Pith. sign in

REVIEW 2 major objections 65 references

DeformGen augments deformable manipulation data by simulating localized disturbances to expand valid states and warping trajectories to match new geometries.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

2026-06-29 04:50 UTC pith:JMHXXUZI

load-bearing objection DeformGen combines localized physics simulation for new states with deformation-field warping for trajectories to augment deformable manipulation data, which targets real gaps but rests on unshown experimental details. the 2 major comments →

arxiv 2606.25939 v2 pith:JMHXXUZI submitted 2026-06-24 cs.RO

DeformGen: Dynamics-Based Topology Augmentation for Deformable Manipulation Policy Learning

Zili Lin , Wenyao Zhang , Yuyang Zhang , Zekun Qi , Junyan Lin , Hanxin Zhu , Jiaolong Yang , Zhibo Chen
show 4 more authors
Yao Mu Xiaokang Yang Xin Jin Wenjun Zeng
This is my paper
classification cs.RO
keywords deformable manipulationdata augmentationpolicy learningdynamics simulationtrajectory transferrobotics
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper presents DeformGen as a way to address two core limits in demonstration augmentation for soft objects: high-dimensional states cannot be reached by simple pose tweaks due to physics, and trajectories do not transfer rigidly when material points deform. It solves the state problem by applying localized physical disturbances then forward-simulating the dynamics to produce topology-coherent new configurations. It solves the trajectory problem by lifting particle displacements into a continuous deformation field that adapts end-effector paths consistently with the changed shape. The result is joint augmentation of both the state distribution and the associated behaviors, which experiments show improves learned policies over raw demonstrations and rigid augmentation baselines on high-fidelity benchmarks.

Core claim

DeformGen achieves topological diversity for deformable objects by expanding the valid state distribution via localized physical disturbances and forward simulation, and transfers trajectories via deformation-field warping, jointly augmenting state and behavior to generally improve policy learning over original demonstrations and rigid baselines.

What carries the argument

dynamics-based state expansion through localized disturbances plus forward simulation, paired with deformation-field warping to adapt trajectories

Load-bearing premise

Forward simulation after localized disturbances produces physically plausible states that actually help policies, and deformation-field warping yields valid non-suboptimal trajectories.

What would settle it

An experiment in which policies trained on DeformGen-augmented data perform no better than or worse than policies trained on the original demonstrations alone across the reported benchmarks.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Policies trained with the augmented states and trajectories outperform those trained on the original demonstrations alone.
  • DeformGen outperforms rigid-style augmentation methods on the same deformable manipulation benchmarks.
  • The generated states maintain physical plausibility and topology coherence while increasing diversity beyond what pose perturbations allow.
  • Trajectory transfer preserves consistency with the deformed object geometry.

Where Pith is reading between the lines

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

  • The same disturbance-plus-simulation pattern might apply to other robotics domains where rigid transformations fail to capture valid configurations.
  • If the warping step reliably avoids suboptimal behaviors, it could reduce the volume of real demonstrations needed for non-rigid tasks.
  • The method implicitly treats simulation as a generator of training distribution rather than only as a verifier.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 0 minor

Summary. The paper proposes DeformGen, a dynamics-based augmentation framework for demonstration data in deformable manipulation policy learning. It targets two challenges: (1) the high-dimensional, physics-constrained state space that cannot be adequately explored via low-dimensional pose perturbations, addressed by applying localized physical disturbances followed by forward simulation to generate topology-coherent states; and (2) non-equivariant trajectory transfer under deformation, addressed by lifting per-particle displacements into a continuous deformation field to warp end-effector trajectories consistently with the new geometry. The central claim is that jointly augmenting states and behaviors in this manner yields policies that outperform those trained on original demonstrations or rigid-style baselines, as shown in experiments on high-fidelity benchmarks.

Significance. If the empirical results hold with proper quantification, the work would offer a concrete mechanism for expanding valid state distributions and adapting behaviors in a physically grounded way, addressing a recognized bottleneck in data-efficient learning for deformable objects. The approach of combining localized disturbances with forward simulation and deformation-field warping is a direct response to the stated challenges and could generalize to other high-dimensional manipulation domains where rigid augmentations fail.

major comments (2)
  1. [Abstract] Abstract: The claim that 'Experiments on high-fidelity deformable manipulation benchmarks show that DeformGen generally improves policy learning compared with training on the original demonstrations alone and with rigid-style augmentation baselines' is presented without any metrics, specific benchmark names, baseline implementations, success rates, variance, or statistical tests. Because the central contribution is an empirical improvement in policy learning, the absence of these details in the abstract (and the lack of any quantitative results visible in the provided text) prevents assessment of whether the method delivers on its claims.
  2. [Method description (paragraphs on state and trajectory challenges)] The description of state augmentation (localized disturbances + forward simulation) and trajectory transfer (deformation-field warping) is presented at a high level without equations, pseudocode, or analysis of failure modes such as simulation instability or warping-induced trajectory invalidity. These components are load-bearing for the joint augmentation claim, yet no verification of physical plausibility or consistency is provided.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the abstract and method description. We address each comment below and will make the indicated revisions to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that 'Experiments on high-fidelity deformable manipulation benchmarks show that DeformGen generally improves policy learning compared with training on the original demonstrations alone and with rigid-style augmentation baselines' is presented without any metrics, specific benchmark names, baseline implementations, success rates, variance, or statistical tests. Because the central contribution is an empirical improvement in policy learning, the absence of these details in the abstract (and the lack of any quantitative results visible in the provided text) prevents assessment of whether the method delivers on its claims.

    Authors: We agree that the abstract should include quantitative support for the central claim. The full manuscript reports results on specific high-fidelity benchmarks (rope, cloth, and bag manipulation tasks) with success rates, comparisons against original demonstrations and rigid augmentation baselines, standard deviations across multiple seeds, and statistical tests. We will revise the abstract to summarize these metrics (e.g., average success rate improvements of X% with p-values) while remaining concise. revision: yes

  2. Referee: [Method description (paragraphs on state and trajectory challenges)] The description of state augmentation (localized disturbances + forward simulation) and trajectory transfer (deformation-field warping) is presented at a high level without equations, pseudocode, or analysis of failure modes such as simulation instability or warping-induced trajectory invalidity. These components are load-bearing for the joint augmentation claim, yet no verification of physical plausibility or consistency is provided.

    Authors: We acknowledge the method section would benefit from greater formality. We will add the explicit equations for lifting particle displacements to a continuous deformation field (via radial basis function interpolation), pseudocode for the full augmentation pipeline, and a dedicated paragraph analyzing failure modes including simulation instability (mitigated by stable integrators) and trajectory invalidity (checked via collision and reachability constraints). Physical plausibility is verified through the forward dynamics step; we will include additional qualitative visualizations and quantitative validity metrics in the revision. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper describes a simulation-driven augmentation pipeline that applies localized disturbances, runs forward dynamics to generate new states, and uses deformation-field warping to adapt trajectories. These operations are implemented via external physics engines and geometric transforms rather than any fitted parameters, self-referential definitions, or self-citation chains that reduce the claimed result to its inputs. The central claim of improved policy learning is presented as an empirical outcome on benchmarks, leaving the derivation self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework depends on domain assumptions about simulation accuracy and warping validity rather than new fitted parameters or invented entities.

axioms (2)
  • domain assumption Forward simulation after localized physical disturbances produces topology-coherent and physically plausible deformable states suitable for policy learning
    Directly invoked to solve the state space challenge in the abstract.
  • domain assumption Deformation-field warping lifts per-particle displacements into a continuous function that preserves manipulation semantics for trajectory transfer
    Central to solving the non-equivariant trajectory challenge.

pith-pipeline@v0.9.1-grok · 5762 in / 1222 out tokens · 33007 ms · 2026-06-29T04:50:59.942911+00:00 · methodology

0 comments
read the original abstract

Demonstration augmentation is proposed for cost-efficient data acquisition, but existing methods are fundamentally limited in deformable manipulation due to two challenges: (1) the state space is high-dimensional with physics-induced constraints, making valid configurations impossible to reach via low-dimensional pose perturbations; and (2) trajectory transfer is non-equivariant, as material points no longer move rigidly together under deformation. We present DeformGen, a dynamics-based augmentation framework that achieves topological diversity for deformable objects. For the state challenge, DeformGen expands the valid state distribution by applying localized physical disturbances and forward-simulating the dynamics to obtain topology-coherent, physically plausible deformable states. For the trajectory challenge, DeformGen transfers source manipulation trajectories via deformation-field warping, which lifts per-particle displacements into a continuous spatial function to adapt the end-effector trajectory consistently with the deformed geometry. In this way, our method jointly augments the state distribution and its associated manipulation behavior. Experiments on high-fidelity deformable manipulation benchmarks show that DeformGen generally improves policy learning compared with training on the original demonstrations alone and with rigid-style augmentation baselines.

Review history (2 revisions) →

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

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

  1. [1]

    pi0: A vision-language-action flow model for general robot control

    Kevin Black, Noah Brown, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, Lachy Groom, Karol Hausman, Brian Ichter, et al. pi0: A vision-language-action flow model for general robot control. arXiv preprint, 2024

    2024

  2. [2]

    pi0.5: a vision-language-action model with open-world generalization.arXiv preprint, 2025

    Physical Intelligence, Kevin Black, Noah Brown, James Darpinian, Karan Dhabalia, Danny Driess, Adnan Es- mail, Michael Equi, Chelsea Finn, Niccolo Fusai, et al. pi0.5: a vision-language-action model with open-world generalization.arXiv preprint, 2025

    2025

  3. [3]

    Openvla: An open-source vision-language-action model.arXiv preprint, 2024

    Moo Jin Kim, Karl Pertsch, Siddharth Karamcheti, Ted Xiao, Ashwin Balakrishna, Suraj Nair, Rafael Rafailov, Ethan Foster, Grace Lam, Pannag Sanketi, et al. Openvla: An open-source vision-language-action model.arXiv preprint, 2024

    2024

  4. [4]

    Gr00t n1: An open foundation model for generalist humanoid robots

    Johan Bjorck, Fernando Castañeda, Nikita Cherniadev, Xingye Da, Runyu Ding, Linxi Fan, Yu Fang, Dieter Fox, Fengyuan Hu, Spencer Huang, et al. Gr00t n1: An open foundation model for generalist humanoid robots. arXiv preprint, 2025

    2025

  5. [5]

    InternVLA-M1: A Spatially Guided Vision-Language-Action Framework for Generalist Robot Policy

    Xinyi Chen, Yilun Chen, Yanwei Fu, Ning Gao, Jiaya Jia, Weiyang Jin, Hao Li, Yao Mu, Jiangmiao Pang, Yu Qiao, et al. Internvla-m1: A spatially guided vision-language-action framework for generalist robot policy. arXiv preprint arXiv:2510.13778, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  6. [6]

    Dreamvla: A vision-language-action model dreamed with comprehensive world knowledge.arXiv preprint, 2025

    Wenyao Zhang, Hongsi Liu, Zekun Qi, Yunnan Wang, Xinqiang Yu, Jiazhao Zhang, Runpei Dong, Jiawei He, He Wang, Zhizheng Zhang, et al. Dreamvla: A vision-language-action model dreamed with comprehensive world knowledge.arXiv preprint, 2025

    2025

  7. [7]

    Disentangled Robot Learning via Separate Forward and Inverse Dynamics Pretraining

    Wenyao Zhang, Bozhou Zhang, Zekun Qi, Wenjun Zeng, Xin Jin, and Li Zhang. Disentangled robot learning via separate forward and inverse dynamics pretraining.arXiv preprint arXiv:2604.16391, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  8. [8]

    Vla-jepa: Enhancing vision-language-action model with latent world model,

    Jingwen Sun, Wenyao Zhang, Zekun Qi, Shaojie Ren, Zezhi Liu, Hanxin Zhu, Guangzhong Sun, Xin Jin, and Zhibo Chen. Vla-jepa: Enhancing vision-language-action model with latent world model.arXiv preprint arXiv:2602.10098, 2026

    work page arXiv 2026

  9. [9]

    Discrete diffusion vla: Bringing discrete diffusion to action decoding in vision- language-action policies.arXiv preprint, 2025

    Zhixuan Liang, Yizhuo Li, Tianshuo Yang, Chengyue Wu, Sitong Mao, Liuao Pei, Xiaokang Yang, Jiangmiao Pang, Yao Mu, and Ping Luo. Discrete diffusion vla: Bringing discrete diffusion to action decoding in vision- language-action policies.arXiv preprint, 2025

    2025

  10. [10]

    Robotwin: Dual-arm robot benchmark with generative digital twins (early version)

    Yao Mu, Tianxing Chen, Shijia Peng, Zanxin Chen, Zeyu Gao, Yude Zou, Lunkai Lin, Zhiqiang Xie, and Ping Luo. Robotwin: Dual-arm robot benchmark with generative digital twins (early version). InECCV, 2025

    2025

  11. [11]

    Mimicgen: A data generation system for scalable robot learning using human demonstrations

    Ajay Mandlekar, Soroush Nasiriany, Bowen Wen, Iretiayo Akinola, Yashraj Narang, Linxi Fan, Yuke Zhu, and Dieter Fox. Mimicgen: A data generation system for scalable robot learning using human demonstrations. In Conference on Robot Learning, pages 1820–1864. PMLR, 2023

    2023

  12. [12]

    Demogen: Syn- thetic demonstration generation for data-efficient visuomotor policy learning.arXiv preprint arXiv:2502.16932, 2025

    Zhengrong Xue, Shuying Deng, Zhenyang Chen, Yixuan Wang, Zhecheng Yuan, and Huazhe Xu. Demogen: Syn- thetic demonstration generation for data-efficient visuomotor policy learning.arXiv preprint arXiv:2502.16932, 2025

    work page arXiv 2025

  13. [13]

    Noveldemonstra- tion generation with gaussian splatting enables robust one-shot manipulation.arXiv preprint arXiv:2504.13175, 2025

    SizheYang, WenyeYu, JiaZeng, JunLv, KeruiRen, CewuLu, DahuaLin, andJiangmiaoPang. Noveldemonstra- tion generation with gaussian splatting enables robust one-shot manipulation.arXiv preprint arXiv:2504.13175, 2025

    work page arXiv 2025

  14. [14]

    Egodemogen: Novel egocentric demonstration generation enables viewpoint-robust manipulation.arXiv preprint arXiv:2509.22578, 2025

    Yuan Xu, Jiabing Yang, Xiaofeng Wang, Yixiang Chen, Zheng Zhu, Bowen Fang, Guan Huang, Xinze Chen, Yun Ye, Qiang Zhang, et al. Egodemogen: Novel egocentric demonstration generation enables viewpoint-robust manipulation.arXiv preprint arXiv:2509.22578, 2025

    work page arXiv 2025

  15. [15]

    Softmimic- gen: A data generation system for scalable robot learning in deformable object manipulation.arXiv preprint arXiv:2603.25725, 2026

    Masoud Moghani, Mahdi Azizian, Animesh Garg, Yuke Zhu, Sean Huver, and Ajay Mandlekar. Softmimic- gen: A data generation system for scalable robot learning in deformable object manipulation.arXiv preprint arXiv:2603.25725, 2026

    work page arXiv 2026

  16. [16]

    SIM1: Physics-Aligned Simulator as Zero-Shot Data Scaler in Deformable Worlds

    Yunsong Zhou, Hangxu Liu, Xuekun Jiang, Xing Shen, Yuanzhen Zhou, Hui Wang, Baole Fang, Yang Tian, Mulin Yu, Qiaojun Yu, et al. Sim1: Physics-aligned simulator as zero-shot data scaler in deformable worlds. arXiv preprint arXiv:2604.08544, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  17. [17]

    Robotic manipulation and sensing of deformable objects in domestic and industrial applications: a survey.The International Journal of Robotics Research, 37(7):688–716, 2018

    Jose Sanchez, Juan-Antonio Corrales, Belhassen-Chedli Bouzgarrou, and Youcef Mezouar. Robotic manipulation and sensing of deformable objects in domestic and industrial applications: a survey.The International Journal of Robotics Research, 37(7):688–716, 2018

    2018

  18. [18]

    Modeling, learning, perception, and control methods for deformable object manipulation.Science Robotics, 6(54):eabd8803, 2021

    Hang Yin, Anastasia Varava, and Danica Kragic. Modeling, learning, perception, and control methods for deformable object manipulation.Science Robotics, 6(54):eabd8803, 2021

    2021

  19. [19]

    Real-to-sim robot policy evaluation with gaussian splatting simulation of soft-body interactions.arXiv preprint arXiv:2511.04665, 2025

    Kaifeng Zhang, Shuo Sha, Hanxiao Jiang, Matthew Loper, Hyunjong Song, Guangyan Cai, Zhuo Xu, Xiaochen Hu, Changxi Zheng, and Yunzhu Li. Real-to-sim robot policy evaluation with gaussian splatting simulation of soft-body interactions.arXiv preprint arXiv:2511.04665, 2025

    work page arXiv 2025

  20. [20]

    High-fidelity simulated data generation for real-world zero-shot robotic manipulation learning with gaussian splatting.IEEE Robotics and Automation Letters, 11(5):5310–5317, 2026

    Haoyu Zhao, Cheng Zeng, Linghao Zhuang, Yaxi Zhao, Shengke Xue, Hao Wang, Xingyue Zhao, Zhongyu Li, Kehan Li, Siteng Huang, Mingxiu Chen, Xin Li, Deli Zhao, and Hua Zou. High-fidelity simulated data generation for real-world zero-shot robotic manipulation learning with gaussian splatting.IEEE Robotics and Automation Letters, 11(5):5310–5317, 2026. doi: 10...

    work page doi:10.1109/lra.2026.3671535 2026

  21. [21]

    Stephen James, Zicong Ma, David Rovick Arrojo, and Andrew J. Davison. RLBench: The Robot Learning Benchmark & Learning Environment.arXiv preprint arXiv:1909.12271, 2019

    work page arXiv 1909

  22. [22]

    RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation

    Yufei Wang, Zhou Xian, Feng Chen, Tsun-Hsuan Wang, Yian Wang, Katerina Fragkiadaki, Zackory Erickson, David Held, and Chuang Gan. RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulation. InInternational Conference on Machine Learning, 2024

    2024

  23. [23]

    Rl-driven data generation for robust vision-based dexterous grasping.arXiv preprint arXiv:2504.18084, 2025

    Atsushi Kanehira, Naoki Wake, Kazuhiro Sasabuchi, Jun Takamatsu, and Katsushi Ikeuchi. Rl-driven data generation for robust vision-based dexterous grasping.arXiv preprint arXiv:2504.18084, 2025

    work page arXiv 2025

  24. [24]

    Semantically controllable augmentations for generalizable robot learning.The International Journal of Robotics Research, 44(10-11):1705–1726, 2025

    Zoey Chen, Zhao Mandi, Homanga Bharadhwaj, Mohit Sharma, Shuran Song, Abhishek Gupta, and Vikash Kumar. Semantically controllable augmentations for generalizable robot learning.The International Journal of Robotics Research, 44(10-11):1705–1726, 2025

    2025

  25. [25]

    Gigabrain-0: A world model-powered vision-language-action model

    GigaAI. Gigabrain-0: A world model-powered vision-language-action model. 2025. URLhttps://arxiv.org/ abs/2510.19430

    work page arXiv 2025

  26. [26]

    Dexmimicgen: Automated data generation for bimanual dexterous manipulation via imitation learning

    Zhenyu Jiang, Yuqi Xie, Kevin Lin, Zhenjia Xu, Weikang Wan, Ajay Mandlekar, Linxi Jim Fan, and Yuke Zhu. Dexmimicgen: Automated data generation for bimanual dexterous manipulation via imitation learning. In2025 IEEE International Conference on Robotics and Automation (ICRA), pages 16923–16930. IEEE, 2025

    2025

  27. [27]

    Dreamgen: Unlocking generalization in robot learning through neural trajectories.arXiv preprint, 2025

    Joel Jang, Seonghyeon Ye, Zongyu Lin, Jiannan Xiang, Johan Bjorck, Yu Fang, Fengyuan Hu, Spencer Huang, Kaushil Kundalia, Yen-Chen Lin, et al. Dreamgen: Unlocking generalization in robot learning through neural trajectories.arXiv preprint, 2025

    2025

  28. [28]

    Manipdreamer3d: Synthesizing plausible robotic manipulation video with occupancy-aware 3d trajectory

    Ying Li, Xiaobao Wei, Xiaowei Chi, Yuming Li, Zhongyu Zhao, Hao Wang, Ningning Ma, Ming Lu, and Sirui Han. Manipdreamer3d: Synthesizing plausible robotic manipulation video with occupancy-aware 3d trajectory. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 6644–6652, 2026

    2026

  29. [29]

    Oxe-auge: Alarge-scalerobotaugmentationofoxeforscalingcross-embodiment policy learning.arXiv preprint arXiv:2512.13100, 2025

    Guanhua Ji, Harsha Polavaram, Lawrence Yunliang Chen, Sandeep Bajamahal, Zehan Ma, Simeon Adebola, ChenfengXu, andKenGoldberg. Oxe-auge: Alarge-scalerobotaugmentationofoxeforscalingcross-embodiment policy learning.arXiv preprint arXiv:2512.13100, 2025

    work page arXiv 2025

  30. [30]

    Robovip: Multi-view video generation with visual identity prompting augments robot manipulation.arXiv preprint arXiv:2601.05241, 2026

    Boyang Wang, Haoran Zhang, Shujie Zhang, Jinkun Hao, Mingda Jia, Qi Lv, Yucheng Mao, Zhaoyang Lyu, Jia Zeng, Xudong Xu, et al. Robovip: Multi-view video generation with visual identity prompting augments robot manipulation.arXiv preprint arXiv:2601.05241, 2026

    work page arXiv 2026

  31. [31]

    One demo is worth a thousand trajectories: Action-view augmentation for visuomotor policies

    Chuer Pan, Litian Liang, Dominik Bauer, Eric Cousineau, Benjamin Burchfiel, Siyuan Feng, and Shuran Song. One demo is worth a thousand trajectories: Action-view augmentation for visuomotor policies. In9th Annual Conference on Robot Learning, 2025

    2025

  32. [32]

    Real2render2real: Scaling robot data without dynamics simulation or robot hardware,

    Justin Yu, Letian Fu, Huang Huang, Karim El-Refai, Rares Andrei Ambrus, Richard Cheng, Muhammad Zubair Irshad, and Ken Goldberg. Real2render2real: Scaling robot data without dynamics simulation or robot hardware,

  33. [33]

    URLhttps://arxiv.org/abs/2505.09601

    work page arXiv

  34. [34]

    Real2edit2real: Generating robotic demonstrations via a 3d control interface.arXiv preprint arXiv:2512.19402, 2025

    Yujie Zhao, Hongwei Fan, Di Chen, Shengcong Chen, Liliang Chen, Xiaoqi Li, Guanghui Ren, and Hao Dong. Real2edit2real: Generating robotic demonstrations via a 3d control interface.arXiv preprint arXiv:2512.19402, 2025

    work page arXiv 2025

  35. [35]

    Skillmimicgen: Automated demonstration gener- ation for efficient skill learning and deployment.arXiv preprint arXiv:2410.18907, 2024

    Caelan Garrett, Ajay Mandlekar, Bowen Wen, and Dieter Fox. Skillmimicgen: Automated demonstration gener- ation for efficient skill learning and deployment.arXiv preprint arXiv:2410.18907, 2024

    work page arXiv 2024

  36. [36]

    3d gaussian splatting for real-time radiance field rendering.ACM Trans

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, George Drettakis, et al. 3d gaussian splatting for real-time radiance field rendering.ACM Trans. Graph., 42(4):139–1, 2023

    2023

  37. [37]

    Deformation constraints in a mass-spring model to describe rigid cloth behaviour

    Xavier Provot et al. Deformation constraints in a mass-spring model to describe rigid cloth behaviour. InGraphics interface, pages 147–147. Canadian Information Processing Society, 1995

    1995

  38. [38]

    Real-time elastic deformations of soft tissues for surgery simulation.IEEE transactions on Visualization and Computer Graphics, 5(1):62–73, 2002

    Stéphane Cotin, Hervé Delingette, and Nicholas Ayache. Real-time elastic deformations of soft tissues for surgery simulation.IEEE transactions on Visualization and Computer Graphics, 5(1):62–73, 2002

    2002

  39. [39]

    A moving least squares material point method with displacement discontinuity and two-way rigid body coupling.ACM Transactions on Graphics (TOG), 37(4):1–14, 2018

    Yuanming Hu, Yu Fang, Ziheng Ge, Ziyin Qu, Yixin Zhu, Andre Pradhana, and Chenfanfu Jiang. A moving least squares material point method with displacement discontinuity and two-way rigid body coupling.ACM Transactions on Graphics (TOG), 37(4):1–14, 2018

    2018

  40. [40]

    Position based dynamics.Journal of Visual Communication and Image Representation, 18(2):109–118, 2007

    Matthias Müller, Bruno Heidelberger, Marcus Hennix, and John Ratcliff. Position based dynamics.Journal of Visual Communication and Image Representation, 18(2):109–118, 2007

    2007

  41. [41]

    May, Tushar Kusnur, George Konidaris, and Laura Herlant

    Sergio Orozco, Brandon B. May, Tushar Kusnur, George Konidaris, and Laura Herlant. Learning equivariant neural-augmented object dynamics from few interactions. InBeyond Rigid Worlds: Representing and Interacting with Non-Rigid Objects, 2025. URLhttps://openreview.net/forum?id=JAiJpFozaD

    2025

  42. [42]

    Tenenbaum, David Held, and Chuang Gan

    Xingyu Lin, Zhiao Huang, Yunzhu Li, Joshua B. Tenenbaum, David Held, and Chuang Gan. Diffskill: Skill ab- straction from differentiable physics for deformable object manipulations with tools. InInternational Conference on Learning Representations (ICLR), 2022

    2022

  43. [43]

    Robocook: Long-horizon elasto-plastic object manipulation with diverse tools

    Haochen Shi, Huazhe Xu, Samuel Clarke, Yunzhu Li, and Jiajun Wu. Robocook: Long-horizon elasto-plastic object manipulation with diverse tools. InConference on Robot Learning (CoRL), 2023

    2023

  44. [44]

    Predicting object interactions with behavior primitives: An application in stowing tasks

    Haonan Chen, Yilong Niu, Kaiwen Hou, Shuijing Liu, Yixuan Wang, Yunzhu Li, and Katherine Driggs-Campbell. Predicting object interactions with behavior primitives: An application in stowing tasks. InConference on Robot Learning (CoRL), 2023

    2023

  45. [45]

    Defgraspsim: Physics-based simulation of grasp outcomes for 3d deformable objects

    Isabella Huang, Yashraj Narang, Ruzena Bajcsy, Fabio Ramos, Tucker Hermans, and Dieter Fox. Defgraspsim: Physics-based simulation of grasp outcomes for 3d deformable objects. InIEEE International Conference on Robotics and Automation (ICRA), 2022

    2022

  46. [46]

    Robotic manipulation of deformable objects: a comprehensive review.Robotic Intelligence and Automation, pages 1–16, 2026

    Lijun Han and Hesheng Wang. Robotic manipulation of deformable objects: a comprehensive review.Robotic Intelligence and Automation, pages 1–16, 2026

    2026

  47. [47]

    A perspective on open challenges in deformable object manipulation

    Ryan Paul McKennaa and John Oyekan. A perspective on open challenges in deformable object manipulation. arXiv preprint arXiv:2602.22998, 2026

    work page arXiv 2026

  48. [48]

    Diffusion policy: Visuomotor policy learning via action diffusion

    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. InThe International Journal of Robotics Research, 2024

    2024

  49. [49]

    Autoencoders as cross-modal teachers: Can pretrained 2d image transformers help 3d representation learning? InICLR, 2023

    Runpei Dong, Zekun Qi, Linfeng Zhang, Junbo Zhang, Jianjian Sun, Zheng Ge, Li Yi, and Kaisheng Ma. Autoencoders as cross-modal teachers: Can pretrained 2d image transformers help 3d representation learning? InICLR, 2023

    2023

  50. [50]

    Tiebot: Learning to knot a tie from visual demonstration through a real-to-sim-to-real approach.arXiv preprint arXiv:2407.03245, 2024

    Weikun Peng, Jun Lv, Yuwei Zeng, Haonan Chen, Siheng Zhao, Jichen Sun, Cewu Lu, and Lin Shao. Tiebot: Learning to knot a tie from visual demonstration through a real-to-sim-to-real approach.arXiv preprint arXiv:2407.03245, 2024

    work page arXiv 2024

  51. [51]

    Robotic assembly of deformable linear objects via curriculum reinforcement learning.IEEE Robotics and Automation Letters, 2025

    Kai Wu, Rongkang Chen, Qi Chen, and Weihua Li. Robotic assembly of deformable linear objects via curriculum reinforcement learning.IEEE Robotics and Automation Letters, 2025

    2025

  52. [52]

    Checheng Yu, Chonghao Sima, Gangcheng Jiang, Hai Zhang, Haoguang Mai, Hongyang Li, Huijie Wang, Jin Chen, Kaiyang Wu, Li Chen, Lirui Zhao, Modi Shi, Ping Luo, Qingwen Bu, Shijia Peng, Tianyu Li, and Yibo Yuan.χ 0: Resource-aware robust manipulation via taming distributional inconsistencies.arXiv preprint arXiv:2602.09021, 2026

    work page arXiv 2026

  53. [53]

    Deep imitation learning of sequential fabric smoothing from an algorithmic supervisor

    Daniel Seita, Aditya Ganapathi, Ryan Hoque, Minho Hwang, Edward Cen, Ajay Kumar Tanwani, Ashwin Balakrishna, Brijen Thananjeyan, Jeffrey Ichnowski, Nawid Jamali, et al. Deep imitation learning of sequential fabric smoothing from an algorithmic supervisor. InIEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020

    2020

  54. [54]

    Fabricflownet: Bimanual cloth manipulation with a flow-based policy

    Thomas Weng, Sujay Bajracharya, Yufei Wang, Khush Agrawal, and David Held. Fabricflownet: Bimanual cloth manipulation with a flow-based policy. InConference on Robot Learning (CoRL), 2022

    2022

  55. [55]

    Dexgarmentlab: Dexterous garment manipulation environment with generalizable policy

    Yuran Wang, Ruihai Wu, Yue Chen, Jiarui Wang, Jiaqi Liang, Ziyu Zhu, Haoran Geng, Jitendra Malik, Pieter Abbeel, and Hao Dong. Dexgarmentlab: Dexterous garment manipulation environment with generalizable policy. arXiv preprint arXiv:2505.11032, 2025

    work page arXiv 2025

  56. [56]

    Flingbot: The unreasonable effectiveness of dynamic manipulation for cloth unfolding

    Huy Ha and Shuran Song. Flingbot: The unreasonable effectiveness of dynamic manipulation for cloth unfolding. InConference on Robot Learning (CoRL), pages 24–33. PMLR, 2022

    2022

  57. [57]

    Phystwin: Physics- informed reconstruction and simulation of deformable objects from videos.ICCV, 2025

    Hanxiao Jiang, Hao-Yu Hsu, Kaifeng Zhang, Hsin-Ni Yu, Shenlong Wang, and Yunzhu Li. Phystwin: Physics- informed reconstruction and simulation of deformable objects from videos.ICCV, 2025

    2025

  58. [58]

    Softgym: Benchmarking deep reinforcement learning for deformable object manipulation

    Xingyu Lin, Yufei Wang, Jake Olkin, and David Held. Softgym: Benchmarking deep reinforcement learning for deformable object manipulation. InConference on Robot Learning, 2020

    2020

  59. [59]

    Taichi: a language for high-performance computation on spatially sparse data structures.ACM Transactions on Graphics (TOG), 38 (6):1–16, 2019

    Yuanming Hu, Tzu-Mao Li, Luke Anderson, Jonathan Ragan-Kelley, and Frédo Durand. Taichi: a language for high-performance computation on spatially sparse data structures.ACM Transactions on Graphics (TOG), 38 (6):1–16, 2019

    2019

  60. [60]

    Warp: A high-performance python framework for gpu simulation and graphics

    Miles Macklin. Warp: A high-performance python framework for gpu simulation and graphics. InNVIDIA GPU Technology Conference (GTC), volume 3, 2022

    2022

  61. [61]

    Interndata-a1: Pioneering high-fidelity synthetic data for pre-training generalist policy.arXiv preprint arXiv:2511.16651, 2025

    Yang Tian, Yuyin Yang, Yiman Xie, Zetao Cai, Xu Shi, Ning Gao, Hangxu Liu, Xuekun Jiang, Zherui Qiu, Feng Yuan, et al. Interndata-a1: Pioneering high-fidelity synthetic data for pre-training generalist policy.arXiv preprint arXiv:2511.16651, 2025

    work page arXiv 2025

  62. [62]

    Learning from demonstrations through the use of non-rigid registration

    John Schulman, Jonathan Ho, Cameron Lee, and Pieter Abbeel. Learning from demonstrations through the use of non-rigid registration. InRobotics Research: The 16th International Symposium ISRR, pages 339–354. Springer, 2016

    2016

  63. [63]

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

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

    2023

  64. [64]

    Diffusion policy: Visuomotor policy learning via action diffusion.The International Journal of Robotics Research, 2023

    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, 2023

    2023

  65. [65]

    Smolvla: A vision-language-action model for affordable and efficient robotics.arXiv preprint, 2025

    Mustafa Shukor, Dana Aubakirova, Francesco Capuano, Pepijn Kooijmans, Steven Palma, Adil Zouitine, Michel Aractingi, Caroline Pascal, Martino Russi, Andres Marafioti, et al. Smolvla: A vision-language-action model for affordable and efficient robotics.arXiv preprint, 2025. Appendix A State Augmentation Details A.1 Formal Assumption Our approach relies on ...

    2025