Jing Yang

Private Data Synthesis for Preference Alignment of Large Language Models

Efficient Multi-objective Prompt Optimization via Pure-exploration Bandits

Lean Clients, Full Accuracy: Hybrid Zeroth- and First-Order Split Federated Learning

Unlabeled Data Can Provably Enhance In-Context Learning of Transformers

On the Learn-to-Optimize Capabilities of Transformers in In-Context Sparse Recovery

Greedy Sampling Is Provably Efficient For RLHF

Data-adaptive Differentially Private Prompt Synthesis for In-Context Learning

On the Training Convergence of Transformers for In-Context Classification

How Transformers Learn Regular Language Recognition: A Theoretical Study on Training Dynamics and Implicit Bias

On the Training Convergence of Transformers for In-Context Classification of Gaussian Mixtures