Pith. sign in

REVIEW 4 major objections 4 minor 24 references

The paper claims that the most useful repair examples for a brittle visuomotor policy are the ones exposing its fragile action responses, and that these can be found offline by measuring action drift under task-preserving counterfactual nui

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 · deepseek-v4-flash

2026-08-01 14:21 UTC pith:DNVQF2KX

load-bearing objection A well-specified, honest paper whose data-efficiency result is solid on candidate-covered conditions but does not survive contact with held-out generalization as cleanly as the abstract implies. the 4 major comments →

arxiv 2607.27261 v1 pith:DNVQF2KX submitted 2026-07-29 cs.RO

It's Not Just More Demos: Counterfactual Action Sensitivity Coverage for Data-Efficient Robust Robot Imitation

Giovanni D'urso , Kaushik Roy , Nicholas Lawrance , Brendan Tidd This is my paper
classification cs.RO
keywords imitation learningvisuomotor policycounterfactual data selectionaction driftrobustness repairbehavior cloningnuisance variationdata efficiency
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 sets out to answer a practical question: when a robot policy works in the lab but breaks under minor visual changes, which additional demonstrations will actually fix it? Its answer is an offline audit: generate pairs of clean and nuisance observations that leave the expert action unchanged, record how much the policy's predicted action moves between the pair (action drift), and use that signal to select a small, response-diverse repair set. In the two simulated manipulation tasks studied, fine-tuning on 20-30 selected candidates lifted all-nuisance success by 0.66 and 0.44, while the same budget of random examples gave 0.26 and 0.28, and hundreds of random examples were needed to approach the selected set's performance. If correct, the result reframes robustness as a coverage problem over a policy's failure modes rather than a data-volume problem.

Core claim

The central claim is that a brittle visuomotor policy can serve as its own probe: comparing its predicted actions on paired clean and nuisance observations, where the expert action is identical by construction, yields an offline fragility signal that ranks visual conditions by how much they break the policy. The paper calls this signal action drift and shows that it correlates with rollout failure across nuisance conditions, with rank correlations of 0.95 and 0.88 on the two tasks. It then argues that selecting repairs by drift score alone is suboptimal because high-drift examples can be redundant, and instead selects candidates that cover diverse response shapes. The reported numbers are th

What carries the argument

The central object is the paired counterfactual observation (clean scene paired with nuisance scene) constructed to preserve the expert action, together with the action-drift vector, the normalized difference between the policy's predicted actions on the pair. Drift magnitude ranks candidate nuisances; drift response features, summary statistics of that difference over time and action dimensions, define distinct response modes. A kernel affinity over those features feeds a greedy coverage objective that picks a budget-sized repair set spanning diverse high-drift modes. The selected nuisance observations are paired with the original demonstration actions and used to fine-tune the policy from

Load-bearing premise

The load-bearing premise is that every nuisance intervention really is task-preserving, meaning the demonstrated expert action remains the correct action after the visual change; if a subtle change alters the intended task, action drift misreads a legitimate action change as fragility and mislabels the selected repairs.

What would settle it

Construct a set of nuisances that deliberately changes the intended goal, violating the equality a*(s_c)=a*(s_n), while keeping the same visual shift magnitudes; if drift ranks these task-changing shifts as high-priority repairs, the signal conflates task change with nuisance fragility. Also, evaluate a held-out nuisance family absent from the candidate pool: if response-guided repair with a budget of 30 cannot beat random selection at the same budget on that condition despite the pool containing useful candidates, the coverage objective has failed.

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

If this is right

  • Robustness repair can be planned offline, before deployment, by auditing the trained policy against task-preserving counterfactuals.
  • At small repair budgets, response-diverse selection is more effective than the same number of random examples, and can match the effect of much larger random budgets.
  • Action drift is a policy-specific fragility signal: the same nuisance condition that is benign for one policy can be destructive for another, so repair data should be chosen per policy rather than per task.
  • Held-out transfer is partial and limited by candidate-pool coverage; large random budgets still transfer best, identifying candidate coverage as the main bottleneck.

Where Pith is reading between the lines

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

  • A testable extension is to average drift over multiple action samples or output distributions for stochastic or diffusion policies; single-sample drift may understate fragility when the policy is multimodal.
  • The same signal could be inverted into a live data-acquisition policy: during a distribution shift, redirect data collection toward the highest-drift scenarios rather than re-collecting random demonstrations.
  • The coverage principle suggests that the best repair set is not intrinsic to the data but depends on the current policy, so the same candidate pool would yield different selections for different checkpoints, which is a checkable prediction.
  • Pairing drift with rollout-derived or ensemble-disagreement signals might overcome the candidate-coverage limit that the paper explicitly acknowledges.

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 / 4 minor

Summary. The paper proposes Counterfactual Nuisance Behaviour Cloning (CFNBC), an offline data-selection framework for repairing visuomotor imitation policies that are brittle to task-preserving visual nuisances. CFNBC generates paired clean/nuisance observations under an action-preserving assumption, measures the change in the policy's predicted action ('action drift'), and selects a small, response-diverse repair set by greedily maximizing weighted response coverage. The selected counterfactual samples, labelled with inherited expert actions, are mixed with clean demonstrations to fine-tune the nominal policy. Experiments on MuJoCo bimanual cube transfer and SimplerEnv cube stacking report that drift strongly ranks nuisance-induced failure and that K=20--30 selected candidates outperform matched-budget random selection, sometimes approaching the performance of much larger random repair budgets. The core idea is clearly presented and the offline, rollout-free selection signal is appealing.

Significance. If the central data-efficiency claim holds, CFNBC would be a useful contribution to the growing literature on data-centric robustness for imitation learning: it turns a policy-specific sensitivity measure into a practical data-selection rule that does not require online rollouts or success labels. The paper explicitly builds on established ideas (counterfactual/action-preserving augmentation, active imitation, distribution-shift benchmarks) and contributes a well-specified selection objective. The reported correlations (Spearman rho=0.95/0.88) are strong and the controlled comparisons between selection strategies are well motivated. However, the significance is tempered by the empirical evidence base: the main quantitative claims rest on three-seed point estimates with no error bars, and the headline 'approaching much larger random budgets' is not supported on held-out nuisance instantiations. The authors are candid about these limitations, but the abstract and introduction present the stronger reading.

major comments (4)
  1. [§IV (Training seeds and uncertainty) and Table II] The central data-efficiency claim—that K=20--30 response-guided candidates substantially outperform matched-budget random selection—is supported by three-seed averages with no variance, no error bars, and no significance test. The paper itself states: 'we use these means to reduce dependence on a single run, but do not claim statistical significance with n=3.' Since the margins in Table II (e.g., Cube transfer All nuis. 0.96 vs 0.56 for random K=20; Cube stacking 0.76 vs 0.60 for random K=30) could plausibly change under typical seed variance for ACT fine-tuning, the quantitative strength of the central claim is not yet established. Please provide per-seed results, confidence intervals, or a larger number of seeds, or explicitly downgrade the claim to a preliminary finding.
  2. [§V-D and Appendix D (Table IV)] The headline claim that selected repair sets 'approach the performance of much larger random repair budgets' is only true on the seen 22-condition evaluation. On held-out nuisance instantiations, Table IV shows a large gap: for Cube transfer, response-guided K=20 attains held-out mean 0.57 vs 0.90 for random K=500; for Cube stacking, held-out worst is 0.02 vs 0.28. The paper acknowledges this in §V-D ('CFNBC is only as good as the candidate response set' and 'held-out transfer is partial'), but the abstract and Section I state the 'approaching larger random budgets' conclusion without this caveat. The central claim should be narrowed to candidate-covered nuisance conditions, or additional evidence is needed that transfer improves with response-guided selection beyond matched budgets.
  3. [§III-A, Eq. (1), and Appendix B] The load-bearing premise is that all generated nuisances are task-preserving, i.e., a*(s_c)=a*(s_n) for every paired observation. Appendix B asserts this by construction, but for local support changes (a cloth patch under the manipulated objects) there is a real risk that contact geometry or friction changes the feasible or intended action, especially in the bimanual cube transfer task. The paper does not validate this assumption, e.g., by checking that the expert action succeeds under the nuisance condition or that the task state remains unchanged. If any candidate intervention changes the intended action without detection, action drift conflates nuisance sensitivity with task-relevant change and the repair set is mislabelled. Please add an explicit validation protocol or at least an ablation showing results are robust to filtering out candidates with large physical-side-effect risk.
  4. [§III-C and Appendix A] The selection objective in Eq. (4) uses response features and an RBF affinity with free parameters (lambda_cf, kernel bandwidth sigma), and the reported drift-weighted score additionally multiplies by normalized drift. The greedy selection procedure is clear, but the choice of summary statistics (mean/std/mean-abs/max-abs) and the median-distance sigma are presented as implementation details without sensitivity analysis. Since the entire method is an offline selection rule, it is important to know how robust the selection is to these choices. A sensitivity study (e.g., varying sigma and the response-feature specification) would strengthen the claim that the gains come from response-guided coverage rather than from incidental properties of the particular affinity function.
minor comments (4)
  1. [Figure 3 and §V-B] The Spearman correlations are computed over nuisance conditions that are also used to build the candidate response set and to evaluate the main repair results. Please clarify the relationship between the plotted conditions and the candidate pool, and state whether the correlation includes all 22 conditions or a subset. This does not invalidate the signal, but it affects how 'offline' and 'prediction' should be interpreted.
  2. [Abstract and Section I] The phrase 'without requiring rollout success labels or online policy execution' is accurate for the selection stage, but the candidate pool is still generated in a simulator with a task-preserving nuisance generator. Consider adding a sentence clarifying that the method is offline with respect to policy rollouts, not necessarily with respect to simulation access.
  3. [Table II] The 'Gain vs nominal' column for Random (high-budget) 500 in Cube transfer reports 0.70, while All nuis. is 1.00 and nominal is 0.30. This is consistent, but the column label could be misread as the gain of the high-budget random method over the low-budget random method. Please rename to 'Gain vs nominal (All nuis.)'.
  4. [Appendix D, Table IV] The 'Held-out gap' column is defined as the difference between seen and held-out all-nuisance performance, but the column header does not make clear which direction is positive. Please add a footnote or caption explaining the sign convention.

Circularity Check

0 steps flagged

No significant circularity found: action drift is an independent policy-output signal, selection does not use success labels, and no load-bearing step reduces to its own inputs.

full rationale

I examined the derivation chain from Eq. (1) (task-preserving counterfactuals), through action drift Eq. (2), response-guided selection Eq. (4), and fine-tuning Eq. (6). No step reduces by construction to its inputs. Action drift is computed from the nominal policy's predicted actions on paired clean/nuisance inputs and is never fitted to rollout success labels; the Spearman correlations in Fig. 3 and the held-out results in Table IV are separate empirical validations. Eq. (4) maximizes response-feature coverage, not observed success, and the evaluation is through independent rollouts. The main evaluation shares the 22-condition nuisance space with the candidate pool, which makes the 'seen-condition' results partly self-referential in a broad sense, but this is an experimental scope concern, not a definitional or fitted-input circularity; the held-out results explicitly mitigate it. The paper also disclaims statistical significance with n=3, which is a robustness concern, not circularity. No load-bearing self-citations, imported uniqueness theorems, or ansatz-by-citation are present. The paper explicitly acknowledges that repair is bounded by candidate-set coverage, which is a stated limitation rather than a circular step. Therefore the derivation is self-contained with respect to circularity.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

The central claim rests mainly on the task-preserving assumption, the drift-failure proxy, candidate coverage, and the heuristic response-feature representation. There are no free physical parameters; the two listed hyperparameters are hand-set or heuristic. The paper does not introduce new physical entities.

free parameters (2)
  • lambda_cf (counterfactual loss weight) = 0.5
    Chosen after preliminary sweeps in the 0.5–0.75 range; balances nuisance repair with clean performance retention. Hand-set, not fit to test success, but a method hyperparameter that affects results.
  • RBF kernel bandwidth sigma = median non-zero pairwise distance in candidate pool
    Heuristic choice for response affinity in Eq. (4); affects which candidates are considered similar and hence selected.
axioms (4)
  • domain assumption Task-preserving counterfactual assumption: a*(s_c)=a*(s_n) in Eq. (1), so inherited expert actions remain valid under nuisances.
    The entire CFNBC pipeline depends on this. Appendix B states nuisances are manually constrained to preserve task state and expert action; if violated, drift is not a nuisance-only signal.
  • domain assumption Local action drift in normalized action space is a usable proxy for closed-loop rollout failure.
    Empirically supported by Fig. 3, but the authors acknowledge local drift may miss delayed failures or overstate recoverable drift; it is not a guarantee.
  • domain assumption Candidate response set covers deployment nuisance response modes.
    Stated in Section V-D: if deployment contains a response mode absent from C, the selector cannot identify or repair it. The headline results are therefore conditional on candidate coverage.
  • ad hoc to paper Response features (mean, std, mean-abs, max-abs over timesteps/action dims) with RBF affinity capture semantically similar policy response modes.
    This operationalization is not derived from theory; it is an engineering choice that defines what 'diverse response modes' means.

pith-pipeline@v1.3.0-daily-deepseek · 12264 in / 9715 out tokens · 97129 ms · 2026-08-01T14:21:49.080379+00:00 · methodology

0 comments
read the original abstract

Visuomotor imitation learning has demonstrated success for manipulation tasks. However, the trained policies remain brittle to visual `nuisances', with even minor task-preserving variations such as lighting, distractions or changes in colour result in heavy degradation of the trained policy's performance. While increasing data diversity can improve robustness, it is unclear which additional demonstrations are informative for a particular trained policy. We propose Counterfactual Nuisance Behaviour Cloning (CFNBC), an offline data-selection framework for targeted robustness repair. Starting from a nominal policy trained on `clean' demonstrations, CFNBC generates paired clean and nuisance observations that preserve the expert action, then measures \emph{action drift}: the change in the policy's predicted action under a nuisance that should not alter the desired behaviour. This provides a policy-specific sensitivity signal for selecting a compact, response-diverse repair set from a larger candidate pool, without requiring rollout success labels or online policy execution. We show in MuJoCo bimanual cube transfer and SimplerEnv cube stacking that action drift correlates with nuisance-induced failure, and that response-guided repair with only $20$--$30$ selected candidates substantially outperforms matched-budget random selection while approaching the performance of much larger random repair budgets. These results support a data-centric view of robustness repair: the most useful data are not necessarily the most numerous, visually diverse, or obviously difficult, but the examples that cover fragile response modes of the current policy.

Figures

Figures reproduced from arXiv: 2607.27261 by Brendan Tidd, Giovanni D'urso, Kaushik Roy, Nicholas Lawrance.

Figure 1
Figure 1. Figure 1: The Counterfactual Nuisance Behaviour Cloning (CFNBC) Pipeline. (A) A nominal visuomotor policy can succeed in the clean domain but fail under task-preserving visual nuisances. We measure action drift: the change in the policy’s predicted action between paired clean and nuisance observations. (B) Response-guided repair selection chooses a compact repair set that covers diverse high-drift response modes, ra… view at source ↗
Figure 2
Figure 2. Figure 2: Representative task-preserving nuisance conditions. We show one example task for visual clarity: clean, appearance [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Action drift identifies fragile nuisance conditions. Each point [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Response-guided repair improves data efficiency. Each point [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Narrow repair data repairs narrow failure modes. Cell colour shows the change in success rate relative to the nominal A200 SimplerEnv [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Single-factor nuisance grid for MuJoCo bimanual cube transfer. The same nuisance families are instantiated for the cube-transfer task [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Multi-factor combination nuisance conditions for MuJoCo bimanual cube transfer. These predefined combinations test robustness [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Single-factor nuisance grid for SimplerEnv cube stacking. The grid shows the appearance, lighting and shadow, distractor proximity, [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Multi-factor combination nuisance conditions for SimplerEnv cube stacking. These conditions combine multiple task-preserving visual [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Representative held-out nuisance conditions for the two evaluation tasks. The top row shows MuJoCo cube-transfer held-out [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗

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

24 extracted references · 8 linked inside Pith

  1. [1]

    Solving rubik’s cube with a robot hand.arXiv preprint arXiv:1910.07113, 2019

    Ilge Akkaya, Marcin Andrychowicz, Maciek Chociej, Mateusz Litwin, Bob McGrew, Arthur Petron, Alex Paino, Matthias Plappert, Glenn Powell, Raphael Ribas, et al. Solving rubik’s cube with a robot hand.arXiv preprint arXiv:1910.07113, 2019

    Pith/arXiv arXiv 1910

  2. [2]

    Rocoda: Counterfactual data augmen- tation for data-efficient robot learning from demon- strations

    Ezra Ameperosa, Jeremy A Collins, Mrinal Jain, and Animesh Garg. Rocoda: Counterfactual data augmen- tation for data-efficient robot learning from demon- strations. In2025 IEEE International Conference on Robotics and Automation (ICRA), pages 13250–13256. IEEE, 2025

    2025

  3. [3]

    Causal confusion in imitation learning.Advances in neural information processing systems, 32, 2019

    Pim De Haan, Dinesh Jayaraman, and Sergey Levine. Causal confusion in imitation learning.Advances in neural information processing systems, 32, 2019

    2019

  4. [4]

    Robust visual sim-to-real transfer for robotic manipulation

    Ricardo Garcia, Robin Strudel, Shizhe Chen, Etienne Arlaud, Ivan Laptev, and Cordelia Schmid. Robust visual sim-to-real transfer for robotic manipulation. In2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 992–999. ieee, 2023

    2023

  5. [5]

    Thriftydagger: Budget-aware novelty and risk gating for interactive imitation learning.arXiv preprint arXiv:2109.08273, 2021

    Ryan Hoque, Ashwin Balakrishna, Ellen Novoseller, Albert Wilcox, Daniel S Brown, and Ken Goldberg. Thriftydagger: Budget-aware novelty and risk gating for interactive imitation learning.arXiv preprint arXiv:2109.08273, 2021

    Pith/arXiv arXiv 2021

  6. [6]

    Hg-dagger: Inter- active imitation learning with human experts

    Michael Kelly, Chelsea Sidrane, Katherine Driggs- Campbell, and Mykel J Kochenderfer. Hg-dagger: Inter- active imitation learning with human experts. In2019 International Conference on Robotics and Automation (ICRA), pages 8077–8083. IEEE, 2019

    2019

  7. [7]

    Evaluating real-world robot manipulation policies in simulation.arXiv preprint arXiv:2405.05941, 2024

    Xuanlin Li, Kyle Hsu, Jiayuan Gu, Karl Pertsch, Oier Mees, Homer Rich Walke, Chuyuan Fu, Ishikaa Lunawat, Isabel Sieh, Sean Kirmani, et al. Evaluating real-world robot manipulation policies in simulation.arXiv preprint arXiv:2405.05941, 2024

    Pith/arXiv arXiv 2024

  8. [8]

    Active domain randomiza- tion

    Bhairav Mehta, Manfred Diaz, Florian Golemo, Christo- pher J Pal, and Liam Paull. Active domain randomiza- tion. InConference on Robot Learning, pages 1162–

  9. [9]

    R3m: A universal visual representation for robot manipulation.arXiv preprint arXiv:2203.12601, 2022

    Suraj Nair, Aravind Rajeswaran, Vikash Kumar, Chelsea Finn, and Abhinav Gupta. R3m: A universal visual representation for robot manipulation.arXiv preprint arXiv:2203.12601, 2022

    Pith/arXiv arXiv 2022

  10. [10]

    Rep- resentation learning with contrastive predictive coding

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Rep- resentation learning with contrastive predictive coding. arXiv preprint arXiv:1807.03748, 2018

    Pith/arXiv arXiv 2018

  11. [11]

    An algo- rithmic perspective on imitation learning.Foundations and Trends® in Robotics, 7(1-2):1–179, 2018

    Takayuki Osa, Joni Pajarinen, Gerhard Neumann, J An- drew Bagnell, Pieter Abbeel, and Jan Peters. An algo- rithmic perspective on imitation learning.Foundations and Trends® in Robotics, 7(1-2):1–179, 2018

    2018

  12. [12]

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

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

    2024

  13. [13]

    The colosseum: A benchmark for evaluating generalization for robotic manipulation.arXiv preprint arXiv:2402.08191, 2024

    Wilbert Pumacay, Ishika Singh, Jiafei Duan, Ranjay Kr- ishna, Jesse Thomason, and Dieter Fox. The colosseum: A benchmark for evaluating generalization for robotic manipulation.arXiv preprint arXiv:2402.08191, 2024

    Pith/arXiv arXiv 2024

  14. [14]

    Un- supervised feature learning for manipulation with con- trastive domain randomization

    Carmel Rabinovitz, Niko Grupen, and Aviv Tamar. Un- supervised feature learning for manipulation with con- trastive domain randomization. In2021 IEEE Interna- tional Conference on Robotics and Automation (ICRA), pages 10153–10159. IEEE, 2021

    2021

  15. [15]

    A reduction of imitation learning and structured prediction to no-regret online learning

    St ´ephane Ross, Geoffrey Gordon, and Drew Bagnell. A reduction of imitation learning and structured prediction to no-regret online learning. InProceedings of the fourteenth international conference on artificial intelli- gence and statistics, pages 627–635. JMLR Workshop and Conference Proceedings, 2011

    2011

  16. [16]

    Realm: A real-to-sim validated benchmark for generalization in robotic manipulation.IEEE Robotics and Automation Letters, 2026

    Martin Sedlacek, Pavlo Yefanov, Georgy Ponimatkin, Jai Bardhan, Simon Pilc, Mederic Fourmy, Evangelos Kazakos, Cees GM Snoek, Josef Sivic, and Vladimir Petrik. Realm: A real-to-sim validated benchmark for generalization in robotic manipulation.IEEE Robotics and Automation Letters, 2026

    2026

  17. [17]

    Time-contrastive networks: Self- supervised learning from video

    Pierre Sermanet, Corey Lynch, Yevgen Chebotar, Jas- mine Hsu, Eric Jang, Stefan Schaal, Sergey Levine, and Google Brain. Time-contrastive networks: Self- supervised learning from video. In2018 IEEE inter- national conference on robotics and automation (ICRA), pages 1134–1141. IEEE, 2018

    2018

  18. [18]

    Domain ran- domization for transferring deep neural networks from simulation to the real world

    Josh Tobin, Rachel Fong, Alex Ray, Jonas Schneider, Wojciech Zaremba, and Pieter Abbeel. Domain ran- domization for transferring deep neural networks from simulation to the real world. In2017 IEEE/RSJ in- ternational conference on intelligent robots and systems (IROS), pages 23–30. IEEE, 2017

    2017

  19. [19]

    Mujoco: A physics engine for model-based control

    Emanuel Todorov, Tom Erez, and Yuval Tassa. Mujoco: A physics engine for model-based control. In2012 IEEE/RSJ international conference on intelligent robots and systems, pages 5026–5033. IEEE, 2012

    2012

  20. [20]

    Chang, Leonidas J

    Fanbo Xiang, Yuzhe Qin, Kaichun Mo, Yikuan Xia, Hao Zhu, Fangchen Liu, Minghua Liu, Hanxiao Jiang, Yifu Yuan, He Wang, Li Yi, Angel X. Chang, Leonidas J. Guibas, and Hao Su. Sapien: A simulated part-based interactive environment. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 11097–11107, June 2020

    2020

  21. [21]

    Decomposing the generalization gap in imitation learning for visual robotic manipulation

    Annie Xie, Lisa Lee, Ted Xiao, and Chelsea Finn. Decomposing the generalization gap in imitation learning for visual robotic manipulation. In2024 IEEE Interna- tional Conference on Robotics and Automation (ICRA), pages 3153–3160. IEEE, 2024

    2024

  22. [22]

    Barlow twins: Self-supervised learning via redundancy reduction

    Jure Zbontar, Li Jing, Ishan Misra, Yann LeCun, and St´ephane Deny. Barlow twins: Self-supervised learning via redundancy reduction. InInternational conference on machine learning, pages 12310–12320. PMLR, 2021

    2021

  23. [23]

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

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

    Pith/arXiv arXiv 2023

  24. [24]

    Aloha unleashed: A simple recipe for robot dexterity.arXiv preprint arXiv:2410.13126, 2024

    Tony Z Zhao, Jonathan Tompson, Danny Driess, Pete Florence, Kamyar Ghasemipour, Chelsea Finn, and Ayzaan Wahid. Aloha unleashed: A simple recipe for robot dexterity.arXiv preprint arXiv:2410.13126, 2024. APPENDIXA CFNBC ALGORITHM ANDIMPLEMENTATIONDETAILS This appendix provides the full CFNBC procedure used in the experiments, including candidate response ...

    Pith/arXiv arXiv 2024