Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Improving Generalization Ability of Robotic Imitation Learning by Resolving Causal Confusion in Observations

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read An imitation-learning policy can learn a binary mask over a convolutional image embedding that selects only the causal parents of the action, and this suffices to make it robust to domain shift without disentangled features.

desk verdict Promising empirical result on one simulated task, but the theory overclaims: unique solvability is not identifiability of the causal graph, so the central 'disentanglement not necessary' claim is not established. read the letter →

arxiv 2507.22380 v1 pith:H5GFITZR submitted 2025-07-30 cs.RO cs.LG

Yifei Chen , Yuzhe Zhang , Giovanni D'urso , Nicholas Lawrance , Brendan Tidd This is my paper
classification cs.ROcs.LG
keywords causalconfusionimitationlearninggeneralizationstructureactionchunkingtransformerout-of-distributionrobustnessroboticmanipulationstructuralmodel
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Imitation learning policies can fail under even mild domain shifts because they learn spurious correlations with task-irrelevant objects. This paper argues that the fix is to learn which parts of the observation causally drive the expert's action, and it shows this can be done directly on a standard convolutional image embedding, without the disentangled representations that earlier causal-confusion methods demanded. The accompanying method, Causal-ACT, trains an Action Chunking Transformer while randomly masking the image features, then searches for the mask that maximizes task reward. In a simulated bimanual cube-transfer task, it raises out-of-distribution transfer success from 0.23 to 0.88, on par with the best domain-randomization baseline but without extra demonstrations or environment sampling. If the claim holds, causal masking is a data-efficient route to generalizable manipulation.

What carries the argument

The load-bearing theoretical object is unique solvability of a structural causal model, defined through the criterion in [4]: an SCM is uniquely solvable with respect to a single variable if and only if that variable has no self-cycle. In an imitation policy causal graph, the action at each time step has in-degrees only from a subset of observations, so no edge $A_t \to A_t$ exists; the proof of Proposition 1 rests on this single condition. The load-bearing algorithmic object is the binary causal mask, an array matching the ResNet embedding shape that is multiplied element-wise with the image features, sampled uniformly during training, and searched after training by fitting a linear energy-based model to episodic rewards. The mask is the mechanism by which the policy switches off observation dimensions that do not cause the action, and its ablation defines the method's behavior: random masks give 0.48 transfer success, fully connected masks give 0.02.

What would settle it

Train Causal-ACT with a distractor that shares the target cube's color and texture so that distractor and target project onto the same embedding units, then remove the distractor at test time while moving the target well outside the 20 cm × 20 cm training range. If the fixed mask still gives roughly 0.88 transfer success, it is selecting causal features; a sharp drop would show that the binary-mask assumption or the fixed test-time mask cannot separate the relevant from the irrelevant information.

Watch

Extended reading notes

Core claim

The paper's central claim is that the causal structure of an imitation policy can be learned without disentangling the observation space. Formally, Proposition 1 states that the structural causal model corresponding to an imitation policy causal graph is uniquely solvable with respect to the action variable $A_t$: a mapping learned from the action's causal parents to the action identifies the true structural function even when observation dimensions have causal connections and cycles among themselves. The proof uses the fact that $A_t$ has in-degrees only from a subset of $X_t$ and therefore has no self-cycle, which is exactly the condition under which unique solvability holds. The practical embodiment is Causal-ACT, which multiplies ResNet image features by a binary causal mask, trains the policy under randomly sampled masks, and performs a reward-guided intervention search to fix the mask at test time. The experiments report that this restores out-of-distribution performance, raising transfer success from 0.23 to 0.88 compared with the ACT baseline.

Load-bearing premise

The method assumes that the task-relevant information in the convolutional embedding can be isolated by switching individual embedding units on and off independently; if causal features are entangled across units or require the joint activation of several units, no binary mask over the raw embedding can represent the true parent set, and the mask found by reward search would only overfit to the training environment.

Editorial extensions

If this is right

  • Out-of-distribution transfer success on the ALOHA cube-transfer task rises from 0.23 (ACT) to 0.88 (Causal-ACT), without additional expert demonstrations.
  • With a fixed budget of demonstrations, Causal-ACT matches the best domain-randomization baseline (k=0, 0.91) while avoiding the need to sample or design training domains.
  • Because the causal graph is learned on top of a convolutional encoder, the same component can be dropped into other imitation learning architectures that use image observations.
  • Ablations show the learned mask matters: a random mask drops transfer success to 0.48, and a fully connected mask (all features passed) collapses it to 0.02.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The theoretical result guarantees unique solvability of the action variable, not that the reward-guided mask search recovers the true parent set; identifying the mask remains an empirical heuristic whose identifiability is unproven.
  • If task-relevant and irrelevant information occupy the same embedding unit (for example, a distractor sharing the target's color and texture), no binary mask can separate them; a soft or grouped mask over channels would be a natural extension.
  • The fixed mask at test time could encode spatial location rather than causal features; evaluating on target positions well outside the training sampling range would reveal whether the mask tracks the cube's causal features or memorizes a region.
  • The intervention search scales exponentially with embedding dimension, so scaling to higher-resolution embeddings will require learned graph samplers rather than the linear energy-based search used here.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes Causal-ACT, an extension of the Action Chunking Transformer that learns a binary mask over image-embedding dimensions to select observation features that are causal parents of the action, with the mask optimized by intervention-based search after training. The authors argue that prior causal-confusion work requiring disentangled representations is unnecessarily restrictive, and they provide Proposition 1 to support the claim that disentanglement is not needed. Experiments in a simulated ALOHA cube-transfer task with distractor cubes report improved out-of-distribution transfer over plain ACT and competitive performance with domain randomization, plus an ablation comparing learned, random, and full-connection masks.

Significance. If the central theoretical claim were established, the paper would make a useful contribution: it would show that causal structure learning for imitation can operate on entangled representations, removing a major practical obstacle. The proposed integration with ACT is simple and the empirical direction—targeting causal confusion rather than data augmentation—is relevant to the robotics imitation-learning community. The paper also provides code and compares against domain randomization, which is a constructive baseline. However, as written, the theoretical support does not deliver the advertised conclusion, and the empirical evidence is too narrow and under-reported to carry the weight of the claim on its own.

major comments (4)
  1. [Section 4.2, Proposition 1 and Appendix A] Proposition 1 does not establish the paper's central claim that disentanglement is unnecessary. The proposition, as proved, only states that the SCM corresponding to an imitation policy causal graph is uniquely solvable with respect to {A_t}; the proof invokes Lemma 1 after observing that A_t has no self-cycle. This is a property of the graph structure, not a property of any estimator or search procedure. The text immediately following Proposition 1 claims that 'if we properly estimate a mapping g_{A_t}, then this mapping uniquely indicates the underlying structural function, i.e., g_{A_t} ⇔ f_{A_t}.' That inference does not follow from unique solvability: unique solvability of a known SCM does not imply that the parent set pa_G(A_t) is identifiable from finite demonstrations, nor that the binary-mask search in Algorithm 1 recovers it. The abstract's assertion that the disentangled representation requirement is 'not necessary in causal relationship learning' is therefore unsupported by the proof. This is a load-bearing issue because the theoretical contribution is explicitly listed as a main contribution.
  2. [Section 4.1 and Algorithm 1] The method assumes that causal parents of the action can be represented by binary masks over individual dimensions of the ResNet image embedding. Section 4.1 defines the causal graph over 'observation dimensions' and allows entanglement, but Algorithm 1 samples masks over the embedding x_t produced by a convolutional encoder. If the true causal features are entangled across channels or spatial positions, no binary mask over the raw embedding can represent the parent set, and the reward-based search may select a mask that merely fits the evaluation environment. The paper does not provide evidence that the embedding dimensions correspond to independent causal features, nor does it test the method on a setting with known ground-truth masks. This assumption is central to the method and needs either justification or a synthetic validation.
  3. [Section 6 and Table 1] The empirical support is limited to a single simulated task, three seeds, and 50 evaluation episodes, with no standard deviations, confidence intervals, or per-seed results reported. The headline OOD transfer numbers are 0.23 for ACT and 0.82 for Causal-ACT in Table 1, yet the text in Section 7 states that Causal-ACT improves OOD performance 'from (0.23)' to '(0.88)'. This inconsistency between the table and the text must be resolved, and the absence of variance reporting makes it impossible to assess whether the improvement is reliable. Given that the paper's central empirical claim rests on this comparison, the reporting needs to be substantially strengthened.
  4. [Section 7, Table 2, ablation study] The full-connection-graph ablation is not interpretable as evidence about causal structure learning. The authors themselves conjecture that the low score (0.02) is due to 'the lack of training in this regime,' since the full mask is never sampled during training. If the policy is never trained with the full mask, then the low performance reflects a train/test mismatch in the mask distribution, not a property of the learned causal structure. The random-graph baseline (0.48) also does not clearly support the claim that the graph-search process is effective, since no error bars are provided and the gap to Causal-ACT (0.82) could be within seed variation. The ablation should be redesigned: for example, train with the full mask included in the sampled distribution, and report variance across seeds.
minor comments (5)
  1. [Section 7, text after Table 1] The reported OOD transfer success rate for Causal-ACT in the text is 0.88, but Table 1 lists 0.82; please correct the inconsistency.
  2. [Section 2, first paragraph] The phrase 'improve generalization2' in the contributions list appears to contain a stray footnote mark; the footnote is not placed at the correct location.
  3. [Section 4.1, paragraph after Eq. (2)] There is a typographical error: 'A=f V (paG(V), ϵV )' should be 'V=f V (paG(V), ϵV )' or the surrounding text should be adjusted for consistency.
  4. [Section 5, Algorithm 1] The notation D is reused: line 4 initializes D=∅ for the intervention dataset, but D is also used for the expert demonstration dataset in the input line. Please use distinct names for clarity.
  5. [Appendix B, Table 3] The table is described as showing hyperparameters, but the 'intervention iteration' row lists only the value 50 without a label for what it controls; adding a brief explanation would help reproducibility.

Circularity Check

2 steps flagged · score 7.0 of 10

Proposition 1 restates the defining no-self-cycle property of the IL graph rather than proving identifiability; the reported OOD gain is produced by the same episodic-reward signal used to select the causal mask.

  1. self definitional [Section 4.2, Proposition 1 and Appendix A proof]
    "Formally, we would like the imitation policy causal graph to satisfy certain conditions, such that if we properly estimate a mapping gAt :pa G(At)→A t, then this mapping uniquely indicates the underlying structural function, i.e., gAt ⇐ ⇒fAt ... Proposition 1. A SCM M=⟨V,U,f,P ξ⟩ corresponding to a imitation policy causal graph is uniquely solvable w.r.t.{A t}."

    In Section 4.1 an imitation policy causal graph is defined so that 'At has only in-degrees from a subset of Xt.' Lemma 1 [4] reduces unique solvability w.r.t. {At} to 'V has no self cycle'. The proof then simply reads this defining property off the definition: 'the causal graph satisfies the condition required in Lemma 1.' So Proposition 1 is true by construction, not a proof that pa_G(At) is identifiable from demonstrations. The subsequent claim that a properly estimated mapping gAt uniquely indicates fAt is an added assertion: unique solvability only guarantees a reduced-form solution for a known graph.

  2. fitted input called prediction [Section 5, Algorithm 1 (lines 14-19); Section 7, Table 1 narrative]
    "After training, we perform the targeted intervention according to Algorithm 1 to search for the graph that performs the best (denoted as g∗) according to the episodic reward. ... return g ∗ = arg max g p(g), h θ1, and π θ2 ... Causal-ACT method maintains strong performance in-distribution (0.96) and substantially improves out-of-distribution performance (0.88) up from (0.23)."

    Algorithm 1 selects g* by computing 'reward Rg by executing π...' and fitting ω via linear regression on (g, R_g), then returning argmax_g p(g). This makes the causal mask a reward-fitted parameter, not the output of a causal-identification procedure from demonstrations. The headline OOD result (0.88 transfer) is then reported for the same selection procedure. Since the paper does not state that the intervention rollouts take place in a held-out environment distinct from the OOD test environment, the reported success is a measure of the reward signal used to choose g*—i.e., a fitted input presented as a predicted causal-structure outcome.

full rationale

Proposition 1 is a valid but trivial application of an external lemma (Bongers et al. [4]); the circularity lies in presenting it as a theoretical justification for identifiability without disentanglement, when unique solvability is entailed by the definition of the IL graph and does not address estimation of pa_G(At). The empirical component has independent content—training ACT with random masks and reward-guided search—so this is not a fully self-referential paper. However, the stated contribution 'theoretically clarify that this requirement is not necessary' reduces to a definitional restatement, and the top-line OOD gain is obtained via episodic-reward selection of the mask. Score 7 reflects partial but substantive circularity: the central theoretical claim is definitional, and the main empirical claim is a reward-fitted selection rather than an independent prediction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central contribution rests on the assumed causal graph structure of the imitation policy, on the stipulation of independent exogenous noise, and on the assumption that binary masks over ResNet features can represent the true parent set. The selected graph g* and the energy-model weight omega are fitted quantities. No new physical entities are introduced.

free parameters (3)
  • causal graph mask g* = binary vector over image feature dimensions, selected by search over 2^dim(x) masks
    Algorithm 1 surveys many masks and uses an energy-based model over episodic rewards to pick the deployed mask; this is the central fitted structure of the method.
  • energy model weight omega = estimated by linear regression on collected (g, R_g) pairs
    Algorithm 1 line 18 fits omega to rank causal graphs, and g* depends on this fit.
  • intervention iterations N = 50
    The number of policy rollouts used to select the graph is chosen by hand in Appendix B and controls how thoroughly the graph space is searched.
assumptions (4)
  • domain assumption A structural causal model with mutually independent exogenous noise terms exists for the imitation policy at each time step.
    Section 4.1, Eq. (2), assumes A_t = f_V(pa_G(V), epsilon_V) with independent noise; this underpins Proposition 1 and is not validated on the actual ALOHA data.
  • domain assumption A_t has incoming edges only from a subset of X_t and no edge from hidden exogenous variables U.
    Section 4.1 and Appendix A: this no-edge structure is what makes Lemma 1 applicable; it is stipulated, not learned from data.
  • ad hoc to paper Binary masks over ResNet embedding dimensions can represent the causal parent set of the action.
    Algorithm 1 samples g over the image feature array, but the paper provides no identifiability or approximation argument for this encoding.
  • standard math Unique solvability criterion of Bongers et al. (2021).
    Used in Appendix A as Lemma 1; the paper cites the source but does not prove it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Improving Generalization Ability of Robotic Imitation Learning by Resolving Causal Confusion in Observations." pith.science (2026). https://pith.science/paper/H5GFITZR

@misc{pith2026250722380,
  author       = {Pith},
  title        = {Pith review of: Improving Generalization Ability of Robotic Imitation Learning by Resolving Causal Confusion in Observations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H5GFITZR}},
  note         = {Machine review of arXiv:2507.22380}
}
read the original abstract

Recent developments in imitation learning have considerably advanced robotic manipulation. However, current techniques in imitation learning can suffer from poor generalization, limiting performance even under relatively minor domain shifts. In this work, we aim to enhance the generalization capabilities of complex imitation learning algorithms to handle unpredictable changes from the training environments to deployment environments. To avoid confusion caused by observations that are not relevant to the target task, we propose to explicitly learn the causal relationship between observation components and expert actions, employing a framework similar to [6], where a causal structural function is learned by intervention on the imitation learning policy. Disentangling the feature representation from image input as in [6] is hard to satisfy in complex imitation learning process in robotic manipulation, we theoretically clarify that this requirement is not necessary in causal relationship learning. Therefore, we propose a simple causal structure learning framework that can be easily embedded in recent imitation learning architectures, such as the Action Chunking Transformer [31]. We demonstrate our approach using a simulation of the ALOHA [31] bimanual robot arms in Mujoco, and show that the method can considerably mitigate the generalization problem of existing complex imitation learning algorithms.

Figures

Figures reproduced from arXiv: 2507.22380 by the authors.

Figure 1
Figure 1. A visualization of the training (left) and the testing (right) environment. Note that the red [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. An example causal graph of an imitation policy. At each time step [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Model Architecture of Causal Structure Learned ACT (Causal-ACT). During training, a causal graph is uniformly sampled to modulate the image features. At test time, the causal graph is fixed to the best-performing graph which is decided by the targeted intervention process. As in ACT, the style variable z is learned by a transformer encoder and set to zero at test time. Proposition 1. A SCM M = ⟨V, U,f, Pξ⟩ correspon… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: The distribution used to sample the number of cubes in domain randomization. [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Shortcut Learning in Generalist Robot Policies: The Role of Dataset Diversity and Fragmentation

    cs.RO 2025-08 conditional novelty 6.0 of 10

    Low within-subdataset diversity and large between-subdataset differences cause shortcut learning in generalist robot policies, and targeted augmentation can mitigate it.

  2. SPARSE Data, Rich Results: Few-Shot Semi-Supervised Learning via Class-Conditioned Image Translation

    cs.CV 2025-08 unverdicted novelty 5.0 of 10

    A GAN framework that translates unlabeled medical images between classes and fuses ensemble, time-averaged pseudo-labels outperforms six prior GAN semi-supervised methods on MedMNIST at 5-50 labels per class.

Reference graph

Works this paper leans on

31 extracted references · 16 canonical work pages · cited by 2 Pith papers

  1. [1]

    Sim-to-real transfer for vision-and-language navigation

    Peter Anderson, Ayush Shrivastava, Joanne Truong, Arjun Majumdar, Devi Parikh, Dhruv Batra, and Stefan Lee. Sim-to-real transfer for vision-and-language navigation. InConference on Robot Learning, pages 671–681. PMLR, 2021

    work page 2021

  2. [2]

    Representation learning: A review and new perspectives.IEEE transactions on pattern analysis and machine intelligence, 35(8): 1798–1828, 2013

    Yoshua Bengio, Aaron Courville, and Pascal Vincent. Representation learning: A review and new perspectives.IEEE transactions on pattern analysis and machine intelligence, 35(8): 1798–1828, 2013

    2013

  3. [3]

    π0: A vision-language-action flow model for general robot control.URL https://arxiv

    Kevin Black, Noah Brown, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, Lachy Groom, Karol Hausman, Brian Ichter, et al. π0: A vision-language-action flow model for general robot control.URL https://arxiv. org/abs/2410.24164, 2024

    arXiv

  4. [4]

    Foundations of structural causal models with cycles and latent variables.The Annals of Statistics, 49(5):2885–2915, 2021

    Stephan Bongers, Patrick Forré, Jonas Peters, and Joris M Mooij. Foundations of structural causal models with cycles and latent variables.The Annals of Statistics, 49(5):2885–2915, 2021

    2021

  5. [5]

    RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control, 2023

    Anthony Brohan, Noah Brown, Justice Carbajal, Yevgen Chebotar, Xi Chen, Krzysztof Choro- manski, Tianli Ding, Danny Driess, Avinava Dubey, Chelsea Finn, Pete Florence, Chuyuan Fu, Montse Gonzalez Arenas, Keerthana Gopalakrishnan, Kehang Han, Karol Hausman, Alexander Herzog, Jasmine Hsu, Brian Ichter, Alex Irpan, Nikhil Joshi, Ryan Julian, Dmitry Kalashnik...

    work page 2023

  6. [6]

    Causal confusion in imitation learning

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

    work page 2019

  7. [7]

    Generative Adversarial Imitation Learning

    Jonathan Ho and Stefano Ermon. Generative Adversarial Imitation Learning. InAdvances in Neural Information Processing Systems, volume 29. Curran Associates, Inc., 2016

    work page 2016

  8. [8]

    On feature learning in the presence of spurious correlations.Advances in Neural Information Processing Systems, 35: 38516–38532, 2022

    Pavel Izmailov, Polina Kirichenko, Nate Gruver, and Andrew G Wilson. On feature learning in the presence of spurious correlations.Advances in Neural Information Processing Systems, 35: 38516–38532, 2022

    work page 2022

Show all 31 references
  1. [9]

    Sim-To-Real via Sim-To- Sim: Data-Efficient Robotic Grasping via Randomized-To-Canonical Adaptation Networks

    Stephen James, Paul Wohlhart, Mrinal Kalakrishnan, Dmitry Kalashnikov, Alex Irpan, Julian Ibarz, Sergey Levine, Raia Hadsell, and Konstantinos Bousmalis. Sim-To-Real via Sim-To- Sim: Data-Efficient Robotic Grasping via Randomized-To-Canonical Adaptation Networks. In Proceeding...

    2019

  2. [10]

    Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  3. [11]

    Learning causally disentangled representations via the principle of independent causal mechanisms

    Aneesh Komanduri, Yongkai Wu, Feng Chen, and Xintao Wu. Learning causally disentangled representations via the principle of independent causal mechanisms. InProceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, pages 4308–4316, 2024

    2024

  4. [12]

    Challenging common assumptions in the unsupervised learning of disentangled representations

    Francesco Locatello, Stefan Bauer, Mario Lucic, Gunnar Raetsch, Sylvain Gelly, Bernhard Schölkopf, and Olivier Bachem. Challenging common assumptions in the unsupervised learning of disentangled representations. Ininternational conference on machine learning, pages 4114–

  5. [13]

    Isaac Gym: High Performance GPU Based Physics Simulation For Robot Learning.Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks, 1, 2021

    Viktor Makoviychuk, Lukasz Wawrzyniak, Yunrong Guo, Michelle Lu, Kier Storey, Miles Macklin, David Hoeller, Nikita Rudin, Arthur Allshire, Ankur Handa, and Gavriel State. Isaac Gym: High Performance GPU Based Physics Simulation For Robot Learning.Proceedings of the Neural Info...

    2021

  6. [14]

    Solving Rubik’s Cube with a Robot Hand, 2019

    OpenAI, Ilge Akkaya, Marcin Andrychowicz, Maciek Chociej, Mateusz Litwin, Bob McGrew, Arthur Petron, Alex Paino, Matthias Plappert, Glenn Powell, Raphael Ribas, Jonas Schneider, Nikolas Tezak, Jerry Tworek, Peter Welinder, Lilian Weng, Qiming Yuan, Wojciech Zaremba, and Lei Zh...

    2019

  7. [15]

    Andrew Bagnell, Pieter Abbeel, and Jan Peters

    Takayuki Osa, Joni Pajarinen, Gerhard Neumann, J. Andrew Bagnell, Pieter Abbeel, and Jan Peters. An Algorithmic Perspective on Imitation Learning.Foundations and Trends® in Robotics, 7(1-2), 2018. ISSN 1935-8253, 1935-8261. doi: 10.1561/2300000053

    2018 doi

  8. [16]

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

    Abby O’Neill, Abdul Rehman, Abhiram Maddukuri, Abhishek 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 Internat...

    2024

  9. [17]

    Cambridge University Press, New York, 2000

    Judea Pearl.Causality: Models, Reasoning and Inference. Cambridge University Press, New York, 2000

    2000

  10. [18]

    The MIT Press, 2017

    Jonas Peters, Dominik Janzing, and Bernhard Schölkopf.Elements of causal inference: founda- tions and learning algorithms. The MIT Press, 2017

    2017

  11. [19]

    Structured Domain Randomization: Bridging the Reality Gap by Context-Aware Synthetic Data

    Aayush Prakash, Shaad Boochoon, Mark Brophy, David Acuna, Eric Cameracci, Gavriel State, Omer Shapira, and Stan Birchfield. Structured Domain Randomization: Bridging the Reality Gap by Context-Aware Synthetic Data. In2019 International Conference on Robotics and Automation (IC...

    2019

  12. [20]

    Behavior Transformers: Cloning k modes with one stone.Advances in Neural Information Processing Systems, 35, 2022

    Nur Muhammad Shafiullah, Zichen Cui, Ariuntuya (Arty) Altanzaya, and Lerrel Pinto. Behavior Transformers: Cloning k modes with one stone.Advances in Neural Information Processing Systems, 35, 2022

    2022

  13. [21]

    Perceiver-Actor: A Multi-Task Transformer for Robotic Manipulation

    Mohit Shridhar, Lucas Manuelli, and Dieter Fox. Perceiver-Actor: A Multi-Task Transformer for Robotic Manipulation. InProceedings of The 6th Conference on Robot Learning. PMLR, 2023

    2023

  14. [22]

    Resnet in resnet: Generalizing residual architectures.arXiv preprint arXiv:1603.08029, 2016

    Sasha Targ, Diogo Almeida, and Kevin Lyman. Resnet in resnet: Generalizing residual architectures.arXiv preprint arXiv:1603.08029, 2016

    2016 arXiv

  15. [23]

    Domain randomization 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 randomization for transferring deep neural networks from simulation to the real world. In2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. doi: ...

    2017

  16. [24]

    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. doi: 10.1109/IROS.2012.6386109

    2012

  17. [25]

    A Survey on Causal Inference.ACM Trans

    Liuyi Yao, Zhixuan Chu, Sheng Li, Yaliang Li, Jing Gao, and Aidong Zhang. A Survey on Causal Inference.ACM Trans. Knowl. Discov. Data, 15(5), 2021. ISSN 1556-4681. doi: 10.1145/3444944

    2021 doi

  18. [26]

    Kebria, Abbas Khosravi, and Saeid Nahavandi

    Maryam Zare, Parham M. Kebria, Abbas Khosravi, and Saeid Nahavandi. A Survey of Imitation Learning: Algorithms, Recent Developments, and Challenges.IEEE Transactions on Cybernetics, 54(12), 2024. ISSN 2168-2275. doi: 10.1109/TCYB.2024.3395626

    2024

  19. [27]

    Transporter Networks: Rearranging the Visual World for Robotic Manipulation

    Andy Zeng, Pete Florence, Jonathan Tompson, Stefan Welker, Jonathan Chien, Maria Attarian, Travis Armstrong, Ivan Krasin, Dan Duong, Vikas Sindhwani, and Johnny Lee. Transporter Networks: Rearranging the Visual World for Robotic Manipulation. InProceedings of the 2020 Conferen...

    2020

  20. [28]

    Invariant causal prediction for block mdps

    Amy Zhang, Clare Lyle, Shagun Sodhani, Angelos Filos, Marta Kwiatkowska, Joelle Pineau, Yarin Gal, and Doina Precup. Invariant causal prediction for block mdps. InInternational Conference on Machine Learning, pages 11214–11224. PMLR, 2020

    2020

  21. [29]

    Learning fine-grained bimanual manipulation with low-cost hardware

    Tony Z Zhao, Vikash Kumar, Sergey Levine, and Chelsea Finn. Learning fine-grained bimanual manipulation with low-cost hardware. InRobotics: Science and Systems XIX, Daegu, Republic of Korea, July 2023. doi: 10.15607/RSS.2023.XIX.016. 11

    2023 doi

  22. [30]

    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

    2024 arXiv

  23. [31]

    Sim-to-real transfer in deep reinforcement learning for robotics: a survey

    Wenshuai Zhao, Jorge Peña Queralta, and Tomi Westerlund. Sim-to-real transfer in deep reinforcement learning for robotics: a survey. In2020 IEEE symposium series on computational intelligence (SSCI), pages 737–744. IEEE, 2020. 12 A Proof of Proposition 1 Proof.To prove Proposi...

    2020

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.