EvoMemNav: Efficient Self-Evolving Fine-Grained Memory for Zero-Shot Embodied Navigation 事件

EvoMemNav: Efficient Self-Evolving Fine-Grained Memory for Zero-Shot Embodied Navigation arXiv:2606.03509v1 Announce Type: new Abstract: Building memory is essential for long-horizon planning in zero-shot embodied navigation. Detector-centric scene graphs often compress observations into sparse nodes, discarding fine-grained visual evidence and accumulating noise, while 3D reconstruction-based methods remain computationally prohibitive. We present EvoMemNav, an efficient, self-evolving, fine-gr