DexMachina: Functional Retargeting for Bimanual Dexterous Manipulation

Zhao Mandi, Yifan Hou, Dieter Fox, Yashraj Narang, Ajay Mandlekar, Shuran Song
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:86198-86213, 2026.

Abstract

We study the problem of functional retargeting: learning dexterous manipulation policies to track object states from human hand-object demonstrations. We focus on long-horizon, bimanual tasks with articulated objects, which are challenging due to large action space, spatiotemporal discontinuities, and the embodiment gap between human and robot hands. We propose DexMachina, a novel curriculum-based algorithm: the key idea is to use virtual object controllers with decaying strength: an object is first driven automatically towards its target states, such that the policy can gradually learn to take over under motion and contact guidance. We release a simulation benchmark with a diverse set of tasks and dexterous hands, and show that DexMachina significantly outperforms baseline methods. Our algorithm and benchmark enable a functional comparison for hardware designs, and we present key findings informed by quantitative and qualitative results. With the recent surge in dexterous hand development, we hope this work will provide a useful platform for identifying desirable hardware capabilities and lower the barrier for contributing to future research. Videos and more at project-dexmachina.github.io

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-mandi26a, title = {{D}ex{M}achina: Functional Retargeting for Bimanual Dexterous Manipulation}, author = {Mandi, Zhao and Hou, Yifan and Fox, Dieter and Narang, Yashraj and Mandlekar, Ajay and Song, Shuran}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {86198--86213}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/mandi26a/mandi26a.pdf}, url = {https://proceedings.mlr.press/v306/mandi26a.html}, abstract = {We study the problem of functional retargeting: learning dexterous manipulation policies to track object states from human hand-object demonstrations. We focus on long-horizon, bimanual tasks with articulated objects, which are challenging due to large action space, spatiotemporal discontinuities, and the embodiment gap between human and robot hands. We propose DexMachina, a novel curriculum-based algorithm: the key idea is to use virtual object controllers with decaying strength: an object is first driven automatically towards its target states, such that the policy can gradually learn to take over under motion and contact guidance. We release a simulation benchmark with a diverse set of tasks and dexterous hands, and show that DexMachina significantly outperforms baseline methods. Our algorithm and benchmark enable a functional comparison for hardware designs, and we present key findings informed by quantitative and qualitative results. With the recent surge in dexterous hand development, we hope this work will provide a useful platform for identifying desirable hardware capabilities and lower the barrier for contributing to future research. Videos and more at project-dexmachina.github.io} }
Endnote
%0 Conference Paper %T DexMachina: Functional Retargeting for Bimanual Dexterous Manipulation %A Zhao Mandi %A Yifan Hou %A Dieter Fox %A Yashraj Narang %A Ajay Mandlekar %A Shuran Song %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-mandi26a %I PMLR %P 86198--86213 %U https://proceedings.mlr.press/v306/mandi26a.html %V 306 %X We study the problem of functional retargeting: learning dexterous manipulation policies to track object states from human hand-object demonstrations. We focus on long-horizon, bimanual tasks with articulated objects, which are challenging due to large action space, spatiotemporal discontinuities, and the embodiment gap between human and robot hands. We propose DexMachina, a novel curriculum-based algorithm: the key idea is to use virtual object controllers with decaying strength: an object is first driven automatically towards its target states, such that the policy can gradually learn to take over under motion and contact guidance. We release a simulation benchmark with a diverse set of tasks and dexterous hands, and show that DexMachina significantly outperforms baseline methods. Our algorithm and benchmark enable a functional comparison for hardware designs, and we present key findings informed by quantitative and qualitative results. With the recent surge in dexterous hand development, we hope this work will provide a useful platform for identifying desirable hardware capabilities and lower the barrier for contributing to future research. Videos and more at project-dexmachina.github.io
APA
Mandi, Z., Hou, Y., Fox, D., Narang, Y., Mandlekar, A. & Song, S.. (2026). DexMachina: Functional Retargeting for Bimanual Dexterous Manipulation. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:86198-86213 Available from https://proceedings.mlr.press/v306/mandi26a.html.

Related Material