Training Generalist Navigation Policies via Action Self-Distillation
One VLM acts as teacher (sees the whole trajectory, infers the goal and reasoning) and student (onboard view only), so language-conditioned navigation reasoning is learned with no human labels. 13 real tasks.同一个 VLM 既当教师（看到完整轨迹，推断导航目标与推理过程）又当学生（只看机载观测），无需人工标注即可学会以语言为条件的导航推理。在 13 个真实任务上验证。
BibTeX
@misc{feng2026training,
title = {{Training Generalist Navigation Policies via Action Self-Distillation}},
author = {Bowen Feng and Keqin Wang and Yulong Yang and Amogh Joshi and Christine Allen-Blanchette and Dhruv Shah and Felix Heide},
year = {2026},
eprint = {2610.17593},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.17593},
}