pith. sign in

arxiv: 2606.10025 · v1 · pith:G6MLHQV3new · submitted 2026-06-08 · 💻 cs.RO · cs.CV· cs.LG

GHOST: Hierarchical Sub-Goal Policies for Generalizing Robot Manipulation

Sriram Krishna , Ben Eisner , Haotian Zhan , Ying Yuan , Haoyu Zhen , Chuang Gan , Shubham Tulsiani , David Held This is my paper

Pith reviewed 2026-06-27 16:08 UTC · model grok-4.3

classification 💻 cs.RO cs.CVcs.LG
keywords robot manipulationhierarchical policiessub-goal predictionvisuomotor controlgeneralizationhuman demonstrationsdiffusion policy3D pose goals
0
0 comments X

The pith

GHOST factorizes robot manipulation into a high-level 3D sub-goal predictor from RGB-D images and a low-level controller to improve generalization.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents GHOST as a way to learn visuomotor policies that work on tasks outside the training distribution. It splits control into a high-level policy that outputs distributions over next 3D end-effector poses from multi-view observations and a low-level policy that turns those goals into robot actions. This split raises performance and robustness over flat policies while letting the high-level part train on human video and the low-level part stay on robot data. The result is that a handful of human demonstrations can adapt the system to new objects and variations.

Core claim

Factorizing control into a high-level policy that predicts the next sub-goal as a distribution over 3D end-effector poses from multi-view RGB-D observations and a low-level goal-conditioned controller that executes embodiment-specific actions improves performance and robustness over flat Diffusion Policies. The same interface lets the high-level policy train on human video without action retargeting because sub-goals remain largely embodiment-agnostic, so the combined system adapts to novel objects and task variations from a small number of human demonstrations.

What carries the argument

Hierarchical split into high-level sub-goal predictor (distribution over 3D poses) and low-level controller, conditioned via projected end-effector heatmaps in the image plane.

If this is right

  • The hierarchical split raises success rate and robustness over a flat Diffusion Policy on the tested manipulation tasks.
  • Human video can supply the high-level policy while the low-level policy stays trained only on robot demonstrations.
  • The system adapts to new objects and task changes after seeing only a small number of human demonstrations.
  • The spatial projection of 3D goals into image heatmaps lets image-based policies condition on those goals without additional machinery.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same split could be tested on tasks where the high-level plan involves longer sequences or multiple objects.
  • If sub-goal distributions prove stable across embodiments, the approach could reduce the amount of robot-specific data needed for new hardware.
  • Adding a third level that predicts sequences of sub-goals might further cut the number of demonstrations required for new tasks.

Load-bearing premise

Sub-goals expressed as 3D end-effector poses are largely independent of the specific robot body so that a high-level policy trained on human video can pair with a low-level policy trained only on robot data.

What would settle it

Train a high-level policy on human video and pair it with a robot low-level policy on the same tasks; measure whether success rate drops below the rate achieved when both levels train on robot data alone.

Figures

Figures reproduced from arXiv: 2606.10025 by Ben Eisner, Chuang Gan, David Held, Haotian Zhan, Haoyu Zhen, Shubham Tulsiani, Sriram Krishna, Ying Yuan.

Figure 1
Figure 1. Figure 1: GHOST learns skills from robot teleoperation data and optionally uses human demonstrations to generalize to novel [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: GHOST High-level sub-goal prediction architecture: RGB-D observations from multiple cameras are processed with a DINOv3 encoder, with patch tokens augmented by 3D coordinates. Additional context (gripper state, language embedding, embodiment name) is encoded via separate MLPs. A decoder-only transformer processes all tokens, with each patch predicting the GMM parameters of a 3D sub-goal distribution over e… view at source ↗
Figure 3
Figure 3. Figure 3: Sub-goal decomposition for robot and human demon [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: End-effector representations for robot and human [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: GHOST Low-level goal-conditioned policy architec￾ture. Images from each camera and the projected end-effector heatmap images are processed independently with ResNet [11] encoders and concatenated along with the proprioceptive input into a global conditioning vector for the Diffusion Policy. Training. We use the standard diffusion loss, training on the robot-only data Drobot: Llo = EDrobot,ϵ h ∥ϵ − ϵθ(a noi… view at source ↗
Figure 7
Figure 7. Figure 7: End-effector heatmap generation for goal conditioning. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: We visualize qualitative results on various tasks with the [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
read the original abstract

We present GHOST, a framework for learning visuomotor manipulation policies that generalize beyond the training distribution. GHOST factorizes control into (i) a high-level policy that predicts the next sub-goal as a distribution over 3D end-effector poses from multi-view RGB-D observations, and (ii) a low-level goal-conditioned controller that executes embodiment-specific actions. To condition image-based policies on 3D goals, we introduce a simple spatial interface that projects predicted goals into the image plane and represents them as end-effector heatmaps. Across a suite of manipulation tasks, this hierarchical factorization consistently improves performance and robustness compared to a flat Diffusion Policy. Further, we show that this hierarchical interface also makes it easy to incorporate human demonstrations without relying on (noisy) action retargeting. As sub-goals are largely embodiment-agnostic, we train the high-level policy on human video to specify how learned skills should be applied and composed, while keeping the low-level policy trained purely on robot data. This hierarchy enables adaptation to novel objects and task variations using a small number of human demonstrations.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper introduces GHOST, a hierarchical framework for learning visuomotor manipulation policies. It factorizes the policy into a high-level component that predicts distributions over 3D end-effector poses as sub-goals from multi-view RGB-D observations and a low-level goal-conditioned controller that executes embodiment-specific actions. A spatial interface projects the 3D goals into image-plane heatmaps. The approach is shown to improve performance and robustness over flat Diffusion Policies and to facilitate the use of human demonstrations for the high-level policy while keeping the low-level policy on robot data, enabling adaptation to novel objects and task variations with few human demos.

Significance. If the empirical results hold, this work could be significant for robot learning by enabling better generalization through hierarchical factorization and easier integration of human video data without action retargeting. The spatial projection interface is a simple contribution that addresses embodiment differences between human and robot data.

major comments (2)
  1. [Abstract and §4 Experiments] The abstract asserts consistent improvement over Diffusion Policy but supplies no quantitative results, error bars, task details, or ablation data. The full paper must provide these (including specific tasks, metrics, and comparisons) to support the central claim of improved performance and robustness.
  2. [§3 Method and §4.3 Human Demonstration Experiments] The premise that sub-goals are largely embodiment-agnostic, allowing the high-level policy trained on human multi-view RGB-D video to be combined with a low-level policy trained only on robot data, lacks direct validation. A comparison of 3D end-effector pose distributions and execution success rates (with and without the human-trained high-level policy) is needed to rule out distribution shift or kinematic mismatch induced by the spatial projection interface.
minor comments (1)
  1. [§3.2 Spatial Interface] Clarify the precise mathematical form of the spatial interface, including how 3D poses are projected to heatmaps and any parameters involved in the projection.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major point below, indicating planned revisions where appropriate.

read point-by-point responses

  1. Referee: [Abstract and §4 Experiments] The abstract asserts consistent improvement over Diffusion Policy but supplies no quantitative results, error bars, task details, or ablation data. The full paper must provide these (including specific tasks, metrics, and comparisons) to support the central claim of improved performance and robustness.

    Authors: Section 4 of the manuscript already contains the requested quantitative results, including task descriptions, metrics, comparisons to Diffusion Policy baselines, error bars, and ablations. To better support the central claims in the abstract itself, we will revise the abstract to incorporate key quantitative findings from the experiments. revision: yes

  2. Referee: [§3 Method and §4.3 Human Demonstration Experiments] The premise that sub-goals are largely embodiment-agnostic, allowing the high-level policy trained on human multi-view RGB-D video to be combined with a low-level policy trained only on robot data, lacks direct validation. A comparison of 3D end-effector pose distributions and execution success rates (with and without the human-trained high-level policy) is needed to rule out distribution shift or kinematic mismatch induced by the spatial projection interface.

    Authors: Section 4.3 reports end-to-end task success rates demonstrating effective combination of the human-trained high-level policy with the robot low-level controller. We agree that explicit analysis of 3D pose distributions would provide stronger direct validation of the embodiment-agnostic assumption. We will add comparisons of predicted 3D sub-goal distributions (human vs. robot) and associated execution metrics in the revised version. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical hierarchy claims rest on performance comparisons, not derivations or self-referential fits

full rationale

The paper describes a hierarchical policy factorization (high-level 3D sub-goal prediction from RGB-D, low-level goal-conditioned controller) and reports empirical gains over flat Diffusion Policy plus human-video transfer. No equations, fitted parameters renamed as predictions, uniqueness theorems, or ansatzes appear in the provided text. Claims are validated via task-suite experiments rather than reducing to inputs by construction. The embodiment-agnostic premise is an assumption tested empirically, not a self-definitional loop.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no information on free parameters, axioms, or invented entities.

pith-pipeline@v0.9.1-grok · 5751 in / 1116 out tokens · 27822 ms · 2026-06-27T16:08:35.705573+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

37 extracted references · 11 linked inside Pith

  1. [1]

    Track2act: Predicting point tracks from internet videos enables diverse zero- shot robot manipulation.CoRR, 2024

    Homanga Bharadhwaj, Roozbeh Mottaghi, Abhinav Gupta, and Shubham Tulsiani. Track2act: Predicting point tracks from internet videos enables diverse zero- shot robot manipulation.CoRR, 2024

    2024

  2. [2]

    arXiv preprint arXiv:2410.24164, 2024

    Kevin Black, Noah Brown, Danny Driess, Adnan Es- mail, Michael Equi, Chelsea Finn, Niccolo Fusai, Lachy Groom, Karol Hausman, Brian Ichter, et al.π 0: A vision- language-action flow model for general robot control. arXiv preprint arXiv:2410.24164, 2024

    Pith/arXiv arXiv 2024

  3. [3]

    Rt-1: Robotics transformer for real-world control at scale.arXiv preprint arXiv:2212.06817, 2022

    Anthony Brohan, Noah Brown, Justice Carbajal, Yev- gen Chebotar, Joseph Dabis, Chelsea Finn, Keerthana Gopalakrishnan, Karol Hausman, Alex Herzog, Jasmine Hsu, et al. Rt-1: Robotics transformer for real-world control at scale.arXiv preprint arXiv:2212.06817, 2022

    Pith/arXiv arXiv 2022

  4. [4]

    Diffusion policy: Visuomotor policy learning via action diffusion.The International Journal of Robotics Research, 44(10-11):1684–1704, 2025

    Cheng Chi, Zhenjia Xu, Siyuan Feng, Eric Cousineau, Yilun Du, Benjamin Burchfiel, Russ Tedrake, and Shuran Song. Diffusion policy: Visuomotor policy learning via action diffusion.The International Journal of Robotics Research, 44(10-11):1684–1704, 2025

    2025

  5. [5]

    Chi, Jeff Dean, Jacob Devlin, Adam Roberts, Denny Zhou, Quoc V

    Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Yunxuan Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, Albert Webson, Shixiang Shane Gu, Zhuyun Dai, Mirac Suzgun, Xinyun Chen, Aakanksha Chowdhery, Alex Castro-Ros, Marie Pellat, Kevin Robinson, Dasha Valter, Sharan Narang, Gaurav Mishra, Adams Yu, Vincent Zhao, Yanping H...

    2024

  6. [6]

    Amplify: Actionless motion priors for robot learning from videos

    Jeremy A Collins, Lor ´and Cheng, Kunal Aneja, Albert Wilcox, Benjamin Joffe, and Animesh Garg. Amplify: Actionless motion priors for robot learning from videos. arXiv preprint arXiv:2506.14198, 2025

    arXiv 2025

  7. [7]

    Vision transformers need registers,

    Timoth ´ee Darcet, Maxime Oquab, Julien Mairal, and Piotr Bojanowski. Vision transformers need registers,

  8. [8]

    URL https://arxiv.org/abs/2309.16588

    Pith/arXiv arXiv

  9. [9]

    The” something something” video database for learning and evaluating visual common sense

    Raghav Goyal, Samira Ebrahimi Kahou, Vincent Michal- ski, Joanna Materzynska, Susanne Westphal, Heuna Kim, Valentin Haenel, Ingo Fruend, Peter Yianilos, Moritz Mueller-Freitag, et al. The” something something” video database for learning and evaluating visual common sense. InProceedings of the IEEE international con- ference on computer vision, pages 5842...

    2017

  10. [10]

    Ego4d: Around the world in 3,000 hours of egocentric video

    Kristen Grauman, Andrew Westbury, Eugene Byrne, Zachary Chavis, Antonino Furnari, Rohit Girdhar, Jack- son Hamburger, Hao Jiang, Miao Liu, Xingyu Liu, et al. Ego4d: Around the world in 3,000 hours of egocentric video. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 18995– 19012, 2022

    2022

  11. [11]

    Point policy: Unify- ing observations and actions with key points for robot manipulation.arXiv preprint arXiv:2502.20391, 2025

    Siddhant Haldar and Lerrel Pinto. Point policy: Unify- ing observations and actions with key points for robot manipulation.arXiv preprint arXiv:2502.20391, 2025

    arXiv 2025

  12. [12]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

    2016

  13. [13]

    Lvsm: A large view synthesis model with minimal 3d inductive bias.arXiv preprint arXiv:2410.17242, 2024

    Haian Jin, Hanwen Jiang, Hao Tan, Kai Zhang, Sai Bi, Tianyuan Zhang, Fujun Luan, Noah Snavely, and Zexiang Xu. Lvsm: A large view synthesis model with minimal 3d inductive bias.arXiv preprint arXiv:2410.17242, 2024

    arXiv 2024

  14. [14]

    Egomimic: Scaling imitation learning via egocentric video

    Simar Kareer, Dhruv Patel, Ryan Punamiya, Pranay Mathur, Shuo Cheng, Chen Wang, Judy Hoffman, and Danfei Xu. Egomimic: Scaling imitation learning via egocentric video. In2025 IEEE International Confer- ence on Robotics and Automation (ICRA), pages 13226– 13233. IEEE, 2025

    2025

  15. [15]

    Droid: A large-scale in-the-wild robot manipulation dataset.arXiv preprint arXiv:2403.12945, 2024

    Alexander Khazatsky, Karl Pertsch, Suraj Nair, Ash- win Balakrishna, Sudeep Dasari, Siddharth Karam- cheti, Soroush Nasiriany, Mohan Kumar Srirama, Lawrence Yunliang Chen, Kirsty Ellis, et al. Droid: A large-scale in-the-wild robot manipulation dataset.arXiv preprint arXiv:2403.12945, 2024

    Pith/arXiv arXiv 2024

  16. [16]

    Phan- tom: Training robots without robots using only human videos.arXiv preprint arXiv:2503.00779, 2025

    Marion Lepert, Jiaying Fang, and Jeannette Bohg. Phan- tom: Training robots without robots using only human videos.arXiv preprint arXiv:2503.00779, 2025

    Pith/arXiv arXiv 2025

  17. [17]

    Learning latent plans from play

    Corey Lynch, Mohi Khansari, Ted Xiao, Vikash Kumar, Jonathan Tompson, Sergey Levine, and Pierre Sermanet. Learning latent plans from play. InConference on robot learning, pages 1113–1132. Pmlr, 2020

    2020

  18. [18]

    Learning to generalize across long-horizon tasks from human demonstrations

    Ajay Mandlekar, Danfei Xu, Roberto Mart ´ın-Mart´ın, Silvio Savarese, and Li Fei-Fei. Learning to generalize across long-horizon tasks from human demonstrations. arXiv preprint arXiv:2003.06085, 2020

    arXiv 2003

  19. [19]

    Open x-embodiment: Robotic learning datasets and rt-x models: Open x-embodiment collaboration 0

    Abby O’Neill, Abdul Rehman, Abhiram Maddukuri, Ab- hishek Gupta, Abhishek Padalkar, Abraham Lee, Acorn Pooley, Agrim Gupta, Ajay Mandlekar, Ajinkya Jain, et al. Open x-embodiment: Robotic learning datasets and rt-x models: Open x-embodiment collaboration 0. In2024 IEEE International Conference on Robotics and Automation (ICRA), pages 6892–6903. IEEE, 2024

    2024

  20. [20]

    Reconstructing hands in 3d with transformers

    Georgios Pavlakos, Dandan Shan, Ilija Radosavovic, Angjoo Kanazawa, David Fouhey, and Jitendra Malik. Reconstructing hands in 3d with transformers. InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9826–9836, 2024

    2024

  21. [21]

    Long-horizon visual planning with goal- conditioned hierarchical predictors.Advances in Neural Information Processing Systems, 33:17321–17333, 2020

    Karl Pertsch, Oleh Rybkin, Frederik Ebert, Shenghao Zhou, Dinesh Jayaraman, Chelsea Finn, and Sergey Levine. Long-horizon visual planning with goal- conditioned hierarchical predictors.Advances in Neural Information Processing Systems, 33:17321–17333, 2020

    2020

  22. [22]

    Wilor: End-to-end 3d hand localization and reconstruction in-the-wild

    Rolandos Alexandros Potamias, Jinglei Zhang, Jiankang Deng, and Stefanos Zafeiriou. Wilor: End-to-end 3d hand localization and reconstruction in-the-wild. InProceed- ings of the Computer Vision and Pattern Recognition Conference, pages 12242–12254, 2025

    2025

  23. [23]

    Motion tracks: A unified representation for human-robot transfer in few- shot imitation learning.arXiv preprint arXiv:2501.06994, 2025

    Juntao Ren, Priya Sundaresan, Dorsa Sadigh, Sanjiban Choudhury, and Jeannette Bohg. Motion tracks: A unified representation for human-robot transfer in few- shot imitation learning.arXiv preprint arXiv:2501.06994, 2025

    arXiv 2025

  24. [24]

    Grounded sam: Assembling open- world models for diverse visual tasks.arXiv preprint arXiv:2401.14159, 2024

    Tianhe Ren, Shilong Liu, Ailing Zeng, Jing Lin, Kun- chang Li, He Cao, Jiayu Chen, Xinyu Huang, Yukang Chen, Feng Yan, et al. Grounded sam: Assembling open- world models for diverse visual tasks.arXiv preprint arXiv:2401.14159, 2024

    Pith/arXiv arXiv 2024

  25. [25]

    Javier Romero, Dimitrios Tzionas, and Michael J. Black. Embodied hands: Modeling and capturing hands and bodies together.ACM Transactions on Graphics, (Proc. SIGGRAPH Asia), 36(6), November 2017

    2017

  26. [26]

    Oriane Sim ´eoni, Huy V . V o, Maximilian Seitzer, Federico Baldassarre, Maxime Oquab, Cijo Jose, Vasil Khali- dov, Marc Szafraniec, Seungeun Yi, Micha ¨el Ramamon- jisoa, Francisco Massa, Daniel Haziza, Luca Wehrstedt, Jianyuan Wang, Timoth ´ee Darcet, Th ´eo Moutakanni, Leonel Sentana, Claire Roberts, Andrea Vedaldi, Jamie Tolan, John Brandt, Camille Co...

    Pith/arXiv arXiv 2025

  27. [27]

    A careful examina- tion of large behavior models for multitask dexterous ma- nipulation

    TRI LBM Team, Jose Barreiros, Andrew Beaulieu, Aditya Bhat, Rick Cory, Eric Cousineau, Hongkai Dai, Ching-Hsin Fang, Kunimatsu Hashimoto, Muham- mad Zubair Irshad, Masha Itkina, Naveen Kuppuswamy, Kuan-Hui Lee, Katherine Liu, Dale McConachie, Ian McMahon, Haruki Nishimura, Calder Phillips-Grafflin, Charles Richter, Paarth Shah, Krishnan Srinivasan, Blake ...

    Pith/arXiv arXiv 2025

  28. [28]

    Mimicplay: Long-horizon imitation learning by watching human play.arXiv preprint arXiv:2302.12422, 2023

    Chen Wang, Linxi Fan, Jiankai Sun, Ruohan Zhang, Li Fei-Fei, Danfei Xu, Yuke Zhu, and Anima Anand- kumar. Mimicplay: Long-horizon imitation learning by watching human play.arXiv preprint arXiv:2302.12422, 2023

    arXiv 2023

  29. [29]

    Articubot: Learning universal articulated object manipulation policy via large scale simulation

    Yufei Wang, Ziyu Wang, Mino Nakura, Pratik Bhowal, Chia-Liang Kuo, Yi-Ting Chen, Zackory Erickson, and David Held. Articubot: Learning universal articulated object manipulation policy via large scale simulation. arXiv preprint arXiv:2503.03045, 2025

    arXiv 2025

  30. [30]

    Any-point trajec- tory modeling for policy learning.arXiv preprint arXiv:2401.00025, 2023

    Chuan Wen, Xingyu Lin, John So, Kai Chen, Qi Dou, Yang Gao, and Pieter Abbeel. Any-point trajec- tory modeling for policy learning.arXiv preprint arXiv:2401.00025, 2023

    Pith/arXiv arXiv 2023

  31. [31]

    Neural task programming: Learning to generalize across hierarchical tasks

    Danfei Xu, Suraj Nair, Yuke Zhu, Julian Gao, Animesh Garg, Li Fei-Fei, and Silvio Savarese. Neural task programming: Learning to generalize across hierarchical tasks. In2018 IEEE international conference on robotics and automation (ICRA), pages 3795–3802. IEEE, 2018

    2018

  32. [32]

    La- tent action pretraining from videos.arXiv preprint arXiv:2410.11758, 2024

    Seonghyeon Ye, Joel Jang, Byeongguk Jeon, Sejune Joo, Jianwei Yang, Baolin Peng, Ajay Mandlekar, Reuben Tan, Yu-Wei Chao, Bill Yuchen Lin, et al. La- tent action pretraining from videos.arXiv preprint arXiv:2410.11758, 2024

    Pith/arXiv arXiv 2024

  33. [33]

    Predicting 4d hand trajectory from monocular videos.arXiv preprint arXiv:2501.08329, 2025

    Yufei Ye, Yao Feng, Omid Taheri, Haiwen Feng, Shub- ham Tulsiani, and Michael J Black. Predicting 4d hand trajectory from monocular videos.arXiv preprint arXiv:2501.08329, 2025

    arXiv 2025

  34. [34]

    Sigmoid loss for language image pre- training

    Xiaohua Zhai, Basil Mustafa, Alexander Kolesnikov, and Lucas Beyer. Sigmoid loss for language image pre- training. InProceedings of the IEEE/CVF international conference on computer vision, pages 11975–11986, 2023

    2023

  35. [35]

    Universal visual decomposer: Long-horizon manipulation made easy

    Zichen Zhang, Yunshuang Li, Osbert Bastani, Abhishek Gupta, Dinesh Jayaraman, Yecheng Jason Ma, and Luca Weihs. Universal visual decomposer: Long-horizon manipulation made easy. In2024 IEEE International Conference on Robotics and Automation (ICRA), pages 6973–6980. IEEE, 2024

    2024

  36. [36]

    Learning fine-grained bimanual manipulation with low-cost hardware.arXiv preprint arXiv:2304.13705, 2023

    Tony Z Zhao, Vikash Kumar, Sergey Levine, and Chelsea Finn. Learning fine-grained bimanual manipulation with low-cost hardware.arXiv preprint arXiv:2304.13705, 2023

    Pith/arXiv arXiv 2023

  37. [37]

    On the continuity of rotation representations in neural networks

    Yi Zhou, Connelly Barnes, Jingwan Lu, Jimei Yang, and Hao Li. On the continuity of rotation representations in neural networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5745–5753, 2019. APPENDIXA IMPLEMENTATIONDETAILS A. High-Level Policy The high-level policy uses a decoder-only transformer op- erating on...

    2019