Jing Xu

RESTRAIN: From Spurious Votes to Signals — Self-Training RL with Self-Penalization

Hybrid Reinforcement: when reward is sparse, better to be dense

The Era of Real-World Human Interaction: RL from User Conversations

Bridging Offline and Online Reinforcement Learning for LLMs

CoT-Self-Instruct: Building high-quality synthetic prompts data for reasoning and non-reasoning tasks

Self-Consistency Preference Optimization

Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge

R.I.P.: Better Models by Survival of the Fittest Prompts