检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Cheng, Yuxing Chang, Yi Wu, Yuan College of Software Jilin University China School of Artificial Intelligence Jilin University China Engineering Research Center of Knowledge-Driven Human-Machine Intelligence MOE China International Center of Future Science Jilin University China

Recent advancements in Large Language Models (LLMs) have demonstrated significant progress in various areas, such as text generation and code synthesis. However, the reliability of performance evaluation has come under scrutiny due to data contamination—the unintended overlap between training and test datasets. This overlap has the potential to artificially inflate model performance, as LLMs are typically trained on extensive datasets scraped from publicly available sources. These datasets often inadvertently overlap with the benchmarks used for evaluation, leading to an overestimation of the models’ true generalization capabilities. In this paper, we first examine the definition and impacts of data contamination. Secondly, we review methods for contamination-free evaluation, focusing on three strategies: data updating-based methods, data rewriting-based methods, and prevention-based methods. Specifically, we highlight dynamic benchmarks and LLM-driven evaluation methods. Finally, we categorize contamination detecting methods based on model information dependency: white-Box, gray-Box, and black-Box detection approaches. Our survey highlights the requirements for more rigorous evaluation protocols and proposes future directions for addressing data contamination challenges. Copyright © 2025, The Authors. All rights reserved.

关键词： Data assimilation

来源：评论

学校读者我要写书评

暂无评论

Rethinking Polyp Segmentation from the Perspectives of Matching Views and Seeking Camouflage

引用

Multimedia Systems 2025年第3期31卷 1-17页

作者： Jiang, Zhengfang Chen, Haipeng Yang, Yongping Liu, Xianzhu Lyu, Yingda College of Software Jilin University Changchun130012 China Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education Jilin University Changchun130012 China College of Computer Science and Technology Jilin University Changchun130012 China Department of Colorectal Anal Surgery The Second Hospital of Jilin University Changchun130022 China National and Local Joint Engineering Research Center of Space Optoelectronics Technology Changchun University of Science and Technology Changchun130022 China College of Opto-Electronic Engineering Changchun University of Science and Technology Changchun130022 China Public Computer Education and Research Center Jilin University Changchun130012 China

Accurate segmentation of polyps from colonoscopy images is of great significance for the prevention of colorectal cancer. Despite notable progress, polyp segmentation remains a challenging task due to the diversity of polyp size and shape, as well as the camouflage property of polyps caused by their high similarity to surrounding tissues. To address these challenges, inspired by human perceptual experience in finding objects, we propose a multi-perspective collaborative network (MPCNet), innovatively identifying the potential polyps from the perspectives of matching views and seeking camouflage. Firstly, from the former perspective, a global–local aware module is proposed to provide various feature receptive fields to match suitable views for polyps of different sizes and shapes. Then, from the latter perspective, a texture enhancement module (TEM) and an interaction refinement module (IRM) are proposed to mine and complement clues essential for breaking the camouflage. Specifically, TEM mines texture-related clues and amplifies the subtle texture differences between polyps and their background, while IRM complements boundary clues and further optimizes feature representation through the interaction between boundaries and regions, thereby generating accurate segmentation results with complete regions and fine boundaries. Extensive experiments on five benchmark polyp datasets show that our MPCNet outperforms other state-of-the-art methods. In particular, we achieve a mean Dice of 81.7% and 82.4% on the most challenging datasets CVC-ColonDB and ETIS, respectively, which are significantly higher than competing methods. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2025.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

DeepSeek: Paradigm Shifts and Technical Evolution in Large AI Models

引用

IEEE/CAA Journal of Automatica Sinica 2025年第5期12卷 841-858页

作者： Xiong, Luolin Wang, Haofen Chen, Xi Sheng, Lu Xiong, Yun Liu, Jingping Xiao, Yanghua Chen, Huajun Han, Qing-Long Tang, Yang Ministry of Education Engineering Research Center of Process System Engineering East China University of Science and Technology Key Laboratory of Smart Manufacturing in Energy Chemical Process Shanghai 200237 China College of Design and Innovation Tongji University Shanghai 200092 China School of Computer Science Fudan University Shanghai Key Laboratory of Data Science Shanghai 200433 China School of Software Beihang University Beijing 100191 China School of Information Science and Engineering East China University of Science and Technology Shanghai 200237 China College of Computer Science and Technology Zhejiang University AZFT Joint Laboratory for Knowledge Engine Hangzhou Innovation Center Hangzhou 310058 China School of Science Computing and Engineering Technologies Swinburne University of Technology Melbourne 3122 VIC Australia

DeepSeek, a Chinese artificial intelligence (AI) startup, has released their V3 and R1 series models, which attracted global attention due to their low cost, high performance, and open-source advantages. This paper begins by reviewing the evolution of large AI models focusing on paradigm shifts, the mainstream large language model (LLM) paradigm, and the DeepSeek paradigm. Subsequently, the paper highlights novel algorithms introduced by DeepSeek, including multi-head latent attention (MLA), mixture-of-experts (MoE), multi-token prediction (MTP), and group relative policy optimization (GRPO). The paper then explores DeepSeek's engineering breakthroughs in LLM scaling, training, inference, and system-level optimization architecture. Moreover, the impact of DeepSeek models on the competitive AI landscape is analyzed, comparing them to mainstream LLMs across various fields. Finally, the paper reflects on the insights gained from DeepSeek's innovations and discusses future trends in the technical and engineering development of large AI models, particularly in data, training, and reasoning. © 2014 Chinese Association of Automation.

关键词： DeepSeek large AI models reasoning capability reinforcement learning test-time scaling

来源：评论

学校读者我要写书评

暂无评论

MLFuse: Multi-Scenario Feature Joint Learning for Multi-Modality Image Fusion

引用

IEEE Transactions on Multimedia 2025年

作者： Lei, Jia Li, Jiawei Liu, Jinyuan Wang, Bin Zhou, Shihua Zhang, Qiang Wei, Xiaopeng Kasabov, Nikola K. Dalian University Key Laboratory of Advanced Design and Intelligent Computing Ministry of Education School of Software Engineering Dalian116622 China University of Science and Technology Beijing School of Computer and Communication Engineering Beijing100083 China Dalian University of Technology School of Mechanical Engineering Dalian116024 China Dalian University of Technology School of Computer Science and Technology Dalian116024 China Auckland University of Technology Knowledge Engineering and Discovery Research Institute Auckland1010 New Zealand Ulster University Intelligent Systems Research Center LondonderryBT52 1SA United Kingdom

Multi-modality image fusion (MMIF) entails synthesizing images with detailed textures and prominent objects. Existing methods tend to use general feature extraction to handle different fusion tasks. However, these methods have difficulty breaking fusion barriers across various modalities owing to the lack of targeted learning routes. In this work, we propose a multi-scenario feature joint learning architecture, MLFuse, that employs the commonalities of multi-modality images to deconstruct the fusion progress. Specifically, we construct a cross-modal knowledge reinforcing network that adopts a multipath calibration strategy to promote information communication between different images. In addition, two professional networks are developed to maintain the salient and textural information of fusion results. The spatial-spectral domain optimizing network can learn the vital relationship of the source image context with the help of spatial attention and spectral attention. The edge-guided learning network utilizes the convolution operations of various receptive fields to capture image texture information. The desired fusion results are obtained by aggregating the outputs from the three networks. Extensive experiments demonstrate the superiority of MLFuse for infrared-visible image fusion and medical image fusion. The excellent results of downstream tasks (i.e., object detection and semantic segmentation) further verify the high-quality fusion performance of our method. The code is publicly available at https://***/jialei-sc/MLFuse. © 1999-2012 IEEE.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

引用

Chinese Journal of Electronics 2021年第1期30卷 55-63页

作者： BO Lili ZHU Xuanrui SUN Xiaobing NI Zhen LI Bin School of Information Engineering Yangzhou University Jiangsu Engineering Research Center of Knowledge Management and Intelligent Service State Key Laboratory for Novel Software Technology Nanjing University

Fine-grained change operations can help software developers fix software bugs more accurately and efficiently. However, the current fine-grained change operations are only used in specific fixing process, such as fixing of If statement. In this paper, we conducted an empirical study to explore the fine-grained change operations for bug fixing. Based on the Mozilla bug data,we examined whether similar bugs are fixed with similar change operations. The results show that: First, for bug reports with similar descriptions or bug-fix commits with similar descriptions, their corresponding fine-grained change operations are not related; Second, in the case where the descriptions of both bug reports and bug-fix commits are similar, the fine-grained change operations in patch code are not related; Third, by classifying bug reports, we find that the change operations in the same bug report category are similar; Finally, by analyzing the fine-grained change operations for each bug, we present some combined patterns that are often used together.

关键词： bug reports bug fixing If statement Mozilla bug data bug-fix commits similar descriptions program debugging fine-grained change operations

来源：评论

学校读者我要写书评

暂无评论

Unsupervised feature selection with evolutionary sparsity

引用

Neural Networks 2025年 189卷 107512页

作者： Zhou, Shixuan Xiang, Yi Huang, Han Huang, Pei Peng, Chaoda Yang, Xiaowei Song, Peng School of Software Engineering South China University of Technology Guangzhou510006 China Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education Jilin University Changchun130012 China Ministry of Education Guangzhou510006 China Guangdong Engineering Center for Large Model and GenAI Technology Guangzhou510006 China School of Information Science Guangdong University of Finance and Economics Guangzhou510006 China School of Mathematics and Informatics South China Agricultural University Guangzhou510006 China School of Computer and Control Engineering Yantai University Yantai264005 China

The 2,0-norm is playing an increasingly important role in unsupervised feature selection. However, existing algorithm for optimization problem with 2,0-norm constraint has two problems: First, they cannot automatically determine the sparsity, also known as the number of key features. Second, they have the risk of converging towards local optima, therefore selecting trivial (less informative) features. To address these problems, this paper proposes an unsupervised feature selection method with evolutionary sparsity (EVSP), which integrates the feature selection process with a sparse projection matrix and population search mechanisms into a unified unsupervised feature selection framework. Specifically, the level of sparsity is encoded as population individuals, and subsequently, a multi-objective evolutionary algorithm based on binary encoding is introduced to recursively determine the optimal level of sparsity, thus unsupervisedly guiding the learning of an optimal row-sparse projection matrix. Moreover, by utilizing the feature weights learned through sparse projection, a two-stage strategy called the mutation-repair operator is designed to steer the evolution of the population, aiming to generate high-quality candidate solutions. Comprehensive experiments on eleven benchmark datasets, with a maximum dimensionality of 10304 features and a maximum size of 9298 samples, demonstrate that the proposed EVSP method can effectively determine the optimal sparsity level, significantly outperforming several state-of-the-art methods. © 2025 Elsevier Ltd

关键词： Evolutionary algorithms

来源：评论

学校读者我要写书评

暂无评论

Leveraging machine learning for software redocumentation—A comprehensive comparison of methods in practice

Leveraging machine learning for software redocumentation—A ...

引用

作者： Geist, Verena Moser, Michael Pichler, Josef Santos, Rodolfo Wieser, Volkmar Software Analytics and Evolution Software Competence Center Hagenberg GmbH Hagenberg Austria Department of Software Engineering University of Applied Sciences Upper Austria Hagenberg Austria Knowledge-Based Vision Systems Software Competence Center Hagenberg GmbH Hagenberg Austria

Source code comments contain key information about the underlying software system. Many redocumentation approaches, however, cannot exploit this valuable source of information. This is mainly due to the fact that not all comments have the same goals and target audience and can therefore only be used selectively for redocumentation. Performing a required classification manually, for example, in the form of heuristics, is usually time-consuming and error-prone and strongly dependent on programming languages and guidelines of concrete software systems. By leveraging machine learning (ML), it should be possible to classify comments and thus transfer valuable information from the source code into documentation with less effort but the same quality. We applied classical ML techniques but also deep learning (DL) approaches to legacy systems by transferring source code comments into meaningful representations using, for example, word embeddings but also novel approaches using quick response codes or a special character-to-image encoding. The results were compared with industry-strength heuristic classification. As a result, we found that ML outperforms the heuristics in number of errors and less effort, that is, we finally achieve an accuracy of more than 95% for an image-based DL network and even over 96% for a traditional approach using a random forest classifier. © 2020 John Wiley & Sons Ltd

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Detecting human features in summaries - Symbol sequence statistical regularity

Detecting human features in summaries - Symbol sequence stat...

引用

7th Hellenic Conference on Artificial Intelligence, SETN 2012

作者： Giannakopoulos, George Karkaletsis, Vangelis Vouros, George A. Software and Knowledge Engineering Laboratory National Center of Scientific Research Demokritos Greece Department of Digital Systems University of Pireaus Greece

ISBN: (纸本)9783642304477

The presented work studies textual summaries, aiming to detect the qualities of human multi-document summaries, in contrast to automatically extracted ones. The measured features are based on a generic statistical regularity measure, named Symbol Sequence Statistical Regularity (SSSR). The measure is calculated over both character and word n-grams of various ranks, given a set of human and automatically extracted multi-document summaries from two different corpora. The results of the experiments indicate that the proposed measure provides enough distinctive power to discriminate between the human and non-human summaries. The results hint on the qualities a human summary holds, increasing intuition related to how a good summary should be generated. © 2012 Springer-Verlag.

关键词： Computers

来源：评论

学校读者我要写书评

暂无评论

Exploiting process patterns and process instances to support adaptability of dynamic business processes 25

Exploiting process patterns and process instances to support...

引用

25th International Workshop on Database and Expert Systems Applications, DEXA 2014

作者： Bogl, Andreas Natschlager, Christine Karlinger, Michael Schrefl, Michael Software Competence Center Hagenberg GmbH Hagenberg Austria Data- and Knowledge Engineering Institute Johannes Kepler University Linz Austria

ISBN: (纸本)9781479957224

The traditional approach for business process modeling is rather static, since all possible process flows must be specified at design-time. This restricts the possibility of the user to react to unexpected situations, so conventional business processes cannot represent the flexible way in which human actors would handle discrepancies between real-world and computerized activities. This paper presents a new approach for dynamic business processes that automatically adapts to changed circumstances based on domain-specific process patterns and an evaluation of related process instances. The presented approach not only finds potential alternatives for failed activities, it also ranks them by exploiting process knowledge implicitly captured by former business process execution. © 2014 IEEE.

关键词： Systems engineering

来源：评论

学校读者我要写书评

暂无评论

Prioritization technique for learning cloud services

引用

International Journal of Intelligent Computing and Cybernetics 2015年第3期8卷 222-231页

作者： Apitchaka Singjai Pradorn Sureephong Software Engineering Department College of Art Media and TechnologyChiang Mai UniversityChiang MaiThailand Knowledge Innovation Center College of Art Media and TechnologyChiang Mai UniversityChiang MaiThailand

Purpose–The purpose of this paper is to propose a combined technique of cumulative voting and numerical assignment to prioritize the services of the learning cloud ***/methodology/approach–The approach starts with requirement elicitation,then analyses of the requirements in terms of prioritization and finally classifies the priority of services into ***–As a result of the case study the requirements of the College of Art Media and Technology students has been classified into three service ***/value–This combined prioritized techniques can involve learners in the decision making process about learning cloud services utilization in the organizations.

关键词： Cloud computing Clustering SaaS Learning cloud service Numeric assignment Requirement prioritization

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：