Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.15483 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910165383512064 |
|---|---|
| author | Intelligence, Physical Ai, Bo Amin, Ali Aniceto, Raichelle Balakrishna, Ashwin Balke, Greg Black, Kevin Bokinsky, George Cao, Shihao Charbonnier, Thomas Choudhary, Vedant Collins, Foster Conley, Ken Connors, Grace Darpinian, James Dhabalia, Karan Dhaka, Maitrayee DiCarlo, Jared Driess, Danny Equi, Michael Esmail, Adnan Fang, Yunhao Finn, Chelsea Glossop, Catherine Godden, Thomas Goryachev, Ivan Groom, Lachlan Habeeb, Haroun Hancock, Hunter Hausman, Karol Hussein, Gashon Hwang, Victor Ichter, Brian Jacobsen, Connor Jakubczak, Szymon Jen, Rowan Jones, Tim Kammerer, Gregg Katz, Ben Ke, Liyiming Khadikov, Mairbek Kuchi, Chandra Lamb, Marinda LeBlanc, Devin LeCount, Brendon Levine, Sergey Li, Xinyu Li-Bell, Adrian Lialin, Vladislav Liang, Zhonglin Lim, Wallace Lu, Yao Luo, Enyu Mano, Vishnu Marwaha, Nandan Mongush, Aikys Murphy, Liam Nair, Suraj Patterson, Tyler Pertsch, Karl Ren, Allen Z. Schelske, Gavin Sharma, Charvi Shi, Baifeng Shi, Lucy Xiaoyang Smith, Laura Springenberg, Jost Tobias Stachowicz, Kyle Stoeckle, Will Tang, Jiaming Tanner, Jimmy Tekeste, Shalom Torne, Marcel Vedder, Kyle Vuong, Quan Walling, Anna Wang, Haohuan Wang, Jason Wang, XuDong Whalen, Chris Whitmore, Samuel Williams, Blake Xu, Charles Yoo, Sukwon Yu, Lili Zhang, Wuming Zhang, Zhuoyang Zhilinsky, Ury |
| author_facet | Intelligence, Physical Ai, Bo Amin, Ali Aniceto, Raichelle Balakrishna, Ashwin Balke, Greg Black, Kevin Bokinsky, George Cao, Shihao Charbonnier, Thomas Choudhary, Vedant Collins, Foster Conley, Ken Connors, Grace Darpinian, James Dhabalia, Karan Dhaka, Maitrayee DiCarlo, Jared Driess, Danny Equi, Michael Esmail, Adnan Fang, Yunhao Finn, Chelsea Glossop, Catherine Godden, Thomas Goryachev, Ivan Groom, Lachlan Habeeb, Haroun Hancock, Hunter Hausman, Karol Hussein, Gashon Hwang, Victor Ichter, Brian Jacobsen, Connor Jakubczak, Szymon Jen, Rowan Jones, Tim Kammerer, Gregg Katz, Ben Ke, Liyiming Khadikov, Mairbek Kuchi, Chandra Lamb, Marinda LeBlanc, Devin LeCount, Brendon Levine, Sergey Li, Xinyu Li-Bell, Adrian Lialin, Vladislav Liang, Zhonglin Lim, Wallace Lu, Yao Luo, Enyu Mano, Vishnu Marwaha, Nandan Mongush, Aikys Murphy, Liam Nair, Suraj Patterson, Tyler Pertsch, Karl Ren, Allen Z. Schelske, Gavin Sharma, Charvi Shi, Baifeng Shi, Lucy Xiaoyang Smith, Laura Springenberg, Jost Tobias Stachowicz, Kyle Stoeckle, Will Tang, Jiaming Tanner, Jimmy Tekeste, Shalom Torne, Marcel Vedder, Kyle Vuong, Quan Walling, Anna Wang, Haohuan Wang, Jason Wang, XuDong Whalen, Chris Whitmore, Samuel Williams, Blake Xu, Charles Yoo, Sukwon Yu, Lili Zhang, Wuming Zhang, Zhuoyang Zhilinsky, Ury |
| contents | We present a new robotic foundation model, called $π_{0.7}$, that can enable strong out-of-the-box performance in a wide range of scenarios. $π_{0.7}$ can follow diverse language instructions in unseen environments, including multi-stage tasks with various kitchen appliances, provide zero-shot cross-embodiment generalization, for example enabling a robot to fold laundry without seeing the task before, and perform challenging tasks such as operating an espresso machine out of the box at a level of performance that matches much more specialized RL-finetuned models. The main idea behind $π_{0.7}$ is to use diverse context conditioning during training. This conditioning information, contained in the prompt, makes it possible to steer the model precisely to perform many tasks with different strategies. It is conditioned not just on a language command that describes what it should do, but on additional multimodal information that also describes the manner or strategy in which it should do it, including metadata about task performance and subgoal images. This enables $π_{0.7}$ to use very diverse data, including demonstrations, potentially suboptimal (autonomous) data including failures, and data from non-robot sources. Our experiments evaluate $π_{0.7}$ across numerous tasks with multiple robot platforms, on tasks that require speed and dexterity, language following, and compositional task generalization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_15483 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | $π_{0.7}$: a Steerable Generalist Robotic Foundation Model with Emergent Capabilities Intelligence, Physical Ai, Bo Amin, Ali Aniceto, Raichelle Balakrishna, Ashwin Balke, Greg Black, Kevin Bokinsky, George Cao, Shihao Charbonnier, Thomas Choudhary, Vedant Collins, Foster Conley, Ken Connors, Grace Darpinian, James Dhabalia, Karan Dhaka, Maitrayee DiCarlo, Jared Driess, Danny Equi, Michael Esmail, Adnan Fang, Yunhao Finn, Chelsea Glossop, Catherine Godden, Thomas Goryachev, Ivan Groom, Lachlan Habeeb, Haroun Hancock, Hunter Hausman, Karol Hussein, Gashon Hwang, Victor Ichter, Brian Jacobsen, Connor Jakubczak, Szymon Jen, Rowan Jones, Tim Kammerer, Gregg Katz, Ben Ke, Liyiming Khadikov, Mairbek Kuchi, Chandra Lamb, Marinda LeBlanc, Devin LeCount, Brendon Levine, Sergey Li, Xinyu Li-Bell, Adrian Lialin, Vladislav Liang, Zhonglin Lim, Wallace Lu, Yao Luo, Enyu Mano, Vishnu Marwaha, Nandan Mongush, Aikys Murphy, Liam Nair, Suraj Patterson, Tyler Pertsch, Karl Ren, Allen Z. Schelske, Gavin Sharma, Charvi Shi, Baifeng Shi, Lucy Xiaoyang Smith, Laura Springenberg, Jost Tobias Stachowicz, Kyle Stoeckle, Will Tang, Jiaming Tanner, Jimmy Tekeste, Shalom Torne, Marcel Vedder, Kyle Vuong, Quan Walling, Anna Wang, Haohuan Wang, Jason Wang, XuDong Whalen, Chris Whitmore, Samuel Williams, Blake Xu, Charles Yoo, Sukwon Yu, Lili Zhang, Wuming Zhang, Zhuoyang Zhilinsky, Ury Machine Learning Robotics We present a new robotic foundation model, called $π_{0.7}$, that can enable strong out-of-the-box performance in a wide range of scenarios. $π_{0.7}$ can follow diverse language instructions in unseen environments, including multi-stage tasks with various kitchen appliances, provide zero-shot cross-embodiment generalization, for example enabling a robot to fold laundry without seeing the task before, and perform challenging tasks such as operating an espresso machine out of the box at a level of performance that matches much more specialized RL-finetuned models. The main idea behind $π_{0.7}$ is to use diverse context conditioning during training. This conditioning information, contained in the prompt, makes it possible to steer the model precisely to perform many tasks with different strategies. It is conditioned not just on a language command that describes what it should do, but on additional multimodal information that also describes the manner or strategy in which it should do it, including metadata about task performance and subgoal images. This enables $π_{0.7}$ to use very diverse data, including demonstrations, potentially suboptimal (autonomous) data including failures, and data from non-robot sources. Our experiments evaluate $π_{0.7}$ across numerous tasks with multiple robot platforms, on tasks that require speed and dexterity, language following, and compositional task generalization. |
| title | $π_{0.7}$: a Steerable Generalist Robotic Foundation Model with Emergent Capabilities |
| topic | Machine Learning Robotics |
| url | https://arxiv.org/abs/2604.15483 |