Sequential recommendation (SR) has received much attention and made promising progress in the past few years due to its high alignment with real recommendation scenarios. It models users' preferences and behavior patterns from their historical behavior sequences and provides personalized recommendations. However, the widespread problem of data sparsity limits the performance of sequential recommendation models. To tackle this, data augmentation (DA) provides a feasible solution by improving the quantity, quality, or diversity of the training samples without the need for additional data collection. In this survey, we present a systematic and timely review of research efforts on data augmentation for sequential recommendation. We start by providing a clear formulation of the problem and task. Then, we develop a unified taxonomy that categorizes existing augmentation methodologies regarding their augmentation objects and principles. Next, we conduct a comparative discussion on the advantages and disadvantages of different categories, supplemented with quantitative performance evaluations, time-complexity analyses, and visual case studies of representative methods, aiming to provide actionable guidance for the selection and development of augmentation methods in real-world scenarios. Finally, we present the future research directions and summarize this survey.

With the rise of Software Defined Networking (SDN) and Network Function Virtualization (NFV), network traffic can be processed sequentially by Virtual Network Functions (VNFs) that run as software on general-purpose servers, creating a Service Function Chain (SFC). However, the dynamic and software-based nature of VNFs presents challenges related to reliability and energy efficiency. In this work, we first formulate a long-term SFC orchestration problem focused on minimizing average energy consumption while satisfying reliability requirements for services. To address this problem, we propose a partial backup strategy that effectively balances reliability and energy overhead under uncertain VNF performance. Additionally, we design a Multi-Agent Reinforcement Learning approach based on Graph Convolutional Deep Q-Networks (MA-GCDQN). This method effectively captures the non-Euclidean topology of physical networks and enhances scalability. Building on the optimized orchestration strategy, we also implement a prototype SFC orchestration platform that aligns with the Management and Orchestration (MANO) architecture. The platform supports physical network virtualization, user-defined service deployment, flow-level traffic processing, and facilitates algorithm training and validation. Finally, we conduct extensive simulations with dynamic service requests across multiple time slots. The results demonstrate the effectiveness of our proposed strategy in reducing long-term energy consumption while maintaining high service reliability, thereby confirming the platform’s capability to support automated and intelligent SFC deployment.

Sketch-based measurement has become a fundamental primitive for monitoring network traffic and supporting performance-critical data center applications. However, existing approaches for deploying sketches on DPUs either underutilize DPU hardware capabilities or suffer from degraded accuracy due to limited Arm-side processing capacity and multi-core contention. In particular, most designs directly port host-side sketch implementations to the DPU Arm subsystem, introducing excessive forwarding overhead and hardware pressure. In this paper, we present Duet, an efficient and accurate sketch-based measurement on DPUs. Duet adopts a layered measurement architecture that dynamically schedules flows between Arm-based sketches and the DPU hardware forwarding pipeline according to traffic characteristics. Heavy flows are offloaded to hardware tables to leverage line-rate processing and relieve Arm-side contention, while light flows are retained in software sketches to avoid excessive hardware table updates. To further address crosslayer aggregation and multi-core concurrency challenges, Duet incorporates a zeroing mechanism and lock-free optimizations to ensure measurement consistency and accuracy. We implement Duet on an NVIDIA BlueField-2 DPU and conduct extensive experiments on a real testbed. Experimental results demonstrate that, compared with traditional sketch-based measurement approaches, Duet improves throughput by a factor of approximately $7.1 \times$, packet rate by a factor of approximately $28.9 \times$, and measurement accuracy by approximately 50.1%.

Model merging is an efficient empowerment technique in the machine learning community that does not require the collection of raw training data and does not require expensive computation. As model merging becomes increasingly prevalent across various fields, it is crucial to understand the available model merging techniques comprehensively. However, there is a significant gap in the literature regarding a systematic and thorough review of these techniques. This survey provides a comprehensive overview of model merging methods and theories, their applications in various domains and settings, and future research directions. Specifically, we first propose a new taxonomic approach that exhaustively discusses existing model merging methods. Secondly, we discuss the application of model merging techniques in large language models, multimodal large language models, and more than ten machine learning subfields, including continual learning, multi-task learning, few-shot learning, and so on. Finally, we highlight the remaining challenges of model merging and discuss future research directions. A comprehensive list of papers about model merging is available at https://github.com/ EnnengYang/Awesome-Model-Merging-Methods-Theories-Applications.

As network complexity and scale continue to grow, ensuring efficient data transmission among nodes and maintaining system stability have become critical challenges for network enterprises. However, traditional distributed data transmission mechanisms often fail to meet performance requirements under complex and dynamic network conditions, resulting in inefficient data transmission, increased latency, and potential node failures. To address the above issues, we propose an intelligent transmission mechanism driven by dynamic chunking, designed to ensure efficient and stable data transmission in distributed systems. Network conditions are monitored in real time, and the data chunking problem is formulated as a Markov Decision Process (MDP). To derive optimal chunking decisions, a deep reinforcement learning framework is designed to autonomously solve the MDP and adaptively learn chunking policies in response to network performance dynamics. Furthermore, to ensure system stability, we employ a peer-to-peer (P2P) mechanism for node discovery, integrate Distributed Hash Tables for efficient chunk location, and leverage P2P protocols to coordinate the exchange and transmission of data chunks among nodes. Extensive simulation results validate that the proposed mechanism achieves substantial improvements over traditional methods in terms of transmission duration and throughput, while exhibiting robust adaptability to dynamic and complex network conditions.

The efficiency of data retrieval is crucial for the lifecycle traceability of large-scale manufacturing products. This article proposes a novel on-chain and off-chain parallel retrieval method based on blockchain and Merkle Patricia Trie (MPT), which is called parallel retrieve index MPT (PRIMPT). In the proposed method, index MPT (InMPT) is designed to store the index information and the off-chain address of transactions based on MPT. In addition, the transaction division and the node division are proposed to realize the parallelization. Among them, the transaction division adds a new stage branch node under the root node of InMPT and divides transactions according to their stages, while the node division groups the nodes in the blockchain according to the derived most efficient group number and assigns retrieval requests to these node groups. Experiment results show that the PRIMPT method improves the efficiency by 44% compared with the state-of-the-art on-chain and off-chain method.

A robotic mobile fulfillment system (RMFS) constitutes a highly intricate and interdependent decision-making system, encompassing numerous closely related and interacting decision challenges. Meanwhile, RMFS is a prime example of a dynamic system in which various types of information, such as inventory levels, order backlogs, robot statuses, and pod locations, continuously change throughout operation. This study focuses on the online decision-making process for joint optimization of order assignment, pod selection, and task allocation within RMFS during the order-picking process. Additionally, it considers the influence of current decisions on future decisions to improve online decision-making efficiency. Initially, a dynamic system model for RMFS is developed. Subsequently, a novel mixed-integer linear programming (MILP) model is formulated to address this critical order-picking challenge effectively. Furthermore, a customized heuristic approach, employing alternating decision-making and greedy selection techniques, is proposed to overcome computational challenges associated with larger-scale RMFS scenarios. In different scales, the proposed approach outperforms a sequential approach commonly used in practice and a MILP model in existing literature for online joint optimization of order assignment and pod selection. The findings highlight the efficacy of the proposed approach in optimizing the RMFS order-picking process, particularly for larger instances, delivering significant benefits in terms of increased throughput and reduced robot travel distances.

Per-flow cardinality measurement in high-speed networks is essential for network security and traffic analysis applications. Flow cardinality refers to the number of distinct elements within a flow, such as the number of unique destination IPs associated with a given source IP. While extensive research has been conducted on single-flow cardinality estimation, achieving accurate per-flow cardinality measurement with real-time performance and low memory overhead remains challenging in large-scale network environments, particularly given the highly skewed distribution of flow cardinalities where mouse flows with smaller cardinalities dominate, and elephant flows with larger cardinalities are fewer. This paper introduces MEC-Sketch, a high-accuracy and memory-efficient cardinality measurement data structure that leverages the inherently skewed distribution of flow cardinalities in network traffic. MEC-Sketch employs a dual-component architecture: a heavy part utilizing a majority vote algorithm for high-accuracy super-spreader detection, and a light part implementing compact cardinality estimators for memory-efficient measurement of mouse flows. We address three fundamental technical challenges: (1) adapting the majority vote algorithm to operate with cardinality estimators that lack native support for real-time queries, (2) designing an effective mapping strategy between large and small estimators, and (3) eliminating noise introduced by hash collisions. Comprehensive evaluations on real-world network traces demonstrate that MEC-Sketch significantly outperforms state-of-the-art solutions in terms of estimation accuracy, memory efficiency, and computational performance for both cardinality estimation and super-spreader detection tasks.

As mobile applications become increasingly computation-intensive, mobile devices (MDs) face growing limitations due to their constrained computational capabilities and battery life. Collaborative Edge Computing (CEC) has emerged as a promising solution to address these challenges by enabling multiple edge service providers (ESPs) to offer computation offloading services to MDs. As such, a CEC resource trading market is essential for efficient interactions between MDs and ESPs. However, jointly determining the offloading ratios, allocating combinatorial computation and communication resources, and designing appropriate pricing strategies in a dynamic market remains a significant challenge. To this end, we propose a truthful online double auction based resource allocation mechanism for partial computation offloading (TRAPO) that explicitly accounts for the stochastic nature of both MDs and ESPs. TRAPO first leverages spatial diversity to construct a set of bids for each MD by mapping their task requirements into resource demands through considering MDs' preferences and partial offloading. Next, we match resource-demanding MDs with resource-supplying ESPs based on adaptive valid price thresholds to maximize social welfare, and calculate the payments of MDs and the rewards of ESPs. Theoretical analyses demonstrate that TRAPO satisfies truthfulness, budget balance, individual rationality, and computational tractability. Simulation experiments further verify the effectiveness and efficiency of TRAPO.

Traditional computation offloading and resource allocation strategies encounter several issues that lead to poor service experience and resource wastage. The resource allocation scheme lacks the flexibility to adapt to the time-varying offloading demands of User Equipment (UEs). Furthermore, there is an imbalance between UEs seeking better service and Service Providers (SPs) aiming to minimize cost expenditures. In this paper, we propose a knowledge-defined networking-based Multi-Layer Computation Offloading and Resource Allocation strategy optimization (ML-CORA) architecture. Based on the ML-CORA, we design a Multi-Layer Local-Global Collaborative computation offloading and resource allocation strategy optimization (ML2GC) algorithm. The basic level of the ML2GC algorithm expresses and optimizes computation offloading demands from the perspective of UE (local), while the meta level optimizes the resource allocation strategy on demand from the perspective of the SP (global), achieving a collaborative multi-objective optimization for a win-win system between UEs and SPs. The two-layer structure of the ML2GC algorithm outputs continuous and discrete actions respectively, which improves the flexibility and efficiency of the algorithm while effectively balancing the interests of all parties and promoting efficient resource utilization. Simulation results based on the real-world dataset of Shanghai Telecom indicate that the ML2GC algorithm significantly improves both social welfare and resource utilization compared to baseline algorithms.

The substantial increase in order volume has resulted in a highly competitive freight market, which requires horizontal collaboration among carriers to guarantee cost-effective and timely delivery in transportation. In this case, the collaborative methods implemented by the carriers obviously impact the freight capacities. Hence, to our best knowledge, this paper investigates the collaborative service network design problem under the frameworks of order sharing and vehicle sharing for the first time. Meanwhile, in response to fluctuations in customer demand, we develop a novel two-stage robust optimization model based on the budget uncertainty set for each collaborative framework. Given the NP-hard nature of solving these formulas, an innovative smart-downscale-scenario column-and-constraint generation algorithm is designed to solve the proposed models efficiently, which embeds a linear classifier that identifies scenarios beneficial for convergence. A comparative analysis with the column-and-constraint generation algorithm shows the advantages of our proposed method, which achieves a 25% improvement in solution speed with only a minimal loss in accuracy. In addition, a qualitative analysis of the solution structures is conducted in various collaborative frameworks to demonstrate their flexibility in managing uncertain demand. Based on generated small-scale instances, the results suggest that when the unit outsourcing cost is high, the order sharing framework is more effective for scenarios characterized by low demand volatility; conversely, the vehicle sharing framework is deemed more appropriate for contexts with high demand volatility. Finally, we explore the grand coalition stability through the cooperative game core allocation method.

The rapid growth of e-commerce platforms has given rise to convenient online shopping methods, but platform operators usually don’t have sufficient capacity to deliver these online orders. Hence, based on the estimated demand for each type of commodity, logistics service providers generally require providing freight services to fulfill customer demands while maintaining their own profitability. However, each origin-destination demand is impacted by various factors and is inherently uncertain. This poses significant challenges for logistics service providers seeking to develop an economically viable and efficient freight service network. Given that most research on service network design problems rarely utilizes historical data, in this paper, we propose a data-driven robust service network design problem, aiming to answer: is it possible to obtain more economically efficient decisions when we design a service network based on historical data? To achieve this, a support vector clustering method is employed to adaptively capture a compact convex uncertainty set from the historical demand data, allowing explicit control over the conservatism of decisions. Based on the obtained uncertainty set, a novel data-driven robust optimization model is developed. Considering the NP-hard nature of this problem, we propose a customized column-and-constraint generation algorithm that effectively solves the model. Compared to the Benders decomposition algorithm, it significantly improves computational efficiency. Numerical experiments demonstrate that the proposed method outperforms traditional robust optimization approaches in mitigating demand uncertainty, exhibiting robustness while being less conservative. Meanwhile, we provide valuable managerial insights to decision-makers through parameter sensitivity analysis. Results reveal that neither overly aggressive nor excessively conservative strategies are optimal, and adjusting the clustering parameter allows decision-makers to tailor the service network to their risk tolerance for more cost-effective operations under demand uncertainty.

Distributed computing continuum systems (DCCSs) integrate heterogeneous end, edge, and cloud resources to support adaptive task execution across tiers. In such environments, collaborative inference with large and small language models offers a promising way to balance inference quality and execution efficiency. However, most existing collaboration strategies are model-centric and often overlook dynamic system conditions such as cloud congestion and device-side load. This paper presents a continuum-aware collaborative offloading (CACO) framework for end-cloud collaborative inference with edge-side scheduling and congestion sensing. The framework follows a cloud-first policy under uncongested conditions to preserve accuracy. When cloud congestion is detected, the edge gateway decides whether to activate end-side collaborative execution according to end-node load conditions and subjectspecific reliability priors. Reliable end nodes are selected for top- $k$ collaborative inference, and majority voting is used to generate the final answer. If consensus fails, the system falls back to cloud inference and updates the reliability priors online. We implement a prototype and evaluate it on the MMLU benchmark under different congestion levels. Experimental results show that the proposed framework preserves near-cloud answer quality under regular conditions while reducing cloud workload pressure, and achieves stronger latency robustness than static offloading under higher congestion.

The integration of Large Language Models (LLMs) into web applications is challenged by dynamic online content that quickly renders their parametric knowledge obsolete. While In-Context Knowledge Editing (ICE) offers a promising solution, existing methods often rely on structured knowledge triples and require training auxiliary models. This restricts their effectiveness in unstructured natural language environments and introduces computational overhead. To address these limitations, we propose DeCO, a training-free framework that operates directly on natural language from external sources. We identify that Chain-of-Thought (CoT) generates reasoning steps in a single forward pass, hindering the incorporation of corrections. DeCO reformulates CoT as an incremental process of generation and editing, where each step is dynamically revised using retrieved knowledge to guide the model with finer granularity. Furthermore, DeCO tackles two critical challenges: (1) the inaccuracy of semantic similarity-based retrieval, and (2) the tendency of LLMs to revert to parametric knowledge or refuse responses under counterfactual conditions. We introduce the Information Seeker, a two-stage module leveraging LLM semantics for precise retrieval, and the Dynamically Prefill Guidance Strategy, which utilizes autoregressive constraints to enforce instruction adherence. Extensive evaluations demonstrate that DeCO outperforms six strong baselines, achieving average gains of 23.3% on multi-hop editing benchmarks and 4.5% on comprehensive benchmarks, underscoring its broad applicability.

Digital twin-enabled vehicular edge computing leverages virtual representations of physical vehicles to facilitate real-time intelligent services such as collaborative perception, trajectory prediction, fault diagnosis, and proactive decision-making through seamless integration of sensing, computing, and communication resources. The effectiveness of these services critically depends on the timely execution of deep learning inference tasks. However, vehicular mobility and dynamic contention for limited edge resources lead to significant delays in inference execution, thereby degrading the accuracy of digital twin-assisted decision-making. To address this challenge, we propose a hierarchical framework for collaborative inference and digital twin deployment. At the upper layer, a mobility-aware strategy for inference task offloading and digital twin deployment is employed, which incorporates dynamic action-space pruning to select edge nodes based on vehicular mobility patterns and link stability. At the lower layer, a delay-resource gradient-based optimization mechanism is designed to adaptively allocate computational resources by matching dynamic inference demands with available edge capacity. Extensive simulations demonstrate the effectiveness of the proposed approach, showing a reduction of at least 15% in Age of Information (AoI) compared to the baseline methods. Moreover, our approach significantly improves the inference efficiency and system resource utilization, and reduces the deadline violation rate.

Rumor blocking approaches in social networks aim to identify a small set of counter-rumor seed nodes and compete with rumor cascades to quickly stop the propagation of rumors. However, current rumor blocking methods assume complete knowledge of rumor node positions, which is often unattainable in real-world scenarios. In this paper, we introduce the concept of Uncertainty Rumor Blocking, where we address the uncertainty surrounding rumor node locations by considering a set of suspicious nodes, each associated with a probability indicating the likelihood of rumor propagation. As traditional node selection algorithms become inadequate under uncertain conditions, we propose a Graph Neural Network-based Inverse Reinforcement Learning (G-IRL) approach to effectively select counter-rumor seed nodes. Through comprehensive experimentation on three datasets, we demonstrate the consistent superiority of our G-IRL over state-of-the-art baseline methods for node selection in the context of uncertainty rumor containment.

The rapid proliferation of heterogeneous cloud services has made Cloud Service Combinatorial Optimization (CSCO) a fundamental challenge for users and cloud service providers (CSPs) in large-scale computing ecosystems. Its complexity arises from the exponential growth of the decision space due to numerous service providers and tightly coupled resource constraints, making it difficult to efficiently identify high-quality CSP combinations. Moreover, intrinsic conflicts among cost, Quality of Service (QoS), and resource utilization further com plicate the search for balanced solutions. In addition, existing multi-attribute decision-making approaches often fail to capture the compensatory trade-offs between price and QoS, resulting in limited interpretability and weak discrimination among candidate solutions. To address these challenges, we propose a two stage cloud service combinatorial optimization framework that integrates evolutionary search and decision-oriented ranking. In Stage I, a cognitive-inspired co-evolutionary dual-population multi-objective optimization algorithm (CDMOA) is developed to efficiently explore the large CSCO search space. By coordinating a Strategist Population that emphasizes feasibility and cost effectiveness with an Explorer Population that promotes solution diversity, CDMOA generates high-quality and well-distributed Pareto-optimal solutions. In Stage II, a bidirectional price–QoS contrastive compensation (BPQC) method is introduced to rank the obtained Pareto solutions. By integrating global–local contrast analysis with a bounded price–QoS compensation mechanism, BPQC operationalizes relative price–QoS trade-offs and pro vides interpretable rankings for decision support. Extensive experiments demonstrate that the proposed framework achieves competitive optimization performance and more effective decision ranking compared with several state-of-the-art baselines.

Sharding, as a Layer-1 scaling technique, is widely recognized as a promising solution to address the scalability limitations faced by blockchains. However, distributing accounts across shards generates numerous cross-shard transactions (CSTs) and inter-shard workload imbalances, potentially compromising scalability. Existing state-of-the-art methods typically optimize transaction distribution for subsequent epochs by periodically reallocating accounts using graph partitioning or community detection to balance loads and minimize CSTs. Nevertheless, these methods often involve computationally expensive account migration and overlook real-world transaction skewness, exacerbating workload imbalances. To address the above problems, we propose a two-phase account group migration service with dynamic load awareness in this paper. This service can monitor and analyze system state to determine the necessity of account migration. Once triggered, it employs a two-phase algorithm that combines account grouping with global group-level re-partitioning to balance workloads and minimize CSTs. Additionally, group-level migration further helps reduce computational overhead. We evaluate the designed service by replaying large-scale real Ethereum transactions. Experimental results demonstrate that, compared with the baselines, our method can not only improves system throughput and reduces transaction confirmation latency, but also enhances overall scalability.

In cloud-edge-end (CEE) collaboration, the resource optimization based on deep reinforcement learning have achieved significant performance improvements in time-slot systems. However, some studies only focus on computing delay and energy consumption in each time slot, ignoring the impact of task backlog queues on system performance. In addition, the delay-oriented optimization tends to offload a large number of tasks to servers, failing to fully utilize the computing resource of mobile devices. To address these issues, we propose the multi-dimensional preference-driven resource optimization framework (MPROF). This study includes several key points: 1) constructing a three-layer heterogeneous architecture that applying the collaboration among edges for CEE; 2) proposing the task backlog estimation mechanism, which mitigates the impact of previous unfinished tasks on the current time slot; 3) proposing the group relative direct-preference policy optimization (GRDPO) that incorporates the preference information for efficient task offloading, and combines it with mathematical programming for the system resource optimization. The simulation experiments are conducted across multiple typical scenarios. The results show that, the proposed framework outperforms existing mainstream methods in task offloading, system delay, task backlog, and energy consumption control, demonstrating certain practical application prospects.