Pith. sign in

REVIEW 4 major objections 3 minor 50 references

The paper claims that a goal-conditioned recurrent state-space model, LaGarNet, can flatten four garment types in simulation and on a real robot using a single policy.

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 ↗

The submission's abstract describes a new garment-flattening robot model, but the body is a different paper about document retrieval, making the submission internally inconsistent.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection The submission is not a coherent paper: the abstract promises a garment-flattening robotics method, but the full text is an unrelated document-retrieval paper, leaving the claimed contribution with zero supporting evidence. the 4 major comments →

arxiv 2508.17070 v1 pith:HAVRYZWC submitted 2025-08-23 cs.RO

LaGarNet: Goal-Conditioned Recurrent State-Space Models for Pick-and-Place Garment Flattening

Halid Abdulrahim Kadi , Kasim Terzi\'c This is my paper
classification cs.RO
keywords LaGarNetgoal-conditioned recurrent state-space modelgarment flatteningpick-and-place manipulationdeformable objectscoverage-alignment rewardrobot manipulation
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 reading

LaGarNet is presented as the first successful application of state-space models to complex garment manipulation. The paper argues that a goal-conditioned recurrent state-space model (GC-RSSM), trained on a coverage-alignment reward, can learn the latent dynamics of pick-and-place fabric flattening well enough to match mesh-based state-of-the-art methods. The central experimental claim is that one LaGarNet policy flattens four different garment types in both simulation and the real world, with fewer hand-built inductive biases. A caveat: the supplied full text is a different paper on multimodal document retrieval, so these claims rest entirely on the abstract and no LaGarNet details appear in the body.

Core claim

The central claim, stated on the paper's own terms, is that a goal-conditioned recurrent state-space model (GC-RSSM) can serve as the latent-dynamics backbone for pick-and-place garment flattening. Whereas previous strong methods lean on mesh-based geometric representations of cloth, LaGarNet learns its own latent state transitions from data. The training signal is a coverage-alignment reward, and the dataset is gathered through a general procedure: a random policy for exploration plus a diffusion policy initialized from a few human demonstrations. The paper reports that this single policy matches mesh-based state-of-the-art performance and flattens four distinct garment types in simulation

What carries the argument

The central object is the goal-conditioned recurrent state-space model (GC-RSSM), a learned latent-dynamics model that represents the fabric state and predicts how pick-and-place actions evolve it toward a goal. It replaces mesh-based inductive biases with a compact latent state, so the same policy can be trained across multiple garment geometries. The two supporting mechanisms are the coverage-alignment reward, which scores how well the fabric covers the target area, and the data-collection pipeline that combines a random policy with a demonstration-initialized diffusion policy.

Load-bearing premise

The load-bearing premise is that the described training recipe—a coverage-alignment reward plus a dataset gathered by a random policy and a few human demonstrations—is sufficient for one policy to flatten four garment types in simulation and reality; the supplied full text is a different paper and provides no LaGarNet experiments to back this up.

What would settle it

Open the LaGarNet paper's experiments section and check whether it reports the coverage-alignment reward, the random-policy plus demonstration dataset, and flattening success for all four garment types in simulation and on the real robot. If those results are absent, the abstract's central claim is unsupported; if a single policy succeeds across all four, the claim is verified.

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

If this is right

  • If the claim holds, state-space models become a viable backbone for deformable-object manipulation, not just rigid-body or locomotion tasks.
  • A single LaGarNet policy across four garment types would mean fabric-flattening skill transfers without per-garment re-engineering.
  • The reduced reliance on mesh representations suggests robot cloth manipulation can be learned from general, relatively cheap data collection.
  • Coverage-alignment rewards would provide a simple universal objective for surface-flattening tasks in simulation and reality.

Where Pith is reading between the lines

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

  • Our inference: if the training recipe is as general as claimed, the same GC-RSSM setup could be pointed at other deformable objects—cables, bags, surgical gauze—but the paper does not report such tests.
  • Our inference: the abstract's claim that this is the 'first successful application of state-space models on complex garments' depends on how 'complex' and 'successful' are measured; a natural check is to compare against mesh-based methods with identical action spaces and evaluation protocols.
  • Our meta-inference: because the manuscript body provided is a different paper, every LaGarNet-specific result should be treated as unverified until the actual experimental section appears; the decisive test is presence of per-garment results.
Share X Bluesky LinkedIn Reddit HN

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

4 major / 3 minor

Summary. The manuscript as submitted purports to present LaGarNet, a goal-conditioned recurrent state-space model (GC-RSSM) for pick-and-place garment flattening. The abstract claims that LaGarNet matches state-of-the-art mesh-based methods, that it is the first successful application of state-space models to complex garments, that it trains on a coverage-alignment reward with a generally collected dataset supported by a random policy and a diffusion policy initialized from few human demonstrations, and that a single policy flattens four garment types in simulation and reality. None of these claims is supported by the submitted full text. The body, from the title through Sections 1–7 and all appendices, is a different paper on zero-shot multimodal document retrieval, presenting the PREMIR framework and its retrieval experiments. The words 'LaGarNet', 'RSSM', 'garment', 'flatten', 'coverage', and 'diffusion policy' do not appear in the manuscript body. There is no architecture description, no derivation, no training protocol, no robot experiment, no simulation setup, and no quantitative result related to garment flattening anywhere in the artifact.

Significance. If the LaGarNet claims were true and fully documented, the paper would make a noteworthy contribution to robot garment manipulation: a single recurrent state-space policy flattening multiple garment types in both simulation and the real world, without mesh-based inductive biases, would be a substantive advance. However, none of the evidence needed to assess that contribution is present in the submitted manuscript. The paper contains no model definition, no reward specification, no data-collection description, no experimental protocol, and no results for the claimed task. The retrieval paper that fills the manuscript is unrelated to the abstract's claims. Because the central claim is entirely unsupported by the submitted content, the manuscript cannot be evaluated on its merits. The stress-test concern therefore lands: this is not a disagreement about interpretation or a hidden assumption, but a complete mismatch between the advertised contribution and the submitted artifact.

major comments (4)
  1. [Abstract vs. full text] The abstract announces LaGarNet, a goal-conditioned recurrent state-space model for garment flattening, with claims of state-of-the-art performance and real/sim experiments. The full text, however, is titled 'Zero-shot Multimodal Document Retrieval via Cross-modal Question Generation' and describes PREMIR, a document-retrieval framework. Sections 1–7 and Appendices A–E contain no mention of LaGarNet, RSSM, garment manipulation, flattening, coverage-alignment rewards, or diffusion policies. This is not a missing detail or a presentation issue; the entire claimed contribution is absent from the manuscript.
  2. [Experiments (all sections)] The abstract promises that a single-policy LaGarNet achieves flattening on four garment types in both real-world and simulation settings. The manuscript contains no tables, figures, or text reporting such experiments. The only experimental results, ablations, and latency tables (Tables 1–12) concern multimodal document retrieval on ViDoSeek, REAL-MM-RAG, CT2C-QA, and Allganize. There are no error bars, no robot hardware description, no simulation environment, no garment categories, and no evaluation metric for flattening. The claimed empirical support cannot be located or checked.
  3. [Method (Section 2)] The manuscript provides no specification of the LaGarNet architecture, the GC-RSSM latent dynamics, the coverage-alignment reward, the general-purpose data collection procedure, or the diffusion-policy initialization mentioned in the abstract. Section 2 of the submitted text defines PREMIR's task and retrieval pipeline, which is unrelated to robot manipulation. Without these components, the central methodological claim is unverifiable, and the paper cannot be reproduced or even partially assessed.
  4. [Section 7 / Limitations] The manuscript's own limitations section discusses generic pre-question generation in the PREMIR retrieval system. This is evidence that the body was written for an entirely different paper. It does not address any limitation of LaGarNet, such as generalization across garment types, reward design, or sim-to-real transfer. The internal mismatch is therefore not confined to a single section but pervades the entire artifact.
minor comments (3)
  1. [Title/authorship] The paper's title and author list correspond to the PREMIR retrieval paper, not to LaGarNet. The arXiv identifier embedded in the full text (2508.17079) also differs from the manuscript's stated identifier (2508.17070). This suggests a submission or packaging error that should be corrected by the authors.
  2. [References] The reference list is entirely composed of information-retrieval, multimodal-LLM, and document-understanding works. It contains no citations to garment manipulation, state-space models for robotics, or deformable-object manipulation. A reader of the claims in the abstract would expect such references to situate the contribution.
  3. [All appendices] Appendices A–E provide implementation details and prompts for the PREMIR retrieval framework. They contain no information relevant to LaGarNet, such as network hyperparameters, reward coefficients, data collection details, or real-robot setup.

Circularity Check

0 steps flagged

No circularity found: the manuscript body is an unrelated retrieval paper, so the LaGarNet claim has no derivation chain to be circular.

full rationale

The submitted artifact is internally inconsistent: the abstract announces LaGarNet, "a novel goal-conditioned recurrent state space (GC-RSSM) model capable of learning latent dynamics of pick-and-place garment manipulation," and claims it "matches the state-of-the-art performance of mesh-based methods" and "achieves flattening on four different types of garments in both real-world and simulation settings." However, the entire full text is the paper "Zero-shot Multimodal Document Retrieval via Cross-modal Question Generation" (PREMIR), a cs.IR retrieval paper with no mention of LaGarNet, RSSM, garments, flattening, coverage rewards, or diffusion policies. There is therefore no derivation chain, no equations, and no fitted parameters in the submitted text that could be checked for equivalence to inputs. The central robotics claim is unverifiable from the artifact, but that is a missing-evidence / completeness problem, not a circularity problem. No self-citation is load-bearing, no imported uniqueness theorem is invoked, and no known result is renamed. Under the review rules, circularity can only be claimed when a specific reduction is exhibited; here there is nothing to exhibit. Accordingly, the circularity score is 0.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract mentions a coverage-alignment reward and a dataset from random and diffusion policies, but since the full text is a different paper, these cannot be audited for free parameters, axioms, or invented entities.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of LaGarNet: Goal-Conditioned Recurrent State-Space Models for Pick-and-Place Garment Flattening." pith.science (2026). https://pith.science/paper/HAVRYZWC

@misc{pith2026250817070,
  author       = {Pith},
  title        = {Pith review of: LaGarNet: Goal-Conditioned Recurrent State-Space Models for Pick-and-Place Garment Flattening},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HAVRYZWC}},
  note         = {Machine review of arXiv:2508.17070}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We present a novel goal-conditioned recurrent state space (GC-RSSM) model capable of learning latent dynamics of pick-and-place garment manipulation. Our proposed method LaGarNet matches the state-of-the-art performance of mesh-based methods, marking the first successful application of state-space models on complex garments. LaGarNet trains on a coverage-alignment reward and a dataset collected through a general procedure supported by a random policy and a diffusion policy learned from few human demonstrations; it substantially reduces the inductive biases introduced in the previous similar methods. We demonstrate that a single-policy LaGarNet achieves flattening on four different types of garments in both real-world and simulation settings.

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

50 extracted references · 43 canonical work pages · 6 internal anchors

  1. [1]

    Doumanoglou, J

    A. Doumanoglou, J. Stria, G. Peleka, I. Mariolis, V. Petrik, A. Kargakos, L. Wagner, V. Hlav \'a c , T.-K. Kim, and S. Malassiotis, ``Folding clothes autonomously: A complete pipeline,'' IEEE Transactions on Robotics, vol. 32, no. 6, pp. 1461--1478, 2016

    work page 2016

  2. [2]

    Ha and J

    D. Ha and J. Schmidhuber, ``World models,'' arXiv preprint arXiv:1803.10122, 2018

    Pith/arXiv arXiv 2018

  3. [3]

    Hafner, T

    D. Hafner, T. Lillicrap, I. Fischer, R. Villegas, D. Ha, H. Lee, and J. Davidson, ``Learning latent dynamics for planning from pixels,'' in International Conference on Machine Learning. 1em plus 0.5em minus 0.4em Long Beach, CA, USA: PMLR, 10--15 June 2019, pp. 2555--2565

    work page 2019

  4. [4]

    Hafner, T

    D. Hafner, T. Lillicrap, J. Ba, and M. Norouzi, ``Dream to control: Learning behaviors by latent imagination,'' in International Conference on Learning Representations, Addis Ababa, Ethiopia, 26--30 April 2020. [Online]. Available: https://openreview.net/forum?id=S1lOTC4tDS

    work page 2020

  5. [5]

    F. Deng, J. Park, and S. Ahn, ``Facing off world model backbones: Rnns, transformers, and s4,'' Advances in Neural Information Processing Systems, vol. 36, pp. 72\,904--72\,930, 2023

    work page 2023

  6. [6]

    X. Ma, D. Hsu, and W. S. Lee, ``Learning latent graph dynamics for deformable object manipulation,'' CoRR, vol. abs/2104.12149, 2021. [Online]. Available: https://arxiv.org/abs/2104.12149

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  7. [7]

    W. Yan, A. Vangipuram, P. Abbeel, and L. Pinto, ``Learning predictive representations for deformable objects using contrastive estimation,'' in Conference on Robot Learning. 1em plus 0.5em minus 0.4em London, UK: PMLR, 8--11 November 2021, pp. 564--574

    work page 2021

  8. [8]

    X. Lin, Y. Wang, J. Olkin, and D. Held, ``Softgym: Benchmarking deep reinforcement learning for deformable object manipulation,'' in Conference on Robot Learning. 1em plus 0.5em minus 0.4em London, UK: PMLR, 8--11 November 2021, pp. 432--448

    work page 2021

  9. [9]

    Bertiche, M

    H. Bertiche, M. Madadi, and S. Escalera, ``Cloth3d: Clothed 3d humans,'' in European Conference on Computer Vision. 1em plus 0.5em minus 0.4em Springer, 2020, pp. 344--359

    work page 2020

  10. [10]

    Canberk, C

    A. Canberk, C. Chi, H. Ha, B. Burchfiel, E. Cousineau, S. Feng, and S. Song, ``Cloth funnels: Canonicalized-alignment for multi-purpose garment manipulation,'' in 2023 IEEE International Conference on Robotics and Automation (ICRA). 1em plus 0.5em minus 0.4em IEEE, 2023, pp. 5872--5879

    work page 2023

  11. [11]

    H. A. Kadi, J. A. Chandy, L. Figueredo, K. Terzi \'c , and P. Caleb-Solly, ``Draper: Towards a robust robot deployment and reliable evaluation for quasi-static pick-and-place cloth-shaping neural controllers,'' 2025. [Online]. Available: https://arxiv.org/abs/2409.15159

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  12. [12]

    H. A. Kadi and K. Terzi \'c , ``Planet-clothpick: Effective fabric flattening based on latent dynamic planning,'' in 2024 IEEE/SICE International Symposium on System Integration (SII). 1em plus 0.5em minus 0.4em IEEE, 2024, pp. 972--979

    work page 2024

  13. [13]

    Avigal, L

    Y. Avigal, L. Berscheid, T. Asfour, T. Kr \"o ger, and K. Goldberg, ``Speedfolding: Learning efficient bimanual folding of garments,'' in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). 1em plus 0.5em minus 0.4em IEEE, 2022, pp. 1--8

    work page 2022

  14. [14]

    Ha and S

    H. Ha and S. Song, ``Flingbot: The unreasonable effectiveness of dynamic manipulation for cloth unfolding,'' in Conference on Robot Learning. 1em plus 0.5em minus 0.4em Auckland, New Zealand: PMLR, 15--18 Decemeber 2022, pp. 24--33

    work page 2022

  15. [15]

    X. Lin, Y. Wang, Z. Huang, and D. Held, ``Learning visible connectivity dynamics for cloth smoothing,'' in Conference on Robot Learning. 1em plus 0.5em minus 0.4em Auckland, New Zealand: PMLR, 5--18 December 2022, pp. 256--266

    work page 2022

  16. [16]

    R. Lee, J. Abou-Chakra, F. Zhang, and P. Corke, ``Learning fabric manipulation in the real world with human videos,'' in 2024 IEEE International Conference on Robotics and Automation (ICRA). 1em plus 0.5em minus 0.4em IEEE, 2024, pp. 3124--3130

    work page 2024

  17. [17]

    Hoque, D

    R. Hoque, D. Seita, A. Balakrishna, A. Ganapathi, A. K. Tanwani, N. Jamali, K. Yamane, S. Iba, and K. Goldberg, ``Visuospatial foresight for physical sequential fabric manipulation,'' Autonomous Robots, vol. 46, no. 1, pp. 175--199, 2022a

    work page

  18. [18]

    R. Lee, D. Ward, V. Dasagi, A. Cosgun, J. Leitner, and P. Corke, ``Learning arbitrary-goal fabric folding with one hour of real robot experience,'' in Conference on Robot Learning. 1em plus 0.5em minus 0.4em London, UK: PMLR, 11 November 2021, pp. 2317--2327

    work page 2021

  19. [19]

    Huang, X

    Z. Huang, X. Lin, and D. Held, ``Mesh-based dynamics with occlusion reasoning for cloth manipulation,'' arXiv preprint arXiv:2206.02881, 2022

    Pith/arXiv arXiv 2022

  20. [20]

    Matas, S

    J. Matas, S. James, and A. J. Davison, ``Sim-to-real reinforcement learning for deformable object manipulation,'' in Conference on Robot Learning. 1em plus 0.5em minus 0.4em Zurich, Switzerland: PMLR, 9--31 October 2018, pp. 734--743

    work page 2018

  21. [21]

    Y. Wu, W. Yan, T. Kurutach, L. Pinto, and P. Abbeel, ``Learning to manipulate deformable objects without demonstrations,'' arXiv preprint arXiv:1910.13439, 2019

    Pith/arXiv arXiv 1910

  22. [22]

    Learning Visual Feedback Control for Dynamic Cloth Folding

    J. Hietala, D. Blanco-Mulero, G. Alcan, and V. Kyrki, ``Closing the sim2real gap in dynamic cloth manipulation,'' arXiv preprint arXiv:2109.04771, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  23. [23]

    C. He, L. Meng, Z. Sun, J. Wang, and M. Q.-H. Meng, ``Fabricfolding: learning efficient fabric folding without expert demonstrations,'' Robotica, vol. 42, no. 4, pp. 1281--1296, 2024

    work page 2024

  24. [24]

    N.-Q. Gu, R. He, and L. Yu, ``Learning to unfold garment effectively into oriented direction,'' IEEE Robotics and Automation Letters, 2023

    work page 2023

  25. [25]

    Cloth Manipulation Planning on Basis of Mesh Representations with Incomplete Domain Knowledge and Voxel-to-Mesh Estimation

    S. Arnold, D. Tanaka, and K. Yamazaki, ``Cloth manipulation planning on basis of mesh representations with incomplete domain knowledge and voxel-to-mesh estimation,'' arXiv preprint arXiv:2103.08137, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  26. [26]

    Lips, V.-L

    T. Lips, V.-L. De Gusseme et al., ``Learning keypoints for robotic cloth manipulation using synthetic data,'' IEEE Robotics and Automation Letters, 2024

    work page 2024

  27. [27]

    T. Weng, S. M. Bajracharya, Y. Wang, K. Agrawal, and D. Held, ``Fabricflownet: Bimanual cloth manipulation with a flow-based policy,'' in Conference on Robot Learning. 1em plus 0.5em minus 0.4em uckland, New Zealand: PMLR, 15–18 December 2022, pp. 192--202

    work page 2022

  28. [28]

    Y. Teng, H. Lu, Y. Li, T. Kamiya, Y. Nakatoh, S. Serikawa, and P. Gao, ``Multidimensional deformable object manipulation based on dn-transporter networks,'' IEEE Transactions on Intelligent Transportation Systems, 2022

    work page 2022

  29. [29]

    K. Mo, C. Xia, X. Wang, Y. Deng, X. Gao, and B. Liang, ``Foldsformer: Learning sequential multi-step cloth manipulation with space-time attention,'' IEEE Robotics and Automation Letters, vol. 8, no. 2, pp. 760--767, 2022

    work page 2022

  30. [30]

    SIS: Seam-Informed Strategy for T-shirt Unfolding

    X. Huang, A. Seino, F. Tokuda, A. Kobayashi, D. Chen, Y. Hirata, N. C. Tien, and K. Kosuge, ``Sis: Seam-informed strategy for t-shirt unfolding,'' arXiv preprint arXiv:2409.06990, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  31. [31]

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

    work page 2023

  32. [32]

    V.-L. D. Gusseme, R. Proesmans, T. Lips, A. Verleysen, and F. Wyffels, ``Insights for robotic cloth manipulation: A comprehensive analysis of a competition-winning system,'' International Journal of Advanced Robotic Systems, vol. 22, no. 2, p. 17298806251322582, 2025

    work page 2025

  33. [33]

    R. Wu, H. Lu, Y. Wang, Y. Wang, and H. Dong, ``Unigarmentmanip: A unified framework for category-level garment manipulation via dense visual correspondence,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 16\,340--16\,350

    work page 2024

  34. [34]

    Doumanoglou, A

    A. Doumanoglou, A. Kargakos, T.-K. Kim, and S. Malassiotis, ``Autonomous active recognition and unfolding of clothes using random decision forests and probabilistic planning,'' in 2014 IEEE international conference on robotics and automation (ICRA). 1em plus 0.5em minus 0.4em Hong Kong, China: IEEE, 1 May–5 June 2 2014a, pp. 987--993

    work page 2014

  35. [35]

    L. Sun, G. Aragon-Camarasa, S. Rogers, and J. P. Siebert, ``Accurate garment surface analysis using an active stereo robot head with application to dual-arm flattening,'' in 2015 IEEE international conference on robotics and automation (ICRA). 1em plus 0.5em minus 0.4em IEEE, 2015, pp. 185--192

    work page 2015

  36. [36]

    Hoque, A

    R. Hoque, A. Balakrishna, C. Putterman, M. Luo, D. S. Brown, D. Seita, B. Thananjeyan, E. Novoseller, and K. Goldberg, ``Lazydagger: Reducing context switching in interactive imitation learning,'' in 2021 IEEE 17th international conference on automation science and engineering (case). 1em plus 0.5em minus 0.4em IEEE, 2021, pp. 502--509

    work page 2021

  37. [37]

    Hafner, J

    D. Hafner, J. Pasukonis, J. Ba, and T. Lillicrap, ``Mastering diverse domains through world models,'' arXiv preprint arXiv:2301.04104, 2023

    Pith/arXiv arXiv 2023

  38. [38]

    Y. Duan, W. Mao, and H. Zhu, ``Learning world models for unconstrained goal navigation,'' Advances in Neural Information Processing Systems, vol. 37, pp. 59\,236--59\,265, 2024

    work page 2024

  39. [39]

    T. Tian, H. Li, B. Ai, X. Yuan, Z. Huang, and H. Su, ``Diffusion dynamics models with generative state estimation for cloth manipulation,'' arXiv preprint arXiv:2503.11999, 2025

    Pith/arXiv arXiv 2025

  40. [40]

    Ebert, C

    F. Ebert, C. Finn, S. Dasari, A. Xie, A. X. Lee, and S. Levine, ``Visual foresight: Model-based deep reinforcement learning for vision-based robotic control,'' CoRR, vol. abs/1812.00568, 2018. [Online]. Available: http://arxiv.org/abs/1812.00568

    Pith/arXiv arXiv 2018

  41. [41]

    Andrychowicz, F

    M. Andrychowicz, F. Wolski, A. Ray, J. Schneider, R. Fong, P. Welinder, B. McGrew, J. Tobin, O. Pieter Abbeel, and W. Zaremba, ``Hindsight experience replay,'' Advances in neural information processing systems, vol. 30, 2017

    work page 2017

  42. [42]

    Ma and J

    M. Ma and J. Zhu, ``Interpretable recurrent variational state-space model for fault detection of complex systems based on multisensory signals,'' Applied Sciences, vol. 14, no. 9, p. 3772, 2024

    work page 2024

  43. [43]

    VisuoSpatial Foresight for Multi-Step, Multi-Task Fabric Manipulation

    R. Hoque, D. Seita, A. Balakrishna, A. Ganapathi, A. K. Tanwani, N. Jamali, K. Yamane, S. Iba, and K. Goldberg, ``Visuospatial foresight for multi-step, multi-task fabric manipulation.'' CoRR, vol. abs/2003.09044, 2020. [Online]. Available: https://arxiv.org/abs/2003.09044

    work page internal anchor Pith review Pith/arXiv arXiv 2003

  44. [44]

    H. A. Kadi and K. Terzi \'c , `` JA - TN : Pick-and-place towel shaping from crumpled states based on transporter net with joint-probability action inference,'' in 8th Annual Conference on Robot Learning, 2024. [Online]. Available: https://openreview.net/forum?id=SW8ntpJl0E

    work page 2024

  45. [45]

    C. Chi, S. Feng, Y. Du, Z. Xu, E. Cousineau, B. Burchfiel, and S. Song, ``Diffusion policy: Visuomotor policy learning via action diffusion,'' arXiv preprint arXiv:2303.04137, 2023

    Pith/arXiv arXiv 2023

  46. [46]

    Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, ``Image quality assessment: from error visibility to structural similarity,'' IEEE transactions on image processing, vol. 13, no. 4, pp. 600--612, 2004

    work page 2004

  47. [47]

    R. F. Prudencio, M. R. Maximo, and E. L. Colombini, ``A survey on offline reinforcement learning: Taxonomy, review, and open problems,'' IEEE Transactions on Neural Networks and Learning Systems, 2023

    work page 2023

  48. [48]

    Agarwal, D

    R. Agarwal, D. Schuurmans, and M. Norouzi, ``An optimistic perspective on offline reinforcement learning,'' in International conference on machine learning. 1em plus 0.5em minus 0.4em PMLR, 2020, pp. 104--114

    work page 2020

  49. [49]

    Pertsch, O

    K. Pertsch, O. Rybkin, F. Ebert, S. Zhou, D. Jayaraman, C. Finn, and S. Levine, ``Long-horizon visual planning with goal-conditioned hierarchical predictors,'' Advances in Neural Information Processing Systems, vol. 33, pp. 17\,321--17\,333, 2020

    work page 2020

  50. [50]

    J. B. Paoletti, Pink and blue: Telling the boys from the girls in America. 1em plus 0.5em minus 0.4em Indiana University Press, 2012

    work page 2012

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.