| license: apache-2.0 | |
| library_name: openwam | |
| pipeline_tag: robotics | |
| tags: | |
| - robotics | |
| - world-action-model | |
| - openwam | |
| # robotwin_dual_system_idm | |
| An **OpenWAM-Study** checkpoint from the Q1 architecture comparison, run on RoboTwin 2.0 under a fixed recipe so that only the composition rule changes. The baseline is `robotwin_dual_system_joint_self_attention`. | |
| **Dual-System / IDM.** Formulates the action module as an inverse-dynamics model conditioned on video features, trained with teacher-forced video states and run in two inference stages: the video trajectory is generated first, then actions are predicted from it. | |
| - Paper: https://arxiv.org/abs/2609.07398 | |
| - Code: https://github.com/OpenWAM-Official/OpenWAM | |
| - Project page: https://openwam-official.github.io/ | |
| ## Citation | |
| ```bibtex | |
| @article{wang2026openwam, | |
| title = {OpenWAM: An Open, Modular Exploration Towards Systematic World-Action Model Pretraining}, | |
| author = {Yuran Wang and Siqiao Huang and Mingleyang Li and Chenhao Zhang and Jiaqi Liang and Weiyang Jin and Yue Chen and Xuemin Chi and Donghao Zhou and Qize Yu and Yu-Kai Wang and Yuhan Rui and Shenzhe Yao and Zhen Yuan and Zhenhao Shen and Kefei Zhu and Zijie Zhu and Ning Gao and Xiaowei Chi and Guanqi He and Shanghang Zhang and Hao Dong and Lin Shao and Hang Zhao}, | |
| year = {2026}, | |
| journal = {arXiv preprint arXiv: 2609.07398} | |
| } | |
| ``` | |