PPL
Zihan Dong
Papers (5)
RUBRIC-ARROW: Alternating Pointwise Rubric Reward Modeling for LLM Post-training in Non-verifiable DomainsEvidence Over Plans: Online Trajectory Verification for Skill DistillationDARE: Difficulty-Adaptive Reinforcement Learning with Co-Evolved Difficulty EstimationConformal Path Reasoning: Trustworthy Knowledge Graph Question Answering via Path-Level CalibrationSelf-Consolidating Language Models: Continual Knowledge Incorporation from Context