Pith. sign in

REVIEW 4 major objections 6 minor 31 references

Modality Selection and Skill Segmentation via Cross-Modality Attention

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper claims that cross-modality attention in a diffusion policy learns to select the most informative sensor at each timestep and, in doing so, learns attention patterns that segment expert trajectories into primitive actions…

desk verdict A plausible but untested hypothesis: the paper shows attention weights differ across manually labeled skills, but it never actually runs an unsupervised segmentation, so the advertised contribution is not yet demonstrated. read the letter →

arxiv 2504.14573 v1 pith:B2UVDJO7 submitted 2025-04-20 cs.RO cs.AI

Jiawei Jiang , Kei Ota , Devesh K. Jha , Asako Kanezaki This is my paper · ORCID
classification cs.ROcs.AI
keywords cross-modalityattentionmodalityselectionunsupervisedskillsegmentationdiffusionpolicyhierarchicalimitationlearningtactilesensingfurnitureassembly
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

This paper tries to show that a cross-modality attention (CMA) mechanism—transformer attention over image, proprioceptive, and tactile embeddings—does two jobs at once in a long-horizon furniture assembly policy. First, it selects which modalities matter at each timestep when generating actions. Second, after imitation training on expert demonstrations, its attention weights and embeddings carry enough skill structure to split trajectories into primitive actions without human segmentation labels. The payoff, if correct, is a route to hierarchical policies for contact-rich tasks that avoids labor-intensive and ambiguous manual labeling.

What carries the argument

The load-bearing object is the Cross-Modality Attention (CMA) module: a transformer-style self-attention block, here 8 heads and 2 layers, applied to embeddings of camera, grip-camera, proprioceptive, and tactile inputs stacked across two consecutive timesteps. Its softmax attention weights $\mathrm{softmax}(QK^\top/\sqrt{d_k})$ are averaged over heads in the last layer, and these averaged weights are the evidence that primitive actions have distinct attention signatures. The module also produces the conditional embedding that drives the 1D-conditional U-Net diffusion policy, so the same learned attention serves both action generation and primitive discovery.

What would settle it

Record force-torque contact events while the robot performs the one-leg assembly task and use them to define objective action boundaries; if unsupervised clustering of CMA embeddings does not place cluster boundaries at these contact events across most trajectories, the claim that CMA segments primitives would be refuted.

Watch

Extended reading notes

Core claim

Working on the one-leg FurnitureSim assembly task with 50 expert demonstrations, the paper reports three experimental findings. After imitation training, the last-layer attention weights of the CMA module, averaged over heads, show distinct patterns for six manually defined primitives; this is taken as evidence that the mechanism has learned primitive characteristics. Training a separate policy for each primitive for 200,000 steps each, matching the 1,200,000-step total, gives lower validation loss than training one policy on whole trajectories, with the screwing primitive as an exception. T-SNE projections of CMA embeddings show clusters that mostly match the manual labels, but with systematic discrepancies—some Grip-and-Move frames land in Reach Base, and Screw splits into screwing and gripper repositioning—which the authors attribute to ambiguity in human labels and cite as motivation for unsupervised segmentation.

Load-bearing premise

The evaluation assumes the authors' manual division of each trajectory into six primitives (Reach Base, Grip and Move Base, Reach Leg, Grip and Move Leg, Insert, Screw) is a reliable and consistent labeling; if those labels are inconsistent, the observed alignment between attention weights and primitives could reflect labeling artifacts rather than real skill structure.

Editorial extensions

If this is right

  • A hierarchical policy can be trained by clustering CMA embeddings into skills and selecting a skill-specific policy, eliminating the need for human-labeled trajectory segmentation.
  • Training separate policies on individual primitives should require fewer total samples than whole-trajectory training, because each primitive has a more consistent input distribution.
  • The attention weights can be used to drop uninformative modalities for each primitive, cutting input dimensionality during execution.
  • The approach is intended to transfer to other long-horizon, contact-rich assembly tasks where tactile or audio feedback is informative.
  • Manual primitive labels are not necessarily the true segment boundaries; learned clustering can reveal structure that human labels miss, such as the split of the screwing action into screwing and gripper repositioning.

Reading between the lines

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

  • A direct test the authors did not run: use the unsupervised clusters themselves as the primitive definitions, train the hierarchical policy on them, and compare task success against policies trained on manual labels; this would separate the segmentation claim from the label choice.
  • If CMA attention weights are stable across seeds and tasks, the mechanism could serve as a generic sensor router for larger multi-modal policies, letting each timestep route only the modalities the current phase of manipulation needs.
  • Because the paper argues human labels are ambiguous, a stronger evaluation would compare clusters against objective phase annotations such as contact onset, gripper release, and insertion completion rather than the manual six-way split.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes a Cross-Modality Attention (CMA) mechanism integrated into a 1D-conditional U-Net diffusion policy for multi-modal imitation learning in furniture assembly. The stated contributions are (1) a CMA architecture for modality selection, (2) an unsupervised segmentation method for expert trajectories into primitive actions, and (3) improved sample efficiency from training per-primitive policies. Experiments use 50 expert trajectories from FurnitureSim one-leg assembly, manually segmented into six primitives (Reach Base, Grip and Move Base, Reach Leg, Grip and Move Leg, Insert, Screw). Results show attention weight patterns that differ across primitives (Fig. 2), lower validation losses for individual per-primitive policies than for a whole-trajectory policy (Fig. 4), and a t-SNE visualization suggesting clustered embeddings (Fig. 5). The paper's own Section 5 states that unsupervised segmentation and hierarchical policy training are future work.

Significance. If the central claim were substantiated, the ability of a cross-modal attention mechanism to discover primitive action structure without human labels would be a valuable advance for long-horizon manipulation. The paper also presents a concrete architecture and a sensible attention-analysis procedure. However, the evidence as presented is almost entirely qualitative: no unsupervised segmentation algorithm is run, no quantitative segmentation evaluation is reported, and the sample-efficiency comparison rests on validation loss from a single seed with an unequal training setup. The paper's abstract and contributions overstate what the experiments demonstrate. The strengths are the use of a standard benchmark, clear architectural description, and a useful illustration of attention-weight variability across manually defined phases.

major comments (4)
  1. [Section 3 and Section 4.4] The advertised contribution of 'unsupervised segmentation of expert trajectories into primitive actions' is not tested. No segmentation algorithm is implemented or run; Figure 5 is a t-SNE visualization, not a segmentation result, and t-SNE can produce visually separated clusters in high-dimensional noise. Moreover, the clusters are interpreted using the same manual labels that Section 4.4 itself calls ambiguous. Section 5 explicitly lists unsupervised segmentation and hierarchical policy training as future work, confirming that the central mechanism was not evaluated. To support the claim, the authors would need to present an actual segmentation method (e.g., clustering with model selection, change-point detection, or attention-boundary detection) and quantitative boundary/segment evaluation on held-out trajectories, or demonstrate that a hierarchical policy trained from the automatically discovered segments solves the task.
  2. [Section 4.3, Figure 4] The sample-efficiency comparison is confounded. The baseline is one policy trained on the full set of 50 trajectories for 1.2M steps, while each of the six primitive policies is trained on a disjoint subset of the data (approximately 8 trajectories) for 200k steps, with total steps matched to the baseline. The per-primitive policies see far less data and fewer distinct trajectories, and the comparison conflates the effect of segmenting the input distribution with the effect of changing dataset size per policy, number of policies, and total optimization budget per dataset item. In addition, only validation loss is reported, not rollout success rate, and no standard deviations or multiple seeds are given, so the claim of improved 'accuracy' (Section 4.3) is unsupported.
  3. [Sections 4.2 and 4.4] The evidence that CMA 'learns characteristics of different primitive actions' is circular. The CMA model is trained on the same demonstration trajectories that were manually segmented in Section 4.1, and these same labels are then used to interpret the attention weights (Fig. 2) and the t-SNE clusters (Fig. 5). Since the model was trained to predict actions from observations that progress through the trajectory, the observed attention differences could reflect temporal phase, labeling artifacts, or spurious correlations rather than emergent skill structure. A non-circular test would require evaluating the attention or embeddings on held-out trajectories, or ablating one modality and showing that attention patterns change in a task-meaningful way.
  4. [Abstract and Section 5] The abstract states that the method 'leverages this segmentation to train a hierarchical policy capable of solving long-horizon, contact-rich manipulation tasks,' but Section 5 states that training a hierarchical policy is future work and no hierarchical policy is trained or evaluated anywhere in the paper. This is a direct discrepancy between the claims and the reported experiments, and it affects the first sentence of the abstract as well as the third listed contribution.
minor comments (6)
  1. [Section 3] The tensor shape is written as '(T· N× B× D' with an unclosed parenthesis; the embedding dimension D is later stated to be 128 in Section 3.3 but should be defined in the architecture description.
  2. [Figure 1] The text 'Thrid stage' contains a typo; it should read 'Third stage.'
  3. [Section 4.4] No hyperparameters for the t-SNE visualization (e.g., perplexity, learning rate, number of iterations) are given, which limits reproducibility.
  4. [Section 4.3] Figure 4 appears to show a single run without error bars or seed information; the paper should state how many seeds were used and include variance estimates.
  5. [Section 4.1] The manual segmentation procedure is described only as 'similar to what's defined in Lin et al. [27];' more detail on the annotation protocol and inter-labeler consistency would help assess the reliability of the ground-truth labels used throughout the evaluation.
  6. [Section 3.2] The use of two consecutive timesteps in CMA is mentioned but not formalized; the paper should specify how temporal and modality dimensions are arranged in the attention computation, and why two timesteps is sufficient.

Circularity Check

0 steps flagged · score 0.0 of 10

No construction-level circularity; the paper's central over-claim is an evidence gap, not a circular derivation.

full rationale

The paper's chain is empirical rather than derivational: train a cross-modality attention (CMA) diffusion policy on 50 expert trajectories, inspect attention weights and t-SNE embeddings, and compare whole-trajectory with per-primitive policies. The manual segmentation into six primitive actions (Sec. 4.1) is used as an evaluation label, not as a training signal or as the output of the claimed 'unsupervised segmentation'; the CMA module is trained by an imitation loss on actions, not on the manual labels. Thus the observation that attention weights differ across manually labeled phases is not forced by construction. The 'unsupervised segmentation' contribution is unsupported because Sec. 4.4 only visualizes t-SNE clusters and Sec. 5 defers actual segmentation to future work, but an unsupported conclusion is an evidence/validity concern, not circularity. There is no fitted parameter renamed as a prediction, no self-citation chain carrying the derivation, and no equation-level identity between inputs and claims. Score 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities. Its central claims rely on a standard diffusion policy backbone, a frozen visual encoder, manually labeled skill segments, and an unvalidated assumption that attention weights directly reflect modality relevance. The only hand-tuned numbers are architecture hyperparameters and training-step counts, none of which are ablated.

free parameters (6)
  • CMA attention heads = 8
    Chosen by hand; the paper says 'we use 8 attention heads' without an ablation.
  • CMA transformer layers = 2
    Chosen by hand; the paper says 'we found sufficient' without an ablation.
  • Embedding dimension = 128
    All modality embeddings are projected to a 128-dimensional space; no ablation.
  • U-Net downsampling dimensions = 256, 512, 1024
    Copied from Diffusion Policy [11]; no task-specific tuning rationale given.
  • Training steps (baseline) = 1,200,000
    Fixed by the authors; no early stopping criterion stated.
  • Training steps per primitive = 200,000
    Chosen so six primitives match the 1.2M baseline; the equivalence assumes linear scaling.
assumptions (5)
  • domain assumption Diffusion policy is an appropriate backbone for visuomotor imitation learning.
    Borrowed from Chi et al. [11]; the paper does not compare to other backbones.
  • domain assumption R3M frozen visual representations are suitable for furniture assembly.
    Borrowed from Nair et al. [30]; no ablation on encoder choice.
  • domain assumption The 50 expert demonstrations from the DPPO-trained state-based policy are high-quality and representative.
    The paper states the policy achieved 96% success (Section 4.1), but does not report variance or diversity of these 50 trajectories.
  • domain assumption Manual segmentation into six primitives corresponds to meaningful skill boundaries.
    Section 4.1: 'We manually segment the trajectories into 6 primitive actions.' The entire evaluation of attention patterns and per-primitive policies depends on these labels; Section 4.4 itself notes ambiguity in the labels.
  • ad hoc to paper Attention weights from the last layer, averaged across heads, faithfully indicate modality relevance.
    Section 4.2 states they use last-layer, head-averaged weights because the layer 'is most directly connected to the final decision making.' This interpretability assumption is not validated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Modality Selection and Skill Segmentation via Cross-Modality Attention." pith.science (2026). https://pith.science/paper/B2UVDJO7

@misc{pith2026250414573,
  author       = {Pith},
  title        = {Pith review of: Modality Selection and Skill Segmentation via Cross-Modality Attention},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B2UVDJO7}},
  note         = {Machine review of arXiv:2504.14573}
}
read the original abstract

Incorporating additional sensory modalities such as tactile and audio into foundational robotic models poses significant challenges due to the curse of dimensionality. This work addresses this issue through modality selection. We propose a cross-modality attention (CMA) mechanism to identify and selectively utilize the modalities that are most informative for action generation at each timestep. Furthermore, we extend the application of CMA to segment primitive skills from expert demonstrations and leverage this segmentation to train a hierarchical policy capable of solving long-horizon, contact-rich manipulation tasks.

Figures

Figures reproduced from arXiv: 2504.14573 by the authors.

Figure 1
Figure 1. Three stages design. First stage: train CMA using imitation learning. Second stage: Use CMA to cluster [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Attention weights obtained through training a multimodal diffusion model. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Primitive Actions [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Validation Loss. the self-attention is formulated as, Attention(Q, K, V ) = softmax  QK⊤ √ dk  | {z } Attention Weights ·V We customize the self-attention block in the CMA to return both the attention weights and the final output. We take the attention weights of the…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 17 canonical work pages

  1. [1]

    Language models are few-shot learners,

    T. B. Brown et al. , “Language models are few-shot learners,” in Proceedings of the 34th International Con- ference on Neural Information Processing Systems , ser. NIPS ’20. Red Hook, NY, USA: Curran Associates Inc., 2020

    work page 2020

  2. [2]

    The llama 3 herd of models,

    A. Grattafioriet al., “The llama 3 herd of models,” 2024. [Online]. Available: https://arxiv.org/abs/2407.21783

    arXiv 2024

  3. [3]

    Visual instruction tuning,

    H. Liu, C. Li, Q. Wu, and Y. J. Lee, “Visual instruction tuning,” 2023

    work page 2023

  4. [4]

    Octo: An open-source generalist robot policy,

    Octo Model Team, D. Ghosh, H. Walke, K. Pertsch, K. Black, O. Mees, S. Dasari, J. Hejna, C. Xu, J. Luo, T. Kreiman, Y. Tan, L. Y. Chen, P. Sanketi, Q. Vuong, T. Xiao, D. Sadigh, C. Finn, and S. Levine, “Octo: An open-source generalist robot policy,” inProceedings of Robotics: Science and Systems , Delft, Netherlands, 2024

    work page 2024

  5. [5]

    Open X-Embodiment: Robotic learn- ing datasets and RT-X models,

    A. O’Neill et al., “Open X-Embodiment: Robotic learn- ing datasets and RT-X models,” https://arxiv.org/abs/ 2310.08864, 2023

    arXiv 2023

  6. [6]

    Do as i can and not as i say: Ground- ing language in robotic affordances,

    M. Ahn et al., “Do as i can and not as i say: Ground- ing language in robotic affordances,” inarXiv preprint arXiv:2204.01691, 2022

    arXiv 2022

  7. [7]

    Beyond Sight: Finetuning Generalist Robot Policies with Heterogeneous Sensors via Language Grounding,

    J. Jones, O. Mees, C. Sferrazza, K. Stachowicz, P. Abbeel, and S. Levine, “Beyond Sight: Finetuning Generalist Robot Policies with Heterogeneous Sensors via Language Grounding,” Jan. 2025, arXiv:2501.04693 [cs]. [Online]. Available: http://arxiv.org/abs/2501. 04693

    arXiv 2025

  8. [8]

    Furniturebench: Reproducible real-world benchmark for long-horizon complex manipulation,

    M. Heo, Y. Lee, D. Lee, and J. J. Lim, “Furniturebench: Reproducible real-world benchmark for long-horizon complex manipulation,” inRobotics: Science and Sys- tems, 2023

    work page 2023

Show all 31 references
  1. [9]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” in Proceedings of the 34th Inter- national Conference on Neural Information Processing Systems, ser. NIPS ’20. Red Hook, NY, USA: Curran Associates Inc., Dec. 2020, pp. 6840–6851

    2020

  2. [10]

    Score-based generative modeling through stochastic differential equations,

    Y. Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole, “Score-based generative modeling through stochastic differential equations,” in International Conference on Learning Representations ,

  3. [11]

    Diffusion policy: Vi- suomotor policy learning via action diffusion,

    C. Chi, Z. Xu, S. Feng, E. Cousineau, Y. Du, B. Burch- fiel, R. Tedrake, and S. Song, “Diffusion policy: Vi- suomotor policy learning via action diffusion,” The International Journal of Robotics Research , 2024

    2024

  4. [12]

    3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representations,

    Y. Ze, G. Zhang, K. Zhang, C. Hu, M. Wang, and H. Xu, “3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representations,” Sep. 2024, arXiv:2403.03954 [cs]. [Online]. Available: http://arxiv.org/abs/2403.03954

    2024 arXiv

  5. [13]

    JUICER: Data-Efficient Imitation Learning for Robotic Assembly,

    L. Ankile, A. Simeonov, I. Shenfeld, and P. Agrawal, “JUICER: Data-Efficient Imitation Learning for Robotic Assembly,” Nov. 2024, arXiv:2404.03729 [cs]. [Online]. Available: http://arxiv.org/abs/2404.03729

    2024 arXiv

  6. [14]

    Goal con- ditioned imitation learning using score-based diffusion policies,

    M. Reuss, M. Li, X. Jia, and R. Lioutikov, “Goal con- ditioned imitation learning using score-based diffusion policies,” in Robotics: Science and Systems , 2023

    2023

  7. [15]

    Diffusion Policy Policy Optimization,

    A. Z. Ren, J. Lidard, L. L. Ankile, A. Simeonov, P. Agrawal, A. Majumdar, B. Burchfiel, H. Dai, and M. Simchowitz, “Diffusion Policy Policy Optimization,” Aug. 2024, arXiv:2409.00588 [cs]. [Online]. Available: http://arxiv.org/abs/2409.00588

    2024 arXiv

  8. [16]

    ObjectFolder: A Dataset of Objects with Implicit Visual, Auditory, and Tactile Representations,

    R. Gao, Y.-Y. Chang, S. Mall, L. Fei-Fei, and J. Wu, “ObjectFolder: A Dataset of Objects with Implicit Visual, Auditory, and Tactile Representations,” Nov. 2021, arXiv:2109.07991 [cs]. [Online]. Available: http://arxiv.org/abs/2109.07991

    2021 arXiv

  9. [17]

    ObjectFolder 2.0: A Multisensory Object Dataset for Sim2Real Transfer,

    R. Gao, Z. Si, Y.-Y. Chang, S. Clarke, J. Bohg, L. Fei-Fei, W. Yuan, and J. Wu, “ObjectFolder 2.0: A Multisensory Object Dataset for Sim2Real Transfer,” Apr. 2022, arXiv:2204.02389 [cs]. [Online]. Available: http://arxiv.org/abs/2204.02389

    2022 arXiv

  10. [18]

    The ObjectFolder Benchmark: Multisensory Learning with Neural and Real Objects,

    R. Gao, Y. Dou, H. Li, T. Agarwal, J. Bohg, Y. Li, L. Fei-Fei, and J. Wu, “The ObjectFolder Benchmark: Multisensory Learning with Neural and Real Objects,” Jun. 2023, arXiv:2306.00956 [cs]. [Online]. Available: http://arxiv.org/abs/2306.00956

    2023 arXiv

  11. [19]

    See, Hear, and Feel: Smart Sensory Fusion for Robotic Manipulation,

    H. Li, Y. Zhang, J. Zhu, S. Wang, M. A. Lee, H. Xu, E. Adelson, L. Fei-Fei, R. Gao, and J. Wu, “See, Hear, and Feel: Smart Sensory Fusion for Robotic Manipulation,” Dec. 2022, arXiv:2212.03858. [Online]. Available: http://arxiv.org/abs/2212.03858

    2022 arXiv

  12. [20]

    Play to the score: Stage-guided dynamic multi-sensory fusion for robotic manipulation,

    R. Feng, D. Hu, W. Ma, and X. Li, “Play to the score: Stage-guided dynamic multi-sensory fusion for robotic manipulation,” in 8th Annual Conference on Robot Learning, 2024. [Online]. Available: https://openreview.net/forum?id=N5IS6DzBmL

    2024

  13. [21]

    Colledanchise and P

    M. Colledanchise and P. Ögren, Behavior Trees in Robotics and AI: An Introduction , Jul. 2018, arXiv:1709.00084 [cs]. [Online]. Available: http: //arxiv.org/abs/1709.00084

    2018 arXiv

  14. [22]

    Integrated Task and Motion Planning,

    C. R. Garrett, R. Chitnis, R. Holladay, B. Kim, T. Silver, L. P. Kaelbling, and T. Lozano-Pérez, “Integrated Task and Motion Planning,” Annual Review of Control, Robotics, and Autonomous Systems, vol. 4, no. 1, pp. 265–293, May 2021. [Online]. Available: https://www.annualrevi...

    2021 doi

  15. [23]

    Long-horizon manipulation of un- known objects via task and motion planning with esti- mated affordances,

    A. Curtis, X. Fang, L. P. Kaelbling, T. Lozano-Pérez, and C. R. Garrett, “Long-horizon manipulation of un- known objects via task and motion planning with esti- mated affordances,” in2022 International Conference on Robotics and Automation (ICRA) , 2022, pp. 1940– 1946

    2022

  16. [24]

    Learning Geometric Reasoning and Control for Long-Horizon Tasks from Visual Input,

    D. Driess, J.-S. Ha, R. Tedrake, and M. Toussaint, “Learning Geometric Reasoning and Control for Long-Horizon Tasks from Visual Input,” in2021 IEEE International Conference on Robotics and Automation (ICRA), May 2021, pp. 14298–14305, iSSN: 2577- 087X. [Online]. Available: htt...

    2021

  17. [25]

    Multistage Cable Routing Through Hierarchical Imitation Learning,

    J. Luo, C. Xu, X. Geng, G. Feng, K. Fang, L. Tan, S. Schaal, and S. Levine, “Multistage Cable Routing Through Hierarchical Imitation Learning,” IEEE Transactions on Robotics , vol. 40, pp. 1476–1491, 2024, conference Name: IEEE Transactions on Robotics. [Online]. Available: ht...

    2024

  18. [26]

    DexSkills: Skill Segmentation Using Haptic Data for Learning Autonomous Long-Horizon Robotic Manipulation Tasks,

    X. Mao, G. Giudici, C. Coppola, K. Althoefer, I. Farkhatdinov, Z. Li, and L. Jamone, “DexSkills: Skill Segmentation Using Haptic Data for Learning Autonomous Long-Horizon Robotic Manipulation Tasks,” in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (...

    2024

  19. [27]

    Generalize by Touching: Tactile Ensemble Skill Transfer for Robotic Furniture Assembly,

    H. Lin, R. Corcodel, and D. Zhao, “Generalize by Touching: Tactile Ensemble Skill Transfer for Robotic Furniture Assembly,” Apr. 2024, arXiv:2404.17684. [Online]. Available: http://arxiv.org/abs/2404.17684

    2024 arXiv

  20. [28]

    From imitation to refinement – residual rl for precise visual assembly,

    L. Ankile, A. Simeonov, I. Shenfeld, M. Torne, and P. Agrawal, “From imitation to refinement – residual rl for precise visual assembly,” 2024. [Online]. Available: https://arxiv.org/abs/2407.16677

    2024 arXiv

  21. [29]

    Residual policy learning,

    T. Silver, K. Allen, J. Tenenbaum, and L. Kaelbling, “Residual policy learning,” 2019. [Online]. Available: https://arxiv.org/abs/1812.06298

    2019 arXiv

  22. [30]

    R3M: A Universal Visual Representation for Robot Manipulation,

    S. Nair, A. Rajeswaran, V. Kumar, C. Finn, and A. Gupta, “R3M: A Universal Visual Representation for Robot Manipulation,” Nov. 2022, arXiv:2203.12601 [cs]. [Online]. Available: http://arxiv.org/abs/2203.12601

    2022 arXiv

  23. [2021]

    Available: https://openreview.net/ forum?id=PxTIG12RRHS

    [Online]. Available: https://openreview.net/ forum?id=PxTIG12RRHS

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.