pith. sign in

arxiv: 2606.30749 · v1 · pith:KRX456TVnew · submitted 2026-06-29 · 💻 cs.RO

From Grasps to Dexterity: Large-Scale Grasp Pretraining for Dexterous Manipulation

Ying Yuan , Xinyu Liu , Sriram Krishna , David Held This is my paper

Pith reviewed 2026-07-01 01:56 UTC · model grok-4.3

classification 💻 cs.RO
keywords dexterous manipulationgrasp pretraininghierarchical imitation learningarticulated tool usedexterous graspingcontact-rich controlsimulation benchmark
0
0 comments X

The pith

Pretraining a low-level controller on 355k grasp trajectories transfers to articulated tool-use tasks and raises real-world success by 33.3 points over diffusion baselines.

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

The paper tests whether large-scale dexterous grasp datasets, normally used only for pick-and-place, can instead supply priors for full functional dexterity with articulated tools. It adapts a hierarchical imitation-learning setup in which a high-level policy predicts hand sub-goals while a low-level goal-conditioned controller, first pretrained on the grasp data, handles contact-rich finger coordination. The controller is then fine-tuned on task demonstrations for six new tool-use scenarios collected in the DexCraft benchmark. Experiments show the pretraining step yields higher success than both end-to-end diffusion policies and hierarchical policies trained from scratch, with the largest gains appearing in real-world trials. The result indicates that grasp corpora can scale pretraining for sustained-contact manipulation beyond their original narrow use.

Core claim

A low-level goal-conditioned controller pretrained on a 355k-trajectory dexterous-grasp dataset, then fine-tuned within a hierarchical imitation-learning framework, produces higher success on articulated tool-use tasks than end-to-end diffusion policies or scratch-trained hierarchical baselines; in real-world tests the method raises full-task success by 33.3 percentage points over DP3.

What carries the argument

Hierarchical imitation learning that pairs high-level hand sub-goal prediction with a low-level goal-conditioned controller first pretrained on large-scale grasp data.

If this is right

  • Grasp datasets become a scalable source of pretraining data for contact-rich dexterous manipulation rather than only for grasp synthesis.
  • The same low-level controller can be reused across multiple downstream tool tasks after brief fine-tuning.
  • Performance gains appear in both simulation and real-world settings, with the largest measured lift in real-robot full-task completion.
  • Hierarchical policies that separate sub-goal planning from low-level control benefit more from grasp pretraining than flat end-to-end policies.

Where Pith is reading between the lines

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

  • If the transfer holds, future work could collect even larger grasp corpora specifically to bootstrap controllers for longer-horizon manipulation sequences.
  • The approach suggests that grasp data may supply useful priors for any task whose low-level motions resemble the finger configurations seen during grasping.
  • One could test whether the same pretraining step accelerates learning when the high-level policy is also learned rather than provided by demonstrations.

Load-bearing premise

A controller trained only on static grasp examples will still produce stable, coordinated finger motion when the robot must keep contact and drive moving parts of a tool.

What would settle it

On the six DexCraft tasks, the grasp-pretrained hierarchical policy shows no improvement over an identical hierarchical policy trained from scratch or over an end-to-end diffusion policy.

Figures

Figures reproduced from arXiv: 2606.30749 by David Held, Sriram Krishna, Xinyu Liu, Ying Yuan.

Figure 1
Figure 1. Figure 1: Left: Our simulation benchmark, DexCraft, with articulated tool use tasks. We visualize object goal poses with green object meshes. Right: With a real-world robot, our policy can perform highly dexterous tasks using proprioception and RGB-D perception as feedback. More videos are available on our project website. Abstract: Large-scale dexterous grasp datasets encode rich priors over hand￾object interaction… view at source ↗
Figure 2
Figure 2. Figure 2: Visualization of DexCraft tasks. For each task, the ini￾tial frame is shown on the left and the target frame is on the right. We visualize target object positions with green object meshes (un￾observed by the policy). Reference objects that the tools will inter￾act with are placed relative to the goal. The robot hand is required to grasp the object, lift it to the target pose, and trigger the object’s artic… view at source ↗
Figure 3
Figure 3. Figure 3: Our method integrates large-scale grasp pretraining with a hierarchical policy framework. (a) A high-level sub-goal prediction policy takes the current point cloud observation as input and predicts the po￾sitions of hand key points. (b) A low-level policy is conditioned on predicted sub-goal key points and current observation and predicts action chunks for the controller. Top: We augment the Dexonomy [2] d… view at source ↗
Figure 4
Figure 4. Figure 4: Real World Setup Environment Setup and Data Collection. The simulation setup is detailed in Section 4. For real world tasks, shown in [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Sample efficiency of our method on the stapler task compared with baselines. Q1: Does hierarchical policy representation benefit perfor￾mance? We study the effect of the hierarchical policy represen￾tation. The results are shown in [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of high-level policy’s sub-goal predictions during real-world deployment. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Point cloud statistics of G2D-Pretrain compared with a downstream task [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Workspace coverage comparison between G2D-Pretrain and a downstream task, visual￾ized as point-cloud projections onto the x-y and x-z planes. and target hand poses. This produces randomized grasping scenes while preserving the relative hand-object grasp geometry. For each grasp instance, we construct a key-frame trajectory consisting of a randomized initial hand pose, an open-hand pose aligned with the tar… view at source ↗
Figure 9
Figure 9. Figure 9: Performance gains from encoder-only transfer with [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 11
Figure 11. Figure 11: Visualization of the 5- keypoint and 16-keypoint condition￾ing schemes. C.3 Transfer Protocols for Pretraining and Fine-Tuning Encoder-only transfer for GraspXL. For GraspXL, we use encoder-only transfer rather than full-checkpoint transfer. Although our conversion and canonical policy interface make GraspXL compatible with the downstream low-level policy, full-checkpoint fine-tuning performs poorly in ou… view at source ↗
Figure 12
Figure 12. Figure 12: Example evaluation episodes of DP3 on the spray bottle task. [PITH_FULL_IMAGE:figures/full_fig_p021_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Example evaluation episodes of hierarchical policy learning from scratch on the spray [PITH_FULL_IMAGE:figures/full_fig_p022_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Example evaluation episodes of our method on the spray bottle task. [PITH_FULL_IMAGE:figures/full_fig_p022_14.png] view at source ↗
read the original abstract

Large-scale dexterous grasp datasets encode rich priors over hand-object interaction, but their use has largely been confined to grasp generation and pick-and-place manipulation. We study whether such data can instead support functional dexterity in articulated tool use, where a robot must acquire a tool, maintain contact, and operate its functional moving parts. We adapt a hierarchical imitation learning framework that combines high-level hand sub-goal prediction with a low-level goal-conditioned controller. We construct a 355k-trajectory grasp-pretraining dataset from large-scale dexterous grasp annotations and use it to pretrain the low-level controller. The controller is then fine-tuned on downstream task demonstrations. To evaluate this setting, we introduce DexCraft, a simulation benchmark with six articulated tool-use tasks requiring coordinated finger motion. Across simulation and real-world experiments, our approach outperforms end-to-end diffusion policy baselines and hierarchical policies trained from scratch. In the real world, it improves full-task success by 33.3 percentage points over DP3. These results show that grasp datasets can serve not only as resources for grasp synthesis, but also as scalable pretraining data for contact-rich dexterous manipulation. Videos are shown on https://yingyuan0414.github.io/grasp2dexterity/ .

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

0 major / 2 minor

Summary. The paper proposes pretraining a low-level goal-conditioned controller on a 355k-trajectory dataset derived from large-scale dexterous grasp annotations, then fine-tuning it within a hierarchical imitation learning framework (high-level sub-goal prediction + low-level controller) on downstream demonstrations. It introduces the DexCraft simulation benchmark consisting of six articulated tool-use tasks and reports that the approach outperforms end-to-end diffusion policy baselines (DP3) and hierarchical policies trained from scratch, with a 33.3 percentage point gain in real-world full-task success over DP3.

Significance. If the results hold, the work is significant because it demonstrates that existing large-scale grasp datasets can provide useful priors for contact-rich, sustained-contact dexterous manipulation beyond pick-and-place, rather than being limited to grasp synthesis. The real-world experiments, direct baseline comparisons, and introduction of DexCraft are concrete strengths; the scale of the pretraining data and explicit description of the fine-tuning protocol support the central claim of transferable low-level control.

minor comments (2)
  1. [Abstract, §4] Abstract and §4: performance deltas (including the 33.3 pp real-world gain) are reported without explicit mention of number of trials, error bars, or statistical tests; adding these to the result tables would strengthen the comparison claims.
  2. [§3.2] §3.2: the construction of the 355k-trajectory grasp-pretraining dataset from annotations is described at a high level; a short additional paragraph on filtering criteria or annotation sources would aid reproducibility.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive review, the recognition of the work's significance in extending grasp datasets to contact-rich dexterous manipulation, and the recommendation for minor revision. The report correctly identifies the core contributions, including the 355k-trajectory pretraining dataset, the hierarchical framework, the DexCraft benchmark, and the 33.3 pp real-world gain over DP3. No major comments were listed in the report.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper pretrains a low-level goal-conditioned controller on a distinct 355k-trajectory grasp dataset constructed from large-scale annotations, then fine-tunes it on separate downstream task demonstrations for the DexCraft benchmark. Reported gains (e.g., 33.3 pp real-world improvement over DP3) are obtained via direct experimental comparisons to end-to-end diffusion policies and from-scratch hierarchical baselines in both simulation and real settings. No equations, fitted parameters, or self-citations are shown to reduce the central claims or performance metrics to quantities defined by the same inputs; the pretraining and evaluation data sources remain independent.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are stated or derivable from the given text.

pith-pipeline@v0.9.1-grok · 5759 in / 967 out tokens · 38634 ms · 2026-07-01T01:56:30.640105+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

56 extracted references · 23 canonical work pages · 3 internal anchors

  1. [1]

    R. Wang, J. Zhang, J. Chen, Y . Xu, P. Li, T. Liu, and H. Wang. Dexgraspnet: A large- scale robotic dexterous grasp dataset for general objects based on simulation.arXiv preprint arXiv:2210.02697, 2022

    work page arXiv 2022

  2. [2]

    J. Chen, Y . Ke, L. Peng, and H. Wang. Dexonomy: Synthesizing all dexterous grasp types in a grasp taxonomy.Robotics: Science and Systems, 2025

    2025

  3. [3]

    Zhang, S

    H. Zhang, S. Christen, Z. Fan, O. Hilliges, and J. Song. GraspXL: Generating grasping motions for diverse objects at scale. InEuropean Conference on Computer Vision (ECCV), 2024

    2024

  4. [4]

    J. Ye, K. Wang, C. Yuan, R. Yang, Y . Li, J. Zhu, Y . Qin, X. Zou, and X. Wang. Dex1b: Learning with 1b demonstrations for dexterous manipulation. InRobotics: Science and Systems (RSS), 2025

    2025

  5. [5]

    Zhang, H

    J. Zhang, H. Liu, D. Li, X. Yu, H. Geng, Y . Ding, J. Chen, and H. Wang. Dexgraspnet 2.0: Learning generative dexterous grasping in large-scale synthetic cluttered scenes. In8th Annual Conference on Robot Learning, 2024

    2024

  6. [6]

    Z. Weng, H. Lu, D. Kragic, and J. Lundell. Dexdiffuser: Generating dexterous grasps with diffusion models.IEEE Robotics and Automation Letters, 9(12):11834–11840, 2024. doi: 10.1109/LRA.2024.3498776

    work page doi:10.1109/lra.2024.3498776 2024

  7. [7]

    Y . Xu, W. Wan, J. Zhang, H. Liu, Z. Shan, H. Shen, R. Wang, H. Geng, Y . Weng, J. Chen, et al. Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned policy.arXiv preprint arXiv:2303.00938, 2023

    work page arXiv 2023

  8. [8]

    W. Wan, H. Geng, Y . Liu, Z. Shan, Y . Yang, L. Yi, and H. Wang. Unidexgrasp++: Improving dexterous grasping policy learning via geometry-aware curriculum and iterative generalist- specialist learning.arXiv preprint arXiv:2304.00464, 2023

    work page arXiv 2023

  9. [9]

    Rajeswaran, V

    A. Rajeswaran, V . Kumar, A. Gupta, G. Vezzani, J. Schulman, E. Todorov, and S. Levine. Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demon- strations. InProceedings of Robotics: Science and Systems (RSS), 2018

    2018

  10. [10]

    C. Bao, H. Xu, Y . Qin, and X. Wang. Dexart: Benchmarking generalizable dexterous manip- ulation with articulated objects. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 21190–21200, 2023

    2023

  11. [11]

    Y . Wang, Z. Wang, M. Nakura, P. Bhowal, C.-L. Kuo, Y .-T. Chen, Z. Erickson, and D. Held. Articubot: Learning universal articulated object manipulation policy via large scale simulation. arXiv preprint arXiv:2503.03045, 2025

    work page arXiv 2025

  12. [12]

    Krishna, B

    S. Krishna, B. Eisner, H. Zhan, Y . Yuan, H. Zhen, C. Gan, S. Tulsiani, and D. Held. Ghost: Hierarchical sub-goal policies for generalizing robot manipulation. InRobotics: Science and Systems (RSS), 2026

    2026

  13. [13]

    M. T. Ciocarlie, C. Goldfeder, and P. K. Allen. Dexterous grasping via eigengrasps : A low-dimensional approach to a high-complexity problem. 2007. URLhttps://api. semanticscholar.org/CorpusID:6853822

    2007

  14. [14]

    Miller and P

    A. Miller and P. Allen. Graspit! a versatile simulator for robotic grasping.IEEE Robotics & Automation Magazine, 11(4):110–122, 2004. doi:10.1109/MRA.2004.1371616

    work page doi:10.1109/mra.2004.1371616 2004

  15. [15]

    Berenson and S

    D. Berenson and S. S. Srinivasa. Grasp synthesis in cluttered environments for dexterous hands. InHumanoids 2008 - 8th IEEE-RAS International Conference on Humanoid Robots, pages 189–196, 2008. doi:10.1109/ICHR.2008.4755944. 9

    work page doi:10.1109/ichr.2008.4755944 2008

  16. [16]

    In: IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2021, virtual, June 19-25, 2021

    P. Grady, C. Tang, C. D. Twigg, M. V o, S. Brahmbhatt, and C. C. Kemp. Contactopt: Optimiz- ing contact to improve grasps. In2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 1471–1481, 2021. doi:10.1109/CVPR46437.2021.00152

    work page doi:10.1109/cvpr46437.2021.00152 2021

  17. [17]

    Mandikal and K

    P. Mandikal and K. Grauman. Learning dexterous grasping with object-centric visual affor- dances.2021 IEEE International Conference on Robotics and Automation (ICRA), pages 6169–6176, 2020. URLhttps://api.semanticscholar.org/CorpusID:233439776

    2021

  18. [18]

    In: 2019 IEEE/RSJ International Conference on Intelli- gent Robots and Systems (IROS)

    S. Brahmbhatt, A. Handa, J. Hays, and D. Fox. Contactgrasp: Functional multi-finger grasp synthesis from contact. In2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 2386–2393, 2019. doi:10.1109/IROS40897.2019.8967960

    work page doi:10.1109/iros40897.2019.8967960 2019

  19. [19]

    P. Li, T. Liu, Y . Li, Y . Geng, Y . Zhu, Y . Yang, and S. Huang. Gendexgrasp: Generalizable dex- terous grasping. In2023 IEEE International Conference on Robotics and Automation (ICRA), pages 8068–8074, 2023. doi:10.1109/ICRA48891.2023.10160667

    work page doi:10.1109/icra48891.2023.10160667 2023

  20. [20]

    Turpin, L

    D. Turpin, L. Wang, E. Heiden, Y .-C. Chen, M. Macklin, S. Tsogkas, S. Dickinson, and A. Garg. Grasp’d: Differentiable contact-rich grasp synthesis for multi-fingered hands. InComputer Vision – ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022, Proceedings, Part VI, page 201–221, Berlin, Heidelberg, 2022. Springer-Verlag. ISBN 978...

    work page doi:10.1007/978-3-031-20068-7 2022

  21. [21]

    Seita, Y

    D. Seita, Y . Wang, S. Shetty, E. Li, Z. Erickson, and D. Held. Toolflownet: Robotic manipu- lation with tools via predicting tool flow from point clouds. InConference on Robot Learning (CoRL), 2022

    2022

  22. [22]

    C. Qi, Y . Wu, L. Yu, H. Liu, B. Jiang, X. Lin, and D. Held. Learning generalizable tool- use skills through trajectory generation. InIEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024

    2024

  23. [23]

    T. Lin, Y . Zhang, Q. Li, H. Qi, B. Yi, S. Levine, and J. Malik. Learning visuotactile skills with two multifingered hands.arXiv:2404.16823, 2024

    work page arXiv 2024

  24. [24]

    Manuelli, W

    L. Manuelli, W. Gao, P. Florence, and R. Tedrake. Kpam: Keypoint affordances for category- level robotic manipulation. In T. Asfour, E. Yoshida, J. Park, H. Christensen, and O. Khatib, editors,Robotics Research, pages 132–157, Cham, 2022. Springer International Publishing

    2022

  25. [25]

    Agarwal, S

    A. Agarwal, S. Uppal, K. Shaw, and D. Pathak. Dexterous functional grasping. In7th An- nual Conference on Robot Learning, 2023. URLhttps://openreview.net/forum?id= 93qz1k6_6h

    2023

  26. [26]

    Hadjivelichkov, S

    D. Hadjivelichkov, S. Zwane, M. Deisenroth, L. Agapito, and D. Kanoulas. One-Shot Transfer of Affordance Regions? AffCorrs! In K. Liu, D. Kulic, and J. Ichnowski, editors,Proceedings of The 6th Conference on Robot Learning (CoRL), volume 205 ofProceedings of Machine Learning Research, pages 550–560, 14–18 Dec 2023

    2023

  27. [27]

    S. Bahl, R. Mendonca, L. Chen, U. Jain, and D. Pathak. Affordances from human videos as a versatile representation for robotics. 2023

    2023

  28. [28]

    Y . Ye, X. Li, A. Gupta, S. De Mellon, S. Birchfield, J. Song, S. Tulsiani, and S. Liu. Affordance diffusion: Synthesizing hand-object interactions. In2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 22479–22489, 2023. doi:10.1109/CVPR52729. 2023.02153

    work page doi:10.1109/cvpr52729 2023

  29. [29]

    Y . Qin, B. Huang, Z.-H. Yin, H. Su, and X. Wang. Dexpoint: Generalizable point cloud reinforcement learning for sim-to-real dexterous manipulation.Conference on Robot Learning (CoRL), 2022. 10

    2022

  30. [30]

    T. Chen, M. Tippur, S. Wu, V . Kumar, E. Adelson, and P. Agrawal. Visual dexterity: In- hand reorientation of novel and complex object shapes.Science Robotics, 8(84):eadc9244,

  31. [31]

    URLhttps://www.science.org/doi/abs/10

    doi:10.1126/scirobotics.adc9244. URLhttps://www.science.org/doi/abs/10. 1126/scirobotics.adc9244

    work page doi:10.1126/scirobotics.adc9244

  32. [32]

    H. Qi, B. Yi, S. Suresh, M. Lambeta, Y . Ma, R. Calandra, and J. Malik. General In-Hand Object Rotation with Vision and Touch. InConference on Robot Learning (CoRL), 2023

    2023

  33. [33]

    J. Wang, Y . Yuan, H. Che, H. Qi, Y . Ma, J. Malik, and X. Wang. Lessons from learning to spin “pens”. InCoRL, 2024

    2024

  34. [34]

    Z.-H. Yin, C. Wang, L. Pineda, F. Hogan, C. Bodduluri, A. Sharma, P. Lancaster, I. Prasad, M. Kalakrishnan, J. Malik, M. Lambeta, T. Wu, P. Abbeel, and M. Mukadam. Dexteritygen: Foundation controller for unprecedented dexterity. 06 2025. doi:10.15607/RSS.2025.XXI.103

    work page doi:10.15607/rss.2025.xxi.103 2025

  35. [35]

    R. S. Sutton, D. Precup, and S. Singh. Between mdps and semi-mdps: A framework for tem- poral abstraction in reinforcement learning.Artificial intelligence, 112(1-2):181–211, 1999

    1999

  36. [36]

    T. G. Dietterich. Hierarchical reinforcement learning with the maxq value function decompo- sition.Journal of artificial intelligence research, 13:227–303, 2000

    2000

  37. [37]

    A. S. Vezhnevets, S. Osindero, T. Schaul, N. Heess, M. Jaderberg, D. Silver, and K. Kavukcuoglu. FeUdal networks for hierarchical reinforcement learning. In D. Precup and Y . W. Teh, editors,Proceedings of the 34th International Conference on Machine Learning, volume 70 ofProceedings of Machine Learning Research, pages 3540–3549. PMLR, 06–11 Aug 2017. URL...

    2017

  38. [38]

    Nachum, S

    O. Nachum, S. S. Gu, H. Lee, and S. Levine. Data-efficient hierarchical reinforcement learning. Advances in neural information processing systems, 31, 2018

    2018

  39. [39]

    Mandlekar, S

    A. Mandlekar, S. Nasiriany, B. Wen, I. Akinola, Y . Narang, L. Fan, Y . Zhu, and D. Fox. Mimicgen: A data generation system for scalable robot learning using human demonstrations. In7th Annual Conference on Robot Learning, 2023

    2023

  40. [40]

    J. A. Collins, L. Cheng, K. Aneja, A. Wilcox, B. Joffe, and A. Garg. Amplify: Actionless motion priors for robot learning from videos.arXiv preprint arXiv:2506.14198, 2025

    work page arXiv 2025

  41. [41]

    C. Chi, S. Feng, Y . Du, Z. Xu, E. Cousineau, B. Burchfiel, and S. Song. Diffusion policy: Visuomotor policy learning via action diffusion. InProceedings of Robotics: Science and Systems (RSS), 2023

    2023

  42. [42]

    Y . Ze, G. Zhang, K. Zhang, C. Hu, M. Wang, and H. Xu. 3d diffusion policy: Generalizable visuomotor policy learning via simple 3d representations. InProceedings of Robotics: Science and Systems (RSS), 2024

    2024

  43. [43]

    J. He, D. Li, X. Yu, Z. Qi, W. Zhang, J. Chen, Z. Zhang, Z. Zhang, L. Yi, and H. Wang. Dexvlg: Dexterous vision-language-grasp model at scale. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 14248–14258, 2025

    2025

  44. [44]

    S. Tao, F. Xiang, A. Shukla, Y . Qin, X. Hinrichsen, X. Yuan, C. Bao, X. Lin, Y . Liu, T. kai Chan, Y . Gao, X. Li, T. Mu, N. Xiao, A. Gurha, V . N. Rajesh, Y . W. Choi, Y .-R. Chen, Z. Huang, R. Calandra, R. Chen, S. Luo, and H. Su. Maniskill3: Gpu parallelized robotics simulation and rendering for generalizable embodied ai.Robotics: Science and Systems, 2025

    2025

  45. [45]

    K. Shaw, A. Agarwal, and D. Pathak. Leap hand: Low-cost, efficient, and anthropomorphic hand for robot learning.Robotics: Science and Systems (RSS), 2023

    2023

  46. [46]

    Xiang, Y

    F. Xiang, Y . Qin, K. Mo, Y . Xia, H. Zhu, F. Liu, M. Liu, H. Jiang, Y . Yuan, H. Wang, L. Yi, A. X. Chang, L. J. Guibas, and H. Su. SAPIEN: A simulated part-based interactive environ- ment. InThe IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2020. 11

    2020

  47. [47]

    K. Mo, S. Zhu, A. X. Chang, L. Yi, S. Tripathi, L. J. Guibas, and H. Su. PartNet: A large-scale benchmark for fine-grained and hierarchical part-level 3D object understanding. InThe IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2019

    2019

  48. [48]

    A. X. Chang, T. Funkhouser, L. Guibas, P. Hanrahan, Q. Huang, Z. Li, S. Savarese, M. Savva, S. Song, H. Su, et al. Shapenet: An information-rich 3d model repository.arXiv preprint arXiv:1512.03012, 2015

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  49. [49]

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space.arXiv preprint arXiv:1706.02413, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  50. [50]

    Sharma, K

    D. Sharma, K. Tokas, A. Puri, and K. Sharda. Shadow hand.Journal of Advance Research in Applied Science (ISSN 2208-2352), 1(1):04–07, Jan. 2014. doi:10.53555/nnas.v1i1.692. URL https://nnpub.org/index.php/AS/article/view/692

    work page doi:10.53555/nnas.v1i1.692 2014

  51. [51]

    T. Feix, J. Romero, H.-B. Schmiedmayer, A. M. Dollar, and D. Kragic. The grasp taxonomy of human grasp types.IEEE Transactions on Human-Machine Systems, 46(1):66–77, 2016. doi:10.1109/THMS.2015.2470657

    work page doi:10.1109/thms.2015.2470657 2016

  52. [52]

    Z. Wei, Z. Xu, J. Guo, Y . Hou, C. Gao, Z. Cai, J. Luo, and L. Shao.D(R,O)grasp: A unified representation of robot and object interaction for cross-embodiment dexterous grasping. In 2025 IEEE International Conference on Robotics and Automation (ICRA), pages 4982–4988,

    2025

  53. [53]

    doi:10.1109/ICRA55743.2025.11127754

    work page doi:10.1109/icra55743.2025.11127754 2025

  54. [54]

    Z. Wei, Y . Yao, and M. Ding. One hand to rule them all: Canonical representations for unified dexterous manipulation, 2026. URLhttps://arxiv.org/abs/2602.16712

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  55. [55]

    P. Wu, Y . Shentu, Z. Yi, X. Lin, and P. Abbeel. Gello: A general, low-cost, and intuitive teleoperation framework for robot manipulators, 2023

    2023

  56. [56]

    Gervet, Z

    T. Gervet, Z. Xian, N. Gkanatsios, and K. Fragkiadaki. Act3d: 3d feature field transformers for multi-task robotic manipulation.arXiv preprint arXiv:2306.17817, 2023. 12 Appendix Table of Contents A Benchmark Details 13 A.1 Details on Decomposing Tasks into Sub-tasks . . . . . . . . . . . . . . . . . . . 13 A.2 Task Details . . . . . . . . . . . . . . . ....

    work page arXiv 2023