Live page · Day archive

ScaleNav: Scalable Internet Video Pretraining for Language-Conditioned Local Navigation

Read the original at HF Daily Papers

Summary

ScaleNav is a framework that uses internet egocentric videos to pretrain robots for language-conditioned navigation, achieving 86.5% success on Short-Horizon OVON and 94.9% on HabitatGS-LN.

Carried by: HF Daily Papers. First seen: .