Enneng Yang, Li Shen, Guibing Guo, Xingwei Wang, Xiaochun Cao, Jie Zhang, Dacheng Tao
Model merging is an efficient empowerment technique in the machine learning community that does not require the collection of raw training data and does not require expensive computation. As model merging becomes increasingly prevalent across various fields, it is crucial to understand the available model merging techniques comprehensively. However, there is a significant gap in the literature regarding a systematic and thorough review of these techniques. This survey provides a comprehensive overview of model merging methods and theories, their applications in various domains and settings, and future research directions. Specifically, we first propose a new taxonomic approach that exhaustively discusses existing model merging methods. Secondly, we discuss the application of model merging techniques in large language models, multimodal large language models, and more than ten machine learning subfields, including continual learning, multi-task learning, few-shot learning, etc. Finally, we highlight the remaining challenges of model merging and discuss future research directions. A comprehensive list of papers about model merging is available at https://github.com/EnnengYang/Awesome-Model-Merging-Methods-Theories-Applications .
{"title":"Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications, and Opportunities","authors":"Enneng Yang, Li Shen, Guibing Guo, Xingwei Wang, Xiaochun Cao, Jie Zhang, Dacheng Tao","doi":"10.1145/3787849","DOIUrl":"https://doi.org/10.1145/3787849","url":null,"abstract":"Model merging is an efficient empowerment technique in the machine learning community that does not require the collection of raw training data and does not require expensive computation. As model merging becomes increasingly prevalent across various fields, it is crucial to understand the available model merging techniques comprehensively. However, there is a significant gap in the literature regarding a systematic and thorough review of these techniques. This survey provides a comprehensive overview of model merging methods and theories, their applications in various domains and settings, and future research directions. Specifically, we first propose a new taxonomic approach that exhaustively discusses existing model merging methods. Secondly, we discuss the application of model merging techniques in large language models, multimodal large language models, and more than ten machine learning subfields, including continual learning, multi-task learning, few-shot learning, etc. Finally, we highlight the remaining challenges of model merging and discuss future research directions. A comprehensive list of papers about model merging is available at <jats:italic toggle=\"yes\">https://github.com/EnnengYang/Awesome-Model-Merging-Methods-Theories-Applications</jats:italic> .","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"94 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2026-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145947217","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}
As large language models (LLMs) continue to evolve, the scope and diversity of data used for training are expanding significantly. However, the training dataset of LLMs may inevitably contain sensitive information such as personal data or copyrighted material, leading to privacy leakage or copyright infringement risks if the model generates highly similar or identical text to these sources. This has drawn attention to the issue of detecting whether the text data is used for LLM training. To date, research on detecting training data usage in artificial intelligence (AI) models has mainly focused on traditional machine learning (ML) models. However, studies on LLMs remain relatively immature. The lack of understanding of research progress in this area has hindered the development of more effective detection methods. Therefore, this article aims to address this gap by conducting the analysis of detecting training data for LLM. Specifically, we analyze the available LLM’s information to the detector, the main detection methods, determination metrics, and discuss the technical challenges and potential directions for future research in this field.
{"title":"Detecting Training Data For Large Language Models: A Survey","authors":"Chen Yang, Junyi Li, Shulin Lan, Yingchao Wang, Hongyang Du, Congcheng Gong, Xingshan Yao, Dusit (Tao) Niyato, Liehuang Zhu","doi":"10.1145/3779430","DOIUrl":"https://doi.org/10.1145/3779430","url":null,"abstract":"As large language models (LLMs) continue to evolve, the scope and diversity of data used for training are expanding significantly. However, the training dataset of LLMs may inevitably contain sensitive information such as personal data or copyrighted material, leading to privacy leakage or copyright infringement risks if the model generates highly similar or identical text to these sources. This has drawn attention to the issue of detecting whether the text data is used for LLM training. To date, research on detecting training data usage in artificial intelligence (AI) models has mainly focused on traditional machine learning (ML) models. However, studies on LLMs remain relatively immature. The lack of understanding of research progress in this area has hindered the development of more effective detection methods. Therefore, this article aims to address this gap by conducting the analysis of detecting training data for LLM. Specifically, we analyze the available LLM’s information to the detector, the main detection methods, determination metrics, and discuss the technical challenges and potential directions for future research in this field.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"45 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2026-01-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145920261","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}
Licheng Zhang, Bach Le, Naveed Akhtar, Siew-Kei Lam, Duc Ngo
Large Language Models (LLMs) have seen rapid advancements in recent years, with models like ChatGPT and DeepSeek, showcasing their remarkable capabilities across diverse domains. While substantial research has been conducted on LLMs in various fields, a comprehensive review focusing on their integration with Computer-Aided Design (CAD) remains notably absent. CAD is the industry standard for 3D modeling and plays a vital role in the design and development of products across different industries. As the complexity of modern designs increases, the potential for LLMs to enhance and streamline CAD workflows presents an exciting frontier. This article presents the first systematic survey exploring the intersection of LLMs and CAD. We begin by outlining the industrial significance of CAD, highlighting the need for Artificial Intelligence (AI)-driven innovation. Next, we provide a detailed overview of the foundation of LLMs. We also examine both closed-source LLMs as well as publicly available models. The core of this review focuses on the various applications of LLMs in CAD, providing a taxonomy of six key areas where these models are making considerable impact. We also provide a comprehensive study of CAD evaluation, reviewing existing methods and metrics in detail. In our analysis, we also examine common data modalities, model usage trends, dataset sources, and industrial application domains to provide a well-rounded picture of the field. Finally, we propose several promising future directions for further advancements, which offer vast opportunities for innovation and are poised to shape the future of CAD technology. Github: https://github.com/lichengzhanguom/LLMs-CAD-Survey-Taxonomy
{"title":"Large Language Models for Computer-Aided Design: A Survey","authors":"Licheng Zhang, Bach Le, Naveed Akhtar, Siew-Kei Lam, Duc Ngo","doi":"10.1145/3787499","DOIUrl":"https://doi.org/10.1145/3787499","url":null,"abstract":"Large Language Models (LLMs) have seen rapid advancements in recent years, with models like ChatGPT and DeepSeek, showcasing their remarkable capabilities across diverse domains. While substantial research has been conducted on LLMs in various fields, a comprehensive review focusing on their integration with Computer-Aided Design (CAD) remains notably absent. CAD is the industry standard for 3D modeling and plays a vital role in the design and development of products across different industries. As the complexity of modern designs increases, the potential for LLMs to enhance and streamline CAD workflows presents an exciting frontier. This article presents the first systematic survey exploring the intersection of LLMs and CAD. We begin by outlining the industrial significance of CAD, highlighting the need for Artificial Intelligence (AI)-driven innovation. Next, we provide a detailed overview of the foundation of LLMs. We also examine both closed-source LLMs as well as publicly available models. The core of this review focuses on the various applications of LLMs in CAD, providing a taxonomy of six key areas where these models are making considerable impact. We also provide a comprehensive study of CAD evaluation, reviewing existing methods and metrics in detail. In our analysis, we also examine common data modalities, model usage trends, dataset sources, and industrial application domains to provide a well-rounded picture of the field. Finally, we propose several promising future directions for further advancements, which offer vast opportunities for innovation and are poised to shape the future of CAD technology. Github: https://github.com/lichengzhanguom/LLMs-CAD-Survey-Taxonomy","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"41 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145908000","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}
Yuntao Hao, Nan Ding, Weiguo Xia, Hongwei Ge, Li Xu
With rapid advancements in artificial intelligence and Internet of Things technologies, the deployment of deep neural network (DNN) models on the edge nodes and the end nodes has become an essential trend. However, the limited computational power, storage capacity, and resource constraints of these devices present significant challenges for deep learning inference. Traditional acceleration methods, such as model compression and hardware optimization, often struggle to balance real-time performance, accuracy, and cost-effectiveness. To address these challenges, collaborative inference through DNN partitioning has emerged as a promising solution. This paper provides a comprehensive overview of architectural frameworks for DNN partitioning in collaborative inference. We establish a unified mathematical framework to describe various architectures, DNN models, and their associated optimization problems. In addition, we systematically classify and analyze existing partitioning strategies based on partition count and granularity. Furthermore, we summarize commonly used experimental setups and tools, offering practical insight into implementation. Finally, we discuss key challenges and open issues in DNN partitioning for collaborative inference, such as ensuring data security and privacy and efficiently partitioning large-scale models, providing valuable guidance for future research.
{"title":"DNN Partitioning for Cooperative Inference in Edge Intelligence: Modeling, Solutions, Toolchains","authors":"Yuntao Hao, Nan Ding, Weiguo Xia, Hongwei Ge, Li Xu","doi":"10.1145/3786145","DOIUrl":"https://doi.org/10.1145/3786145","url":null,"abstract":"With rapid advancements in artificial intelligence and Internet of Things technologies, the deployment of deep neural network (DNN) models on the edge nodes and the end nodes has become an essential trend. However, the limited computational power, storage capacity, and resource constraints of these devices present significant challenges for deep learning inference. Traditional acceleration methods, such as model compression and hardware optimization, often struggle to balance real-time performance, accuracy, and cost-effectiveness. To address these challenges, collaborative inference through DNN partitioning has emerged as a promising solution. This paper provides a comprehensive overview of architectural frameworks for DNN partitioning in collaborative inference. We establish a unified mathematical framework to describe various architectures, DNN models, and their associated optimization problems. In addition, we systematically classify and analyze existing partitioning strategies based on partition count and granularity. Furthermore, we summarize commonly used experimental setups and tools, offering practical insight into implementation. Finally, we discuss key challenges and open issues in DNN partitioning for collaborative inference, such as ensuring data security and privacy and efficiently partitioning large-scale models, providing valuable guidance for future research.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"29 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2025-12-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145894664","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}
Intelligent navigation is essential for unmanned systems. Yet nowadays navigation technologies still fall short of animals’ innate navigation prowess, characterized by continuous, efficient, adaptive, low-power navigating across complex terrains, despite technological advancements. Neuroscience's half-century exploration has revealed the brain's innate “Global Positioning System (GPS),” instigating research into Brain-Inspired Navigation (BIN). BIN, is a cutting-edge navigation technology, that bridges disciplines but lacks a cohesive guide for its interdisciplinary study. In this paper, we offer a comprehensive BIN review, mapping its neural basis, computational foundations, current progress, and implementation conditions, providing a general framework for researchers alongside forward-looking recommendations for future development in the domain. The highlights of this paper can be available at https://binucoe.github.io/Awesome-Brain-inspired-Navigation/ .
{"title":"A Comprehensive Review of Brain-inspired Navigation","authors":"Xu He, Xiaolin Meng, Lingfei Mo, Youdong Zhang, Fangwen Yu, Jingnan Liu","doi":"10.1145/3786344","DOIUrl":"https://doi.org/10.1145/3786344","url":null,"abstract":"Intelligent navigation is essential for unmanned systems. Yet nowadays navigation technologies still fall short of animals’ innate navigation prowess, characterized by continuous, efficient, adaptive, low-power navigating across complex terrains, despite technological advancements. Neuroscience's half-century exploration has revealed the brain's innate “Global Positioning System (GPS),” instigating research into Brain-Inspired Navigation (BIN). BIN, is a cutting-edge navigation technology, that bridges disciplines but lacks a cohesive guide for its interdisciplinary study. In this paper, we offer a comprehensive BIN review, mapping its neural basis, computational foundations, current progress, and implementation conditions, providing a general framework for researchers alongside forward-looking recommendations for future development in the domain. The highlights of this paper can be available at <jats:ext-link xmlns:xlink=\"http://www.w3.org/1999/xlink\" ext-link-type=\"uri\" xlink:href=\"https://binucoe.github.io/Awesome-Brain-inspired-Navigation/\">https://binucoe.github.io/Awesome-Brain-inspired-Navigation/</jats:ext-link> .","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"8 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145829942","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}
The subject of one’s opinions expressed in textual data provides rich information regarding their attitudes and behaviors. Many natural language processing tasks leverage such information to, for example, study product purchasing behaviors or extract insights during global events. The task of identifying these subjects is referred to as aspect extraction . Aspect extraction approaches typically focus on the identification of explicitly stated aspects in a text sample. However, it is suggested that implicit aspects , or those that must be inferred by the context provided in the text, comprise more than 20% of all aspects in a given dataset and that identification of implicit aspects is important for accurate aspect-based analyses such as aspect-based sentiment analysis. As such, this paper surveys recent work in implicit aspect extraction. We define and describe commonly used datasets and algorithmic approaches and detail various challenges which have thus far led to limited research in implicit aspect extraction as compared to explicit aspect extraction, like fewer benchmark datasets and limited use of powerful attention models.
{"title":"Implicit Aspect Extraction: A Systematic Review","authors":"Meghna Chaudhary, Tempestt Neal","doi":"10.1145/3786590","DOIUrl":"https://doi.org/10.1145/3786590","url":null,"abstract":"The subject of one’s opinions expressed in textual data provides rich information regarding their attitudes and behaviors. Many natural language processing tasks leverage such information to, for example, study product purchasing behaviors or extract insights during global events. The task of identifying these subjects is referred to as <jats:italic toggle=\"yes\">aspect extraction</jats:italic> . Aspect extraction approaches typically focus on the identification of explicitly stated aspects in a text sample. However, it is suggested that <jats:italic toggle=\"yes\">implicit aspects</jats:italic> , or those that must be inferred by the context provided in the text, comprise more than 20% of all aspects in a given dataset and that identification of implicit aspects is important for accurate aspect-based analyses such as aspect-based sentiment analysis. As such, this paper surveys recent work in implicit aspect extraction. We define and describe commonly used datasets and algorithmic approaches and detail various challenges which have thus far led to limited research in implicit aspect extraction as compared to explicit aspect extraction, like fewer benchmark datasets and limited use of powerful attention models.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"4 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2025-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145829943","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}
Single image blind motion deblurring, a cornerstone of low-level computer vision, seeks to recover a sharp image from a single blurred observation, addressing challenges posed by motion-induced degradation. This survey provides a comprehensive review of the field, spanning traditional methodologies and deep learning (DL) methods. We begin by defining the problem, outlining its significance, and tracing its research evolution. The paper systematically examines traditional approaches–including prior-based, edge-detection, patch-based, and specialized deblurring techniques–followed by an in-depth exploration of DL-based methods, categorized into hybrid model-driven/data-driven frameworks and fully data-driven architectures. Key datasets, loss functions, and quantitative performance evaluations of both classic and state-of-the-art methods on benchmarks are presented to offer practical insights. We conclude by summarizing advancements, identifying persistent challenges such as handling complex real-world data and computational efficiency, and proposing future research directions. This survey serves as a valuable resource for researchers, providing a holistic understanding of blind motion deblurring and fostering innovation in this dynamic domain.
{"title":"A Survey of Single Image Blind Motion Deblurring from Traditional to Deep Learning","authors":"Tingting Zhang, Jiawei Lu, Qiyu Jin, Tieyong Zeng","doi":"10.1145/3785655","DOIUrl":"https://doi.org/10.1145/3785655","url":null,"abstract":"Single image blind motion deblurring, a cornerstone of low-level computer vision, seeks to recover a sharp image from a single blurred observation, addressing challenges posed by motion-induced degradation. This survey provides a comprehensive review of the field, spanning traditional methodologies and deep learning (DL) methods. We begin by defining the problem, outlining its significance, and tracing its research evolution. The paper systematically examines traditional approaches–including prior-based, edge-detection, patch-based, and specialized deblurring techniques–followed by an in-depth exploration of DL-based methods, categorized into hybrid model-driven/data-driven frameworks and fully data-driven architectures. Key datasets, loss functions, and quantitative performance evaluations of both classic and state-of-the-art methods on benchmarks are presented to offer practical insights. We conclude by summarizing advancements, identifying persistent challenges such as handling complex real-world data and computational efficiency, and proposing future research directions. This survey serves as a valuable resource for researchers, providing a holistic understanding of blind motion deblurring and fostering innovation in this dynamic domain.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"47 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2025-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145829946","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}
With the evolution of mobile embodied intelligence, agents such as drones and autonomous robots are transitioning toward high agility. This shift imposes stringent demands on embodied perception, requiring high-accuracy and low-latency feedback loops for reliable interaction. Event-based vision has emerged as a transformative paradigm. Its microsecond-level temporal resolution and high dynamic range render it ideal for embodied perception tasks on high-agility mobile platforms. However, asynchronous nature, substantial noise, lack of persistent semantic information, and large data volume pose challenges for processing on resource-constrained mobile agents. This paper surveys the literature from 2014-2025 and presents a comprehensive overview of event-based mobile embodied perception. We organize review around four key pillars: event abstraction methods, perception algorithm advancements, hardware and software acceleration strategies, and mobile applications . We discuss critical tasks including visual odometry, object tracking, optical flow, and 3D reconstruction, while highlighting challenges associated with sensor fusion and real-time deployment. Furthermore, we outline future research directions, such as improving event cameras with advanced optics and leveraging neuromorphic computing for efficient processing. To support ongoing research, we provide an open-source Online Sheet with recent developments. We hope this survey serves as a reference, facilitating adoption of event-based vision across diverse mobile embodied applications.
{"title":"Event Camera Meets Mobile Embodied Perception: Abstraction, Algorithm, Acceleration, Application","authors":"Haoyang Wang, Ruishan Guo, Pengtao Ma, Ciyu Ruan, Xinyu Luo, Wenhua Ding, Tianyang Zhong, Jingao Xu, Yunhao Liu, Xinlei Chen","doi":"10.1145/3786332","DOIUrl":"https://doi.org/10.1145/3786332","url":null,"abstract":"With the evolution of mobile embodied intelligence, agents such as drones and autonomous robots are transitioning toward high agility. This shift imposes stringent demands on embodied perception, requiring high-accuracy and low-latency feedback loops for reliable interaction. Event-based vision has emerged as a transformative paradigm. Its microsecond-level temporal resolution and high dynamic range render it ideal for embodied perception tasks on high-agility mobile platforms. However, asynchronous nature, substantial noise, lack of persistent semantic information, and large data volume pose challenges for processing on resource-constrained mobile agents. This paper surveys the literature from 2014-2025 and presents a comprehensive overview of event-based mobile embodied perception. We organize review around four key pillars: event <jats:italic toggle=\"yes\">abstraction</jats:italic> methods, perception <jats:italic toggle=\"yes\">algorithm</jats:italic> advancements, hardware and software <jats:italic toggle=\"yes\">acceleration</jats:italic> strategies, and mobile <jats:italic toggle=\"yes\">applications</jats:italic> . We discuss critical tasks including visual odometry, object tracking, optical flow, and 3D reconstruction, while highlighting challenges associated with sensor fusion and real-time deployment. Furthermore, we outline future research directions, such as improving event cameras with advanced optics and leveraging neuromorphic computing for efficient processing. To support ongoing research, we provide an open-source <jats:italic toggle=\"yes\">Online Sheet</jats:italic> with recent developments. We hope this survey serves as a reference, facilitating adoption of event-based vision across diverse mobile embodied applications.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"16 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2025-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145829944","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}
Mathematical reasoning has long represented one of the most fundamental and challenging frontiers in artificial intelligence research. In recent years, large language models (LLMs) have achieved significant advances in this area. This survey examines the development of mathematical reasoning abilities in LLMs through two high-level cognitive phases: comprehension, where models gain mathematical understanding via diverse pretraining strategies, and answer generation, which has progressed from direct prediction to step-by-step Chain-of-Thought (CoT) reasoning. We review methods for enhancing mathematical reasoning, ranging from training-free prompting to fine-tuning approaches such as supervised fine-tuning and reinforcement learning, and discuss recent work on extended CoT and “test-time scaling”. Despite notable progress, fundamental challenges remain in terms of capacity, efficiency, and generalization. To address these issues, we highlight promising research directions, including advanced pretraining and knowledge augmentation techniques, formal reasoning frameworks, and meta-generalization through principled learning paradigms. This survey tries to provide some insights for researchers interested in enhancing reasoning capabilities of LLMs and for those seeking to apply these techniques to other domains.
{"title":"A Survey on Large Language Models for Mathematical Reasoning","authors":"Peng-Yuan Wang, Tian-Shuo Liu, Chenyang Wang, Ziniu Li, Yidi Wang, Shu Yan, Chengxing Jia, Xu-Hui Liu, Xinwei Chen, Jiacheng Xu, Yang Yu","doi":"10.1145/3786333","DOIUrl":"https://doi.org/10.1145/3786333","url":null,"abstract":"Mathematical reasoning has long represented one of the most fundamental and challenging frontiers in artificial intelligence research. In recent years, large language models (LLMs) have achieved significant advances in this area. This survey examines the development of mathematical reasoning abilities in LLMs through two high-level cognitive phases: comprehension, where models gain mathematical understanding via diverse pretraining strategies, and answer generation, which has progressed from direct prediction to step-by-step Chain-of-Thought (CoT) reasoning. We review methods for enhancing mathematical reasoning, ranging from training-free prompting to fine-tuning approaches such as supervised fine-tuning and reinforcement learning, and discuss recent work on extended CoT and “test-time scaling”. Despite notable progress, fundamental challenges remain in terms of capacity, efficiency, and generalization. To address these issues, we highlight promising research directions, including advanced pretraining and knowledge augmentation techniques, formal reasoning frameworks, and meta-generalization through principled learning paradigms. This survey tries to provide some insights for researchers interested in enhancing reasoning capabilities of LLMs and for those seeking to apply these techniques to other domains.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"56 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2025-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145829945","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}
The use of Unmanned Aerial Vehicles (UAVs) in various domains is continuously increasing. The benefits offered by UAVs enable the facilitation of functions or execution of tasks, such as search and rescue, which would otherwise be challenging or financially restrictive. Furthermore, UAVs are important for the military domain being considered as efficient weapons. Therefore, many countries and retail companies are investing in the massive production of UAVs for market as well as government purposes. Nevertheless, most UAVs in all domains, lack even basic cybersecurity mechanisms, thus, posing cybersecurity threats that significantly affect cybersecurity aspects or result in physical damage and even loss of human life. Many cybersecurity solutions have been proposed in the literature, encompassing both reactive and proactive measures, with the latter being preferable. Added to this, threat modeling emerges as an effective proactive measure with minimal cost and complexity. The current work conducts extensive and comprehensive research concerning the domain of UAVs. Initially, the paper outlines the historical development of UAVs and the different taxonomies in which they are classified according to different characteristics (e.g., number of rotors). The next sections detail a comprehensive description of the Internet of Drones (IoD) environment and the various relevant cybersecurity issues. Subsequently, the paper examines a wide range of diverse threat modeling approaches found in the literature, while it categorises relevant papers based on whether they describe solutions that integrate threat modeling or propose novel approaches for threat modeling.
{"title":"A Comprehensive Survey and Taxonomy of Cybersecurity Challenges and Proactive Measures for IoD","authors":"Arnolnt Spyros, Periklis Chatzimisios, Dimitrios Kavallieros, Theodora Tsikrika, Stefanos Vrochidis, Yiannis Kompatsiaris","doi":"10.1145/3785658","DOIUrl":"https://doi.org/10.1145/3785658","url":null,"abstract":"The use of Unmanned Aerial Vehicles (UAVs) in various domains is continuously increasing. The benefits offered by UAVs enable the facilitation of functions or execution of tasks, such as search and rescue, which would otherwise be challenging or financially restrictive. Furthermore, UAVs are important for the military domain being considered as efficient weapons. Therefore, many countries and retail companies are investing in the massive production of UAVs for market as well as government purposes. Nevertheless, most UAVs in all domains, lack even basic cybersecurity mechanisms, thus, posing cybersecurity threats that significantly affect cybersecurity aspects or result in physical damage and even loss of human life. Many cybersecurity solutions have been proposed in the literature, encompassing both reactive and proactive measures, with the latter being preferable. Added to this, threat modeling emerges as an effective proactive measure with minimal cost and complexity. The current work conducts extensive and comprehensive research concerning the domain of UAVs. Initially, the paper outlines the historical development of UAVs and the different taxonomies in which they are classified according to different characteristics (e.g., number of rotors). The next sections detail a comprehensive description of the Internet of Drones (IoD) environment and the various relevant cybersecurity issues. Subsequently, the paper examines a wide range of diverse threat modeling approaches found in the literature, while it categorises relevant papers based on whether they describe solutions that integrate threat modeling or propose novel approaches for threat modeling.","PeriodicalId":50926,"journal":{"name":"ACM Computing Surveys","volume":"4 1","pages":""},"PeriodicalIF":16.6,"publicationDate":"2025-12-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145829947","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：