NSTL国家科技图书文献中心

1. Pre-Trained Models for Search and Recommendation: Introduction to the Special Issue—Part 1 NSTL国家科技图书文献中心

Wenjie Wang | Zheng Liu ... - 《ACM transactions on information systems》 - 2025,43(2) - 27.1～27.6 - 共6页

摘要： variety of studies on the application of pre-trained | , but are not limited to, leveraging pre-trained | models to search and recommendation, covering both | models for fine-grained user representation learning | In summary, this special issue contains a

关键词： Pre-trained Models | Search | Recommendation | Large Language Models | Trustworthiness

2. Revisiting Class-Incremental Learning with Pre-Trained Models: Generalizability and Adaptivity are All You Need NSTL国家科技图书文献中心

Zhou, Da-Wei | Cai, Zi-Wen ... - 《International Journal of Computer Vision》 - 2025,133(3) - 1012～1032 - 共21页

摘要： pre-trained models (PTMs) accessible for CIL | . Traditional CIL models* are trained from scratch to* | pre-trained and downstream datasets, PTM can be | continually acquire knowledge as data evolves. Recently, pre | the embeddings of PTM and adapted models* for*

关键词： Class-incremental learning | Pre-trained models | Continual learning | Catastrophic forgetting

3. Using Embeddings of Pre-trained Models for Cross-Database Dysarthria Detection: Supervised vs. Self-supervised Approach NSTL国家科技图书文献中心

Sally Ismail | Margarita Anastassov...... - 《Machine Learning,Optimization,and Data Science,Part II》 - International Conference on Machine Learning,Optimization,and Data Science - 2025, - 229～239 - 共11页

摘要： weakly supervised, pre-trained* models, as feature* | : SVM and a neural network. The pre-trained models | One of the main obstacles facing the research | in the automatic detection of voice pathology is | the lack of data. In this work, we try to tackle

关键词： Self-supervised learning | Voice pathology | Deep learning | Pre-trained models | Machine learning | Wav2vec2 | HuBERT | Whisper

4. SDPT: Synchronous Dual Prompt Tuning for Fusion-Based Visual-Language Pre-trained Models NSTL国家科技图书文献中心

Yang Zhou | Yongjian Wu ... - 《Computer Vision - ECCV 2024,Part XLIV》 - European Conference on Computer Vision - 2025, - 340～356 - 共17页

摘要： fusion-based visual-language pre-trained* models (VLPMs* | -trained* models. However, their application to dual-modal* | success in parameter-efficient fine-tuning on large pre | Prompt tuning methods have achieved remarkable | ), such as GLIP, has encountered issues. Existing prompt

关键词： Prompt tuning | Parameter-efficient fine-tuning | Visual-language pre-trained models

5. Adapt and Refine: A Few-Shot Class-Incremental Learner via Pre-Trained Models NSTL国家科技图书文献中心

Sunyuan Qiang | Zhu Xiong ... - 《Pattern Recognition and Computer Vision,Part I》 - Chinese Conference on Pattern Recognition and Computer Vision - 2025, - 431～444 - 共14页

摘要： powerful representational capabilities of pre-trained | models (PTMs). Subsequently, we further adapt and | The intricate and ever-changing nature of the | real world imposes greater demands on neural networks | , necessitating the rapid assimilation of fleeting new concepts

关键词： Few-shot class-incremental learning (FSCIL) | Catastrophic forgetting | Pre-trained models (PTMs)

6. Acoustic Classification of Bird Species Using Improved Pre-trained Models NSTL国家科技图书文献中心

Jie Xie | Mingying Zhu ... - 《PRICAI 2024,Part I》 - Pacific Rim International Conference on Artificial Intelligence - 2025, - 375～382 - 共8页

摘要：-trained* ImageNet models for acoustic classification of* | bird species. However, applying pre-trained* ImageNet* | novel classification framework based on a pre-trained | identification. Previous studies have explored various pre | models directly to audio classification tasks presents

关键词： Bird sound classification | Pre-trained model | Attention | GRU

7. Introducing Structural Information of Argumentative Essays Into Pre-trained Models NSTL国家科技图书文献中心

Chuhan Wang | Dailin Li ... - 《Natural Language Processing and Chinese Computing,Part V》 - CCF International Conference on Natural Language Processing and Chinese Computing - 2025, - 365～376 - 共12页

摘要： significant differences in predictions between models | simultaneously among models* with varying parameter sizes* | inference from large language models. By integrating these | Argument Mining (AM) is a significant task in | the field of Natural Language Processing (NLP

关键词： Argument components identification | Pre-trained language model | Ensemble learning

8. Exploiting Pre-Trained Models and Low-Frequency Preference for Cost-Effective Transfer-based Attack NSTL国家科技图书文献中心

MINGYUAN FAN | CEN CHEN ... - 《ACM transactions on knowledge discovery from data》 - 2025,19(2) - 52.1～52.18 - 共18页

摘要： employing pre-trained* models as proxy models, which are* | ), which addresses the optimization task in pre-trained | prices of training proxy models* also leads to* | models. CTA can be unleashed against broad applications | target models, regardless of their architecture or

关键词： Deep Neural Networks | Adversarial Examples | Black-box Adversarial Attacks | Transferability

9. Leveraging Pre-trained Models for Robust Federated Learning for Kidney Stone Type Recognition NSTL国家科技图书文献中心

Ivan Reyes-Amezcua | Michael Rojas-Ruiz ... - 《Advances in Computational Intelligence,Part II》 - Mexican International Conference on Artificial Intelligence - 2025, - 168～181 - 共14页

摘要： degradation. Using pre-trained* models, this research* | . This highlights the potential of merging pre-trained | data privacy. However, FL models* are susceptible to* | models with FL to address privacy and performance | Deep learning developments have improved

关键词： Federated learning | Robustness | Medical imaging

10. A System of Multimodal Image-Text Retrieval Based on Pre-Trained Models Fusion NSTL国家科技图书文献中心

Qiang Li | Feng Zhao ... - 《Concurrency and computation: practice and experience》 - 2025,37(3) - e8345.1～e8345.11 - 共11页

摘要： fusion of pre-trained* models. Firstly, we enhance the* | challenge for the models* generalization performance and* | to improve the models* generalization performance* | three different models. Finally, we construct a | models based on the specific characteristics of the

摘要： variety of studies on the application of pre-trained | , but are not limited to, leveraging pre-trained | models to search and recommendation, covering both | models for fine-grained user representation learning | In summary, this special issue contains a

关键词： Pre-trained Models | Search | Recommendation | Large Language Models | Trustworthiness

摘要： pre-trained models (PTMs) accessible for CIL | . Traditional CIL models* are trained from scratch to* | pre-trained and downstream datasets, PTM can be | continually acquire knowledge as data evolves. Recently, pre | the embeddings of PTM and adapted models* for*

关键词： Class-incremental learning | Pre-trained models | Continual learning | Catastrophic forgetting

摘要： weakly supervised, pre-trained* models, as feature* | : SVM and a neural network. The pre-trained models | One of the main obstacles facing the research | in the automatic detection of voice pathology is | the lack of data. In this work, we try to tackle

关键词： Self-supervised learning | Voice pathology | Deep learning | Pre-trained models | Machine learning | Wav2vec2 | HuBERT | Whisper

摘要： fusion-based visual-language pre-trained* models (VLPMs* | -trained* models. However, their application to dual-modal* | success in parameter-efficient fine-tuning on large pre | Prompt tuning methods have achieved remarkable | ), such as GLIP, has encountered issues. Existing prompt

关键词： Prompt tuning | Parameter-efficient fine-tuning | Visual-language pre-trained models

摘要： powerful representational capabilities of pre-trained | models (PTMs). Subsequently, we further adapt and | The intricate and ever-changing nature of the | real world imposes greater demands on neural networks | , necessitating the rapid assimilation of fleeting new concepts

关键词： Few-shot class-incremental learning (FSCIL) | Catastrophic forgetting | Pre-trained models (PTMs)

摘要：-trained* ImageNet models for acoustic classification of* | bird species. However, applying pre-trained* ImageNet* | novel classification framework based on a pre-trained | identification. Previous studies have explored various pre | models directly to audio classification tasks presents

关键词： Bird sound classification | Pre-trained model | Attention | GRU

摘要： significant differences in predictions between models | simultaneously among models* with varying parameter sizes* | inference from large language models. By integrating these | Argument Mining (AM) is a significant task in | the field of Natural Language Processing (NLP

关键词： Argument components identification | Pre-trained language model | Ensemble learning

摘要： employing pre-trained* models as proxy models, which are* | ), which addresses the optimization task in pre-trained | prices of training proxy models* also leads to* | models. CTA can be unleashed against broad applications | target models, regardless of their architecture or

关键词： Deep Neural Networks | Adversarial Examples | Black-box Adversarial Attacks | Transferability

摘要： degradation. Using pre-trained* models, this research* | . This highlights the potential of merging pre-trained | data privacy. However, FL models* are susceptible to* | models with FL to address privacy and performance | Deep learning developments have improved

关键词： Federated learning | Robustness | Medical imaging

摘要： fusion of pre-trained* models. Firstly, we enhance the* | challenge for the models* generalization performance and* | to improve the models* generalization performance* | three different models. Finally, we construct a | models based on the specific characteristics of the

关键词： Chinese-CLIP | image-text retrieval | MixGen | model fusion | multimodal

4008-161-200 800-990-8900

国家科技图书文献中心

摘要： variety of studies on the application of pre-trained | , but are not limited to, leveraging pre-trained | models to search and recommendation, covering both | models for fine-grained user representation learning | In summary, this special issue contains a

关键词： Pre-trained Models | Search | Recommendation | Large Language Models | Trustworthiness

摘要： pre-trained models (PTMs) accessible for CIL | . Traditional CIL models are trained from scratch to | pre-trained and downstream datasets, PTM can be | continually acquire knowledge as data evolves. Recently, pre | the embeddings of PTM and adapted models for

关键词： Class-incremental learning | Pre-trained models | Continual learning | Catastrophic forgetting

摘要： weakly supervised, pre-trained models, as feature | : SVM and a neural network. The pre-trained models | One of the main obstacles facing the research | in the automatic detection of voice pathology is | the lack of data. In this work, we try to tackle

关键词： Self-supervised learning | Voice pathology | Deep learning | Pre-trained models | Machine learning | Wav2vec2 | HuBERT | Whisper

摘要： fusion-based visual-language pre-trained models (VLPMs | -trained models. However, their application to dual-modal | success in parameter-efficient fine-tuning on large pre | Prompt tuning methods have achieved remarkable | ), such as GLIP, has encountered issues. Existing prompt

关键词： Prompt tuning | Parameter-efficient fine-tuning | Visual-language pre-trained models

摘要： powerful representational capabilities of pre-trained | models (PTMs). Subsequently, we further adapt and | The intricate and ever-changing nature of the | real world imposes greater demands on neural networks | , necessitating the rapid assimilation of fleeting new concepts

关键词： Few-shot class-incremental learning (FSCIL) | Catastrophic forgetting | Pre-trained models (PTMs)

摘要：-trained ImageNet models for acoustic classification of | bird species. However, applying pre-trained ImageNet | novel classification framework based on a pre-trained | identification. Previous studies have explored various pre | models directly to audio classification tasks presents

关键词： Bird sound classification | Pre-trained model | Attention | GRU

摘要： significant differences in predictions between models | simultaneously among models with varying parameter sizes | inference from large language models. By integrating these | Argument Mining (AM) is a significant task in | the field of Natural Language Processing (NLP

关键词： Argument components identification | Pre-trained language model | Ensemble learning

摘要： employing pre-trained models as proxy models, which are | ), which addresses the optimization task in pre-trained | prices of training proxy models also leads to | models. CTA can be unleashed against broad applications | target models, regardless of their architecture or

关键词： Deep Neural Networks | Adversarial Examples | Black-box Adversarial Attacks | Transferability

摘要： degradation. Using pre-trained models, this research | . This highlights the potential of merging pre-trained | data privacy. However, FL models are susceptible to | models with FL to address privacy and performance | Deep learning developments have improved

关键词： Federated learning | Robustness | Medical imaging

摘要： fusion of pre-trained models. Firstly, we enhance the | challenge for the models generalization performance and | to improve the models generalization performance | three different models. Finally, we construct a | models based on the specific characteristics of the

关键词： Chinese-CLIP | image-text retrieval | MixGen | model fusion | multimodal

4008-161-200 800-990-8900

国家科技图书文献中心

摘要： pre-trained models (PTMs) accessible for CIL | . Traditional CIL models* are trained from scratch to* | pre-trained and downstream datasets, PTM can be | continually acquire knowledge as data evolves. Recently, pre | the embeddings of PTM and adapted models* for*

摘要： weakly supervised, pre-trained* models, as feature* | : SVM and a neural network. The pre-trained models | One of the main obstacles facing the research | in the automatic detection of voice pathology is | the lack of data. In this work, we try to tackle

摘要： fusion-based visual-language pre-trained* models (VLPMs* | -trained* models. However, their application to dual-modal* | success in parameter-efficient fine-tuning on large pre | Prompt tuning methods have achieved remarkable | ), such as GLIP, has encountered issues. Existing prompt

摘要：-trained* ImageNet models for acoustic classification of* | bird species. However, applying pre-trained* ImageNet* | novel classification framework based on a pre-trained | identification. Previous studies have explored various pre | models directly to audio classification tasks presents

摘要： significant differences in predictions between models | simultaneously among models* with varying parameter sizes* | inference from large language models. By integrating these | Argument Mining (AM) is a significant task in | the field of Natural Language Processing (NLP

摘要： employing pre-trained* models as proxy models, which are* | ), which addresses the optimization task in pre-trained | prices of training proxy models* also leads to* | models. CTA can be unleashed against broad applications | target models, regardless of their architecture or

摘要： degradation. Using pre-trained* models, this research* | . This highlights the potential of merging pre-trained | data privacy. However, FL models* are susceptible to* | models with FL to address privacy and performance | Deep learning developments have improved

摘要： fusion of pre-trained* models. Firstly, we enhance the* | challenge for the models* generalization performance and* | to improve the models* generalization performance* | three different models. Finally, we construct a | models based on the specific characteristics of the