Evaluating Imitation Learning Policies for Robotic Manipulation Success
This research evaluates 281 imitation-learning policies for robotic manipulation, revealing critical design choices that affect performance. Key findings indicate that representation techniques significantly enhance success rates in real-world tasks.