Pith. sign in

REVIEW 5 major objections 5 minor 60 references

CAAT: Contact-Aware Attention Scaling and Tactile Masking for Data-Efficient Contact-Rich Manipulation

T0 review · 5 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read CAAT hard-codes a contact phase switch into Transformer-based policies, emphasizing vision before contact and touch during contact, and reports gains of up to 21.1 percentage points over standard visuo-tactile fusion.

desk verdict A modular, well-tested fusion idea with a valuable hardware contribution, but the contact-state-dependence claim is undercut by the paper's own ablations. read the letter →

arxiv 2608.01102 v1 pith:FWA7LFKH submitted 2026-08-02 cs.RO

Jiaming Jiang , Yuzhe Huang , Hao Liang , Pei Lin , Shengcheng Luo , Fanrong Dong , Jiaping Wu , Chenxi Xiao
show 2 more authors
Wanlin Li Ziyuan Jiao
This is my paper · ORCID
classification cs.RO
keywords contact-richmanipulationvisuo-tactilefusionimitationlearningTransformerpoliciesattentionscalingtactilemaskingdataefficiencysensing
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 argues that visuo-tactile policies fail at contact-rich manipulation under limited data because they leave the hard part—deciding when touch matters—to be learned from demonstrations, where contact occupies only a tiny fraction of the trajectory. CAAT instead hard-codes the phase structure: a binary contact estimate switches attention scaling so that vision dominates before contact and touch dominates during contact, while a masking step deletes tactile background regions that have not changed since the start of the episode. The paper reports that this explicit prior improves average success by 18.0 percentage points over direct visuo-tactile concatenation and 10.0 over learned gating in simulation, and by 21.1 over the strongest baseline in real-world trials, across ACT, Diffusion Policy, and $\pi_0$. A sympathetic reader should care because the components are modular and add negligible overhead, which makes the claimed gains a plausible recipe for making tactile feedback usable in low-data regimes.

What carries the argument

The load-bearing machinery is the contact-aware attention readout: after joint self-attention, the action query computes separate cross-attention readouts over the visual, tactile, and proprioceptive token subsets and combines them as $\hat{r}_t = \gamma_v(z_t) r^v_t + \gamma_\tau(z_t) r^\tau_t + r^\Upsilon_t$, where $z_t$ is a binary contact state from a CNN-MLP classifier on $|T_t - T_{\rm ref}|$. The scaling factors are user-fixed functions satisfying $\gamma_v(0) > \gamma_\tau(0)$ and $\gamma_\tau(1) > \gamma_v(1)$, so vision is weighted before contact and touch after it. The mechanism's job is to replace a demonstration-hungry learned modality weighting with a structural prior, while retaining full cross-modal representation learning in the shared self-attention layers; Dynamic Tactile Masking then supplies the spatial counterpart by zeroing tactile tokens similar to the first-frame reference.

What would settle it

A decisive experiment is to train CAAT on the same demonstrations but replace the first-frame reference image with one taken mid-contact, or to start episodes already in contact, and measure success; if the average success rate does not collapse toward the gated-fusion baseline, the reference-image assumption is not the source of the gain, but the paper reports no such sensitivity check.

Watch

Extended reading notes

Core claim

The central discovery is that a task-level prior about contact—rather than more capacity or more data—is what unblocks visuo-tactile fusion. After joint self-attention, the action query reads out visual, tactile, and proprioceptive tokens separately, and the readouts are combined under a contact-conditioned weight $\hat{r}_t = \gamma_v(z_t) r^v_t + \gamma_\tau(z_t) r^\tau_t + r^\Upsilon_t$, with the weights flipped between $(0.7,0.3)$ and $(0.3,0.7)$ as the estimated contact state switches. This is paired with Dynamic Tactile Masking, which compares each current tactile patch token with the corresponding first-frame reference token and zeroes tokens whose cosine similarity exceeds $\rho=0.8$, suppressing static background. The paper shows, in simulation and on a real two-finger tactile gripper, that this combination improves success rates over direct concatenation and learned gating, and that the fixed numerical scaling beats binary and learnable scaling variants. In the authors' own terms, the finding is that explicit contact-phase priors make Transformer-based visuo-tactile policies data-efficient without changing their action decoders.

Load-bearing premise

The whole switching mechanism rests on the assumption that a reliable binary contact state can be computed by comparing the current tactile image with a fixed non-contact reference image taken at the first timestep; if contact is already present at the start, the sensor drifts during an episode, or contact changes the tactile image only subtly, the scaling and masking will switch at the wrong time, and the paper reports no failure analysis for these cases.

Editorial extensions

If this is right

  • Adding CAAT to an existing Transformer policy requires no change to its action decoder, so the same contact prior can be layered onto future policy backbones beyond the three tested.
  • The prior pays off most where demonstrations are scarce: with 25, 50, and 100 demonstrations in simulation, CAAT outperforms both baselines at every budget, so the gain is a data-efficiency gain, not just a ceiling-raising gain.
  • Dynamic Tactile Masking helps all fusion strategies, not only CAAT, which implies that suppressing static tactile background is an independently useful preprocessing step for visuo-tactile policy learning.
  • The fixed numerical scaling $(0.7,0.3)$ to $(0.3,0.7)$ beats both binary and learned scaling, suggesting that the exact weights are less important than having a smooth, phase-dependent preference rather than a hard switch.
  • In real-world trials the largest gains appear on open-box and powerbank-extraction tasks, where contact forces and local interaction matter most, so the mechanism's value scales with how contact-dominated the task is.

Reading between the lines

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

  • Editorial inference: because the paper fixes Dynamic Tactile Masking across all baselines, the reported margins are joint effects; the individual contribution of contact-aware scaling alone is not isolated by a full ablation.
  • Editorial inference: the reference-image dependence suggests a natural robustness test—episodes beginning in contact or a corrupted reference frame—that the paper does not report; a reader should not assume the gains survive those conditions.
  • Editorial inference: the readout-scaling mechanism is not tied to binary contact; the same design could grade emphasis continuously from a force estimate or grasp-state signal, which the paper does not explore.
  • Editorial inference: the fixed numerical weights introduce a user choice; a small per-task search over $\gamma$ values would test whether the exact constants matter, an experiment the paper leaves open.
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

5 major / 5 minor

Summary. The paper proposes CAAT, a modular framework for Transformer-based visuo-tactile policies. CAAT has two components: Dynamic Tactile Masking (DTM), which suppresses tactile tokens whose cosine similarity to a non-contact reference exceeds a threshold (Eqs. 3–6), and Contact-Aware Attention Scaling, which computes modality-specific attention readouts after joint self-attention and reweights the visual and tactile readouts according to an estimated binary contact state (Eqs. 8–13). The contact state is produced by a CNN–MLP classifier trained with ground-truth contact labels (Eqs. 15–17). The authors report that CAAT with fixed numerical scaling (0.7/0.3 before contact, 0.3/0.7 during contact) improves average success rate by 18.0 percentage points over direct concatenation and 10.0 points over gated fusion on the UniVTAC benchmark with ACT, and achieves a 60.0% average real-world success rate across ACT, Diffusion Policy, and pi0, outperforming the strongest baseline by 21.1 percentage points. The paper also presents data-efficiency curves, ablations of DTM, and attention visualizations. The central claim is that explicit contact priors improve data-efficient visuo-tactile policy learning.

Significance. If the result holds, CAAT is a useful inductive bias: it is architecture-agnostic, adds negligible compute, and its data-efficiency gains are consistent across demonstration budgets. The real-world UMI-style tactile gripper and the evaluation across three policy families are also valuable assets. However, the current evidence does not establish the central claim. The headline numbers come from the single variant selected after seeing the same benchmark's results, the contact-state signal uses ground-truth labels not available to baselines, and no error bars or statistical tests are reported. These are fixable with additional experiments and analysis, but they are load-bearing for the paper's conclusions.

major comments (5)
  1. [§4.2, Table 1 and 'Comparison of scaling strategies'] The headline gain of 18.0 percentage points is reported only for the 'Ours (Numerical)' variant, whose weights gamma_v=0.7/gamma_tau=0.3 before contact and 0.3/0.7 during contact were adopted after comparing binary, learnable, and numerical scaling on the same UniVTAC benchmark. The two other CAAT variants perform essentially at the level of the concatenation baseline (51.8% and 50.6% vs. 51.6%), so the reported advantage is not robust to the choice of scaling parameterization. Please report results for all variants with variance estimates, and either specify the numerical weights in advance using a validation split or a separate held-out set, or correct for the multiple variants tested.
  2. [§3.3, Eq. (16) and Appendix 'CAAT-Specific Settings'] The contact classifier is trained with ground-truth contact labels y_t, but the paper never states how these labels are obtained in real-world experiments. In simulation, UniVTAC may provide them, but the real-world acquisition procedure is absent. Because the concatenation and gating baselines do not receive contact labels, CAAT effectively uses privileged supervision for its switching signal, which confounds the comparison. Please specify the real-world labeling protocol, or compare against a baseline that also receives the same contact-state information, and additionally report the performance of a rule-based contact detector (e.g., thresholding the tactile difference D_t) to show that the learned classifier is not the source of the gain.
  3. [§4.4 'Ablation Studies', paragraph on Contact-Aware Attention Scaling] The ablation that removes contact-aware scaling is described only qualitatively: 'Removing attention scaling substantially degrades performance in both settings,' but no success rates are given. Table 3 reports CAAT with and without Dynamic Tactile Masking, but it does not report the fixed-fusion baseline (DTM with fixed, contact-independent scaling). Please provide the actual numbers for this control in simulation and real-world, together with the exact readout weights used, so the contribution of the contact-state-dependent switching can be quantified separately from the contribution of the hand-picked numerical weights.
  4. [§Appendix 'Evaluation Protocol' and Tables 1–2] No error bars, confidence intervals, or statistical tests are reported anywhere. The simulation protocol uses 100 rollouts across 100 random seeds and the real-world protocol uses 20 trials per task, which is sufficient for a difference-of-proportions test or a bootstrap confidence interval. With such granularities, the reported 18.0 and 21.1 percentage point gaps need error estimates to rule out run-to-run variation. Please include per-task standard errors or confidence intervals and, where possible, a paired or per-task significance test.
  5. [§4.2, Table 1, and Appendix 'CAAT-Specific Settings'] The fixed numerical weights (0.7, 0.3) and (0.3, 0.7) are user-specified and not swept. Given that the binary and learnable variants fail to beat the gated baseline, the method's success appears to hinge on a narrow choice of the scaling coefficients. Please provide a sensitivity analysis of the numerical weights (e.g., varying each gamma by 0.1 or 0.2) to show that the performance gain is not a narrow peak, and justify the chosen values from first principles or from a validation set.
minor comments (5)
  1. [Eq. (15)] The notation D_t = |T_t - T_ref| is ambiguous: the text says the classifier input is the difference between tactile images, but the equation could be read as a scalar norm; please clarify the exact input representation.
  2. [Eqs. (20)–(22)] The learnable variant is said to use state-dependent scaling, but the definition gamma_L_tau(z_t) = 1 - gamma_L_v(z_t) forces the two weights to sum to one; this is a strong constraint that may explain why the learnable variant underperforms. Please state whether the numerical variant also imposes this constraint and discuss the modeling choice.
  3. [Appendix 'CAAT-Specific Settings'] The contact estimator is described as a 'frozen CNN–MLP binary classifier,' but Eq. (16) implies it is trained; please specify on which data it is trained, when it is frozen, and how the reference image T_ref is selected in episodes where the first timestep already involves contact.
  4. [Figure 4 and Table 4] The data-efficiency curves are informative, but the 25- and 50-demonstration points for ACT+DTM in Table 4 are non-monotonic (21.6, 25.4, 22.2); please comment on this variance or add error bars to Figure 4.
  5. [Table 2 and Figure 5] The real-world table reports only point estimates, and the attention heatmap analysis is purely qualitative; please state the number of trials visualized and whether the attention-shift pattern was consistent across rollouts.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: CAAT's contact-scaling weights and tactile mask are explicit user-specified priors, and the reported gains are empirical evaluations rather than derived predictions.

full rationale

CAAT does not present a first-principles derivation; Eq. (2)-(18) define a forward architecture. The contact state z_t is estimated from a tactile difference image (Eq. 15-17), and the scaling factors gamma_v(0)=0.7, gamma_tau(0)=0.3, gamma_v(1)=0.3, gamma_tau(1)=0.7 are openly user-specified in Appendix CAAT-Specific Settings, so no fitted parameter is renamed as a prediction. The sensorimotor prior (vision before contact, touch during contact) is supported by external citations (Lei et al. 2026; Li et al. 2026), and the authors' own prior tactile-hardware papers (e.g., PP-Tac, Lin et al. 2025) are background hardware citations, not load-bearing evidence for the main claim. The main experimental concerns, selecting the numerical scaling variant after seeing Table 1 and not specifying how ground-truth contact labels for Eq. (16) are obtained in the real world, are selection-bias and reproducibility issues rather than circular reasoning. The ablation also lacks an explicit constant-scaling control, which weakens attribution of the gain to contact-state dependence, but this is an experimental-control gap rather than a derivation that reduces to its own inputs. Therefore the paper's success-rate claims are empirical and not circular by construction.

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

The central dependence of the method is on the reliability of a reference-based contact signal and on the availability of ground-truth contact labels. No new physical entities are introduced; the method's gains hinge on hand-chosen scaling and masking thresholds whose sensitivity is not reported.

free parameters (3)
  • DTM cosine similarity threshold rho = 0.8
    Fixed manually; no sensitivity analysis is reported, yet it determines which tactile tokens are masked as static background.
  • Contact-aware scaling weights (gamma_v, gamma_tau) = (0.7, 0.3) before contact; (0.3, 0.7) during contact
    Hand-selected after comparing binary and learnable variants on the same benchmark; the headline result depends on this choice.
  • Contact probability threshold = 0.5
    Standard but arbitrary; converts the classifier output into a binary contact state used for scaling.
assumptions (4)
  • domain assumption The difference between the current tactile image and a fixed non-contact reference reliably indicates contact state.
    Used in Eq. (15)-(17) and the CAAT-specific settings; no analysis covers initial contact, sensor drift, or non-contact scene changes.
  • domain assumption Ground-truth contact labels y_t are available for training the contact classifier.
    Eq. (16) assumes such labels exist, but the paper never specifies how they are collected for real-world tasks or whether baselines have access to them.
  • domain assumption Patch-token cosine similarity between current and reference tactile observations separates static background from contact deformations.
    Eq. (4)-(5) assumes the sensor and reference view remain aligned and that contact-induced changes are local appearance shifts.
  • domain assumption A pretrained tactile ViT encoder is frozen and transferable across tasks and sensors.
    The appendix describes a ViT-VQGAN pretrained on public tactile datasets; the method relies on its features being comparable across time via cosine similarity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CAAT: Contact-Aware Attention Scaling and Tactile Masking for Data-Efficient Contact-Rich Manipulation." pith.science (2026). https://pith.science/paper/FWA7LFKH

@misc{pith2026260801102,
  author       = {Pith},
  title        = {Pith review of: CAAT: Contact-Aware Attention Scaling and Tactile Masking for Data-Efficient Contact-Rich Manipulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FWA7LFKH}},
  note         = {Machine review of arXiv:2608.01102}
}
abstract

In contact-rich manipulation, visual observations primarily guide motion in free space, whereas tactile observations become particularly informative during contact. However, standard Transformer-based visuo-tactile policies typically rely on either token concatenation or learnable gating. These approaches lack explicit contact-aware priors, making it difficult to efficiently learn effective cross-modal representations from demonstrations. To address this limitation, we propose CAAT, a lightweight contact-aware framework that explicitly incorporates contact priors through attention scaling and dynamic tactile masking. Specifically, CAAT emphasizes visual information before contact and tactile information during contact. It also suppresses static background tokens by comparing the current tactile observation with a non-contact reference. CAAT can be integrated into commonly used Transformer-based policies without modifying their action decoders. In simulation, integrating CAAT with ACT improves the average success rate by 18.0 percentage points over direct visuo-tactile fusion and by 10.0 percentage points over gated fusion. In real-world experiments using a visuo-tactile UMI platform, CAAT achieves an average success rate of 60.0% across ACT, Diffusion Policy, and $\pi_0$, outperforming the strongest baseline by an average of 21.1 percentage points. These results demonstrate that explicit contact priors and dynamic tactile masking are effective in improving visuo-tactile policy learning and task performance of diverse policy architectures. https://mrjiangjm.github.io/caat/

Figures

Figures reproduced from arXiv: 2608.01102 by the authors.

Figure 1
Figure 1. Overview of visuo-tactile fusion strategies and their performance comparisons: (A) Direct token concatenation and (B) learnable gating lack explicit priors for switching between contact and non-contact phases. (C) CAAT introduces contact-aware attention scaling to emphasize vision tokens before contact and tactile tokens during contact. (D) With dynamic tactile masking applied to all methods, CAAT consistently outpe… view at source ↗
Figure 2
Figure 2. Overview of CAAT. CAAT introduces contact-aware priors into Transformer-based visuo-tactile action prediction. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Data efficiency on the UniVTAC benchmark. The [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Attention heatmap visualization. Panel (a) presents one representative real-world rollout, while panel (b) presents two [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Hardware design of the tactile gripper. The dexter [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 20 canonical work pages

  1. [1]

    Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education

    Clancey, William J. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83)

  2. [2]

    Classification Problem Solving

    Clancey, William J. Classification Problem Solving. Proceedings of the Fourth National Conference on Artificial Intelligence

  3. [3]

    , title =

    Robinson, Arthur L. , title =. 1980 , doi =. https://science.sciencemag.org/content/208/4447/1019.full.pdf , journal =

    1980

  4. [4]

    New Ways to Make Microcircuits Smaller---Duplicate Entry

    Robinson, Arthur L. New Ways to Make Microcircuits Smaller---Duplicate Entry. Science

  5. [5]

    Clancey and Glenn Rennels , abstract =

    Diane Warner Hasling and William J. Clancey and Glenn Rennels , abstract =. Strategic explanations for a diagnostic consultation system , journal =. 1984 , issn =. doi:https://doi.org/10.1016/S0020-7373(84)80003-6 , url =

    doi:10.1016/s0020-7373(84)80003-6 1984

  6. [6]

    and Rennels, Glenn R

    Hasling, Diane Warner and Clancey, William J. and Rennels, Glenn R. and Test, Thomas. Strategic Explanations in Consultation---Duplicate. The International Journal of Man-Machine Studies

  7. [7]

    Poligon: A System for Parallel Problem Solving

    Rice, James. Poligon: A System for Parallel Problem Solving

  8. [8]

    Transfer of Rule-Based Expertise through a Tutorial Dialogue

    Clancey, William J. Transfer of Rule-Based Expertise through a Tutorial Dialogue

Show all 60 references
  1. [9]

    The Engineering of Qualitative Models

    Clancey, William J. The Engineering of Qualitative Models

  2. [10]

    2023 , eprint=

    Attention Is All You Need , author=. 2023 , eprint=

    2023

  3. [11]

    Pluto: The 'Other' Red Planet

    NASA. Pluto: The 'Other' Red Planet

  4. [12]

    arXiv preprint arXiv:2304.13705 , year=

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

    arXiv

  5. [13]

    The International Journal of Robotics Research , volume=

    Diffusion policy: Visuomotor policy learning via action diffusion , author=. The International Journal of Robotics Research , volume=. 2025 , publisher=

    2025

  6. [14]

    Black, Kevin and Brown, Noah and Driess, Danny and Esmail, Adnan and Equi, Michael and Finn, Chelsea and Fusai, Niccolo and Groom, Lachy and Hausman, Karol and Ichter, Brian and others , journal=

  7. [15]

    arXiv preprint arXiv:2601.20321 , year=

    Tactile-Force Alignment in Vision-Language-Action Models for Force-aware Manipulation , author=. arXiv preprint arXiv:2601.20321 , year=

  8. [16]

    arXiv preprint arXiv:2602.10093 , year=

    UniVTAC: A Unified Simulation Platform for Visuo-Tactile Manipulation Data Generation, Learning, and Benchmarking , author=. arXiv preprint arXiv:2602.10093 , year=

  9. [17]

    arXiv preprint arXiv:2210.00121 , year=

    Visuo-tactile transformers for manipulation , author=. arXiv preprint arXiv:2210.00121 , year=

    arXiv

  10. [18]

    IEEE Transactions on Robotics , volume=

    A review of tactile information: Perception and action through touch , author=. IEEE Transactions on Robotics , volume=. 2020 , publisher=

    2020

  11. [19]

    arXiv preprint arXiv:2503.02881 , year=

    Reactive diffusion policy: Slow-fast visual-tactile policy learning for contact-rich manipulation , author=. arXiv preprint arXiv:2503.02881 , year=

    arXiv

  12. [20]

    IEEE Transactions on Robotics , volume=

    Slip detection for grasp stabilization with a multifingered tactile robot hand , author=. IEEE Transactions on Robotics , volume=. 2020 , publisher=

    2020

  13. [21]

    arXiv preprint arXiv:2505.07455 , year=

    GelFusion: Enhancing robotic manipulation under visual constraints via visuotactile fusion , author=. arXiv preprint arXiv:2505.07455 , year=

    arXiv

  14. [22]

    arXiv preprint arXiv:2606.08737 , year=

    Dream-Tac: A Unified Tactile World Action Model for Contact-Rich Robot Manipulation , author=. arXiv preprint arXiv:2606.08737 , year=

    arXiv

  15. [23]

    arXiv preprint arXiv:2606.11184 , year=

    TacForeSight: Force-Guided Tactile World Model for Contact-Rich Manipulation , author=. arXiv preprint arXiv:2606.11184 , year=

    arXiv

  16. [24]

    2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=

    Adaptive visuo-tactile fusion with predictive force attention for dexterous manipulation , author=. 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=. 2025 , organization=

    2025

  17. [25]

    arXiv preprint arXiv:2503.08548 , year=

    Tla: Tactile-language-action model for contact-rich manipulation , author=. arXiv preprint arXiv:2503.08548 , year=

    arXiv

  18. [26]

    arXiv preprint arXiv:2504.16649 , year=

    Pp-tac: Paper picking using tactile feedback in dexterous robotic hands , author=. arXiv preprint arXiv:2504.16649 , year=

    arXiv

  19. [27]

    The Fourteenth International Conference on Learning Representations , year=

    DexMove: Learning Tactile-Guided Non-Prehensile Manipulation with Dexterous Hands , author=. The Fourteenth International Conference on Learning Representations , year=

  20. [28]

    arXiv preprint arXiv:2606.08765 , year=

    RGB-S: Image-Aligned Tactile Saliency for Robust Dexterous Manipulation , author=. arXiv preprint arXiv:2606.08765 , year=

    arXiv

  21. [29]

    arXiv preprint arXiv:2606.11767 , year=

    Blind Dexterous Grasping via Real2Sim2Real Tactile Policy Learning , author=. arXiv preprint arXiv:2606.11767 , year=

    arXiv

  22. [30]

    arXiv preprint arXiv:2606.19161 , year=

    HT-Bench: Benchmarking and Learning Dexterous Full-Hand Tactile Representations with Egocentric Vision , author=. arXiv preprint arXiv:2606.19161 , year=

    arXiv

  23. [31]

    IEEE/ASME Transactions on Mechatronics , year=

    Tactile-Guided Exploration and Positioning for High-Precision Robotic Peg-in-Hole Tasks , author=. IEEE/ASME Transactions on Mechatronics , year=

  24. [32]

    Biomimetic Intelligence and Robotics , pages=

    Vtla: Vision-tactile-language-action model with preference learning for insertion manipulation , author=. Biomimetic Intelligence and Robotics , pages=. 2026 , publisher=

    2026

  25. [33]

    arXiv preprint arXiv:2406.13640 , year=

    Transferable tactile transformers for representation learning across diverse sensors and tasks , author=. arXiv preprint arXiv:2406.13640 , year=

    arXiv

  26. [34]

    arXiv preprint arXiv:2603.09565 , year=

    ReTac-ACT: A State-Gated Vision-Tactile Fusion Transformer for Precision Assembly , author=. arXiv preprint arXiv:2603.09565 , year=

  27. [35]

    Conference on Robot Learning , pages=

    Towards learning to detect and predict contact events on vision-based tactile sensors , author=. Conference on Robot Learning , pages=. 2020 , organization=

    2020

  28. [36]

    arXiv preprint arXiv:2603.19201 , year=

    Omnivta: Visuo-tactile world modeling for contact-rich robotic manipulation , author=. arXiv preprint arXiv:2603.19201 , year=

    arXiv

  29. [37]

    arXiv preprint arXiv:2110.04627 , year=

    Vector-quantized image modeling with improved vqgan , author=. arXiv preprint arXiv:2110.04627 , year=

    arXiv

  30. [38]

    arXiv preprint arXiv:2410.24090 , year=

    Sparsh: Self-supervised touch representations for vision-based tactile sensing , author=. arXiv preprint arXiv:2410.24090 , year=

    arXiv

  31. [39]

    2024 , eprint=

    Binding Touch to Everything: Learning Unified Multimodal Tactile Representations , author=. 2024 , eprint=

    2024

  32. [40]

    Frontiers in Robotics and AI , volume=

    Visuo-tactile feedback policies for terminal assembly facilitated by reinforcement learning , author=. Frontiers in Robotics and AI , volume=. 2025 , publisher=

    2025

  33. [41]

    arXiv e-prints , pages=

    ManiFeel: Benchmarking and Understanding Visuotactile Manipulation Policy Learning , author=. arXiv e-prints , pages=

  34. [42]

    arXiv preprint arXiv:2405.12213 , year=

    Octo: An open-source generalist robot policy , author=. arXiv preprint arXiv:2405.12213 , year=

    arXiv

  35. [43]

    arXiv preprint arXiv:2506.16211 , year=

    Controlvla: Few-shot object-centric adaptation for pre-trained vision-language-action models , author=. arXiv preprint arXiv:2506.16211 , year=

    arXiv

  36. [44]

    Conference on Robot Learning , pages=

    Rt-2: Vision-language-action models transfer web knowledge to robotic control , author=. Conference on Robot Learning , pages=. 2023 , organization=

    2023

  37. [45]

    arXiv preprint arXiv:2603.12665 , year=

    Tacvla: Contact-aware tactile fusion for robust vision-language-action manipulation , author=. arXiv preprint arXiv:2603.12665 , year=

  38. [46]

    arXiv preprint arXiv:2606.13102 , year=

    FTP-1: A Generalist Foundation Tactile Policy Across Tactile Sensors for Contact-Rich Manipulation , author=. arXiv preprint arXiv:2606.13102 , year=

    arXiv

  39. [47]

    arXiv preprint arXiv:2506.15953 , year=

    ViTacFormer: Learning cross-modal representation for visuo-tactile dexterous manipulation , author=. arXiv preprint arXiv:2506.15953 , year=

    arXiv

  40. [48]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    AT-VLA: Adaptive Tactile Injection for Enhanced Feedback Reaction in Vision-Language-Action Models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  41. [49]

    arXiv preprint arXiv:2604.01414 , year=

    Learning when to see and when to feel: Adaptive vision-torque fusion for contact-aware manipulation , author=. arXiv preprint arXiv:2604.01414 , year=

    arXiv

  42. [50]

    IEEE Robotics and Automation Letters , year=

    Vla-touch: Enhancing vision-language-action model with dual-level tactile feedback , author=. IEEE Robotics and Automation Letters , year=

  43. [51]

    The International Journal of Robotics Research , volume=

    Cable manipulation with a tactile-reactive gripper , author=. The International Journal of Robotics Research , volume=. 2021 , publisher=

    2021

  44. [52]

    2025 , booktitle=

    TacTape: Real-time High-accuracy Tactile Fiducial System with Structured 3D Texture for Vision-based Tactile Sensors , author=. 2025 , booktitle=

    2025

  45. [53]

    Sensors , volume=

    Gelsight: High-resolution robot tactile sensors for estimating geometry and force , author=. Sensors , volume=. 2017 , publisher=

    2017

  46. [54]

    arXiv preprint arXiv:2508.08706 , year=

    Omnivtla: Vision-tactile-language-action model with semantic-aligned tactile sensing , author=. arXiv preprint arXiv:2508.08706 , year=

  47. [55]

    Advances in neural information processing systems , volume=

    Attention bottlenecks for multimodal fusion , author=. Advances in neural information processing systems , volume=

  48. [56]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Multi-modal gated mixture of local-to-global experts for dynamic image fusion , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  49. [57]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Trusted multi-view classification with dynamic evidential fusion , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2022 , publisher=

    2022

  50. [58]

    arXiv preprint arXiv:2402.10329 , year=

    Universal manipulation interface: In-the-wild robot teaching without in-the-wild robots , author=. arXiv preprint arXiv:2402.10329 , year=

    arXiv

  51. [59]

    IEEE Robotics and Automation Letters , volume=

    Digit: A novel design for a low-cost compact high-resolution tactile sensor with application to in-hand manipulation , author=. IEEE Robotics and Automation Letters , volume=. 2020 , publisher=

    2020

  52. [60]

    IEEE Robotics and Automation Letters , volume=

    9dtact: A compact vision-based tactile sensor for accurate 3d shape reconstruction and generalizable 6d force estimation , author=. IEEE Robotics and Automation Letters , volume=. 2023 , publisher=

    2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.