A Behavior Foundation Model That Sees the Terrain

Loading video
Loading videoSeeBFM post-trains a behavior foundation model with terrain height scans, then uses two depth cameras to predict the scan without retraining the controller.
Behavior Foundation ModelTerrain AdaptationHumanoidDepth CameraRobot LearningBehavioral foundation models
Category: humanoid
Author: @an_thai_le
Date: 2026-11-09T00:00:00
Duration: 179.029s
Reference: https://arxiv.org/abs/2610.15833





