Distilling Realizable Students from Unrealizable Teachers
This addresses inefficiencies in policy distillation for robotics and AI systems with information asymmetry, though it appears incremental as it builds on existing teacher-student frameworks.
The paper tackles policy distillation under privileged information, where a student with partial observations learns from a teacher with full-state access, by introducing methods that strategically query the teacher and initialize training for efficient exploration, resulting in significant improvements in training efficiency and final performance over baselines.
We study policy distillation under privileged information, where a student policy with only partial observations must learn from a teacher with full-state access. A key challenge is information asymmetry: the student cannot directly access the teacher's state space, leading to distributional shifts and policy degradation. Existing approaches either modify the teacher to produce realizable but sub-optimal demonstrations or rely on the student to explore missing information independently, both of which are inefficient. Our key insight is that the student should strategically interact with the teacher --querying only when necessary and resetting from recovery states --to stay on a recoverable path within its own observation space. We introduce two methods: (i) an imitation learning approach that adaptively determines when the student should query the teacher for corrections, and (ii) a reinforcement learning approach that selects where to initialize training for efficient exploration. We validate our methods in both simulated and real-world robotic tasks, demonstrating significant improvements over standard teacher-student baselines in training efficiency and final performance. The project website is available at : https://portal-cornell.github.io/CritiQ_ReTRy/