Closer Look

"Closer Look" research encompasses a broad range of investigations into the inner workings and limitations of various machine learning models. Current efforts focus on analyzing model behavior under different conditions, such as distribution shifts and limited data, and improving model performance through techniques like pruning, calibration, and architectural modifications (e.g., incorporating convolutional layers into transformers). These studies aim to enhance the reliability, efficiency, and fairness of machine learning systems across diverse applications, from object detection and natural language processing to human activity recognition and face recognition. The ultimate goal is to build more robust, interpretable, and trustworthy AI systems.