The field of biometric authentication and machine learning is moving towards more robust and reliable methods for identity verification and bias detection. Researchers are exploring new approaches to improve the accuracy and stability of authentication systems, particularly in open-set scenarios where unseen data is common. One notable direction is the use of multi-modal pretraining frameworks and self-constraint learning to enhance the representational capacity of signal encoders. Additionally, there is a growing interest in detecting and analyzing visual biases in generative models and classifiers, with a focus on developing interactive tools and unsupervised methods for bias discovery. These advances have the potential to significantly improve the performance and fairness of biometric authentication systems and machine learning models. Noteworthy papers include:

• A proposed ECG identity authentication system that achieves high accuracy and stability in open-set settings through multi-modal pretraining and self-constraint learning.
• The introduction of the Visual Bias Explorer, an interactive tool for discovering and analyzing visual biases in generative Text-to-Image models.
• The development of Classifier-to-Bias, a training-free bias discovery framework that can identify biases in pre-trained classification models without requiring labeled data.