Ziyang Cai

R&B: Domain Regrouping and Data Mixture Balancing for Efficient Foundation Model Training

h1: Bootstrapping Models to Reason over Longer Horizons via Reinforcement Learning

Extrapolation by Association: Length Generalization Transfer In Transformers

Everything Everywhere All at Once: LLMs can In-Context Learn Multiple Tasks in Superposition

Self-Improving Transformers Overcome Easy-to-Hard and Length Generalization Challenges

Everything Everywhere All at Once: LLMs can In-Context Learn Multiple Tasks in Superposition