Simple-to-Complex Structured Demonstrations for Vision-Language-Action Learning

2026-07-05

AI Summary

This work addresses a critical yet overlooked aspect in vision-language-action (VLA) learning: the organization of demonstration data, which significantly impacts policy learning efficiency, stability, and generalization. The study pioneers the treatment of demonstration organization as a key design factor in VLA systems and introduces three general principles—task decomposition, environmental standardization, and curriculum-based data ordering by complexity—to construct a structured demonstration collection strategy. Implemented on a dual-arm robotic platform, this approach enables the model to progressively acquire fundamental skills from simple to complex scenarios and subsequently compose them to accomplish intricate tasks. Evaluated on block manipulation and towel folding benchmarks, the proposed method substantially outperforms end-to-end trajectory collection baselines in both task success rate and training stability.

Abstract

Vision-Language-Action (VLA) models have demonstrated strong capabilities in robotic manipulation by integrating visual perception, language understanding, and robot action generation. Existing research has primarily focused on improving model architectures, training strategies, and dataset scale, while little attention has been paid to how demonstrations are collected and organized. We identify demonstration organization as a fundamental yet overlooked aspect of imitation learning, as it directly affects policy learning efficiency, training stability, and policy generalization. To address this gap, we propose a simple-to-complex structured demonstration collection strategy for VLA learning using a dual-arm robotic platform. Our approach systematically organizes data through three general principles: (i) decomposing complex manipulation tasks into progressively learnable sub-skills, (ii) standardizing the interaction environment to reduce unnecessary variability, and (iii) organizing demonstrations according to progressively increasing task complexity. This structured design enables VLA models to first acquire fundamental manipulation skills before learning increasingly complex task compositions, facilitating more effective learning of long-horizon manipulation tasks. We evaluate the proposed strategy on two representative robotic manipulation tasks: block grasping and sorting, and towel folding. Experimental results show consistent improvements in task success rate and training stability compared with the baseline method of directly collecting end-to-end complete task trajectories. These findings highlight demonstration organization as a previously underexplored but important factor in VLA learning and provide practical insights into efficient skill acquisition, scalable dataset construction, and long-horizon robotic manipulation.

Problem

• Vision-Language-Action
• demonstration organization
• imitation learning
• robotic manipulation
• long-horizon tasks

Innovation

• structured demonstrations
• vision-language-action learning
• imitation learning
• task decomposition
• robotic manipulation