Multi-view clustering (MVC), as an important machine learning task, aims to group data into distinct groups by leveraging complementary and consistent information across multiple views. During the last two decades, it has been widely studied, and many methods have been proposed, which has brought incredible development to this field. However, few works comprehensively summarize existing methods and point out the potential challenges in this field for the next decades. To this end, our survey thoroughly reviews existing MVC methods according to three taxonomies, i.e., techniques, fusion strategies, and scenarios. Specifically, seven typical techniques, four fusion strategies, and five typical scenarios are included. Besides, we also collect the commonly used datasets and analyze the performance of typical MVC methods. Moreover, we summarize six application scenarios of existing MVC methods ranging from computer vision, and information retrieval tasks to medical diagnosis and bio-informatics. In particular, we point out seven interesting future directions in this field, which will definitely enlighten the readers.

Mobile software engineering has been a hot research topic for decades. Our fellow researchers have proposed various approaches (with over 7,000 publications for Android alone) in this field that essentially contributed to the great success of the current mobile ecosystem. Existing research efforts mainly focus on popular mobile platforms, namely Android and iOS. OpenHarmony, a newly open-sourced mobile platform, has rarely been considered, although it is the one requiring the most attention as OpenHarmony is expected to occupy one-third of the market in China (if not in the world). To fill the gap, we present to the mobile software engineering community a research roadmap for encouraging our fellow researchers to contribute promising approaches to OpenHarmony. Specifically, we start by presenting a literature review of mobile software engineering, attempting to understand what problems have been targeted by the mobile community and how they have been resolved. We then summarize the existing (limited) achievements of OpenHarmony and subsequently highlight the research gap between Android/iOS and OpenHarmony. This research gap eventually helps in forming the roadmap for conducting software engineering research for OpenHarmony.

Multimodal Continual Instruction Tuning (MCIT) is essential for sequential task adaptation of Multimodal Large Language Models (MLLMs) but is severely restricted by catastrophic forgetting. While existing literature focuses on the reasoning language backbone, in this work, we expose a critical yet neglected dual-forgetting phenomenon across both perception drift in Cross-modal Projection Space and reasoning collapse in Low-rank Parameter Space. To resolve this, we present MAny (Merge Anything), a framework that merges task-specific knowledge through Cross-modal Projection Merging (CPM) and Low-rank Parameter Merging (LPM). Specifically, CPM recovers perceptual alignment by adaptively merging cross-modal visual representations via visual-prototype guidance, ensuring accurate feature recovery during inference. Simultaneously, LPM eliminates mutual interference among task-specific low-rank modules by recursively merging low-rank weight matrices. By leveraging recursive least squares, LPM provides a closed-form solution that mathematically guarantees an optimal fusion trajectory for reasoning stability. Notably, MAny operates as a training-free paradigm that achieves knowledge merging via efficient CPU-based algebraic operations, eliminating additional gradient-based optimization beyond initial tuning. Our extensive evaluations confirm the superior performance and robustness of MAny across multiple MLLMs and benchmarks. Specifically, on the UCIT benchmark, MAny achieves significant leads of up to 8.57% and 2.85% in final average accuracy over state-of-the-art methods across two different MLLMs, respectively.

Recent findings suggest that even modest amounts of well-crafted instruction data can drive Large Language Models (LLMs) to achieve competitive performance through fine-tuning. This has sparked growing research interest in instruction data mining, which focuses on automatically extracting high-quality instruction samples from vast datasets. Among the existing approaches, the Adversarial Instruction-Following Difficulty approach has emerged as a prominent technique, highlighting that considering the adversarial robustness of LLMs can more effectively identify high-quality instruction data. However, this approach introduces significant computational overhead, primarily arising from the highly demanding procedures required to generate and comprehensively evaluate adversarial examples across large-scale datasets. In this work, we introduce a novel model compression-driven framework for high-quality instruction data mining, enhanced with integrated adversarial attack strategies. Model compression via structured pruning reduces computational overhead, while three types of adversarial perturbations including granular, multilingual and out-of-distribution attacks are applied to analyze the robustness characteristics of LLMs. This process helps identify more challenging and informative instruction data in an efficient manner. Notably, the compressed proxy model runs on a single GPU without requiring large-scale clusters, enhancing scalability and deployability in resource-constrained settings. Furthermore, we propose the Lightweight Adversarial Instruction-Following Difficulty metric, a surrogate metric that quantifies the divergence between a compressed model’s outputs with and without the instruction, thereby guiding the extraction of high-quality instruction data. We validate the effectiveness of our framework on widely adopted benchmark datasets and evaluate the performance of LLMs from LLaMA, Mistral, Gemma, and DeepSeek model families across the code generation and general domain tasks.

Speculative decoding (SD) accelerates LLM inference by verifying draft tokens in parallel. However, this method presents a critical trade-off: it improves throughput in low-load, memory-bound systems but degrades performance in high-load, compute-bound environments due to verification overhead. Existing speculative decoding methods use fixed lengths and cannot adapt to workload changes or decide when to stop speculation. The cost of restarting speculative inference also remains unquantified. Under high load, the benefit of speculation diminishes, while retaining the draft model reduces KV cache capacity, limiting batch size and degrading throughput. To overcome this, we propose Nightjar, a resource-aware adaptive speculative framework. It first adjusts to the request load by dynamically selecting the optimal speculative length for different batch sizes. Crucially, Nightjar proactively disables speculative decoding when the MAB planner determines that speculation is no longer beneficial, and during the disabled phase, offloads the draft model to the CPU only under GPU memory pressure. This reclaims memory for the KV cache, thereby facilitating larger batch sizes and maximizing overall system throughput. Experiments show that Nightjar achieves up to 14.76% higher throughput than standard speculative decoding and up to 20.18% lower latency in the main benchmark suite under dynamic request arrival rates for real-time LLM serving scenarios.

Traceability link recovery between requirements and source code is vital for software quality assurance and evolution analysis. Although automated traceability techniques have advanced greatly, the large semantic gap between vague natural-language requirements and precise source code still hinders accurate link recovery. Most existing approaches optimize traceability algorithms yet ignore the inherent quality of requirement descriptions, which prevents fundamental reduction of the semantic gap. This work proposes a requirement-oriented normalization method. Using controlled natural language and large-language-model-based prompt engineering, raw requirements are decomposed and converted into standardized use-case specifications to strengthen semantic representation and mitigate semantic divergence. Evaluated on four public datasets under two typical traceability frameworks, the normalized specifications improve tracing performance for semantically ambiguous raw requirements. However, over-normalization may degrade results for already high-quality requirements closely aligned with code semantics. The results validate source-side requirement normalization as a promising strategy for traceability improvement and reveal its applicable boundaries for practical usage.

Software technology is undergoing a paradigm shift driven by two converging trends. First, the scope of software responsibility has expanded significantly: as “software-defined everything” becomes a reality, software has evolved into the integration core of sociocyber-physical systems (SCPSs). Second, the capabilities and development methods of software are being greatly enhanced by recent breakthroughs in artificial intelligence (AI). This article presents perspectives and observations on software engineering in this era of rapid progress. We aim to outline a set of foundational challenges in engineering SCPSs and highlight the need for innovative software solutions that extend beyond AI technologies alone. Specifically, we examine the need for a new paradigm that can address the complexities introduced by SCPSs, which challenge conventional paradigms through the blurring of system boundaries, continuous lifecycle evolution, and the embracing of inherent uncertainty. We highlight new engineering principles of socio-technical co-design, cyber-physical integration, and development-operation convergence, and a knowledge- and data-driven approach to taming uncertainty. Emerging proposals, including digital humanism, agentic SCPS, ubiquitous operating system, and continuous quality assurance, are discussed alongside possible extensions to existing technologies. We then outline key research directions for both runtime support and quality assurance. On the runtime side, we argue for a new generation of software infrastructure for SCPSs, including unified hardware abstractions, scalable and resilient runtime systems, AI-enabled system management, and human-centric operating system primitives. On the assurance side, we highlight the need for new approaches combining unified socio-cyber-physical modeling, specification of both technical and non-technical properties, data-driven simulation and testing, continuous verification, and runtime monitoring under uncertainty. Furthermore, we present specific challenges within key application domains, including intelligent vehicles, smart manufacturing, and smart cities. In doing so, we aim to stimulate discussion within the software engineering community and encourage support from industry and government to address the critical engineering and governance issues inherent to this new generation of systems.

Knowledge Graphs (KGs) can enhance Large Language Models (LLMs) by providing a structured reasoning foundation, thereby improving accuracy in knowledge-intensive question answering. However, LLMs struggle to identify valuable reasoning paths within the complex and often incomplete structures of KGs. To address this, we propose Search-on-Graph (SoG), a retrieval-enhanced framework for faithful LLM reasoning over KGs. SoG introduces a dynamic entity value metric to guide and iteratively refine the search process based on path rewards assessed by the LLM. Additionally, it incorporates real-time retrieval of unstructured information to compensate for KG incompleteness, supporting both answer generation and path evaluation. Extensive experiments show that SoG achieves state-of-the-art performance on four of five real-world datasets with GPT-3.5 and on all five with GPT-4.

Introduction Existing serverless workflow orchestration systems are predominantly designed for a single-cloud FaaS system, leading to vendor lock-in. This dependency restricts performance optimization, cost reduction, and the overall availability of applications. However, orchestrating serverless workflows on Jointcloud FaaS systems faces two main challenges: (1) additional overhead caused by centralized cross-cloud orchestration; and (2) a lack of reliable failover and fault-tolerant mechanisms for cross-cloud serverless workflows.Methods To address these challenges, we propose Joint lambda, a distributed runtime system designed to orchestrate serverless workflows on multiple FaaS systems without relying on a centralized orchestrator. Joint lambda introduces a compatibility layer, Backend-Shim, which leverages inter-cloud heterogeneity to optimize makespan and reduce costs with on-demand billing. By using function-side orchestration instead of centralized nodes, it enables independent function invocations and data transfers, thereby minimizing cross-cloud communication overhead. For high availability, it ensures exactly-once execution via datastores and failover mechanisms for serverless workflows on Jointcloud FaaS systems.Results We validate Joint lambda on two heterogeneous FaaS systems, AWS and Aliyun, using four representative workflows. Compared to the most advanced commercial orchestration services for single-cloud serverless workflows, Joint lambda reduces makespan by up to 3.3 & times; while saving up to 65% in cost. Furthermore, Joint lambda is up to 4.0 & times; faster than state-of-the-art orchestrators for cross-cloud serverless workflows, while achieving competitive cost performance in representative scenarios.Conclusions The evaluation demonstrates that Joint lambda effectively eliminates vendor lock-in and mitigates cross-cloud communication overhead without sacrificing economic efficiency. By incorporating decentralized function-side orchestration and robust failover mechanisms, it provides strong execution guarantees and high availability for complex serverless workflows across heterogeneous Jointcloud FaaS environments.

JointCloud computing enables multiple cloud domains to collaboratively process and share data across organizational boundaries. In such settings, sensitive data are distributed among mutually distrustful participants and require strong protection during storage and retrieval. Blockchain offers a decentralized trust foundation for JointCloud systems through tamperevident logging and coordinated data management, yet efficient search over encrypted on-chain records remains insufficiently supported. Existing authenticated blockchain query structures are largely designed for plaintext predicates, whereas searchableencryption solutions typically rely on off-chain indexes or hybrid execution, which may enlarge the trust boundary and complicate result verification. This paper presents OnChainSE, a blockchain-native architecture for verifiable search over encrypted blockchain data in JointCloud environments. Unlike conventional authenticated query frameworks for plaintext attributes, OnChainSE targets settings in which both searchable attributes and query inputs must remain protected across collaborating cloud domains. It integrates encrypted keywords, encrypted numerical metadata, and authenticated index commitments into a unified blockchainnative structure to support privacy-preserving keyword and range queries. It further enables succinct per-block verification through proof aggregation, allowing lightweight clients to validate the authenticity of returned encrypted pointers and check their consistency with committed on-chain search state without reconstructing the full search state. Compact checkpoint summaries are also introduced to narrow the search scope and reduce historical retrieval overhead before proof generation. We implement an OnChainSE prototype and evaluate it under diverse data scales and query workloads. Experimental results show that OnChainSE reduces query overhead and proof cost under the evaluated workloads while preserving verifiable retrieval for encrypted data. These results suggest that OnChainSE is a practical approach to authenticated encrypted retrieval in blockchain-enabled JointCloud systems.

The advent of advanced technologies, such as large models, has precipitated a surging demand for computational resources, thereby driving the evolution of cloud computing services from single-cloud architectures to multi-cloud paradigms. The accompanying question is how to enable resources from different cloud service providers to collaborate efficiently. Although existing works have explored multi-cloud scheduling methods, most of these works are centralized scheduling, where the decision-maker can schedule computing resources across all clouds. However, this is nearly impossible given the current situation in which computing resources in different clouds come from various cloud service providers. In response to this challenge, JointCloud, a multi-cloud cooperation architecture, has been proposed, which aims at enhancing the cooperation among multiple clouds to provide efficient multi-cluster services. Following the idea of JointCloud, proposes a multi-objective evolutionary algorithm (MOEA) based method for task scheduling in multi-cluster cloud computing environments, without intervening in the intra-cluster scheduling scheme. In the proposed method, we construct a mathematical model with the optimization objectives of minimizing overall waiting time and load imbalance between clusters based on the actual operation data in China Computing NET ((CNET)-N-2). In addition, we also develop an MOEA specifically tailored to address this problem. The performance of the proposed MOEA and existing state-of-the-art MOEAs is examined on the proposed problems. Comparison results highlight the promising performance of the proposed MOEA, the specifically tailored algorithm in effectively addressing the multi-cluster task scheduling problem. In addition, we also compared the results of the MOEAs with the results of three classical scheduling methods, the results proved the effectiveness of the MOEA-based method on this problem.

This paper examines audio self-supervised learning (SSL) through the alignment between pretraining objectives, architectural inductive biases, and downstream applications. Rather than treating SSL methods as a chronological sequence of pretext tasks or model families, we ask how different supervisory signals shape the representations that models are expected to learn. The discussion is organized around five paradigms: auxiliary tasks, contrastive learning, generative reconstruction, discrete token prediction, and multimodal alignment. These objectives place different demands on the model, from local structural sensitivity and contrastive invariance to contextual inference, discrete semantic abstraction, and multimodal grounding. We relate these demands to the biases of CNNs, recurrent and State Space Models, Transformers, and hybrid architectures, showing how local acoustic compression, sequential state propagation, content-dependent global routing, and local–global integration support different forms of audio SSL. The same view is then used to interpret downstream applications in speech processing, environmental sound analysis, music information retrieval, medical and bioacoustic analysis, and multimodal audio understanding as practical tests of whether learned representations and architectural choices generalize across domains. We also review benchmark protocols and open challenges, including tokenization bottlenecks, long-context efficiency, robustness, and secure multimodal deployment, and discuss how codec-based tokenization and audio-language modeling extend this objective–architecture–application pipeline. The accompanying repository is released at https://github.com/colaudiolab/Awesome-Self-Supervised-Audio-Learning.

The Fork and Pull-Request model is widely used in collaborative development of open source software (OSS), fostering innovation through independent repository copies, but it can also lead to inefficiencies and fragmentation. A key underexplored aspect is the integration effectiveness—the degree to which distributed original commits across forks are effectively integrated back into the main repository. It plays a critical role in OSS project productivity but remains poorly understood. In response, we introduce convergence entropy, a novel metric that quantifies the integration effectiveness by measuring the similarity between distributions of original and merged commits across forks, adjusted for integration ratio. This metric highlights not only the volume of contributions but also their diversity and coordination, offering a unique lens to understand forking practices. Moreover, we explore the relationship between convergence entropy and three dimensions of OSS project productivity, showing significant correlations. We also observe that other factors can alter this dynamic.

Root cause analysis (RCA) in microservice systems is challenging because observable symptoms often propagate along service dependencies and become separated from the actual underlying cause. Recent large language model (LLM)-based agent frameworks have shown promise for automating RCA, but existing methods still suffer from uncontrolled context growth, weakly constrained diagnostic workflows, and unreliable conclusions under inconsistent evidence from multiple sources. To address these issues, we propose Context-RCA, a context-controlled multi-agent framework for microservice fault diagnosis. Context-RCA has three key components. First, it uses selective context injection to organize agent inputs through domain isolation, stage awareness, and dynamic context updates. Second, it uses an SOP-driven workflow to enforce structured diagnosis across stages and prevent premature conclusions before sufficient evidence is collected. Third, it uses iterative consensus verification to cross-check hypotheses against logs, metrics, and traces and refine judgments from observed symptoms to deeper root causes. Experiments on 400 fault cases from the 2025 International AIOps Challenge dataset show that Context-RCA achieves an overall score of 65.93 on DeepSeek-V3.2 and 60.06 on GPT-4o, outperforming the strongest baseline by 14.21% and 24.30%, respectively. Context-RCA also improves fault localization and root cause description quality. Ablation results further show that selective context injection contributes a 5.51% gain, confirming context control as a primary factor in reliable RCA with LLMs.

Real-world constrained multi-objective optimization problems (CMOPs) commonly involve multiple constraints, and understanding and exploiting their coupling relationships is crucial for efficient optimization. Recent constraint-decoupling methods handle individual constraints separately, but they generally search only in the evolutionary direction to approximate single-constraint Pareto fronts (SCPFs). In this study, we show that part or all of the constrained Pareto front (CPF) may be unrelated to any SCPF and instead be shaped by the boundaries of infeasible regions. We refer to such a portion as the independent CPF (ICPF) and introduce the reverse CPF (RCPF) to characterize its associated informative infeasible boundaries. Based on these observations, we propose a bidirectional constraint-decoupling coevolutionary algorithm named DCF2D. DCF2D dynamically identifies the constraints obstructing the main population and activates constraint-specific auxiliary populations. These populations adaptively search in the evolutionary direction for the corresponding SCPFs or in the reverse evolutionary direction for the corresponding RCPFs. Its three-stage framework integrates unconstrained global exploration, event-driven bidirectional coevolution, and final convergence refinement. Experiments on 87 benchmark instances from seven test suites and 28 real-world engineering CMOPs demonstrate that DCF2D achieves the best overall performance among nine algorithms. Code available at: https://github.com/RuiqingS/DCF2D.

Reinforcement learning from human feedback (RLHF) emerges as a promising paradigm for aligning large language models (LLMs). However, a notable challenge in RLHF is overoptimization, where beyond a certain threshold, the pursuit of higher rewards leads to a decline in human preferences. In this paper, we observe the weakness of KL regularization which is commonly employed in existing RLHF methods to address overoptimization. To mitigate this limitation, we scrutinize the RLHF objective in the offline dataset and propose uncertainty-penalized RLHF (UP-RLHF), which incorporates uncertainty regularization during RL-finetuning. To enhance the uncertainty quantification abilities for reward models, we first propose a diverse low-rank adaptation (LoRA) ensemble by maximizing the nuclear norm of LoRA matrix concatenations. Then we optimize policy models utilizing penalized rewards, determined by both rewards and uncertainties provided by the diverse reward LoRA ensembles. Our experimental results, based on two real human preference datasets, showcase the effectiveness of diverse reward LoRA ensembles in quantifying reward uncertainty. Additionally, uncertainty regularization in UP-RLHF proves to be pivotal in mitigating overoptimization, thereby contributing to the overall performance.

Continual Generalized Category Discovery (C-GCD) seeks to incrementally discover new categories from unlabeled data and memorize old categories’ knowledge, fostering model adaptability in real-world scenarios. Especially, the unlabeled data is from both old and new classes, requiring the model to recognize previously learned classes while discovering. In response, recent efforts focus on devising specific frameworks and various anti-forgetting strategies, striving for a typical stability-plasticity trade-off. Unlike previous studies, in this work, we first revisit these methods and identify that most of these methods over-protect old classes, hampering the accurate discovery of novel ones. To address this challenge, we introduce the Decouple Your Discovery and Memory (DYDM), a dual-branch architecture that decouples the discovery of new classes and the memorization of old classes. The discovery branch is focused on accurately recognizing new classes, while the memory branch consolidates all identified categories in a recursive manner and functions as the inference branch. Importantly, benefiting from the strong knowledge retention ability of the memory branch, the discovery branch can facilitate the recognition of novel classes from the unlabeled data, achieving a win-win outcome between plasticity and stability. Extensive experiments on various datasets and settings demonstrate the superiority of our approach, achieving leads of up to 9.87%, 7.30%, 3.18%, and 8.25%. Furthermore, our framework can integrate with existing approaches, consistently enhancing their performance.

Existing evolutionary algorithms for Constrained Multi-objective Optimization Problems (CMOPs) typically treat all constraints uniformly, overlooking their distinct geometric relationships with the true Constrained Pareto Front (CPF). In reality, constraints play different roles: some directly shape the final CPF, some create infeasible obstacles, while others are irrelevant. To exploit this insight, we propose a novel algorithm named RCCMO, which sequentially performs unconstrained exploration, single-constraint exploitation, and full-constraint refinement. The core innovation of RCCMO lies in a constraint prioritization method derived from these geometric insights, seamlessly coupled with a unique dual-directional search mechanism. Specifically, RCCMO first prioritizes constraints that constitute the final CPF, approaching them from the evolutionary direction (optimizing objectives) to locate the CPF directly shaped by single-constraint boundaries. Subsequently, for constraints that merely hinder the population's progress, RCCMO searches from the anti-evolutionary direction (targeting the infeasible boundaries where hindering constraints intersect with the CPF) to effectively discover how these constraints obstruct and form the final CPF. Meanwhile, irrelevant constraints are intentionally bypassed. Furthermore, a series of specialized mechanisms are proposed to accelerate the algorithm's execution, reduce heuristic misjudgments, and dynamically adjust search directions in real time. Extensive experiments on 5 benchmark test suites and 29 real-world CMOPs demonstrate that RCCMO significantly outperforms seven state-of-the-art algorithms.

Active Learning (AL) aims to reduce data annotation costs by selecting the most informative samples from an unlabeled data pool. Traditional AL methods often rely on a single snapshot to identify uncertain or representative samples, often overlooking the poor generalization of a single model. Recent AL studies have attempted to address this issue by tracking a broader range of training dynamics for data selection, typically using averaging or accumulating manner. However, both our theoretical and experimental analyses reveal that these methods obscure the variability inherent in the training process, potentially prioritizing hard-to-learn samples that result in poor generalization. In this paper, we propose a novel AL method termed as Dynamic Confidence Variance (DCoV), that seamlessly integrates variability with the training dynamic to effectively identify a well-generalized Coreset. DCoV leverages the variance of the model’s prediction confidence throughout the training process for active sampling and model training. Our theoretical analysis demonstrates that DCoV provides a lower bound on the population risk of the model learned from selected labeled subset, spanning the entire training process. Extensive experiments demonstrate that our approach significantly outperforms existing state-of-the-art AL methods on various balanced and imbalanced benchmark datasets across various modalities.