Xingyu Zhou

A Unified Theoretical Analysis of Private and Robust Offline Alignment: from RLHF to DPO

Square$\chi$PO: Differentially Private and Robust $\chi^2$-Preference Optimization in Offline Direct Alignment

On the Sample Complexity of Differentially Private Policy Optimization

On Differentially Private Federated Linear Contextual Bandits

Follow-the-Perturbed-Leader for Adversarial Bandits: Heavy Tails, Robustness, and Privacy

Taming Heavy-Tailed Losses in Adversarial Bandits and the Best-of-Both-Worlds Setting

Locally Private and Robust Multi-Armed Bandits