PPL
Jindong Li
Papers (5)
CoSPlay: Cooperative Self-Play at Test-Time with Self-Generated Code and Unit TestFrom Generic Correlation to Input-Specific Credit in On-Policy Self DistillationAnti-Self-Distillation for Reasoning RL via Pointwise Mutual InformationWhere Does Long-Context Supervision Actually Go? Effective-Context Exposure BalancingLearning Less Is More: Premature Upper-Layer Attention Specialization Hurts Language Model Pretraining