Interleave-VLA
Published in ICRA (International Conference on Robotics and Automation) 2025 Safe-VLM Workshop Spotlight, 2025
This paper introduces Interleave-VLA, a novel robot learning paradigm that leverages interleaved image-text instructions to enhance robot manipulation capabilities in unseen scenarios.
Recommended citation: @misc{fan2025interleavevlaenhancingrobotmanipulation, title={Interleave-VLA: Enhancing Robot Manipulation with Interleaved Image-Text Instructions}, author={Cunxin Fan and Xiaosong Jia and Yihang Sun and Yixiao Wang and Jianglan Wei and Ziyang Gong and Xiangyu Zhao and Masayoshi Tomizuka and Xue Yang and Junchi Yan and Mingyu Ding}, year={2025}, eprint={2505.02152}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2505.02152}, }
Download Paper