检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Na, CheolWon Choi, YunSeok Lee, Jee-Hyong College of Computing and Informatics Sungkyunkwan University Korea Republic of

Many adversarial attack approaches are proposed to verify the vulnerability of language models. However, they require numerous queries and the information on the target model. Even black-box attack methods also require the target model’s output information. They are not applicable in real-world scenarios, as in hard black-box settings where the target model is closed and inaccessible. Even the recently proposed hard black-box attacks still require many queries and demand extremely high costs for training adversarial generators. To address these challenges, we propose Q-faker (Query-free Hard Black-box Attacker), a novel and efficient method that generates adversarial examples without accessing the target model. To avoid accessing the target model, we use a surrogate model instead. The surrogate model generates adversarial sentences for a target-agnostic attack. During this process, we leverage controlled generation techniques. We evaluate our proposed method on eight datasets. Experimental results demonstrate our method’s effectiveness including high transferability and the high quality of the generated adversarial examples, and prove its practical in hard black-box settings. © 2025, CC BY-SA.

关键词：

来源：评论

学校读者我要写书评

暂无评论

DeCAP: Context-Adaptive Prompt Generation for Debiasing Zero-shot Question Answering in Large Language Models

arXiv

引用

arXiv 2025年

作者： Bae, Suyoung Choi, YunSeok Lee, Jee-Hyong College of Computing and Informatics Sungkyunkwan University Korea Republic of

While Large Language Models (LLMs) excel in zero-shot Question Answering (QA), they tend to expose biases in their internal knowledge when faced with socially sensitive questions, leading to a degradation in performance. Existing zero-shot methods are efficient but fail to consider context and prevent bias propagation in the answers. To address this, we propose DeCAP, a method for debiasing LLMs using Context-Adaptive Prompt Generation. DeCAP leverages a Question Ambiguity Detection to take appropriate debiasing actions based on the context and a Neutral Answer Guidance Generation to suppress the LLMs make objective judgments about the context, minimizing the propagation of bias from their internal knowledge. Our various experiments across eight LLMs show that DeCAP achieves state-of-the-art zero-shot debiased QA performance. This demonstrates DeCAP’s efficacy in enhancing the fairness and accuracy of LLMs in diverse QA settings. Our code is publicly available at https://***/BaeSuyoung/DeCAP. © 2025, CC BY.

关键词： Zero-shot learning

来源：评论

学校读者我要写书评

暂无评论

Empirical Study on Data Attributes Insufficiency of Evaluation Benchmarks for LLMs 31

Empirical Study on Data Attributes Insufficiency of Evaluati...

引用

31st International Conference on Computational Linguistics, COLING 2025

作者： Liu, Chuang Jin, Renren Yao, Zheng Li, Tianyi Cheng, Liang Steedman, Mark Xiong, Deyi College of Intelligence and Computing Tianjin University China School of Electrical Engineering and Computer Science University of Queensland Australia School of Informatics University of Edinburgh United Kingdom

ISBN: (纸本)9798891761964

Previous benchmarks for evaluating large language models (LLMs) have primarily emphasized quantitative metrics, such as data volume. However, this focus may neglect key qualitative data attributes that can significantly impact the final rankings of LLMs, resulting in unreliable leaderboards. In this paper, we investigate whether current LLM benchmarks adequately consider these data attributes. We specifically examine three attributes: diversity, redundancy, and difficulty. To explore these attributes, we propose a framework with three separate modules, each designed to assess one of the attributes. Using a method that progressively incorporates these attributes, we analyze their influence on the benchmark. Our experimental results reveal a meaningful correlation between LLM rankings on the revised benchmark and the original benchmark when these attributes are accounted for. These findings indicate that existing benchmarks often fail to meet all three criteria, highlighting a lack of consideration for multifaceted data attributes in current evaluation datasets. © 2025 Association for Computational Linguistics.

关键词： Computational linguistics

来源：评论

学校读者我要写书评

暂无评论

AI-Powered Cloud computing for Performance Optimization and Scalability in Distributed Systems

AI-Powered Cloud Computing for Performance Optimization and ...

引用

2025 International Conference on computing for Sustainability and Intelligent Future, COMP-SIF 2025

作者： Selvam, Muthu Kishan, B.S. College of Computing and Informatics University of North Carolina at Charlotte CharlotreNC United States Bms Institute of Technology and Management Department of Mechanical Engineering Bangalore India

ISBN: (纸本)9798331538538

The IT profession has observed the ingenious application of Artificial Intelligence in the field of cloud computing as a solution to increase performance along with scalability within the distributed computing environment. Conventional cloud computing context has issues with management of resources, responsiveness, and finally security. The following paper will discuss an AI based architecture which will incorporate ML algorithms to optimize the usage of the resources available in the computer cloud to balance the load between different users and ensure security. Thus, the proposed model incorporates reinforcement learning for the dynamic load distribution, predictive analytics for workload forecasting, and deep learning based on anomaly detection to improve the security features of the system. The performance comparison with the baseline and the experiments conducted show satisfactory results in more than one performance indicators. For CPU usage, it rose to 26.4%, which was achieved alongside the provision of the 48% higher request throughput and the 40% better response time under the scenarios of the heavy workload. Moreover, the security framework using CNNs enhanced the detection accuracy of intrusion to 96.7% and a reduction of the false positives by sixty-six percent. These are some of the implications suggesting that the application of artificial intelligence can bring positive changes in the goals regarding cloud performance, resource utilization and overall security features. In this way, using AI-driven automation and prediction integrated in this work is offered a powerful, adaptive, and safe cloud model for expanded usage. With the proposed approach that has been developed the businesses and the service providers will be in a position to attain efficient business operational costs together with strong system reliability that will assist in developing the next generation of intelligent cloud computing. © 2025 IEEE.

关键词： Distributed cloud

来源：评论

学校读者我要写书评

暂无评论

Improved Xception with Local Dense Connections and Transition Layer for Facial Expression Recognition 24

Improved Xception with Local Dense Connections and Transitio...

引用

3rd International Conference on Artificial Intelligence and Intelligent Information Processing, AIIIP 2024

作者： Zhang, Zhiping Abdul-Rahman, Shuzlina Sabri, Norlina Mohd College of Computing Informatics and Media Universiti Teknologi MARA Selangor Malaysia College of Computer and Mathematics Xinyu University Jiangxi Xinyu China Universiti Teknologi MARA Cawangan Terengganu Kuala Terengganu Malaysia

ISBN: (纸本)9798400707308

Traditional deep convolutional neural networks are used for facial expression recognition, which makes the number of neurons and parameters huge, wastes computing resources, and even causes problems such as overfitting and network degradation. Meanwhile, single-scale expression features cannot describe rich facial expression information and are not evenly distributed on face images. We proposed an improved Xception method based on the original Xception network, it retains the depth-separable convolution and residual structure, adds multi-scale convolution and local dense connection. Our proposed method achieved an accuracy rate of 89.54% on the FER2013 dataset, strengthened feature reuse, and enhanced generalization ability. © 2024 Copyright held by the owner/author(s).

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Underwater Sonar Image Targets Detection Based on Improved RT-DETR

引用

IEEE Geoscience and Remote Sensing Letters 2025年 22卷

作者： Li, Ang Hamzah, Raseeda Gao, Yousheng Jiujiang Vocational University College of Information Engineering Jiangxi Jiujiang332000 China College of Computing Informatics and Mathematics Selangor Shah Alam40450 Malaysia Melaka College of Computing Informatics and Mathematics Jasin Melaka77300 Malaysia

Underwater sonar imagery is characterized by small target sizes and low resolution, which can result in detection failures or false positives. To counteract these challenges, we introduce the underwater sonar detection transformer (US-DETR), an underwater sonar object detection model derived from the real-time detection transformer (RT-DETR) framework, incorporating attention-based feature fusion. US-DETR includes a novel enhanced feature interaction (EFI) module, which enhances the feature extraction network’s ability to perceive global information of the detected target. In addition, we propose a novel nonlocal attention feature fusion (NAFF) module to heighten the network’s sensitivity to the spatial relationships between feature channels across different scales, thereby enhancing its channel position and global information awareness. Experiments are conducted on a benchmark underwater sonar image dataset. Experimental results show that compared with RT-DETR, US-DETR achieves a 2.2% higher mean average precision (mAP) and a 2.1% higher F1 score compared with RT-DETR. The model also strikes an effective balance between detection speed and accuracy, achieving real-time performance of 126 FPS, which can meet the real-time requirements in industrial production. © 2004-2012 IEEE.

关键词： Underwater acoustics

来源：评论

学校读者我要写书评

暂无评论

Knowledge augmented significant language model-based chatbot for explainable diabetes mellitus prediction 19

Knowledge augmented significant language model-based chatbot...

引用

19th International Conference on Ubiquitous Information Management and Communication, IMCOM 2025

作者： Elfayoumi, Mazen AbouElazm, Mohamed Mohamed, Omar Abuhmed, Tamer El-Sappagh, Shaker Faculty of Computer Science and Engineering Galala University Egypt Faculty of Computer Science Goldsmith UOL United Kingdom College of Computing and Informatics Sungkyunkwan University Korea Republic of

ISBN: (纸本)9798331507817

This study proposes an innovative diabetes prediction chatbot that utilizes large language models (LLMs) to determine the likelihood of diabetes based on specific patient inputs. Unlike conventional machine learning models and in addition to providing precise, individualized, robust prediction of diabetes augmented by the percentage of its confidence, this chatbot provides detailed text-based explanations for its individual predictions and enhances user’s understanding by retrieving and presenting similar cases using case-based reasoning techniques. Utilizing key health indicators such as hypertension status, heart disease presence, smoking history, BMI, HbA1c level, and blood glucose levels, the chatbot not only predicts the presence of diabetes but also educates users about the underlying reasons for each prediction. The system's explanation module promotes transparency and trust in the predictive process. The proposed architecture integrates the retrieval augmented generation (RAG) technique with prompt engineering across multiple LLMs, including Llama 3.1, GPT-3.5, and Gemma2. This approach augments the chatbot’s responses with contextually relevant data from similar past cases, thereby enhancing both relevance and accuracy. Furthermore, RAG minimizes LLM hallucinations while enhancing them with up-to-date, precise, personalized medical information. The objective of this chatbot is to offer healthcare professionals and patients a dependable, educational instrument for early detection of diabetes supported by transparent AI principles. This will have a positive effect on preventative healthcare measures. © 2025 IEEE.

关键词： Prediction models

来源：评论

学校读者我要写书评

暂无评论

Facilitating Gender Diverse Authorship: A Comparative Analysis of Academic Publisher Name Change Policies 25

Facilitating Gender Diverse Authorship: A Comparative Analys...

引用

Proceedings of the 17th ACM Web Science Conference 2025

作者： Deanna Zarrillo College of Computing & Informatics Drexel University Philadelphia PA USA

ISBN: (纸本)9798400714832

Post-publication name change policies are vital for safeguarding privacy and equity for authors navigating identity changes, including gender transitions, within academic publishing. Before the introduction of these policies in 2019, trans, non-binary, and gender diverse authors faced significant barriers, often risking privacy violations and disruptions to their academic records. This study employs thematic content analysis to assess the publicly available name change policies of nine academic journal publishers, examining their structure, discoverability, and alignment with inclusivity principles. Key findings reveal a lack of standardization across policies, with notable variation in content and accessibility. While privacy and correction mechanisms are commonly addressed, critical themes such as author engagement and broader industry context remain underdeveloped. The policies’ discoverability on publisher websites also varies widely, potentially limiting their utility to those who need them most. These gaps highlight covert marginalization embedded in policy design and communication. By situating this analysis within an ethic of care and the broader context of digital identity management, this study reveals how publishing policies intersect with web-based systems of scholarly communication. The findings urge academic publishers, technologists, and policymakers to co-create inclusive solutions that align with emerging metadata standards and ethical frameworks. This research lays a foundation for understanding how academic infrastructure can evolve to better serve diverse author communities in a connected and equitable web ecosystem.

关键词： Design justice

来源：评论

学校读者我要写书评

暂无评论

Deep Reinforcement Learning-Based Framework for Enhancing Cybersecurity

International Journal of Interactive Mobile Technologies

引用

International Journal of Interactive Mobile Technologies 2025年第3期19卷 170-190页

作者： Al-Nawashi, Malek M. Al-Hazaimeh, Obaida M. Tahat, Nedal M. Gharaibeh, Nasr Abu-Ain, Waleed A. Abu-Ain, Tarik Al-Huson University College Al-Balqa Applied University Irbid Jordan Faculty of Information Technology Department of Information Security and Cybersecurity Philadelphia University Amman Jordan Faculty of Science The Hashemite University Zarqa Jordan College of Science and Computer Engineering Taibah University Yanbu Saudi Arabia College of Computing and Informatics Saudi Electronic University Riyadh Saudi Arabia

The detection of cyberattacks has been increasingly emphasized in recent years, focusing on both infrastructure and people. Conventional security measures such as intrusion detection, firewalls, and encryption are insufficient in protecting cyber systems against growing and changing threats. In order to address this problem, scholars have explored reinforcement learning (i.e., RL) as a potential solution for intricate cybersecurity decision-making difficulties. Nevertheless, the use of RL faces several obstacles, including dynamic attack scenarios, insufficient training data, and the challenge of replicating real-world complexities. This study presents a novel framework that uses deep reinforcement learning (i.e., DRL) to simulate harmful cyberattacks and improve cybersecurity. This study presents an agent-based framework that is capable of ongoing learning and adaptation in a dynamic network security environment. The agent determines the optimal course of action by considering the current state of the network and the rewards it receives for its decisions. The CIC-IDS-2018 database, constructed using Python 3.7 programming, was used. The conducted studies yielded outstanding results, with a detection accuracy of 98.82% achieved for the CIC-IDS-2018 database in cyber-attack classification. © 2025 by the authors of this article. Published under CC-BY.

关键词： Cybersecurity

来源：评论

学校读者我要写书评

暂无评论

LLM+RAG Driven Topic Modeling 2025

LLM+RAG Driven Topic Modeling

引用

Proceedings of the 56th ACM Technical Symposium on Computer Science Education V. 2

作者： Frank Ley Garcia College of Computing and Informatics UNC Charlotte Charlotte NC USA

ISBN: (纸本)9798400705328

This paper explores the use of Large Language Models (LLMs) combined with Retrieval-Augmented Generation (RAG) to assist instructors in identifying course-wide student challenges through topic modeling. Unlike previous studies that primarily generate personalized resources for individual students, this research focuses on analyzing reflections from an entire class to inform curriculum design and intervention strategies. Using the LLaMa-3.1-8B model, experiments across varying cosine similarity thresholds reveal both the strengths and limitations of integrating retrieval-based models. While RAG did not consistently outperform standalone LLMs, it offers key insights into the complexities of applying retrieval-augmented approaches in educational settings.

关键词： large language models

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：