Artificial Intelligence Review最新文献

{"title":"A survey on cutting-edge relation extraction techniques based on language models","authors":"Jose A. Diaz-Garcia,&nbsp;Julio Amador Diaz Lopez","doi":"10.1007/s10462-025-11280-0","DOIUrl":"10.1007/s10462-025-11280-0","url":null,"abstract":"<div><p>This comprehensive survey examines the latest advancements in Relation Extraction (RE), a pivotal task in natural language processing essential for applications across biomedical, financial, and legal sectors. This study highlights the evolution and current state of RE techniques by analyzing 137 papers presented at the Association for Computational Linguistics (ACL) conferences from 2020 to 2023, focusing on models that leverage language models. Our findings underscore the dominance of BERT-based methods in achieving state-of-the-art results for RE while also noting the promising capabilities of emerging Large Language Models (LLMs) like T5, especially in few-shot relation extraction scenarios where they excel in identifying previously unseen relations.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-025-11280-0.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145142000","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From concept to manufacturing: evaluating vision-language models for engineering design","authors":"Cyril Picard,&nbsp;Kristen M. Edwards,&nbsp;Anna C. Doris,&nbsp;Brandon Man,&nbsp;Giorgio Giannone,&nbsp;Md Ferdous Alam,&nbsp;Faez Ahmed","doi":"10.1007/s10462-025-11290-y","DOIUrl":"10.1007/s10462-025-11290-y","url":null,"abstract":"<div><p>Engineering design is undergoing a transformative shift with the advent of AI, marking a new era in how we approach product, system, and service planning. Large language models have demonstrated impressive capabilities in enabling this shift. Yet, with text as their only input modality, they cannot leverage the large body of visual artifacts that engineers have used for centuries and are accustomed to. This gap is addressed with the release of multimodal vision-language models (VLMs), such as GPT-4V, enabling AI to impact many more types of tasks. Our work presents a comprehensive evaluation of VLMs across a spectrum of engineering design tasks, categorized into four main areas: Conceptual Design, System-Level and Detailed Design, Manufacturing and Inspection, and Engineering Education Tasks. Specifically in this paper, we assess the capabilities of two VLMs, GPT-4V and LLaVA 1.6 34B, in design tasks such as sketch similarity analysis, CAD generation, topology optimization, manufacturability assessment, and engineering textbook problems. Through this structured evaluation, we not only explore VLMs’ proficiency in handling complex design challenges but also identify their limitations in complex engineering design applications. Our research establishes a foundation for future assessments of vision language models. It also contributes a set of benchmark testing datasets, with more than 1000 queries, for ongoing advancements and applications in this field.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-025-11290-y.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145142100","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Generative AI: a double-edged sword in the cyber threat landscape","authors":"Werisha Ibrar,&nbsp;Danish Mahmood,&nbsp;Ahmad Sami Al-Shamayleh,&nbsp;Ghufran Ahmed,&nbsp;Salman Z. Alharthi,&nbsp;Adnan Akhunzada","doi":"10.1007/s10462-025-11285-9","DOIUrl":"10.1007/s10462-025-11285-9","url":null,"abstract":"<div><p>Generative AI’s swift progress advances onto profound cybersecurity dilemmas. Its usage by malevolent entities to automate intricate malware creation poses a significant threat, circumventing conventional defensive measures. This paradigm shift enables the generation of polymorphic malware, eluding signature-based detection and facilitating precision-targeted assaults. The democratization of Generative AI exacerbates these threats by extending advanced capabilities to a broader spectrum of malicious actors. A comprehensive examination of AI-generated malware’s prevalence, and its repercussions is imperative to fortify cyber resilience. Such scrutiny informs proactive defense strategies vital for safeguarding digital assets within increasingly interconnected systems. Robust threat intelligence frameworks and AI-centric defensive mechanisms emerge as imperative shields against evolving cyber perils. Addressing this emergent challenge stands as an indispensable endeavor in contemporary cybersecurity discourse.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-025-11285-9.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145160753","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Comprehensive review of reinforcement learning for medical ultrasound imaging","authors":"Hanae Elmekki,&nbsp;Saidul Islam,&nbsp;Ahmed Alagha,&nbsp;Hani Sami,&nbsp;Amanda Spilkin,&nbsp;Ehsan Zakeri,&nbsp;Antonela Mariel Zanuttini,&nbsp;Jamal Bentahar,&nbsp;Lyes Kadem,&nbsp;Wen-Fang Xie,&nbsp;Philippe Pibarot,&nbsp;Rabeb Mizouni,&nbsp;Hadi Otrok,&nbsp;Shakti Singh,&nbsp;Azzam Mourad","doi":"10.1007/s10462-025-11268-w","DOIUrl":"10.1007/s10462-025-11268-w","url":null,"abstract":"<div><p>Medical Ultrasound (US) imaging has seen increasing demands over the past years, becoming one of the most preferred imaging modalities in clinical practice due to its affordability, portability, and real-time capabilities. However, it faces several challenges that limit its applicability, such as operator dependency, variability in interpretation, and limited resolution, which are amplified by the low availability of trained experts. This calls for the need of autonomous systems that are capable of reducing the dependency on humans for increased efficiency and throughput. Reinforcement Learning (RL) comes as a rapidly advancing field under Artificial Intelligence (AI) that allows the development of autonomous and intelligent agents through rewarded interactions with their environments. Several existing surveys on advancements in US imaging predominantly focus on partially autonomous AI solutions. However, none of these surveys explore the intersection between the stages of the US process and the recent advancements in RL solutions. To bridge this gap, this survey proposes a comprehensive taxonomy that integrates the stages of the US process with the RL development pipeline -including data preparation, problem formulation, simulation environment, RL training, validation and finetuning- and reviews current research efforts under this taxonomy. This work aims to highlight the potential of RL in building autonomous US solutions while identifying limitations and opportunities for further advancements in this field.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12185671/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144493826","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Tianji’s horse racing optimization (THRO): a new metaheuristic inspired by ancient wisdom and its engineering optimization applications","authors":"Liying Wang,&nbsp;Haiping Du,&nbsp;Zhenxing Zhang,&nbsp;Gang Hu,&nbsp;Seyedali Mirjalili,&nbsp;Nima Khodadadi,&nbsp;Abdelazim G. Hussien,&nbsp;Yingying Liao,&nbsp;Weiguo Zhao","doi":"10.1007/s10462-025-11269-9","DOIUrl":"10.1007/s10462-025-11269-9","url":null,"abstract":"<div><p>In this study, we introduce a novel metaheuristic algorithm named Tianji’s horse racing optimization (THRO), inspired by the Chinese historical story of Tianji’s horse racing. The story illustrates how Tianji leveraged his strengths to counteract his opponent’s weaknesses, ultimately leading to his victory in the competition. This strategic principle, which led to Tianji’s victory, forms the foundation of THRO’s design. The need for such a proposal arises from the limitations of existing optimization algorithms, which often struggle with convergence speed and solution accuracy when solving complex problems. THRO addresses these challenges by employing a unique dynamic individual matching strategy that enhances the algorithm’s convergence rate and solution precision. In this algorithm, an effective greedy strategy is employed to maximize benefits by selecting individuals from its population and matching them with individuals from the opponent’s population, thereby facilitating individual updates. This paper provides mathematically grounded explanations and analysis of how the algorithm converges to the global optimum with probability 1. To validate the efficacy of THRO, comparative experiments with 12 popular algorithms are conducted on 23 classical benchmark functions and the CEC2017 test suite. For the 29 CEC2017 functions across 10, 30, 50, and 100 dimensions, THRO achieves the slowest Friedman average ranking values among all competing methods, which are 2.052, 2.500, 2.293, and 2.259, respectively. Additionally, we conduct a comprehensive comparison with several advanced algorithms, including high-performance hybrid optimizers and the CEC winners, across the CEC2014, CEC2017, CEC2020, and CEC2022 suites, where THRO again achieves the slowest Friedman average ranking value of 1.729. Furthermore, six engineering design problems are employed to comprehensively check the applicability of THRO. Eventually, THRO’s proficiency extends to the application of identifying damping parameters of magnetorheological damper (MRD) models in mechanical systems. The results confirm that THRO exhibits remarkable competitiveness in solving various complex problems.The source code of THRO is publicly available at https://github.com/zwg770123/THRO.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-025-11269-9.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145168600","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adaptive anomaly detection for identifying attacks in cyber-physical systems: A systematic literature review","authors":"Pablo Moriano,&nbsp;Steven C. Hespeler,&nbsp;Mingyan Li,&nbsp;Maria Mahbub","doi":"10.1007/s10462-025-11292-w","DOIUrl":"10.1007/s10462-025-11292-w","url":null,"abstract":"<div><p>Modern cyberattacks in cyber-physical systems (CPS) rapidly evolve and cannot be deterred effectively with most current methods, which focus on characterizing past threats. Adaptive anomaly detection (AAD) is among the most promising techniques to detect evolving cyberattacks, with an emphasis on fast data processing and model adaptation. AAD has been researched extensively; however, to the best of our knowledge, our work is the first systematic literature review (SLR) on current research in this field. We present a comprehensive SLR, gathering 397 relevant papers and systematically analyzing 65 of them (47 research and 18 survey papers) on AAD in CPS from 2013 to November 2023. We introduce a novel taxonomy considering attack types, CPS application, learning paradigm, data management, and algorithms. Our findings show that most studies addressed either model adaptation or data processing, but rarely both simultaneously. This indicates a research gap in fully adaptive solutions. We also categorize algorithms, datasets, and attack characteristics, and summarize strengths and weaknesses across the literature. Our review provides a structured and accessible reference for researchers and practitioners, offering insights into key trends and highlighting limitations in current approaches. Finally, we outline several future research directions, including the need for integrated real-time processing and adaptive learning, explainability, and uncertainty quantification in AAD for CPS.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-025-11292-w.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145168461","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Self-organizing fuzzy neural network with adaptive evolution strategy for nonlinear and nonstationary processes","authors":"Xi Meng,&nbsp;Qizheng Hou,&nbsp;Limin Quan,&nbsp;Junfei Qiao","doi":"10.1007/s10462-025-11283-x","DOIUrl":"10.1007/s10462-025-11283-x","url":null,"abstract":"<div><p>Fuzzy neural networks, which combine the strengths of fuzzy logic systems and artificial neural networks, prove to be effective in modeling industrial processes. However, because of the nonlinearity and nonstationarity exhibited in complex industrial processes, constructing an accurate model and maintaining its performance in uncertain environments have remained challenging. Hence, a self-organizing fuzzy neural network with an adaptive evolution strategy (AE-SOFNN) is proposed for nonlinear and nonstationary process modeling. First, a self-organizing mechanism based on the network learning accuracy and the activity of rules is developed to achieve a compact structure. Meanwhile, by integrating the least squares method and an improved second-order algorithm, a hybrid learning algorithm is applied to adjust network parameters. Then, an adaptive evolution strategy is proposed to enable the AE-SOFNN to better adapt to changes, aiming to ensure the accuracy and robustness of the constructed network in nonstationary environments. Specifically, an adaptive activation threshold based on generalization ability is developed to determine how to update, namely by either local updating or global updating. The variation of linear parameters during local updating is taken as an indicator of concept drift, helping to improve the global updating performance via the selection of appropriate samples. Finally, the effectiveness of the AE-SOFNN is evaluated by a chaotic time-series prediction problem and an industrial application, demonstrating the superiority of AE-SOFNN in modeling nonlinear and nonstationary processes.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-025-11283-x.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145165900","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Physics-informed stochastic configuration network promoted model predictive control with multi-objective optimization","authors":"Lei Xu,&nbsp;Chunhua Yang,&nbsp;Xiaodong Xu,&nbsp;Biao Luo,&nbsp;Tingwen Huang","doi":"10.1007/s10462-025-11278-8","DOIUrl":"10.1007/s10462-025-11278-8","url":null,"abstract":"<div><p>Model predictive control(MPC) has attracted much attention for its superior control performance in industrial processes. However, due to the challenges in building models for industrial processes and the necessary multiple optimization objectives during the MPC optimization steps, it is difficult to achieve satisfactory control results. In this work, we propose a physics-informed stochastic configuration network(PISCN) modeling method, and a predictive control scheme based on PISCN combined with multi-objective optimization(MOO) for a class of nonlinear dynamic systems. We first develop a data-driven and physically guided hybrid modeling method that embeds physical knowledge into the loss function of stochastic configuration networks(SCN) to improve model accuracy. During the model training, we employ a parallel configuration method(PCM) to randomly assign input weights and bias of hidden nodes, reducing the number of training iterations. Secondly, the PISCN model is incorporated into MPC framework and multiple optimization objectives are considered simultaneously. Particularly, the corresponding closed-loop stability is analyzed and proven. Finally, the proposed method is applied in the dehydration reaction stage in sintering process of ternary cathode materials. The results show that compared with SCN based MPC, PISCN can obtain a more accurate model and achieve better control performance by considering multiple objectives. The sintering time and energy consumption are significantly reduced.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-025-11278-8.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145165897","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A survey of deep learning for industrial visual anomaly detection","authors":"Zhuo Li,&nbsp;Yuhao Yan,&nbsp;Xiangheng Wang,&nbsp;Yifei Ge,&nbsp;Lin Meng","doi":"10.1007/s10462-025-11287-7","DOIUrl":"10.1007/s10462-025-11287-7","url":null,"abstract":"<div><p>Industrial visual anomaly detection is critical for ensuring system reliability, safety, and efficiency. This paper presents a comprehensive survey of state-of-the-art anomaly detection techniques, analyzing methodologies, implementations, and recent advancements. Our survey aims to accelerate researchers’ understanding of emerging trends while providing a structured foundation for newcomers. We systematically review 196 recent papers covering five learning strategies, including fully supervised, semi-supervised, self-supervised, weakly supervised, and unsupervised approaches. This paper provides a detailed introduction to twelve industrial anomaly detection methods, revealing their theoretical foundations, technical principles, and practical applications. Additionally, we provide a detailed overview to 2D and 3D datasets for industrial visual anomaly detection. In addition, we critically analyze and summarize the experimental results, identify key performance indicators, and discuss the latest trends in the field of industrial anomaly detection. Beyond analysis, we contribute actionable insights for selecting optimal models for real-world deployment. Finally, we highlight open challenges and outline future research directions to drive innovation in this evolving field. The detailed resources are available at https://github.com/IHPCRits/IAD-Survey.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-06-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-025-11287-7.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145165893","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Comprehensive review of recent developments in visual object detection based on deep learning","authors":"Enerst Edozie,&nbsp;Aliyu Nuhu Shuaibu,&nbsp;Ukagwu Kelechi John,&nbsp;Bashir Olaniyi Sadiq","doi":"10.1007/s10462-025-11284-w","DOIUrl":"10.1007/s10462-025-11284-w","url":null,"abstract":"<div><p>This comprehensive review looks into the recent developments in visual object detection, focusing on the transformative effect of deep learning (DL) technologies. In object detection, computer vision is a basic issue. This involves object detection and location in the video and image frames, which has notable advantages in robotics, autonomous driving, medical imaging, and surveillance. This review, therefore, presents a thorough integration analysis in visual object detection of the latest developments, providing both the historical context and state-of-the-art analysis. This review categorizes current methods into one-stage and two-stage frameworks, studying their architectural innovations, detection accuracy, computational speed, and deployment readiness. This review further scrutinizes the performance measures, emphasizes the inevitability of large-scale annotated datasets, and provides a curated overview of the widely used datasets in the field. Notable features include a discussion of practical applications and current research trends, and a comprehensive comparative analysis that compares models based on accuracy, speed, and trade-offs. A unique addition of this work is a thorough comparative analysis table that benchmarks traditional and modern models in terms of mean Average Precision (mAP), frames per second (FPS), advantages, limitations, and the coverage of transformer-based models and real-time deployments. The review’s holistic approach provides significant insights for researchers and practitioners seeking to understand, benchmark, develop, or benchmark object detection systems.</p></div>","PeriodicalId":8449,"journal":{"name":"Artificial Intelligence Review","volume":"58 9","pages":""},"PeriodicalIF":13.9,"publicationDate":"2025-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s10462-025-11284-w.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145164624","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}