Tag
PlatonicNav introduces a training-free framework for embodied navigation that uses vision-only semantic maps and blind matching to ground language goals, achieving generalization across tasks and embodiments without explicit cross-modal training.