Pub Date : 2026-01-23DOI: 10.1016/j.cosrev.2026.100905
Rafał Stachowiak, Tomasz P. Pawlak
Mathematical Programming (MP) is a well-established framework for formulating optimization problems using variables, constraints, and an objective function. The task of developing an MP model involves addressing subproblems such as discovering an MP model from domain knowledge, conformance checking of a candidate MP model with domain knowledge, and enhancing an invalid MP model based on domain knowledge. Traditionally, experts manually perform these tasks, leading to iterative processes that are both labor-intensive and error-prone. Recent literature highlights an emerging field of algorithms focused on automating MP model development using domain knowledge artifacts, which we jointly term MP model mining and divide into discovery, conformance checking, and enhancement problems. This study organizes and analyzes existing knowledge on MP model mining, aiming to elucidate the state of the art and pinpoint current gaps and challenges. Through a systematic review via an acknowledged literature search engine, we address 29 research questions concerning various dimensions, identify 15 knowledge gaps, and propose a future research agenda.
{"title":"Mathematical programming model mining: A systematic field survey","authors":"Rafał Stachowiak, Tomasz P. Pawlak","doi":"10.1016/j.cosrev.2026.100905","DOIUrl":"10.1016/j.cosrev.2026.100905","url":null,"abstract":"<div><div>Mathematical Programming (MP) is a well-established framework for formulating optimization problems using variables, constraints, and an objective function. The task of developing an MP model involves addressing subproblems such as discovering an MP model from domain knowledge, conformance checking of a candidate MP model with domain knowledge, and enhancing an invalid MP model based on domain knowledge. Traditionally, experts manually perform these tasks, leading to iterative processes that are both labor-intensive and error-prone. Recent literature highlights an emerging field of algorithms focused on automating MP model development using domain knowledge artifacts, which we jointly term MP model mining and divide into discovery, conformance checking, and enhancement problems. This study organizes and analyzes existing knowledge on MP model mining, aiming to elucidate the state of the art and pinpoint current gaps and challenges. Through a systematic review via an acknowledged literature search engine, we address 29 research questions concerning various dimensions, identify 15 knowledge gaps, and propose a future research agenda.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100905"},"PeriodicalIF":12.7,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-23DOI: 10.1016/j.cosrev.2026.100912
Allegra De Filippo , Giuseppe Spillo , Ludovico Boratto , Michela Milano , Cataldo Musto , Giovanni Semeraro
The concept of sustainability, as outlined by the United Nations’ Sustainable Development Goals (SDGs), refers to the ability to meet the needs of the present without compromising the ability of future generations to meet their own needs. This vision is addressed by combining goals concerning the environmental, social, and economic spheres. In this context, Recommender Systems (RS) have emerged as tools that can foster these principles by nudging responsible user behavior and promoting sustainable decision-making. However, the interplay between RS and sustainability is inherently complex since it can be analyzed from two different perspectives: (i) RS for Sustainability, which focuses on how recommendation algorithms can support the achievement of SDGs, and (ii) Sustainability of RS, which focuses on developing recommendation models that inherently adhere to sustainability principles. While the integration of both these perspectives is beneficial and crucial, unfortunately, the current literature has addressed these aspects independently. Accordingly, in this survey, we first provide a comprehensive review of the existing literature on RS that either promotes sustainable behaviors aligned with the SDGs or embeds sustainability principles into their algorithmic design. Next, we identify current gaps and propose key research directions toward an integrated, holistic approach that concurrently addresses both aspects to advance the development of sustainable RS.
{"title":"Recommender systems and sustainability: a dual perspective","authors":"Allegra De Filippo , Giuseppe Spillo , Ludovico Boratto , Michela Milano , Cataldo Musto , Giovanni Semeraro","doi":"10.1016/j.cosrev.2026.100912","DOIUrl":"10.1016/j.cosrev.2026.100912","url":null,"abstract":"<div><div>The concept of sustainability, as outlined by the United Nations’ Sustainable Development Goals (SDGs), refers to the ability to meet the needs of the present without compromising the ability of future generations to meet their own needs. This vision is addressed by combining goals concerning the <em>environmental</em>, <em>social</em>, and <em>economic</em> spheres. In this context, Recommender Systems (RS) have emerged as tools that can foster these principles by nudging responsible user behavior and promoting sustainable decision-making. However, the interplay between RS and sustainability is inherently complex since it can be analyzed from two different perspectives: <em>(i) RS for Sustainability</em>, which focuses on how recommendation algorithms can support the achievement of SDGs, and <em>(ii) Sustainability of RS</em>, which focuses on developing recommendation models that inherently adhere to sustainability principles. While the integration of both these perspectives is beneficial and crucial, unfortunately, the current literature has addressed these aspects independently. Accordingly, in this survey, we first provide a comprehensive review of the existing literature on RS that either promotes sustainable behaviors aligned with the SDGs or embeds sustainability principles into their algorithmic design. Next, we identify current gaps and propose key research directions toward an integrated, holistic approach that concurrently addresses both aspects to advance the development of sustainable RS.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100912"},"PeriodicalIF":12.7,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-23DOI: 10.1016/j.cosrev.2026.100907
Begum Mutlu , Yunus Can Bilge
Natural Language Processing and Sign Language Processing share common goals and challenges, as both fields focus on enabling computers to understand and generate modes of communication to enhance interaction between humans and computers. The interaction happens differently, with one relying on spoken or written text and the other on visual-gestural input. Although sign languages possess significant linguistic complexity and expressiveness, they have traditionally been rarely addressed in the fields of computational linguistics and natural language processing research. The fields share similarities (sequence modeling, contextual understanding, representation learning) as well as face similar challenges; annotated data sparsity, ambiguity resolution, and multilingual understanding. In this paper, the key tasks that can be addressed in sign language processing, particularly from a natural language processing perspective, are identified and deeply examined. Sign language translation and production, machine translation, part of speech tagging, named entity resolution, coreference resolution, sentiment analysis, and sign language models in sign languages are included. An overview of these sign language tasks, as well as previously unexplored tasks that are very apparent in natural language processing but not in sign language processing, is provided. Moreover, possible reuses of already available sign language data from a linguistic perspective are also shared. Limitations and open challenges are identified to direct future research toward the linguistic aspects of sign languages, recognizing that more language-based methodologies may be necessary for improved understanding and communication in it.
{"title":"An overview of sign language processing from natural language processing perspective","authors":"Begum Mutlu , Yunus Can Bilge","doi":"10.1016/j.cosrev.2026.100907","DOIUrl":"10.1016/j.cosrev.2026.100907","url":null,"abstract":"<div><div>Natural Language Processing and Sign Language Processing share common goals and challenges, as both fields focus on enabling computers to understand and generate modes of communication to enhance interaction between humans and computers. The interaction happens differently, with one relying on spoken or written text and the other on visual-gestural input. Although sign languages possess significant linguistic complexity and expressiveness, they have traditionally been rarely addressed in the fields of computational linguistics and natural language processing research. The fields share similarities (sequence modeling, contextual understanding, representation learning) as well as face similar challenges; annotated data sparsity, ambiguity resolution, and multilingual understanding. In this paper, the key tasks that can be addressed in sign language processing, particularly from a natural language processing perspective, are identified and deeply examined. Sign language translation and production, machine translation, part of speech tagging, named entity resolution, coreference resolution, sentiment analysis, and sign language models in sign languages are included. An overview of these sign language tasks, as well as previously unexplored tasks that are very apparent in natural language processing but not in sign language processing, is provided. Moreover, possible reuses of already available sign language data from a linguistic perspective are also shared. Limitations and open challenges are identified to direct future research toward the linguistic aspects of sign languages, recognizing that more language-based methodologies may be necessary for improved understanding and communication in it.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100907"},"PeriodicalIF":12.7,"publicationDate":"2026-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033305","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-22DOI: 10.1016/j.cosrev.2026.100897
Noor ul Ain , Sajid Ali Khan , Suliman Aladhadh , Usama Mir , Muhammad Ramzan
Early and accurate detection of brain tumors in MRI images is important for increasing patient survival rates. The latest trends in Deep Learning (DL) have revolutionized medical imaging analysis. This paper presents a systematic review of state-of-the-art methodologies, including Convolutional Neural Networks (CNNs), Transformers, Graph Neural Networks (GNNs), and Explainable AI (XAI). It also summarizes the information on pre-processing techniques, commonly available datasets, and evaluation metrics. Following that, experimental validation is performed on various DL models, including CNNs, a Custom Deep CNN (for Ablation), transfer learning models (VGG16, ResNet50, EfficientNetB0), and a Swin Transformer. The transformer achieved the superior Mean ± Std accuracy (0.97 ± 0.02), Precision 99%, Recall 99% and F1-score (98%) across 5 Runs. Moreover, the Attention Maps of the Swin Transformer are evaluated, providing insight into the decision-making process of DL models. Finally, the study also highlights existing challenges and outlines future research directions, including federated learning, self-supervised approaches, and lightweight hybrid architectures to build scalable, interpretable diagnostic models.
早期和准确地发现脑肿瘤的MRI图像是提高患者存活率的重要。深度学习（DL）的最新趋势已经彻底改变了医学成像分析。本文对最先进的方法进行了系统的回顾，包括卷积神经网络（cnn），变压器，图神经网络（GNNs）和可解释的人工智能（XAI）。它还总结了有关预处理技术、常用数据集和评估指标的信息。随后，在各种深度学习模型上进行实验验证，包括CNN， Custom Deep CNN（用于消融），迁移学习模型（VGG16, ResNet50, EfficientNetB0）和Swin Transformer。该变压器在5次运行中取得了优异的Mean±Std准确度（0.97±0.02），精密度99%，召回率99%和f1评分（98%）。此外，还对Swin变压器的注意图进行了评估，从而深入了解DL模型的决策过程。最后，该研究还强调了现有的挑战，并概述了未来的研究方向，包括联邦学习、自我监督方法和轻量级混合架构，以构建可扩展、可解释的诊断模型。
{"title":"Transformers meet CNNs: A comprehensive review and benchmarking of deep learning architectures for brain tumor classification in MRI","authors":"Noor ul Ain , Sajid Ali Khan , Suliman Aladhadh , Usama Mir , Muhammad Ramzan","doi":"10.1016/j.cosrev.2026.100897","DOIUrl":"10.1016/j.cosrev.2026.100897","url":null,"abstract":"<div><div>Early and accurate detection of brain tumors in MRI images is important for increasing patient survival rates. The latest trends in Deep Learning (DL) have revolutionized medical imaging analysis. This paper presents a systematic review of state-of-the-art methodologies, including Convolutional Neural Networks (CNNs), Transformers, Graph Neural Networks (GNNs), and Explainable AI (XAI). It also summarizes the information on pre-processing techniques, commonly available datasets, and evaluation metrics. Following that, experimental validation is performed on various DL models, including CNNs, a Custom Deep CNN (for Ablation), transfer learning models (VGG16, ResNet50, EfficientNetB0), and a Swin Transformer. The transformer achieved the superior Mean ± Std accuracy (0.97 ± 0.02), Precision 99%, Recall 99% and F1-score (98%) across 5 Runs. Moreover, the Attention Maps of the Swin Transformer are evaluated, providing insight into the decision-making process of DL models. Finally, the study also highlights existing challenges and outlines future research directions, including federated learning, self-supervised approaches, and lightweight hybrid architectures to build scalable, interpretable diagnostic models.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100897"},"PeriodicalIF":12.7,"publicationDate":"2026-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033308","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-21DOI: 10.1016/j.cosrev.2026.100908
Arpan Mahara, Naphtali Rishe
The proliferation of generative models, such as Generative Adversarial Networks (GANs), Diffusion Models, and Variational Autoencoders (VAEs), has enabled the synthesis of high-quality multimedia data. However, these advancements have also raised significant concerns regarding adversarial attacks, unethical usage, and societal harm. Recognizing these challenges, researchers have increasingly focused on developing methodologies to detect synthesized data effectively, aiming to mitigate potential risks. Prior reviews have predominantly focused on deepfake detection and often overlook recent advancements in synthetic image forensics, particularly approaches that incorporate multimodal frameworks, reasoning-based detection, and training-free methodologies. To bridge this gap, this survey provides a comprehensive and up-to-date review of state-of-the-art techniques for detecting and classifying synthetic images generated by advanced generative AI models. The review systematically examines core detection paradigms, categorizes them into spatial-domain, frequency-domain, fingerprint-based, patch-based, training-free, and multimodal reasoning-based frameworks, and offers concise descriptions of their underlying principles. We further provide detailed comparative analyses of these methods on publicly available datasets to assess their generalizability, robustness, and interpretability. Finally, the survey highlights open challenges and future directions, emphasizing the potential of hybrid frameworks that combine the efficiency of training-free approaches with the semantic reasoning of multimodal models to advance trustworthy and explainable synthetic image forensics.
{"title":"Methods and trends in detecting AI-generated images: A comprehensive review","authors":"Arpan Mahara, Naphtali Rishe","doi":"10.1016/j.cosrev.2026.100908","DOIUrl":"10.1016/j.cosrev.2026.100908","url":null,"abstract":"<div><div>The proliferation of generative models, such as Generative Adversarial Networks (GANs), Diffusion Models, and Variational Autoencoders (VAEs), has enabled the synthesis of high-quality multimedia data. However, these advancements have also raised significant concerns regarding adversarial attacks, unethical usage, and societal harm. Recognizing these challenges, researchers have increasingly focused on developing methodologies to detect synthesized data effectively, aiming to mitigate potential risks. Prior reviews have predominantly focused on deepfake detection and often overlook recent advancements in synthetic image forensics, particularly approaches that incorporate multimodal frameworks, reasoning-based detection, and training-free methodologies. To bridge this gap, this survey provides a comprehensive and up-to-date review of state-of-the-art techniques for detecting and classifying synthetic images generated by advanced generative AI models. The review systematically examines core detection paradigms, categorizes them into spatial-domain, frequency-domain, fingerprint-based, patch-based, training-free, and multimodal reasoning-based frameworks, and offers concise descriptions of their underlying principles. We further provide detailed comparative analyses of these methods on publicly available datasets to assess their generalizability, robustness, and interpretability. Finally, the survey highlights open challenges and future directions, emphasizing the potential of hybrid frameworks that combine the efficiency of training-free approaches with the semantic reasoning of multimodal models to advance trustworthy and explainable synthetic image forensics.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100908"},"PeriodicalIF":12.7,"publicationDate":"2026-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146014540","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Deeply empowered by artificial intelligence (AI) technologies, integrated sensing, communication, and computation (ISCC) architectures have emerged as a critical component in the sixth-generation (6G) mobile communication networks. This survey is conducted to provide a thorough analysis of AI-empowered task-oriented sensing, communication, and computation (AI-TSCC) in 6G networks. At first, the intrinsic relationships among sensing, communication, and computation are systematically reviewed, and the limitations of current technologies are identified. Then, the AI-empowered methods for performance improvement in ISCC are analyzed, including AI-based TSCC and AI-assisted TSCC. Next, the tasks are classified into three categories according to their characteristics. Focusing on the optimization performance of AI-TSCC systems, model-driven metrics, task execution metrics, and other relevant metrics are introduced. Furthermore, existing technical bottlenecks and future research directions are summarized to provide theoretical and practical guidelines for building efficient 6G networks.
{"title":"A survey on AI-empowered task-oriented sensing, communication, and computation in 6G networks","authors":"Yuxin Zhang, Xingwei Wang, Xuewen Luo, Xinyue Pei, Fuliang Li, Donghong Han, Tianyu Li","doi":"10.1016/j.cosrev.2026.100899","DOIUrl":"10.1016/j.cosrev.2026.100899","url":null,"abstract":"<div><div>Deeply empowered by artificial intelligence (AI) technologies, integrated sensing, communication, and computation (ISCC) architectures have emerged as a critical component in the sixth-generation (6G) mobile communication networks. This survey is conducted to provide a thorough analysis of AI-empowered task-oriented sensing, communication, and computation (AI-TSCC) in 6G networks. At first, the intrinsic relationships among sensing, communication, and computation are systematically reviewed, and the limitations of current technologies are identified. Then, the AI-empowered methods for performance improvement in ISCC are analyzed, including AI-based TSCC and AI-assisted TSCC. Next, the tasks are classified into three categories according to their characteristics. Focusing on the optimization performance of AI-TSCC systems, model-driven metrics, task execution metrics, and other relevant metrics are introduced. Furthermore, existing technical bottlenecks and future research directions are summarized to provide theoretical and practical guidelines for building efficient 6G networks.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100899"},"PeriodicalIF":12.7,"publicationDate":"2026-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146014531","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-21DOI: 10.1016/j.cosrev.2026.100906
Congqing He , Haichuan Hu , Yanli Li , Hao Zhang , Quanjun Zhang
Recent advances in large language models (LLMs) have unlocked new opportunities for machine learning and deep learning applications in the legal domain. LLMs demonstrate remarkable capabilities in comprehending complex legal language, analyzing lengthy documents, and generating contextually relevant legal text. This survey provides a task-oriented overview of the application of LLMs in the legal domain, focusing on the categorization of tasks and a review of associated methods, datasets and benchmarks. We first review both general and legal-domain LLMs, summarize their foundational architectures and adaptation methods, and briefly analyze their suitability for various legal tasks. Subsequently, we categorize LLM-based legal tasks along the typical legal workflow, including legal information retrieval, legal document analysis, judicial decision prediction, legal question answering, legal document generation, legal agent-based modeling, and legal education and training. Within each task, we primarily analyze specific methodologies such as retrieval-augmented generation, prompting strategies, and reasoning. Furthermore, a comprehensive collection of legal datasets, benchmarks, and model resources is presented as a practical reference for researchers and practitioners. Finally, we outline open challenges and future directions, addressing issues such as bias, interpretability, data privacy, and regulatory compliance in legal LLMs. This survey provides a structured and comprehensive overview to facilitate the adoption and further development of LLMs in the legal domain. The collection is available at https://github.com/hecongqing/Awesome-LLM4Law.
{"title":"A survey of large language models for legal tasks: Progress, prospects and challenges","authors":"Congqing He , Haichuan Hu , Yanli Li , Hao Zhang , Quanjun Zhang","doi":"10.1016/j.cosrev.2026.100906","DOIUrl":"10.1016/j.cosrev.2026.100906","url":null,"abstract":"<div><div>Recent advances in large language models (LLMs) have unlocked new opportunities for machine learning and deep learning applications in the legal domain. LLMs demonstrate remarkable capabilities in comprehending complex legal language, analyzing lengthy documents, and generating contextually relevant legal text. This survey provides a task-oriented overview of the application of LLMs in the legal domain, focusing on the categorization of tasks and a review of associated methods, datasets and benchmarks. We first review both general and legal-domain LLMs, summarize their foundational architectures and adaptation methods, and briefly analyze their suitability for various legal tasks. Subsequently, we categorize LLM-based legal tasks along the typical legal workflow, including legal information retrieval, legal document analysis, judicial decision prediction, legal question answering, legal document generation, legal agent-based modeling, and legal education and training. Within each task, we primarily analyze specific methodologies such as retrieval-augmented generation, prompting strategies, and reasoning. Furthermore, a comprehensive collection of legal datasets, benchmarks, and model resources is presented as a practical reference for researchers and practitioners. Finally, we outline open challenges and future directions, addressing issues such as bias, interpretability, data privacy, and regulatory compliance in legal LLMs. This survey provides a structured and comprehensive overview to facilitate the adoption and further development of LLMs in the legal domain. The collection is available at <span><span>https://github.com/hecongqing/Awesome-LLM4Law</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100906"},"PeriodicalIF":12.7,"publicationDate":"2026-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146014532","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-19DOI: 10.1016/j.cosrev.2026.100898
Amir Masoud Rahmani , Amir Haider , Farhad Soleimanian Gharehchopogh , Komeil Moghaddasi , Aso Darwesh , Mehdi Hosseinzadeh
Autonomous vehicles, most notably self-driving cars, are seen by many as a generational shift in transportation, offering the potential to reduce crash rates and relieve congestion. In order to operate, autonomous vehicles combine data from cameras, radar, and LiDAR and need to act on this data in real-time creating a heavy and bursty computational demand. Processing these heavy and bursty demands within the autonomy system's power and thermal envelope will require considerable effort in determining which tasks to perform when and in what order. In some cases, offloading workloads to edge or cloud servers creates an opportunity to offload compute, reduce end-to-end latency, and enhance overall responsiveness if workloads are offloaded under strict latency constraints. In this work, we survey state-of-the-art offloading methods, identify significant challenges such as latency management, network reliability, and security, and outline future improvements within the area of vehicular systems. From an analysis of the state of the literature, we also objectively evaluate when and how offloading can enhance multi-faceted computational demands of autonomous driving stacks, with the overall goal of support safer and more capable vehicular systems.
{"title":"Strategic offloading in autonomous vehicles: A systematic survey of current schemes, challenges, and future prospects","authors":"Amir Masoud Rahmani , Amir Haider , Farhad Soleimanian Gharehchopogh , Komeil Moghaddasi , Aso Darwesh , Mehdi Hosseinzadeh","doi":"10.1016/j.cosrev.2026.100898","DOIUrl":"10.1016/j.cosrev.2026.100898","url":null,"abstract":"<div><div>Autonomous vehicles, most notably self-driving cars, are seen by many as a generational shift in transportation, offering the potential to reduce crash rates and relieve congestion. In order to operate, autonomous vehicles combine data from cameras, radar, and LiDAR and need to act on this data in real-time creating a heavy and bursty computational demand. Processing these heavy and bursty demands within the autonomy system's power and thermal envelope will require considerable effort in determining which tasks to perform when and in what order. In some cases, offloading workloads to edge or cloud servers creates an opportunity to offload compute, reduce end-to-end latency, and enhance overall responsiveness if workloads are offloaded under strict latency constraints. In this work, we survey state-of-the-art offloading methods, identify significant challenges such as latency management, network reliability, and security, and outline future improvements within the area of vehicular systems. From an analysis of the state of the literature, we also objectively evaluate when and how offloading can enhance multi-faceted computational demands of autonomous driving stacks, with the overall goal of support safer and more capable vehicular systems.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100898"},"PeriodicalIF":12.7,"publicationDate":"2026-01-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146000913","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-16DOI: 10.1016/j.cosrev.2026.100895
Michał Kalinowski , Bożena Kostek
When exploring concepts such as sign language and machine learning, it is clear that this important, rapidly growing field bridges communication gaps, improves accessibility, and promotes inclusion for deaf and hard-of-hearing communities. Over the past five years, advanced machine learning has made significant progress, driven by innovative methods and emerging datasets. This paper presents recent advances in sign language translation, focusing on input methods including camera-based approaches and deep learning techniques. Key contributions from the reviewed works are identified and highlighted, showing trends in research on sign language recognition and translation. The metrics used to evaluate sign language recognition and translation are also examined. The critical features and differences of selected datasets relevant to sign language recognition are explained. The paper ends with a discussion of promising future research directions.
{"title":"Machine learning in sign language: A comprehensive analysis and trend survey","authors":"Michał Kalinowski , Bożena Kostek","doi":"10.1016/j.cosrev.2026.100895","DOIUrl":"10.1016/j.cosrev.2026.100895","url":null,"abstract":"<div><div>When exploring concepts such as sign language and machine learning, it is clear that this important, rapidly growing field bridges communication gaps, improves accessibility, and promotes inclusion for deaf and hard-of-hearing communities. Over the past five years, advanced machine learning has made significant progress, driven by innovative methods and emerging datasets. This paper presents recent advances in sign language translation, focusing on input methods including camera-based approaches and deep learning techniques. Key contributions from the reviewed works are identified and highlighted, showing trends in research on sign language recognition and translation. The metrics used to evaluate sign language recognition and translation are also examined. The critical features and differences of selected datasets relevant to sign language recognition are explained. The paper ends with a discussion of promising future research directions.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100895"},"PeriodicalIF":12.7,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145976554","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-14DOI: 10.1016/j.cosrev.2026.100893
Wei Ai , Yilong Tan , Yuntao Shou , Tao Meng , Haowen Chen , Zhixiong He , Keqin Li
In recent years, the rapid evolution of large vision–language models (LVLMs) has driven a paradigm shift in multimodal fake news detection (MFND), transforming it from traditional feature-engineering approaches to unified, end-to-end multimodal reasoning frameworks. Early methods primarily relied on shallow fusion techniques to capture correlations between text and images, but they struggled with high-level semantic understanding and complex cross-modal interactions. The emergence of LVLMs has fundamentally changed this landscape by enabling joint modeling of vision and language with powerful representation learning, thereby enhancing the ability to detect misinformation that leverages both textual narratives and visual content. Despite these advances, the field lacks a systematic survey that traces this transition and consolidates recent developments. To address this gap, this paper provides a comprehensive review of MFND through the lens of LVLMs. We first present a historical perspective, mapping the evolution from conventional multimodal detection pipelines to foundation model-driven paradigms. Next, we establish a structured taxonomy covering model architectures, datasets, and performance benchmarks. Furthermore, we analyze the remaining technical challenges, including interpretability, temporal reasoning, and domain generalization. Finally, we outline future research directions to guide the next stage of this paradigm shift. To the best of our knowledge, this is the first comprehensive survey to systematically document and analyze the transformative role of LVLMs in combating multimodal fake news. The summary of existing methods mentioned is in our Github: https://github.com/Tan-YiLong/Overview-of-Fake-News-Detection.
{"title":"The paradigm shift: A comprehensive survey on large vision language models for multimodal fake news detection","authors":"Wei Ai , Yilong Tan , Yuntao Shou , Tao Meng , Haowen Chen , Zhixiong He , Keqin Li","doi":"10.1016/j.cosrev.2026.100893","DOIUrl":"10.1016/j.cosrev.2026.100893","url":null,"abstract":"<div><div>In recent years, the rapid evolution of large vision–language models (LVLMs) has driven a paradigm shift in multimodal fake news detection (MFND), transforming it from traditional feature-engineering approaches to unified, end-to-end multimodal reasoning frameworks. Early methods primarily relied on shallow fusion techniques to capture correlations between text and images, but they struggled with high-level semantic understanding and complex cross-modal interactions. The emergence of LVLMs has fundamentally changed this landscape by enabling joint modeling of vision and language with powerful representation learning, thereby enhancing the ability to detect misinformation that leverages both textual narratives and visual content. Despite these advances, the field lacks a systematic survey that traces this transition and consolidates recent developments. To address this gap, this paper provides a comprehensive review of MFND through the lens of LVLMs. We first present a historical perspective, mapping the evolution from conventional multimodal detection pipelines to foundation model-driven paradigms. Next, we establish a structured taxonomy covering model architectures, datasets, and performance benchmarks. Furthermore, we analyze the remaining technical challenges, including interpretability, temporal reasoning, and domain generalization. Finally, we outline future research directions to guide the next stage of this paradigm shift. To the best of our knowledge, this is the first comprehensive survey to systematically document and analyze the transformative role of LVLMs in combating multimodal fake news. The summary of existing methods mentioned is in our Github: <span><span>https://github.com/Tan-YiLong/Overview-of-Fake-News-Detection</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":48633,"journal":{"name":"Computer Science Review","volume":"60 ","pages":"Article 100893"},"PeriodicalIF":12.7,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145961765","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号