StepOPSD: Step-Aware Online Preference Distillation for Agent Reinforcement Learning 文章

ArXiv CS.AI2026-05-27NEWSen作者: Yanfei Zhang, Xu Lin, Chenglin Wu
人工智能
查看原文 →
关系图谱

StepOPSD: Step-Aware Online Preference Distillation for Agent Reinforcement Learning · 相关技术

相关技术

RedisORM远程代码执行（RCE）reinforcement learningdivide-and-conquer partitioningTokenTeraTURNStraight-Through EstimatorSPAParts-of-Speech (POS) tagsOn-policy distillationNLIMaliGrouped Memorization EvaluationGranular Alignment ParadigmGRPOForDiTDenseCISANNAgentAPI