Preprint
Jul 2026
CLIFT: Turning Gemini Robotics On-Device into Humanoid Specialists via Non-Invasive Closed-Loop Iterative Fine-Tuning
CLIFT: Closed-Loop Iterative Fine-Tuning is introduced, which turns deployment-time reward feedback into API-compatible supervised data and enables closed-loop policy improvement without accessing weights, gradients, likelihoods, or losses-pushing GROD to near-perfect success after two flywheel cycles, all without opening the model box.
Yuxin Chen, H. Srikanth, N. Jew et al.
· 0 citations