Sungyoon Lee

Inconsistency-Aware Minimization: Improving Generalization with Unlabeled Data

Stepwise Feature Learning in Self-Supervised Learning

Inconsistency-Aware Minimization: Improving Generalization with Unlabeled Data

Prior Forgetting and In-Context Overfitting

Prediction Risk and Estimation Risk of the Ridgeless Least Squares Estimator under General Assumptions on Regression Errors

How Classifier Features Transfer to Downstream: An Asymptotic Analysis in a Two-Layer Model

Stepwise Feature Learning in Self-Supervised Learning

How Classifiers Extract General Features for Downstream Tasks: An Asymptotic Analysis in Two-Layer Models

Emergence of Alignment and Local Elasticity in Two-Layer Neural Networks

Transformer Training Instability of Softmax and Lipschitz-Kernel Attentions