Plan in Sandbox, Navigate in Open Worlds: Learning Physics-Grounded Abstracted Experience for Embodied Navigation

Vision-Language Models (VLMs) have demonstrated exceptional general reasoning capabilities. However, their performance in embodied navigation remains hindered by a scarcity of aligned open-world vision and robot control data. Despite simulators providing a cost-effective alternative for data collection, the inherent reliance on photorealistic simulations often limits the transferability of learned policies. To this end, we propose $\mathit{\text{S}}\mathit{\text{andbox-}}\mathit{\text{A}}\mathit{\text{bstracted}}\mathit{\text{G}}\mathit{\text{rounded}}\mathit{\text{E}}\mathit{\text{xperience}}$ ($\mathit{\text{SAGE}}$), a framework that enables agents to learn within a physics-grounded semantic abstraction rather than a photorealistic simulation, mimicking the human capacity for mental simulation where plans are rehearsed in simplified physics abstractions before execution. $\mathit{\text{SAGE}}$ system operates via three synergistic phases: (1) $\mathit{\text{Genesis}}$: constructing diverse, physics-constrained semantic environments to bootstrap experience; (2) $\mathit{\text{Evolution}}$: distilling experiences through Reinforcement Learning (RL), utilizing a novel asymmetric adaptive clipping mechanism to stabilize updates; (3) $\mathit{\text{Navigation}}$: bridging the abstract policy to open-world control. We demonstrate that $\mathit{\text{SAGE}}$ significantly improves planner-assisted embodied navigation, achieving a 53.21% LLM-Match Success Rate on A-EQA (+9.7% over baseline), while showing encouraging transfer to physical indoor robot deployment.