RDD: Retrieval-Based Demonstration Decomposer for Planner Alignment in Long-Horizon Tasks - Mingxuan Yan, Yuping Wang,…

A training-free method that automatically decomposes demonstrations into sub-tasks by aligning visual features with low-level visuomotor policy training data for improved hierarchical robot learning.

关键词：robotics, hierarchical learning, vision-language-action models, demonstration decomposition, retrieval-based learning, visuomotor policy, machine learning, computer vision, AI