판 이력 — Planner-Centric Reinforcement Learning for Deep Research with Structure-Aware Reward | AIChainDay