Hancheng Min

LoRA Provably Reduces Forgetting and Enables Adapter Merging in Multiclass Linear Classification

Softmax-Induced Ill-Conditioning in Transformer Models

Neural Collapse under Gradient Flow on Shallow ReLU Networks for Orthogonally Separable Data

Convergence Rates for Gradient Descent on the Edge of Stability for Overparametrised Least Squares

Gradient Flow Provably Learns Robust Classifiers for Data from Orthonormal Clusters

Gradient Flow Provably Learns Robust Classifiers for Orthonormal GMMs

Gradient Descent and Attention Models: Challenges Posed by the Softmax Function