Xuchao Zhang

Provable and Practical In-Context Policy Optimization for Self-Improvement

Multi-Preference Optimization: Generalizing DPO via Set-Level Contrasts

ACTS: Adaptive Control for Test-time Scaling

Generative Caching for Structurally Similar Prompts and Responses

CREAM: Consistency Regularized Self-Rewarding Language Models

Anyprefer: An Agentic Framework for Preference Data Synthesis

REFA: Reference Free Alignment with Fine-Grained Length Control

Synergistic Weak-Strong Collaboration by Aligning Preferences

AMPO: Active Multi Preference Optimization for Self-play Preference Selection

Unveiling Context-Aware Criteria in Self-Assessing LLMs