Sim-and-Human Co-training for Data-Efficient and Scene-Generalizable Bimanual Manipulation
Co-trains a bimanual policy on sim data (for robot-valid actions) and human videos (for real visuals), then fine-tunes on 80 real episodes per task. Reports 62.5% on held-out scenes, 53.7 points above real-only.用仿真数据（提供机器人可执行动作）和人类视频（提供真实视觉）联合预训练双臂策略，再在每任务 80 条真机数据上微调。报告在未见场景上成功率 62.5%，比纯真机数据高 53.7 个百分点。
BibTeX
@misc{fang2026sim,
title = {{Sim-and-Human Co-training for Data-Efficient and Scene-Generalizable Bimanual Manipulation}},
author = {Kaipeng Fang and Weiqing Liang and Yuyang Li and Ji Zhang and Pengpeng Zeng and Heng Tao Shen and Jingkuan Song and Lianli Gao},
year = {2026},
eprint = {2601.19406},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2601.19406},
}