1
Reward Shaping and Action Masking for Compositional Tasks using Behavior Trees and LLMs
融合行为树与大语言模型，实现组合任务的智能奖励塑形与动作掩码，提升多步推理效率。
arXiv:2605.05795v2 Announce Type: replace Abstract: Decomposing complex tasks into a sequence of simpler subtasks can improve learning efficiency for …