检索结果-内蒙古大学图书馆

Recent Advances in Discrete Speech Tokens: A Review

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Guo, Yiwei Li, Zhihan Wang, Hankun Li, Bohan Shao, Chongtian Zhang, Hanglei Du, Chenpeng Chen, Xie Liu, Shujie Yu, Kai MoE Key Lab of Artificial Intelligence AI Institute X-LANCE Lab Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai China Beijing100080 China

The rapid advancement of speech generation technologies in the era of large language models (LLMs) has established discrete speech tokens as a foundational paradigm for speech representation. These tokens, characterized by their discrete, compact, and concise nature, are not only advantageous for efficient transmission and storage, but also inherently compatible with the language modeling framework, enabling seamless integration of speech into text-dominated LLM architectures. Current research categorizes discrete speech tokens into two principal classes: acoustic tokens and semantic tokens, each of which has evolved into a rich research domain characterized by unique design philosophies and methodological approaches. This survey systematically synthesizes the existing taxonomy and recent innovations in discrete speech tokenization, conducts a critical examination of the strengths and limitations of each paradigm, and presents systematic experimental comparisons across token types. Furthermore, we identify persistent challenges in the field and propose potential research directions, aiming to offer actionable insights to inspire future advancements in the development and application of discrete speech tokens. Copyright © 2025, The Authors. All rights reserved.

关键词： Semantics

Advanced Zero-Shot Text-to-Speech for Background Removal and Preservation with Controllable Masked Speech Prediction

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Zhang, Leying Zhang, Wangyou Chen, Zhengyang Qian, Yanmin Auditory Cognition and Computational Acoustics Lab MoE Key Lab of Artificial Intelligence AI Institute China Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai China

The acoustic background plays a crucial role in natural conversation. It provides context and helps listeners understand the environment, but a strong background makes it difficult for listeners to understand spoken words. The appropriate handling of these backgrounds is situation-dependent: Although it may be necessary to remove background to ensure speech clarity, preserving the background is sometimes crucial to maintaining the contextual integrity of the speech. Despite recent advancements in zero-shot Text-to-Speech technologies, current systems often struggle with speech prompts containing backgrounds. To address these challenges, we propose a Controllable Masked Speech Prediction strategy coupled with a dual-speaker encoder, utilizing a task-related control signal to guide the prediction of dual background removal and preservation targets. Experimental results demonstrate that our approach enables precise control over the removal or preservation of background across various acoustic conditions and exhibits strong generalization capabilities in unseen scenarios. © 2025, CC BY.

关键词： Background noise

Federated and transfer learning for cancer detection based on image analysis

学校读者我要写书评

暂无评论

Neural Computing and Applications 2025年第4期37卷 2239-2284页

作者： Bechar, Amine Medjoudj, Rafik Elmir, Youssef Himeur, Yassine Amira, Abbes Laboratoire LITAN École supérieure en Sciences et Technologies de l’Informatique et du Numérique RN 75 Amizour Bejaia06300 Algeria SGRE-Lab University Tahri Mohammed of Bechar Bechar08000 Algeria College of Engineering and Information Technology University of Dubai Dubai United Arab Emirates Department of Computer Science University of Sharjah sharjah United Arab Emirates Institute of Artificial Intelligence De Montfort University Leicester United Kingdom

This review highlights the efficacy of combining federated learning (FL) and transfer learning (TL) for cancer detection via image analysis. By integrating these techniques, research has shown improvements in diagnostic accuracy and efficiency. Specifically, the use of FL and TL has led to a measurable improvement in the precision of cancer diagnoses, with some studies reporting up to a 20% increase in accuracy compared to traditional methods. This synthesis of FL and TL optimizes distributed data usage while leveraging existing models to expedite learning and application in cancer detection tasks. A concrete assessment of the two methods, including their strengths and weaknesses, is presented. Moving on, their applications in cancer detection are discussed, including potential directions for the future. Finally, this article offers a thorough description of the functions of TL and FL in image-based cancer detection. The authors also make insightful suggestions for additional study in this rapidly developing area. The findings underscore the potential of these combined approaches to significantly advance medical imaging and cancer diagnosis, setting a promising direction for future research. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Federated learning

MobileSteward: Integrating Multiple App-Oriented Agents with Self-Evolution to Automate Cross-App Instructions

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Liu, Yuxuan Sun, Hongda Liu, Wei Luan, Jian Du, Bo Yan, Rui Xiaomi AI Lab Beijing China Gaoling School of Artificial Intelligence Renmin University of China Beijing China School of Computer Science Wuhan University Wuhan China

Mobile phone agents can assist people in automating daily tasks on their phones, which have emerged as a pivotal research spotlight. However, existing procedure-oriented agents struggle with cross-app instructions, due to the following challenges: (1) complex task relationships, (2) diverse app environment, and (3) error propagation and information loss in multi-step execution. Drawing inspiration from object-oriented programming principles, we recognize that object-oriented solutions is more suitable for cross-app instruction. To address these challenges, we propose a self-evolving multi-agent framework named MobileSteward, which integrates multiple app-oriented StaffAgents coordinated by a centralized StewardAgent. We design three specialized modules in MobileSteward: (1) Dynamic Recruitment generates a scheduling graph guided by information flow to explicitly associate tasks among apps. (2) Assigned Execution assigns the task to app-oriented StaffAgents, each equipped with app-specialized expertise to address the diversity between apps. (3) Adjusted Evaluation conducts evaluation to provide reflection tips or deliver key information, which alleviates error propagation and information loss during multi-step execution. To continuously improve the performance of MobileSteward, we develop a Memory-based Self-evolution mechanism, which summarizes the experience from successful execution, to improve the performance of MobileSteward. We establish the first English Cross-APP Benchmark (CAPBench) in the real-world environment to evaluate the agents’ capabilities of solving complex cross-app instructions. Experimental results demonstrate that MobileSteward achieves the best performance compared to both single-agent and multi-agent frameworks, highlighting the superiority of MobileSteward in better handling user instructions with diverse complexity. Copyright © 2025, The Authors. All rights reserved.

关键词： Mobile agents

AI-based Framework for Robust Model-Based Connector Mating in Robotic Wire Harness Installation

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Kienle, Claudius Alt, Benjamin Schneider, Finn Pertlwieser, Tobias Jäkel, Rainer Rayyes, Rania ArtiMinds Robotics Karlsruhe Germany IAS Lab Computer Science Department TU Darmstadt Germany AICOR Institute for Artificial Intelligence University of Bremen Germany Karlsruhe Germany

Despite the widespread adoption of industrial robots in automotive assembly, wire harness installation remains a largely manual process, as it requires precise and flexible manipulation. To address this challenge, we design a novel AI-based framework that automates cable connector mating by integrating force control with deep visuotactile learning. Our system optimizes search-and-insertion strategies using first-order optimization over a multimodal transformer architecture trained on visual, tactile, and proprioceptive data. Additionally, we design a novel automated data collection and optimization pipeline that minimizes the need for machine learning expertise. The framework optimizes robot programs that run natively on standard industrial controllers, permitting human experts to audit and certify them. Experimental validations on a center console assembly task demonstrate significant improvements in cycle times and robustness compared to conventional robot programming approaches. Videos are available under https: //***/*** Codes 68T40 © 2025, CC BY.

关键词： Robot programming

Delusions of Large Language Models

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Xu, Hongshen Yang, Zixv Zhu, Zichen Lan, Kunyao Wang, Zihan Wu, Mengyue Ji, Ziwei Chen, Lu Fung, Pascale Yu, Kai X-LANCE Lab Department of Computer Science and Engineering MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University Shanghai China Hong Kong University of Science and Technology Hong Kong

Large Language Models (LLMs) often generate factually incorrect but plausible outputs, known as hallucinations. We identify a more insidious phenomenon, LLM delusion, defined as high-belief hallucinations—incorrect outputs with abnormally high confidence, making them harder to detect and mitigate. Unlike ordinary hallucinations, delusions persist with low uncertainty, posing significant challenges to model reliability. Through empirical analysis across different model families and sizes on several Question-Answering tasks, we show that delusions are prevalent and distinct from hallucinations. LLMs exhibit lower honesty with delusions, which are harder to override via fine-tuning or self-reflection. We link delusion formation with training dynamics and dataset noise and explore mitigation strategies such as retrieval-augmented generation and multi-agent debating to mitigate delusions. By systematically investigating the nature, prevalence, and mitigation of LLM delusions, our study provides insights into the underlying causes of this phenomenon and outlines future directions for improving model reliability. Copyright © 2025, The Authors. All rights reserved.

关键词：

MVLA-Net:A Multi-View Lesion Attention Network for Advanced Diagnosis and Grading of Diabetic Retinopathy

学校读者我要写书评

暂无评论

computers, Materials & Continua 2025年第4期83卷 1173-1193页

作者： Tariq Mahmood Tanzila Saba Faten S.Alamri Alishba Tahir Noor Ayesha Artificial Intelligence and Data Analytics(AIDA)Lab College of Computer&Information SciencesPrince Sultan UniversityRiyadh11586Saudi Arabia Faculty of Information Sciences University of EducationVehari CampusVehari61100Pakistan Department of Mathematical Sciences College of SciencePrincess Nourah bint Abdulrahman UniversityRiyadh84428Saudi Arabia Shifa College of Medicine Shifa Tameer-E-Millat UniversityIslamabad44000Pakistan Center of Excellence in Cyber Security(CYBEX) Prince Sultan UniversityRiyadh11586Saudi Arabia

Innovation in learning algorithms has made retinal vessel segmentation and automatic grading tech-niques crucial for clinical diagnosis and prevention of diabetic *** traditional methods struggle with accuracy and reliability due to multi-scale variations in retinal blood vessels and the complex pathological relationship in fundus images associated with diabetic *** the single-modal diabetic retinopathy grading network addresses class imbalance challenges and lesion representation in fundus image data,dual-modal diabetic retinopathy grading methods offer superior ***,the scarcity of dual-modal data and the lack of effective feature fusion methods limit their potential due to multi-scale *** paper addresses these issues by focusing on multi-scale retinal vessel segmentation,dual feature fusion,data augmentation,and attention-based *** proposed model aims to improve comprehensive segmentation for retinal images with varying vessel *** employs a dual-branch parallel architecture that integrates a transformer encoder with a convolutional neural network encoder to extract local and global information for synergistic saliency *** that,the model uses residual structures and attention modules to extract critical lesions,enhancing the accuracy and reliability of diabetic retinopathy *** evaluate the efficacy of the proposed approach,this study compared it with other pre-trained publicly open models,ResNet152V2,ConvNext,Efficient Net,DenseNet,and Swin Transform,with the same developmental *** models achieved approximately 85%accuracy with the same image preparation ***,the proposed approach outperforms and optimizes existing models by achieving an accuracy of 99.17%,99.04%,and 99.24%,on Kaggle APTOS19,IDRiD,and EyePACS datasets,*** results support the model’s utility in helping ophthalmologists diagnose diabetic retinopathy more rapidly and accurat

关键词： Diabetic retinopathy grading retinal vessel segmentation dual-modal deep learning attention mecha-nism health risks

Latent-space adversarial training with post-aware calibration for defending large language models against jailbreak attacks

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Yi, Xin Li, Yue Wang, Linlin Wang, Xiaoling He, Liang Lab of Artificial Intelligence for Education East China Normal University Shanghai200062 China Shanghai Institute of Artificial Intelligence for Education East China Normal University Shanghai200062 China School of Computer Science and Technology East China Normal University Shanghai200062 China

Ensuring safety alignment has become a critical requirement for large language models (LLMs), particularly given their widespread deployment in real-world applications. However, LLMs remain susceptible to jailbreak attacks, which exploit system vulnerabilities to bypass safety measures and generate harmful outputs. Although numerous defense mechanisms based on adversarial training have been proposed, a persistent challenge lies in the exacerbation of over-refusal behaviors, which compromise the overall utility of the model. To address these challenges, we propose a Latentspace Adversarial Training with Post-aware Calibration (LATPC) framework. During the adversarial training phase, LATPC compares harmful and harmless instructions in the latent space and extracts safety-critical dimensions to construct refusal features attack, precisely simulating agnostic jailbreak attack types requiring adversarial mitigation. At the inference stage, an embedding-level calibration mechanism is employed to alleviate over-refusal behaviors with minimal computational overhead. Experimental results demonstrate that, compared to various defense methods across five types of jailbreak attacks, LATPC framework achieves a superior balance between safety and utility. Moreover, our analysis underscores the effectiveness of extracting safety-critical dimensions from the latent space for constructing robust refusal feature attacks. © 2025, CC BY.

关键词： Adversarial machine learning

Development of HMM Based Parts of Speech Tagger for Hadoti 1st

学校读者我要写书评

暂无评论

Development of HMM Based Parts of Speech Tagger for Hadoti

1st International Conference on Computation of artificial intelligence and Machine Learning, ICCAIML 2024

作者： Nagar, Anushka Joshi, Nisheeth Katyayan, Pragya Arora, Palak Department of Computer Science Banasthali Vidyapith Rajasthan Radha Kishnpura India Speech and Language Processing Lab Center for Artificial Intelligence Banasthali Vidyapith Rajasthan Radha Kishnpura India

ISBN: (纸本)9783031714832

In this paper, we have shown the development of a Part of Speech (POS) tagger for Hadoti - a prominent language spoken in Rajasthan, India - despite its limited resources. For this, we manually tagged a corpus of 50,000 POS-tagged sentences and trained it using a Hidden Markov Model (HMM). Since no prior work had been reported in this area, we couldn't compare our results to any other system. This paper documents the efforts made to create an HMM POS tagger for Hadoti, to stimulate further research in this field. This work is expected to serve as a foundation for preserving the language and as a resource for aspiring researchers who wish to explore this area of Hadoti Language Processing. The system was evaluated for accuracy and produced 99.87% accurate results on seen data and 98.78% on unseen data. The system was able to produce an accuracy of 99.33% on the entire test corpus. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Hidden Markov models