ADEPT: Accelerating Dexterity via Pre-Training and Post-Training using Reinforcement Learning
2026-08-19 • Robotics
RoboticsArtificial Intelligence
AI summaryⓘ
The authors present ADEPT, a method to teach robots with many fingers to do complex tasks by first training them on a simple object-moving task. This pretraining helps robots learn new skills faster without starting from scratch each time. They also developed special techniques to keep the robot’s initial skills intact while learning new ones. Using this approach, robots with cameras and touch sensors can transfer what they learned in simulations directly to the real world and perform tasks quickly, similar to human speed.
reinforcement learningsim-to-real transferdexteritypretrainingrobot manipulationmulti-fingered robotsbehavior cloningtactile perceptionlong-horizon tasksjoint-space control
Authors
Jayjun Lee, Jessica Yin, Asif Rana, Nicholas Blauch, Sam Mady, Mohak Bhardwaj, Nima Fazeli, Nathan Ratliff, Karl Van Wyk, Ankur Handa
Abstract
We introduce Accelerating Dexterity via Pre-Training (ADEPT), a large-scale reinforcement learning (RL) framework for learning sim-to-real transferable dexterity across high degree-of-freedom (DoF) robot embodiments that can solve long-horizon tasks directly from raw visuo-tactile perception. ADEPT pretrains a dexterous policy on a generic object reposing task, then post-trains downstream policies with this pretrained behavior as a prior. ADEPT enables learning new behaviors that are otherwise difficult to discover from scratch on multi-fingered robots and avoids learning the same set of skills over again for every new downstream task. The pretrained policy zero-shots the reposing phase of downstream tasks, but naïve RL fine-tuning rapidly degrades this capability during transfer. We address this with a stable post-training recipe combining behavior-cloning distillation, critic warm-up, and conservative on-policy updates. To safely exploit the full kinematic dexterity, we introduce a joint-space Geometric Fabric that mediates between the RL policy and the robot. We distill post-trained teachers into perceptive students that zero-shot sim-to-real transfer on two embodiments: a 23 DoF Kuka-Allegro with two RGB cameras, and a 29 DoF Flexiv-Sharpa with two RGB cameras and five vision-based tactile sensors, and can solve long-horizon tasks from challenging initial states with dexterity at human-level speed.