makes decisions independently, leading to high conflicts. Although some studies use a centralized scheduler to eliminate conflicts by gathering full shard data, but incur security risks. To address this, we propose Concordia, a distributed transaction scheduling scheme via cooperative perception. Each shard predicts future cross-shard interactions via a Graph Neural Network (GNN) based algorithm, generates low-dimensional perception signals, and shares them through incremental personalized broadcasts. Shards then use a popularity-driven transaction packing algorithm to schedule and pack transactions effectively. With 16 shards, Concordia achieves 2.39x higher throughput than non-perception scheduling while cutting conflicts from 62.3% to 1.5%, and improves throughput 1.67x over centralized scheduling while reducing perception overhead by 95.2%.
High-quality GPS trajectories are essential for location-based web services and smart city applications, including navigation, ride-sharing and delivery. However, due to low sampling rates and limited infrastructure coverage during data collection, real-world trajectories are often sparse and feature unevenly distributed location points. Recovering these trajectories into dense and continuous forms is essential but challenging, given their complex and irregular spatio-temporal patterns. In this paper, we introduce a novel diffusion model for TRA jectory rEC overy named TRACE, which reconstruct dense and continuous trajectories from sparse and incomplete inputs. At the core of TRACE, we propose a State Propagation Diffusion Model (SPDM), which integrates a novel memory mechanism, so that during the denoising process, TRACE can retain and leverage intermediate results from previous steps to effectively reconstruct those hard-to-recover trajectory segments. Extensive experiments on multiple real-world datasets show that TRACE outperforms the state-of-the-art, offering >26% accuracy improvement without significant inference overhead. Our work strengthens the foundation for mobile and web-connected location services, advancing the quality and fairness of data-driven urban applications. Code is available at:~ https://github.com/JinmingWang/TRACE
As critical infrastructure for Web 3.0, DAG-based blockchains promise high throughput for DeFi, IoT, and DApps. However, realizing this potential is challenging, as system performance is dictated by a multitude of interdependent parameters across network, node, and consensus layers. Manual configuration fails to adapt to dynamic workloads, leading to suboptimal performance. We introduce Alzo, a novel auto-tuner that employs hierarchical reinforcement learning (HRL) to navigate this complex configuration space. By decomposing the DAG blockchain's workflow into distinct stages, Alzo's HRL policy learns from stage-level performance metrics to control critical parameters governing consensus, execution, and graph topology in real-time. Furthermore, we employ a shadow-control loop to ensure the safety of all parameter adjustments. Our experiments show that Alzo significantly outperforms other configurations, achieving higher throughput and lower latency under variable workloads with minimal overhead.
The rise of Internet of Vehicles (IoV) technologies is transforming traffic management from isolated control to a collective, multi-vehicle process. At the heart of this shift is multi-vehicle dynamic navigation, which requires simultaneously routing large fleets under evolving traffic conditions. Existing path search algorithms and reinforcement learning methods struggle to scale to city-wide networks, often failing to capture the nonlinear, stochastic, and coupled dynamics of urban traffic. To address these challenges, we propose CityNav, a hierarchical, LLM-powered framework for large-scale multi-vehicle navigation. CityNav integrates a global traffic allocation agent, which coordinates strategic traffic flow distribution across regions, with local navigation agents that generate locally adaptive routes aligned with global directives. To enable effective cooperation, we introduce a cooperative reasoning optimization mechanism, in which agents are jointly trained with a dual-reward structure: individual rewards promote per-vehicle efficiency, while shared rewards encourage network-wide coordination and congestion reduction. Extensive experiments on four real-world road networks of varying scales (up to 1.6 million roads and 430,000 intersections) and traffic datasets demonstrate that CityNav consistently outperforms nine classical path search and RL-based baselines in city-scale travel efficiency and congestion mitigation. Our results highlight the potential of LLMs to enable scalable, adaptive, and cooperative city-wide traffic navigation, providing a foundation for intelligent, large-scale vehicle routing in complex urban environments. Our project is available at https://github.com/usail-hkust/CityNav.
Metro flow prediction is a critical application in smart city and Web of Things infrastructures, essential for optimizing urban mobility. However, building such predictive systems faces three key challenges: (1) the fragmentation of multimodal spatiotemporal data, (2) the inefficiency of existing models in capturing long-range dependencies, and (3) the data silos and privacy concerns inherent in distributed station infrastructures. To address these challenges, a multimodal federated learning framework named MF3 (Mamba-Transformer-Federated Metro Flow Prediction) is proposed. First, a multimodal alignment (MA) module is designed, where cross-modal alignment attention bridges visual and spatiotemporal features, thereby enhancing feature complementarity and alignment. Second, a dual-path Mamba-Transformer (DMT) module is designed, in which Mamba's linear long-range memory and the Transformer's global perception operate in parallel, reducing information loss. Third, a blockchain-based federated reputation (BFR) module is established to perform personalized federated learning, thereby enhancing privacy protection. Finally, extensive experiments on real metro datasets from Hangzhou and Shanghai demonstrate that MF3 achieves superior performance in terms of prediction accuracy. In summary, the proposed MF3 framework provides a new feasible paradigm for metro flow prediction, supporting urban traffic optimization, metro operation and scheduling, and the development of smart city and Web of Things infrastructures.
Web-based participatory urban sensing has emerged as a vital approach for modern urban management by leveraging mobile individuals as distributed sensors. However, existing urban sensing systems struggle with limited generalization across diverse urban scenarios and poor interpretability in decision-making. In this work, we introduce AgentSense, a hybrid, training-free framework that integrates large language models (LLMs) into participatory urban sensing through a multi-agent refinement system. AgentSense initially employs a classical planner to generate baseline solutions and then iteratively refines them to adapt sensing task assignments to dynamic urban conditions and heterogeneous worker preferences, while producing natural language explanations that enhance transparency and trust. Extensive experiments across two large-scale mobility datasets and seven types of dynamic disturbances demonstrate that AgentSense offers distinct advantages in adaptivity and explainability over traditional methods. Furthermore, compared to single-agent LLM baselines, our approach outperforms in both performance and robustness, while delivering more reasonable and transparent explanations. These results position AgentSense as a significant advancement towards deploying adaptive and explainable urban sensing systems on the web.
The rapid evolution of the Web of Things (WoT) has created new opportunities for connectivity and standardization across heterogeneous devices, enabling the development of increasingly complex systems. However, edge devices deployed in resource-constrained scenarios face significant challenges. These devices require lightweight and efficient models to achieve high accuracy while operating within strict memory constraints. Typical approaches to model generation, Neural Architecture Search (NAS), have proven effective in automating the search for optimal architectures. However, existing NAS methods suffer from two critical limitations: (1) they fail to incorporate quantization into the search space, which can result in overlooking larger models that might perform better after quantization; and (2) current model evaluation methods struggle to provide accurate assessments within a short time period. To address these challenges, we propose EdgeGen, a novel NAS framework that integrates multiple quantization methods into the search process, enabling the discovery of larger models that need quantization to satisfy the constraint. EdgeGen employs a multi-beam Monte Carlo Tree Search (MCTS) algorithm and a constraint validator to explore the expanded search space efficiently, searching vast original and quantized models. Furthermore, EdgeGen evaluates the model performance following a GNN-based performance predictor, which provides a rapid and precise prediction. Across multiple benchmarks, EdgeGen consistently outperforms state-of-the-art NAS methods. Code available at: https://doi.org/10.5281/zenodo.18323232.
The rapid growth of Web, mobile, and Web of Things (WoT) applications increasingly relies on Federated Learning (FL) to enable privacy-preserving intelligence across distributed and heterogeneous devices. However, the global model in FL often suffers from performance degradation due to global knowledge forgetting caused by highly non-IID client data. While client-side Knowledge Distillation (KD) has emerged as a promising paradigm for transferring global knowledge to local models, existing approaches inadvertently distort global knowledge through local data priors, fundamentally limiting their effectiveness in real-world Web and mobile environments. We further find that this knowledge forgetting is asymmetric, with locally under-represented classes suffering significantly more severe performance degradation than over-represented ones. While the conventional KD inherently couples these two distinct kinds of knowledge, thereby hindering the targeted preservation of critical global knowledge for under-represented classes. To address these challenges, we propose Federated Adaptive Knowledge Distillation (FedAKD), a novel framework that adaptively calibrates and decouples global knowledge according to the local data distribution, enabling precise and flexible distillation across heterogeneous clients. Extensive experiments on benchmark datasets and diverse data heterogeneity settings demonstrate that FedAKD achieves state-of-the-art performance, offering a scalable and sustainable solution for Web, mobile, and WoT intelligence deployment.
Large Language Models (LLMs) are increasingly deployed on edge devices to address privacy and latency concerns in modern Web applications. While numerous studies focus on inference frameworks, the critical problem of tuning runtime configurations remains largely underexplored. This endeavor is particularly challenging on edge devices due to severe budget limitations and the dynamic variability of system resources. To address these challenges, we draw upon key insights regarding parameter sensitivity, configuration transferability, and rank stability to propose LaTune, a lightweight and adaptive tuning framework. LaTune is designed to efficiently find optimal runtime configurations by incorporating three complementary components: parameter selection to focus on the most impactful parameters, knowledge transfer to leverage historical data for accelerated search, and two-stage optimization to dynamically select the best configuration based on real-time resource constraints. Experiments across four edge devices and LLMs show that LaTune achieves up to 3.93x higher hypervolume and 6.90x throughput gains over baselines. It accelerates tuning efficiency by 2-3x, converging within 10-20 iterations, and ensures robust execution under heavy contention where static methods fail. Our code is open-sourced at https://github.com/pkuaiweb/LaTune.
Time series anomaly detection (TSAD) has been a long-standing pillar problem in Web-scale systems and online infrastructures, such as service reliability monitoring, system fault diagnosis, and performance optimization. Large language models (LLMs) have demonstrated unprecedented capabilities in time series analysis, the potential of multimodal LLMs (MLLMs), particularly vision-language models, in TSAD remains largely under-explored. One natural way for humans to detect time series anomalies is through visualization and textual description. It motivates our research question: Can multimodal LLMs perform time series anomaly detection? Existing studies often oversimplify the problem by treating point-wise anomalies as special cases of range-wise ones or by aggregating point anomalies to approximate range-wise scenarios. They limit our understanding for realistic scenarios such as multi-granular anomalies and irregular time series. To address the gap, we build a VisualTimeAnomaly benchmark to comprehensively investigate zero-shot capabilities of MLLMs for TSAD, progressively from point-, range-, to variate-wise anomalies, and extends to irregular sampling conditions. Our study reveals several key insights. 1) MLLMs and traditional TSAD methods are complementary: MLLMs excel at coarse-grained anomalies while traditional methods are effective at fine-grained anomalies. 2) MLLMs are resilient to irregular time series. 3) Input time series modality changing from text to image makes information focus shift from quantitative variations to qualitative patterns while significantly reducing hallucinations. Built on the findings, we propose a MLLMs-based multi-agent framework TSAD-Agents to achieve automatic TSAD. Our framework comprises scanning, planning, detection, and checking agents that synergistically collaborate to reason, plan, and self-reflect to enable automatic TSAD. These agents adaptively invoke tools such as traditional methods and MLLMs and dynamically switch between text and image modalities to optimize detection performance.
Starlink, the low-Earth orbit (LEO) satellite constellation developed by SpaceX, has rapidly become the world's largest commercial satellite network and a key component of the global Internet infrastructure. Despite its growing prominence, critical aspects of its terrestrial operations, including its internal network architecture, user behavior patterns, and potentially vulnerable exposed services, remain largely unexamined. This paper presents a multi-perspective measurement that characterizes Starlink's ground-side infrastructure and ecosystem from a DNS-centric viewpoint. First, we leverage internal DNS leakage to infer the structure of Starlink's private networks. Second, we analyze passive DNS data to identify user behavior patterns and service usage trends. Finally, we perform large-scale active scanning of Starlink's IP address space to evaluate its service deployment and security posture, utilizing DNS records to isolate infrastructure-related IPs from those of end-users. Collectively, our findings provide novel insights into the architecture, operation, and security of Starlink's terrestrial network.
Adaptive traffic signal control (ATSC) is essential for mitigating urban congestion in modern smart cities, where traffic infrastructure is evolving into interconnected Web-of-Things (WoT) environments with thousands of sensing-and-control nodes. However, existing methods face a critical scalability-coordination tradeoff: centralized approaches optimize global objectives but become computationally intractable at city scale, while decentralized multi-agent methods scale efficiently yet lack network-level coherence, resulting in suboptimal performance. In this paper, we present HALO, a hierarchical reinforcement learning framework that addresses this tradeoff for large-scale ATSC. HALO decouples decision-making into two levels: a high-level global guidance policy employs Transformer-LSTM encoders to model spatio-temporal dependencies across the entire network and broadcast compact guidance signals, while low-level local intersection policies execute decentralized control conditioned on both local observations and global context. To ensure better alignment of global-local objectives, we introduce an adversarial goal-setting mechanism where the global policy proposes challenging-yet-feasible network-level targets that local policies are trained to surpass, fostering robust coordination. We evaluate HALO extensively on multiple standard benchmarks, and a newly constructed large-scale Manhattan-like network with 2,668 intersections under real-world traffic patterns, including peak transitions, adverse weather and holiday surges. Results demonstrate HALO shows competitive performance and becomes increasingly dominant as network complexity grows across small-scale benchmarks, while delivering the strongest performance in all large-scale regimes, offering up to 6.8% lower average travel time and 5.0% lower average delay than the best state-of-the-art.