Latest Articles for Pre-trained Models

TransTroj enhances backdoor attacks, improving effectiveness and resistance to detection.

A new dataset reveals insights into the use of pre-trained models in software projects.

A strategy to enhance performance and fairness in federated learning models.

Research shows improved accuracy in recognizing emotions from speech across languages.

A new method enhances active learning efficiency in machine learning.

New training method helps language models remember information better.

A method for assessing the transferability of pre-trained models for object detection.

This article presents a new method for improving visual grounding by merging data and model knowledge.

New methods enhance ASR for underrepresented languages using data from similar languages.

A new method enhances deep learning models for imbalanced datasets.

A novel approach to classifying unseen audio-visual content.

Exploring new methods to improve speech emotion recognition using natural data.

Idefics2 showcases improvements in vision-language processing through innovative design choices.

This study investigates the effectiveness of multilingual models in detecting audio deepfakes.

A new framework enhances AI's grasp of 3D spaces.

Preference Flow Matching offers a new way to align AI outputs with user preferences.

A new framework improves named entity recognition efficiency and performance.

Analyzing the effectiveness of ViTs for texture recognition compared to traditional methods.

This research focuses on improving methods for removing unwanted information from language models.

This article discusses a new model combining visual and language processing.

A new framework minimizes human effort while addressing model biases.

A new method improves guidance in diffusion models using reinforcement learning.

Analyzing effective clean-label backdoor attack techniques in machine learning.

A novel method improves learning new classes while retaining old knowledge.

This article discusses methods to improve how robots identify the uses of objects.

Advanced models predict heart conditions using PPG and ECG signals.

A new approach to enhance tagging accuracy using pre-trained models.

A framework guides the selection of effective pre-trained models for vulnerability detection.

Using Darwinian concepts to improve neural network training methods.

Study assesses pre-trained AI models for MRI, CT, and X-ray interpretation.

A new framework enhances data pruning by focusing on pre-trained models for molecular tasks.

A new method for music tagging using few-shot learning shows promising results.

This study explores using Transfer Learning for effective quality control in CFRP tape laying.

Enhancing robots' ability to adapt and learn from experiences.

A novel approach to address memory issues in machine learning.

A novel approach uses real-time MRI to visualize speech production movements.

A new method streamlines audio and video creation for better synchronization.

A look at continual learning and innovative methods to retain knowledge in AI models.

Exploring how AI can learn while keeping user data private.

Researchers enhance computers' ability to recognize functional objects in 3D environments.