Tag
Introduces a three-stage training paradigm to internalize world model planning in LLM agents, enabling future-aware decision-making. Outperforms baselines on search and mathematical reasoning tasks.