Wei Hu

Provable Low-Frequency Bias of In-Context Learning of Representations

RaanA: A Fast, Flexible, and Data-Efficient Post-Training Quantization Algorithm

What Happens During the Loss Plateau? Understanding Abrupt Learning in Transformers

Benign Overfitting in Single-Head Attention

Swing-by Dynamics in Concept Learning and Compositional Generalization

Linear Projections of Teacher Embeddings for Few-Class Distillation

Benign Overfitting in Single-Head Attention

Let Me Grok for You: Accelerating Grokking via Embedding Transfer from a Weaker Model

HERTA: A High-Efficiency and Rigorous Training Algorithm for Unfolded Graph Neural Networks