Large language models (LLMs) are increasingly powering web-based applications, whose effectiveness relies on fine-tuning with large-scale instruction data. However, such data often contains valuable or sensitive information that limits its public sharing among business organizations. Federated learning (FL) enables collaborative fine-tuning of LLMs without accessing raw data. Existing approaches to federated LLM fine-tuning usually adopt a uniform model architecture, making it challenging to fit highly heterogeneous client-side data in varying domains and tasks, e.g., hospitals and financial institutions conducting federated fine-tuning may require different LLM architectures due to the distinct nature of their domains and tasks. To address this, we propose FedAMoLE, a lightweight personalized FL framework that enables data-driven heterogeneous model architectures. It features a heterogeneous mixture of low-rank adaptation (LoRA) experts module to aggregate architecturally heterogeneous models and a reverse selection-based expert assignment strategy to tailor model architectures for each client based on data distributions. Experiments across seven scenarios demonstrate that FedAMoLE improves client-side performance by an average of 5.97

Diffusion Transformers (DiTs) are costly at high resolution because self-attention scales quadratically with token sequence length. Existing pruning methods do not jointly provide end-to-end learnability, low training overhead, and deterministic token counts for predictable token-dependent computation. We propose Shiva-DiT, based on Residual-Based Differentiable Top-k Selection. Its forward pass executes hard top-k selection, while a residual-aware straight-through estimator propagates gradients to both token scores and the budget k without evaluating a second backbone path. A Context-Aware Router and Adaptive Ratio Policy learn layer- and timestep-dependent retention schedules under a target average budget. Experiments on SD3-Medium, Flux.1-dev, and PixArt-Σ show consistent reductions in FLOPs and measured latency. On SD3-Medium, Shiva-DiT provides four fidelity-latency operating points and reaches a 1.54x wall-clock speedup with competitive fidelity.

Open-source software vulnerability patch detection is a critical component for maintaining software security and ensuring software supply chain integrity. Traditional manual detection methods face significant scalability challenges when processing large volumes of commit histories, while being prone to human errors and omissions. Existing automated approaches, including heuristic-based methods and pre-trained model solutions, suffer from limited accuracy, poor generalization capabilities, and inherent methodological constraints that hinder their practical deployment. To address these fundamental challenges, this paper conducts a comprehensive empirical study of existing vulnerability patch detection methods, revealing three key insights that guide the design of effective solutions: the critical impact of search space reduction, the superiority of pre-trained semantic understanding over architectural complexity, and the temporal limitations of web crawling approaches. Based on these insights, we propose a novel two-stage framework that combines version-driven candidate filtering with large language modelbased multi-turn dialogue voting to achieve accurate and efficient vulnerability patch identification. Extensive experiments on a dataset containing 750 real vulnerabilities demonstrate that our method outperforms current approaches.

Memory disaggregation has become a research trend in data centers. Existing studies fall into two paths: Network-based logical memory disaggregation (LMD) and Compute Express Link (CXL)-based physical memory disaggregation (PMD). However, LMD suffers from network overhead, while PMD incurs expensive hardware costs and lacks flexibility. This paper advocates for building LMD systems on shared CXL memory, taking advantage of its low latency and cache-coherent memory access. However, shared CXL memory has severe scalability issues due to its strict coherence model. This paper introduces Swarm, an efficient LMP system built on shared CXL memory. Swarm divides shared CXL memory into small cacheable memory and large non-cacheable memory. Hardware only needs to maintain coherence for cacheable memory, while software handles coherence for non-cacheable memory, thereby enabling all CXL memory to be shared. Then, Swarm implements cross-node RPC and dynamic global memory allocation on shared CXL memory to improve performance and memory utilization. Swarm also proposes distributed computing offloading to fully leverage compute power on memory nodes for acceleration. Our evaluation shows that Swarm not only improves the throughput (e.g., by 3.6 & times; and 1.8 & times;) compared with representative network-based LMD system, AIFM and CXL-based PMD system when computing offloading is enabled but also achieves an advantage in TCO.

Retrieval-augmented generation (RAG) enables large language models (LLMs) to produce evidence-based responses, and its performance hinges on the matching between the retriever and LLMs. Retriever optimization has emerged as an efficient alternative to fine-tuning LLMs. However, existing solutions suffer from objective mismatch between retriever optimization and the goal of RAG pipeline. Reinforcement learning (RL) provides a promising solution to address this limitation, yet applying RL to retriever optimization introduces two fundamental challenges: 1) the deterministic retrieval is incompatible with RL formulations, and 2) state aliasing arises from query-only retrieval in multi-hop reasoning. To address these challenges, we replace deterministic retrieval with stochastic sampling and formulate RAG as a Markov decision process, making retriever optimizable by RL. Further, we incorporate retrieval history into the state at each retrieval step to mitigate state aliasing. Extensive experiments across diverse RAG pipelines, datasets, and retriever scales demonstrate consistent improvements of our approach in RAG performance.

With the development of Federated Learning (FL) in IoT Edge-Cloud Computing environments, mobile terminals are able to cooperate without the leakage on raw data. However, factors including the terminals' high mobility and the network fluctuations make the cooperator selection during FL training extremely complex. Under the distributed cooperation, traditional FL strategies show certain limitations and cannot always select the available nodes when training, leading to the difficulties in energy and latency optimization. In this paper, we propose a Hierarchical Reinforcement Learning (HRL)-based federated knowledge distillation (HRL-FedKD) framework in which both high-level and low-level controllers utilize the Double Deep Q-Network (DDQN) algorithm. The high-level controller selects the nodes participating in FL training, while the low-level controller determines the number of local training epochs for each node. After training, the global model will be compressed into a lightweight model by knowledge distillation (KD) in deployment while preserving the personalization of local models. The experiments were conducted using Chest X-Ray and Brain Tumor MRI datasets to validate the proposed FL strategy. The results demonstrate that the HRL-FedKD framework can effectively optimize latency and energy consumption in complex state spaces.

Spiking neural networks (SNNs) are gaining attention for energy-efficient, event-driven computing due to their ability to encode information through discrete spikes over multiple time steps. Their use of a biologically plausible neuron model makes them well-suited for asynchronous and sparse data processing in edge workloads. However, this temporal and stateful behavior introduces burst-driven execution patterns and inter-step dependencies, posing significant challenges for GPU-based deployment. Existing GPU schedulers, optimized for dense and feedforward DNNs, fail to fully exploit the irregularity and sparsity of SNNs, leading to poor resource utilization and unpredictable latency under dynamic workloads. To address these challenges, we propose MPL-schedule, a multi-preemptive, priority-aware scheduling scheme that dynamically coordinates heterogeneous SNN tasks on GPUs. Our approach formulates the scheduling process as a reward maximization problem under resource constraints, integrating adaptive priority computation, overhead-aware time-slicing, and state-preserving preemption. We evaluate MPL-schedule across diverse scenarios, including task complexities, spike arrival patterns, and time-slice configurations. Compared to state-of-the-art baselines, our method improves throughput by up to 15.0

The proliferation of intelligent consumer electronics necessitates a service architecture capable of delivering personalized services while protecting privacy. Federated Learning (FL), as a privacy-preserving paradigm, enables multiple consumer electronic devices to collaboratively train a model without exposing sensitive user data. However, FL performance (i.e., convergence performance, global generalization, and local personalization) is severely compromised by heterogeneous data distribution across consumer electronic devices, which is prevalent in most real-world scenarios. Prompt learning, as an efficient approach, is introduced to improve either global generalization or local personalization. They typically struggle to balance both directions, as optimizing one often leads to failure in another. To address this problem, we design a novel FL framework named PGA-FL, which fuses local and global prompts by Genetic Algorithm (GA) to update the local prompts on the clients and aggregates the uploaded local prompts based on their semantic similarity to update the global prompts on the server, to achieve a superior trade-off between the global generalization and the local personalization. To demonstrate the effectiveness of our PGA-FL framework, we theoretically prove its convergence performance and perform extensive experiments. Experimental results demonstrate that the proposed PGA-FL framework can achieve superior performance in terms of convergence performance, global generalization, and local personalization with respect to other state-of-the-art baseline algorithms.

Modern microservice systems exhibit continuous structural evolution in their runtime call graphs due to workload fluctuations, fault responses, and deployment activities. Despite this complexity, our analysis of over 500,000 production traces from ByteDance reveals a latent regularity: execution paths concentrate around a small set of recurring invocation patterns. However, existing resource management approaches fail to exploit this structure. Industrial autoscalers like Kubernetes HPA ignore inter-service dependencies, while recent academic methods often assume static topologies, rendering them ineffective under dynamic execution contexts. In this work, we propose Morphis, a dependency-aware provisioning framework that unifies pattern-aware trace analysis with global optimization. It introduces structural fingerprinting that decomposes traces into a stable execution backbone and interpretable deviation subgraphs. Then, resource allocation is formulated as a constrained optimization problem over predicted pattern distributions, jointly minimizing aggregate CPU usage while satisfying end-to-end tail-latency SLOs. Our extensive evaluations on the TrainTicket benchmark demonstrate that Morphis reduces CPU consumption by 35-38

Efficient container image distribution is crucial for enabling machine learning inference at the network edge, where resource limitations and dynamic network conditions create significant challenges. In this paper, we present PeerSync, a decentralized P2P-based system designed to optimize image distribution in edge environments. PeerSync employs a popularity- and network-aware download engine that dynamically adapts to content popularity and real-time network conditions. PeerSync further integrates automated tracker election for rapid peer discovery and dynamic cache management for efficient storage utilization. We implement PeerSync with 8000+ lines of Rust code and test its performance extensively on both large-scale Docker-based emulations and physical edge devices. Experimental results show that PeerSync delivers a remarkable speed increase of 2.72 & times;, 1.79 & times;, and 1.28 & times; compared to the Baseline solution, Dragonfly, and Kraken, respectively, while significantly reducing cross-network traffic by 90.72% under congested and varying network conditions.

Digital society increasingly relies on complex service ecosystems, formed by interconnected services from technology giants. However, the growing scale and intricate dependencies of these ecosystems pose significant challenges to their evolution, frequently leading to systemic failures during upgrades or restructuring. To address these challenges, current research is shifting from the perspective of single services to the ecosystem. Drawing upon the synthesis and induction of current research, we present a comprehensive survey focused on the study of service ecosystem evolution, employing a novel three-stage analytical framework. This framework structures the evolutionary lifecycle and provides a systematic way to organize and review existing research, filling the gap caused by the lack of surveys specifically focused on service ecosystem evolution. Additionally, we pioneer the application of complex network theory to analyze service ecosystem evolution, providing a novel perspective to capture the inherent connectivity, dynamics, and emergent properties often missed by traditional approaches. Building on our analysis, we point out critical research gaps and propose three specific future directions. We provide a robust theoretical foundation and methodological guidance for understanding and guiding the sustainable evolution of modern service ecosystems.

The wide spreading of Internet of Things (IoT) sensors generates vast spatio-temporal data streams, but ensuring data credibility is a critical yet unsolved challenge for applications like smart homes. While spatio-temporal graph (STG) models are a leading paradigm for such data, they often fall short in dynamic, human-centric environments due to two fundamental limitations: (1) their reliance on static graph topologies, which fail to capture physical, event-driven dynamics, and (2) their tendency to confuse spurious correlations with true causality, undermining robustness in human-centric environments. To address these gaps, we propose the Dynamic Causal Spatio-Temporal Graph Network (DyC-STG), a novel framework designed for real-time data credibility analysis in IoT. Our framework features two synergistic contributions: an event-driven dynamic graph module that adapts the graph topology in real-time to reflect physical state changes, and a causal reasoning module to distill causally-aware representations by strictly enforcing temporal precedence. To facilitate the research in this domain we release two new real-world datasets. Comprehensive experiments show that DyC-STG establishes a new state-of-the-art, outperforming the strongest baselines by 1.4 percentage points and achieving an F1-Score of up to 0.930.

Diffusion models have demonstrated exceptional generative capabilities but are computationally intensive, posing significant challenges for deployment in resource-constrained or latency-sensitive environments.Quantization offers an effective means to reduce model size and computational cost, with post-training quantization (PTQ) being particularly appealing due to its compatibility with pre-trained models without requiring retraining or training data.However, existing PTQ methods for diffusion models often rely on manual, architecture-specific heuristics that limit their generalizability and hinder integration with industrial deployment pipelines.To address these limitations, we propose SegQuant, a deployment-aware quantization framework that adaptively combines complementary techniques to enhance cross-model versatility.SegQuant consists of a segment-aware, graph-based quantization strategy (SegLinear) that captures structural semantics and spatial heterogeneity, along with a dual-scale quantization scheme (DualScale) that preserves polarity-asymmetric activations using a hardware-native dual-path computation, avoiding performance penalties from custom implementations, which is crucial for maintaining visual fidelity in generated outputs.SegQuant is broadly applicable beyond Transformer-based diffusion models, achieving strong performance while ensuring seamless compatibility with mainstream deployment tools.

The growth of large models demands multinode cooperation during training and inference processes. The computing node failures can interrupt these processes, subsequently causing information loss and prolonging the execution time. To reduce the prohibitively large overhead incurred by the computing nodes failures, the accurate prediction of computing node failure is vital, which can help to avert potential large overhead, service interruptions, and negative customer experiences. Existing solutions of computing nodes failure prediction mainly focus on utilizing state-of-the-art time-series models to enhance the performance of computing node failure prediction. However, on the one hand, they could not capture the causal relationship between device over-utilization and node failures; On the other hand, they fail to extract the complex spatial-temporal cascading correlations among computing node failure events. These limits can degrade the performance of computing node failure prediction. To address these above problems, this article makes an effort to focus on designing a continuous-time dynamic graphs-based computing node failures prediction (CTDG-NFP) scheme, to accurately predict in dynamic cluster environments. Specifically, the CTDG-NFP scheme first designs a novel multiple-dimensional feature-biased neighbor sampling method, which jointly considers CPU utilization-biased, memory utilization-biased, temporal-biased and spatial-biased, to sample relevant context. Then, the CTDG-NFP scheme extracts diverse computing node failure motifs by multiple-dimensional feature-biased-based long-short-path walk method and set-based anonymization method. Finally, the CTDG-NFP scheme adopts time encoder to encode these motifs, and thereby extracting the complex spatial-temporal correlations among computing node failure events. On this basis, contrastive learning is adopted to train the computing node failure prediction model. Extensive evaluations with various real-world failure traces demonstrate the CTDG-NFP scheme can achieve superior performance in terms of six widely used performance metrics compared with the SOTA node failure prediction methods.

Embodied Multi-Agent Systems have proven highly effective in addressing complex tasks through coordinated collaboration among heterogeneous agents. However, real-world environments and task specifications are inherently dynamic, exhibiting frequent changes, uncertainty, and variability. Despite these characteristics, most existing frameworks employ static architectures with fixed agent capabilities and rigid task allocation strategies, which substantially constrain their adaptability to evolving conditions. This inflexibility presents significant challenges to maintaining robust and efficient multi-agent cooperation in dynamic and unpredictable settings.To address these limitations, we propose DRAMA, short for Dynamic Orchestration for Resilient Multi-Agent Ecosystems, tailored for rapidly changing environments. DRAMA adopts a multilayer architecture that incorporates three principal mechanisms: adaptive scheduling through an affinity-driven mechanism, fault-tolerant continuity via hierarchical trust-chain task takeover, and collective spatial intelligence that consolidates distributed observations for predictive reasoning. Together, these components enable event-triggered rescheduling and decentralized fault recovery, ensuring uninterrupted task execution amid agent arrivals, dropouts, or recoveries. Extensive experiments in the embodied VirtualHome-Social environment demonstrate that DRAMA achieves a 7% improvement in runtime efficiency and a 10% increase in throughput compared with state-of-the-art baselines, while maintaining superior stability and robustness under dynamic agent populations.

Large-scale multimodal pre-trained models like CLIP rely heavily on high-quality training data, yet raw web-crawled datasets are often noisy, misaligned, and redundant, leading to inefficient training and suboptimal generalization. Existing data selection methods are either heuristic-based, suffering from bias and limited diversity, or data-driven but task-agnostic, failing to optimize for multi-task scenarios. To address these gaps, we introduce TADS (Task-Aware Data Selection), a novel framework for multi-task multimodal pre-training that integrates Intrinsic Quality, Task Relevance, and Distributional Diversity into a learnable value function. TADS employs a comprehensive quality assessment system with unimodal and cross-modal operators, quantifies task relevance via interpretable similarity vectors, and optimizes diversity through cluster-based weighting. A feedback-driven meta-learning mechanism adaptively refines the selection strategy based on proxy model performance across multiple downstream tasks. Experiments on CC12M demonstrate that TADS achieves superior zero-shot performance on benchmarks like ImageNet, CIFAR-100, MS-COCO, and Flickr30K, using only 36

With the rapid development of on-board applications, edge computing is now widely used in Internet of vehicles, enabling vehicles with limited resources to offload tasks to the edge for execution via computation offloading. However, current research methods are often hard to adapt to dynamic scenarios due to model training costs and vehicle mobility and also lack consideration for load balancing in high-load situations. To improve the quality of experience of users and balance the load of edge servers simultaneously, this paper proposes an improved greedy strategy method for computation offloading. First, to mitigate potential communication overload during peak hours, this study analyzes the relationship between transmission scheduling and execution queues, and investigates a dynamic task transmission control method. Second, explicit modeling of round-trip communication reliability in mobile environments is provided to extend the vehicle interconnection model. Subsequently, analyzing the structure of the optimal solution for total latency optimization, the priority of offloaded tasks classified. A multi-perspective analysis of task offloading is then conducted, and a greedy strategy is adopted ensure both the quality of user experience and load balancing at the edge. Finally, comparative experiments real-world datasets validate the efficiency of the proposed method and model under high-mobility and high-load experimental scenarios.

Facial expression recognition (FER) has gained significant attention due to its diverse applications. Achieving accurate facial expression recognition requires the consideration of both global central features and subtle local features. To this end, we propose the Multi-level and Multi-scale Network (MM-Net), an FER network that leverages both multi-level and multi-scale attention mechanisms. Specifically, we design a multi-hierarchical feature learning mechanism to facilitate the FER task with Multi-Level Attention Block (MLAB) and Multi-scale Attention Block (MSAB). The MLAB focuses on learning fine-grained features with adaptive attention across different blocks in the shallow network. Meanwhile, the MSAB facilitates the multi-scale fusion of deep features, enabling the network to capture richer semantic information and expression of features. In addition, we propose Limited Center Loss, which optimizes the network by minimizing the distance between the same classes while increasing the gap between different classes. Experimental results on public datasets show that our proposed MM-Net outperforms current state-of-the-art methods, achieving results of 90.42% on the RAF-DB dataset, 90.05% on FERPlus, 65.91% on AffectNet, and 57.52% on SFEW.

Learning-augmented paging has been extensively studied in recent years. A key advantage over naive ML-based approaches is bounded robustness, which guarantees worst-case performance even when predictions are inaccurate, making these algorithms valuable for real-world systems. Prior work achieves robustness bounds of 2H_k + O(1) in the randomized setting, leaving a gap to the optimal competitive ratio H_k. In this paper, we study how to close this gap. We begin by reviewing online optimality and proving a new property of the latest H_k-competitive algorithm, which facilitates our analysis in the learning-augmented setting. Then, we review existing learning-augmented paging algorithms and introduce a unifying primitive, the relative prediction budget, which captures the essence of establishing robustness and reveals that prior algorithms either overuse or underutilize predictions. Guided by the above analysis, we develop a new framework that achieves the best-possible robustness up to an additive constant for learning-augmented paging: H_k + O(1). Experiments further demonstrate strong practical performance.