Jiarui Jiang

Trained Mamba Emulates Online Gradient Descent in In-Context Linear Regression

Unveil Benign Overfitting for Transformer in Vision: Training Dynamics, Convergence, and Generalization