Beyond Trajectory-Level Attribution: Graph-Based Credit Assignment for Agentic Reinforcement Learning 文章

ArXiv CS.AI2026-05-27NEWSen作者: Xin Cheng, Shuo He, Lang Feng, HaiYang Xu, Ming Yan, Lei Feng, Bo An
大语言模型
查看原文 →
关系图谱

Beyond Trajectory-Level Attribution: Graph-Based Credit Assignment for Agentic Reinforcement Learning · 相关技术

相关技术

LLMORMODElanguage model远程代码执行（RCE）信用分配reinforcement learningdivide-and-conquer partitioningUnifiedStraight-Through EstimatorStanParts-of-Speech (POS) tagsNarrative Abstraction BenchmarkLMMlarge language modelsGroupForFFIEffort Metric AttentionENADiTCamouflaged object detectionARGANNAgentAPI