Zhongxiang Dai

ActiveDPO: Active Direct Preference Optimization for Sample-Efficient Alignment

FedPOB: Sample-Efficient Federated Prompt Optimization via Bandits

Dual-Robust Cross-Domain Offline Reinforcement Learning Against Dynamics Shifts

Meta-Prompt Optimization for LLM-Based Sequential Decision Making

Provably Sample-Efficient Active Preference Data Collection

General Risk Measure meets Offline RL: Provably Efficient Risk-Sensitive Offline RL via Optimized Certainty Equivalent

Large Language Model-Enhanced Multi-Armed Bandits

Self-Reflective Generation at Test Time

Reinforcement Reward Model with Policy Feedback

T-POP: Test-Time Personalization with Online Preference Feedback

Combinatorial Dueling Bandits

EduDial: Constructing a Large-scale Multi-turn Teacher–Student Dialogue Corpus

Convergence Rates of Constrained Expected Improvement

Refining Adaptive Zeroth-Order Optimization at Ease

Adaptive Batch-Wise Sample Scheduling for Direct Preference Optimization

Neural Dueling Bandits: Preference-Based Optimization with Human Feedback

Online Clustering of Dueling Bandits

Source Attribution for Large Language Model-Generated Data

Prompt Optimization with Human Feedback