MimicX refines humanoid tracking with policy-in-the-loop feedback: 25.7% error cut
Shuaijun Liu · hf · 2026-10-15
MimicX is a policy-in-the-loop framework that uses execution feedback to refine video-driven humanoid motion tracking. Human videos offer rich motion targets, but visually plausible references can still fail persistently under physics-based execution — and those failures reveal where supervision should change.
The framework localizes failure windows and affected body regions/channels from reconstructed and retargeted motion, jointly adapts tracking objectives and the reset curriculum for policy continuation, and uses repeated rollout verification to select execution-priority improvements under tracking guards. Across four core video tasks, MimicX cuts body-tracking error by 25.7% and boosts robust execution horizon by 255.6%; MimicX-HLoop accelerates feedback via heterogeneous execution.
More from Embodied
- Horizon Robotics' VITAL hits 90.8 PDMS with a single denoising step at 3.29x inference speed — HorizonRobotics · 2026-10-15
- SafeDriveVLA: world model 'dreaming' guards driving VLAs against unsafe navigation commands — UCI6055 · 2026-10-15
- Grain-levelling robot keeps workers out of ship holds, cutting engulfment and fall risks — LexiLove · 2026-10-15
- Pi-0.5 can post-train to grab a pen, but legs-balance coordination is the hard part — Darpinian · 2026-10-15
- Why robots walk easier than they pick pens: a VLA training debate — Darpinian · 2026-10-15
- NVIDIA VP to discuss open models and physical AI at GTC Berlin — RevLebaredian · 2026-10-15