Pith. sign in

REVIEW 3 major objections 5 minor 35 references

Robots overcome vision-dominated failure in contact tasks by sensing only the unexpected touch residuals that vision cannot predict.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-12 02:50 UTC pith:M2BNSL6W

load-bearing objection Solid residual-tactile VLA recipe with real gains on five contact tasks; predictive-coding framing is mostly packaging and residual orthogonality is only shown end-to-end. the 3 major comments →

arxiv 2607.03387 v2 pith:M2BNSL6W submitted 2026-07-03 cs.RO

Feeling the Unexpected: ResTacVLA for Contact-Rich Manipulation via Residual Tactile Representation

Pengwei Zhang , Bin Xie , Ce Hao , Xinpan Meng , Xinyu Guo , Fang Deng , Long Cheng , Tiancai Wang This is my paper
classification cs.RO
keywords residual tactile representationvision-language-actioncontact-rich manipulationpredictive codingmodality collapsesurprise-aware gatinglatent contact primitives
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Simply bolting tactile sensors onto vision-language-action robot policies usually fails: continuous high-bandwidth vision drowns out the sparse, event-driven touch signals. This paper claims the fix is to stop feeding raw touch and instead feed only the residual—the difference between what the wrist camera predicts the contact should feel like and what the tactile sensor actually reports. Those residuals are turned into a small set of discrete Latent Contact Primitives and admitted into the policy only when the visual prediction itself is uncertain. On five real contact-rich tasks (screwing, insertion, wiping) the resulting policy raises average success from roughly 28 % to 63 % and stays robust when grasps are noisy or targets move mid-execution. A reader who cares about generalist robots should care because contact skills that humans take for granted remain the clearest bottleneck for today’s vision-centric models.

Core claim

ResTacVLA shows that reformulating tactile input as the residual between a visual prior and physical sensation, quantizing that residual into Latent Contact Primitives, and gating it by visual-prediction uncertainty systematically prevents modality collapse and produces large, consistent gains on contact-rich manipulation.

What carries the argument

Residual Tactile Representation: the discrepancy r_t = z_t − ẑ_t produced by a Cross-Modal Predictor, discretized by a vector-quantized bottleneck into Latent Contact Primitives, then modulated by a Surprise-Aware Gate driven by the predictor’s own uncertainty σ_t. It converts sparse touch into dense information gain that is injected only when vision is unreliable.

Load-bearing premise

The method assumes a wrist-camera predictor can extract residuals that carry enough orthogonal physical information for a simple uncertainty gate to decide, reliably, when touch should override vision.

What would settle it

Train the identical residual-gated policy and a strong naïve-touch baseline on the same five tasks, then evaluate both on a held-out contact-rich task under occlusion; if residual gating no longer yields a clear success-rate advantage, or if the gate stays near zero throughout contact phases, the central claim fails.

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

If this is right

  • Contact-rich VLA policies can obtain large success gains without redesigning the vision-language backbone.
  • Tactile input becomes useful exactly in the occluded or force-critical phases where vision fails, rather than acting as constant noise.
  • The residual-plus-gate pattern is architecture-agnostic and already improves a Diffusion Policy baseline by the same mechanism.
  • Robustness to grasp noise, mid-execution target shifts and surface-height changes follows directly from prioritizing residual surprise.
  • The learned Latent Contact Primitives form a compact, cross-task vocabulary of physical events that vision alone cannot see.

Where Pith is reading between the lines

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

  • The same residual-coding idea could keep other sparse modalities (force, audio, proprioceptive spikes) from being drowned by vision or language.
  • If the visual prior is trained on broader data, the residual stream may surface rarer contact events without extra human labels.
  • Failure under lighting shifts or tactile-sensor recalibration would be the natural next stress test of the orthogonality claim.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes ResTacVLA, a VLA policy for contact-rich manipulation that reformulates tactile input as a Residual Tactile Representation r_t = z_t − ẑ_t, where ẑ_t is a wrist-camera visual prior produced by a Cross-Modal Predictor (CMP). Residuals are quantized via a VQ bottleneck into Latent Contact Primitives and injected into a flow-matching action expert through a Surprise-Aware Gate driven by the predictor’s aleatoric uncertainty σ_t. Built on π0.5 and evaluated on five real-robot tasks (lightbulb screwing, plug insertion, peg-in-hole, peg transfer, plate wiping), ResTacVLA reports 62.8% average success (+34.6% over vision-only π0.5), phase-wise gains on interaction stages, ablations of VQ and gating (Table II), and robustness to grasp noise, mid-execution displacements, and surface-height changes (Table III).

Significance. Contact-rich manipulation remains a clear bottleneck for vision-centric VLAs, and modality collapse under naive tactile fusion is a recognized practical failure mode. The residual-plus-surprise-gate formulation is a clean, biologically motivated systems idea that is architecture-agnostic (also improves Diffusion Policy) and is supported by real-robot results, phase-wise metrics, interpretability plots (t-SNE of primitives, gate traces), and three robustness suites. If the residual truly supplies orthogonal information gain rather than task-specific re-encoding, the approach offers a reusable recipe for tactile integration in generalist policies. Strengths include multi-task CMP pretraining, frozen-then-finetune protocol, and explicit comparison against both ResNet and UniT tactile baselines.

major comments (3)
  1. [§III-B, Eq. (1), Table I] §III-B and Abstract claim that r_t = z_t − ẑ_t is dense, high-value, and orthogonal to vision, thereby ‘inherently resolving the bandwidth mismatch.’ The only supporting evidence is end-to-end success (Table I), t-SNE (Fig. 4), and gate traces (Fig. 5) on the same five-task distribution used to train the CMP. There is no mutual-information comparison I(r_t; multi-view vision) vs I(z_t; ·), and no ablation that routes raw (non-residual) tactile latents through the identical VQ + SAG pipeline. Without that control, the performance gap versus π0.5 w/ T-UniT (42.3% → 62.8%) cannot be attributed specifically to residualization rather than to VQ discretization or gating alone. A residual-vs-raw ablation (or an explicit MI / reconstruction probe) is load-bearing for the central mechanistic claim.
  2. [Table I, Table III, §IV-A] All quantitative claims rest on 15–25 physical trials per condition with point estimates only (Tables I–III). No standard errors, confidence intervals, or statistical tests are reported. Given the absolute gains claimed (+34.6% average, up to +46.7% on individual phases) and the known trial-to-trial variance of contact-rich real-robot evaluation, the absence of uncertainty quantification weakens the strength of the superiority statements. At minimum, bootstrap or binomial CIs (or repeated seeds) should be added for the main table and the robustness suite.
  3. [§III-B, Fig. 4, Table III] The CMP is pre-trained and frozen on multi-task interaction data drawn from the same five tasks later used for policy evaluation (§III-A, §IV-A). Consequently, residual quality and the semantics of the VQ codebook (Fig. 4) are never probed on held-out objects, novel contact dynamics, or sensor placements outside the training distribution. The robustness suite (Table III) perturbs execution conditions but not the residual extractor itself. A modest OOD residual probe (e.g., novel object geometry or different GelSight mounting) would substantially strengthen the claim that Latent Contact Primitives capture general physical events rather than task-specific patterns.
minor comments (5)
  1. [§III-B, Eq. (1)] Eq. (1) writes L_pred = λ_σ log σ²_t + ∥z_t − μ_t∥² / σ²_t; the conventional Gaussian NLL also includes a ½ factor and often a constant. Clarify whether the omitted constants are absorbed into λ_σ or whether the loss is intentionally unnormalized.
  2. [Table II] Table II reports a single average over only Plug Insertion and Plate Wiping; per-task numbers (or at least the two individual rates) would make the −26.7% / −13.3% deltas easier to interpret.
  3. [Fig. 1, §III-C] Fig. 1 caption and the main text both use ‘Surprise Signal’ / ‘surprise-aware gate’; ensure consistent capitalization and that g_t is defined before first use in the figure.
  4. [Abstract, Table I] The abstract states ‘up to 86.7% task success’; Table I shows 86.7% only for Peg-A (alignment). Clarify that the peak is phase-specific, not full-task success, to avoid overstatement.
  5. [§II-B] Related Work §II-B correctly notes that contrastive alignment can suppress residual information; a short citation to predictive-coding robotics or forward-model residual work (beyond the neuroscience refs) would better situate the contribution for the robotics audience.

Circularity Check

0 steps flagged

No significant circularity: residual definition and gating are engineered representations validated by independent real-robot success rates, not predictions that reduce to their inputs by construction.

full rationale

ResTacVLA is an empirical systems paper. The residual r_t = z_t - μ_t (Eq. 1, Sec. III-B) is defined by construction as the visual-tactile discrepancy and then quantized; this is an intentional design choice inspired by predictive coding, not a claimed first-principles derivation whose output is forced to equal its input. The Surprise-Aware Gate g_t = Sigmoid(MLP(σ_t)) (Eq. 3) is likewise a learned modulator whose utility is measured, not assumed. Central claims (Table I average 62.8 % success, +34.6 % over π_0.5; Table III robustness) are obtained from held-out physical trials (15–25 rollouts per task) against vision-only and naïve-tactile baselines, with ablations (Table II) that remove VQ or gating. No parameter is fitted to a subset and then re-reported as a prediction of a closely related quantity; no uniqueness theorem or load-bearing result is imported solely via self-citation; the project-page link is non-load-bearing. The derivation chain therefore does not collapse into its own premises.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 4 invented entities

The central performance claim rests on a small set of architectural free parameters, standard robotics and deep-learning assumptions, and three invented modules whose only evidence is the paper’s own ablations and success rates.

free parameters (4)
  • λ_σ (variance penalty in NLL)
    Hyper-parameter that prevents uncertainty collapse in the Cross-Modal Predictor (Eq. 1); value not reported, yet controls the surprise signal that drives the gate.
  • λ_p (prediction-loss weight)
    Balances reconstruction, prediction and VQ losses in L_CMP (Eq. 2); chosen by hand.
  • VQ codebook size K and dimension
    Discretization capacity of Latent Contact Primitives; reduced dimension and EMA updates are used to avoid collapse, but exact K is unreported.
  • number of expert demonstrations (~100 per task)
    Data volume that defines the multi-task pre-training distribution of the CMP and the fine-tuning set of the policy.
axioms (4)
  • domain assumption Visual priors from a wrist camera can predict the majority of tactile latents, so the residual r_t = z_t − ẑ_t isolates high-value, vision-orthogonal information.
    Stated as the core of Residual Tactile Representation Learning (§III-B); never independently measured outside the five tasks.
  • ad hoc to paper Aleatoric uncertainty σ_t of the visual predictor is a reliable proxy for tactile information gain and can be mapped by a small MLP into a useful gate g_t.
    Introduced in §III-C (Eq. 3); the mapping is learned end-to-end and validated only by the ablation in Table II.
  • ad hoc to paper Vector quantization of residuals yields discrete Latent Contact Primitives that are more robust for policy learning than continuous residuals.
    Justified by the −26.7 % drop when VQ is removed (Table II); treated as an architectural axiom rather than a derived necessity.
  • domain assumption Standard conditional flow-matching objective and PaliGemma-style VLM backbone remain optimal once residual tokens are concatenated.
    Inherited from the π0.5 framework (§III-A) without re-derivation.
invented entities (4)
  • Residual Tactile Representation (r_t = z_t − ẑ_t) no independent evidence
    purpose: Convert sparse tactile into dense information-gain signal that avoids modality collapse.
    Defined in §III-B; independent evidence is limited to the paper’s own success-rate gains and t-SNE plots.
  • Latent Contact Primitives (VQ codebook entries) no independent evidence
    purpose: Discretize residuals into semantic contact events (collision, alignment success, free-space).
    Emergent clusters shown in Fig. 4; no external validation or public codebook release.
  • Surprise-Aware Gate (SAG) no independent evidence
    purpose: Adaptively inject primitives only when visual uncertainty is high.
    Eq. 3–4 and temporal plot in Fig. 5; necessity shown only by the fixed-fusion ablation.
  • Cross-Modal Predictor (CMP) no independent evidence
    purpose: Produce visual tactile prior μ_t, σ_t and residual for the rest of the pipeline.
    Pre-trained then frozen; architecture is a standard ResNet-18 + UniT encoder pair with no external benchmark.

pith-pipeline@v1.1.0-grok45 · 17240 in / 3366 out tokens · 29776 ms · 2026-07-12T02:50:51.200859+00:00 · methodology

0 comments
read the original abstract

Tactile perception is indispensable for contact-rich manipulation, yet integrating it into Vision-Language-Action (VLA) models often induces modality collapse, where high-bandwidth visual features overshadow sparse tactile cues. Inspired by Predictive Coding, a neural mechanism where the brain attenuates predictable inputs to prioritize surprising stimuli, we propose ResTacVLA. Rather than treating tactile data as raw input, we reformulate it as a Residual Tactile Representation capturing the discrepancy between visual priors and physical sensations. By filtering out visually predictable dynamics, this formulation transforms sparse tactile signals into dense, high-value information gain, thereby inherently resolving the bandwidth mismatch. These residuals are discretized through a Vector Quantized (VQ) bottleneck into Latent Contact Primitives that capture critical events missed by vision. Analogous to the neural surprise signal, we leverage the uncertainty of the visual prior to adaptively gate tactile integration, prioritizing residuals specifically during visually unreliable phases to explicitly prevent visual dominance. Experimental results show that ResTacVLA consistently outperforms all baselines on a diverse set of contact-rich manipulation tasks, while remaining robust to unexpected dynamic disturbances. Project page: https://awilekong.github.io/ResTacVLA/

Figures

Figures reproduced from arXiv: 2607.03387 by Bin Xie, Ce Hao, Fang Deng, Long Cheng, Pengwei Zhang, Tiancai Wang, Xinpan Meng, Xinyu Guo.

Figure 1
Figure 1. Figure 1: Predictive Coding-Inspired Residual Tactile Fusion in [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overall architecture of ResTacVLA. (a) Step 1: Residual Tactile Representation Learning. The Cross-Modal Predictor (CMP) is pre-trained to estimate tactile latents from wrist-camera observations, extracting residual representations that capture the discrepancy between visual priors and physical sensations. These residuals are discretized into Latent Contact Primitives via a VQ bottleneck. (b) Step 2: Surpr… view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the five contact-rich manipulation tasks designed for evaluation. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: t-SNE Visualization of Latent Contact Primitives. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Temporal Evolution of the Surprise-Aware Gate [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Model Robustness Evaluation. Three perturbation scenarios are designed: (1) initial grasp perturbation applied to Plug Insertion, (2) dynamic target displacement during Peg-in-Hole execution, and (3) surface height variation applied to Plate Wiping. V. CONCLUSIONS In this work, we introduced ResTacVLA, a biologically inspired framework designed to efficiently integrate tactile feedback in VLA models for co… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

35 extracted references · 17 linked inside Pith

  1. [1]

    π 0: A vision- language-action flow model for general robot control,

    K. Black, N. Brown, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fusai, L. Groom, K. Hausman, B. Ichteret al., “π 0: A vision- language-action flow model for general robot control,”arXiv preprint arXiv:2410.24164, 2024

    Pith/arXiv arXiv 2024

  2. [2]

    π0.5: A vision-language-action model with open-world generalization,

    P. Intelligence, K. Black, N. Brown, J. Darpinian, K. Dhabalia, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fusaiet al., “π0.5: A vision-language-action model with open-world generalization,”arXiv preprint arXiv:2504.16054, 2025

    Pith/arXiv arXiv 2025

  3. [3]

    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,”The International Journal of Robotics Research, vol. 44, no. 10-11, pp. 1863–1891, 2025

    2025

  4. [4]

    Forge: Force-guided exploration for robust contact-rich manipulation under uncertainty,

    M. Noseworthy, B. Tang, B. Wen, A. Handa, C. Kessens, N. Roy, D. Fox, F. Ramos, Y . Narang, and I. Akinola, “Forge: Force-guided exploration for robust contact-rich manipulation under uncertainty,” IEEE Robotics and Automation Letters, 2025

    2025

  5. [5]

    Adaptive compliance policy: Learning approximate compliance for diffusion guided control,

    Y . Hou, Z. Liu, C. Chi, E. Cousineau, N. Kuppuswamy, S. Feng, B. Burchfiel, and S. Song, “Adaptive compliance policy: Learning approximate compliance for diffusion guided control,” in2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025, pp. 4829–4836

    2025

  6. [6]

    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,” in2025 IEEE Inter- national Conference on Robotics and Automation (ICRA). IEEE, 2025, pp. 5961–5968

    2025

  7. [7]

    Vla-touch: Enhanc- ing vision-language-action models with dual-level tactile feedback,

    J. Bi, K. Y . Ma, C. Hao, M. Z. Shou, and H. Soh, “Vla-touch: Enhanc- ing vision-language-action models with dual-level tactile feedback,” arXiv preprint arXiv:2507.17294, 2025

    Pith/arXiv arXiv 2025

  8. [8]

    Vtla: Vision- tactile-language-action model with preference learning for insertion manipulation,

    C. Zhang, P. Hao, X. Cao, X. Hao, S. Cui, and S. Wang, “Vtla: Vision- tactile-language-action model with preference learning for insertion manipulation,”arXiv preprint arXiv:2505.09577, 2025

    Pith/arXiv arXiv 2025

  9. [9]

    A closer look at multimodal representation collapse,

    A. Chaudhuri, A. Dutta, T. Bui, and S. Georgescu, “A closer look at multimodal representation collapse,”arXiv preprint arXiv:2505.22483, 2025

    Pith/arXiv arXiv 2025

  10. [10]

    Multi-modal manipulation via multi- modal policy consensus,

    H. Chen, J. Xu, H. Chen, K. Hong, B. Huang, C. Liu, J. Mao, Y . Li, Y . Du, and K. Driggs-Campbell, “Multi-modal manipulation via multi- modal policy consensus,”arXiv preprint arXiv:2509.23468, 2025

    Pith/arXiv arXiv 2025

  11. [11]

    Implicitrdp: An end-to-end visual-force diffusion policy with structural slow-fast learning,

    W. Chen, H. Xue, Y . Wang, F. Zhou, J. Lv, Y . Jin, S. Tang, C. Wen, and C. Lu, “Implicitrdp: An end-to-end visual-force diffusion policy with structural slow-fast learning,”arXiv preprint arXiv:2512.10946, 2025

    arXiv 2025

  12. [12]

    Predictive attenuation of touch and tactile gating are distinct perceptual phenomena,

    K. Kilteni and H. H. Ehrsson, “Predictive attenuation of touch and tactile gating are distinct perceptual phenomena,”Iscience, vol. 25, no. 4, 2022

    2022

  13. [13]

    Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive- field effects,

    R. P. Rao and D. H. Ballard, “Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive- field effects,”Nature neuroscience, vol. 2, no. 1, pp. 79–87, 1999

    1999

  14. [14]

    The free-energy principle: a unified brain theory?

    K. Friston, “The free-energy principle: a unified brain theory?”Nature reviews neuroscience, vol. 11, no. 2, pp. 127–138, 2010

    2010

  15. [15]

    Central cancellation of self-produced tickle sensation,

    D. Wolpert, “Central cancellation of self-produced tickle sensation,” Nature Neuroscience, vol. 1, pp. 635–640, 1998

    1998

  16. [16]

    Memoryvla: Perceptual-cognitive memory in vision-language-action models for robotic manipulation,

    H. Shi, B. Xie, Y . Liu, L. Sun, F. Liu, T. Wang, E. Zhou, H. Fan, X. Zhang, and G. Huang, “Memoryvla: Perceptual-cognitive memory in vision-language-action models for robotic manipulation,”arXiv preprint arXiv:2508.19236, 2025

    Pith/arXiv arXiv 2025

  17. [17]

    Rt-1: Robotics transformer for real-world control at scale,

    A. Brohan, N. Brown, J. Carbajal, Y . Chebotar, J. Dabis, C. Finn, K. Gopalakrishnan, K. Hausman, A. Herzog, J. Hsuet al., “Rt-1: Robotics transformer for real-world control at scale,”arXiv preprint arXiv:2212.06817, 2022

    Pith/arXiv arXiv 2022

  18. [18]

    Rt-2: Vision- language-action models transfer web knowledge to robotic control, 2023,

    A. Brohan, N. Brown, J. Carbajal, Y . Chebotar, X. Chen, K. Choro- manski, T. Ding, D. Driess, A. Dubey, C. Finnet al., “Rt-2: Vision- language-action models transfer web knowledge to robotic control, 2023,”URL https://arxiv. org/abs/2307.15818, vol. 1, p. 2, 2024

    Pith/arXiv arXiv 2023

  19. [19]

    Tla: Tactile-language-action model for contact-rich manipulation,

    P. Hao, C. Zhang, D. Li, X. Cao, X. Hao, S. Cui, and S. Wang, “Tla: Tactile-language-action model for contact-rich manipulation,”arXiv preprint arXiv:2503.08548, 2025

    Pith/arXiv arXiv 2025

  20. [20]

    Sparsh: Self-supervised touch representations for vision-based tactile sensing,

    C. Higuera, A. Sharma, C. K. Bodduluri, T. Fan, P. Lancaster, M. Kalakrishnan, M. Kaess, B. Boots, M. Lambeta, T. Wuet al., “Sparsh: Self-supervised touch representations for vision-based tactile sensing,”arXiv preprint arXiv:2410.24090, 2024

    Pith/arXiv arXiv 2024

  21. [21]

    Unit: Data efficient tactile representation with generalization to unseen objects,

    Z. Xu, R. Uppuluri, X. Zhang, C. Fitch, P. G. Crandall, W. Shou, D. Wang, and Y . She, “Unit: Data efficient tactile representation with generalization to unseen objects,”IEEE Robotics and Automation Letters, 2025

    2025

  22. [22]

    Touch begins where vision ends: Generalizable policies for contact-rich manipula- tion,

    Z. Zhao, S. Haldar, J. Cui, L. Pinto, and R. Bhirangi, “Touch begins where vision ends: Generalizable policies for contact-rich manipula- tion,”arXiv preprint arXiv:2506.13762, 2025

    Pith/arXiv arXiv 2025

  23. [23]

    Forcevla: Enhancing vla models with a force-aware moe for contact-rich manipulation,

    J. Yu, H. Liu, Q. Yu, J. Ren, C. Hao, H. Ding, G. Huang, G. Huang, Y . Song, P. Caiet al., “Forcevla: Enhancing vla models with a force-aware moe for contact-rich manipulation,”arXiv preprint arXiv:2505.22159, 2025

    arXiv 2025

  24. [24]

    Ta-vla: Elucidating the design space of torque-aware vision- language-action models,

    Z. Zhang, H. Xu, Z. Yang, C. Yue, Z. Lin, H.-a. Gao, Z. Wang, and H. Zhao, “Ta-vla: Elucidating the design space of torque-aware vision- language-action models,”arXiv preprint arXiv:2509.07962, 2025

    Pith/arXiv arXiv 2025

  25. [25]

    Tactile- vla: unlocking vision-language-action model’s physical knowledge for tactile generalization,

    J. Huang, S. Wang, F. Lin, Y . Hu, C. Wen, and Y . Gao, “Tactile- vla: unlocking vision-language-action model’s physical knowledge for tactile generalization,”arXiv preprint arXiv:2507.09160, 2025

    Pith/arXiv arXiv 2025

  26. [26]

    Flow matching for generative modeling,

    Y . Lipman, R. T. Chen, H. Ben-Hamu, M. Nickel, and M. Le, “Flow matching for generative modeling,”arXiv preprint arXiv:2210.02747, 2022

    Pith/arXiv arXiv 2022

  27. [27]

    Paligemma: A versatile 3b vlm for transfer,

    L. Beyer, A. Steiner, A. S. Pinto, A. Kolesnikov, X. Wang, D. Salz, M. Neumann, I. Alabdulmohsin, M. Tschannen, E. Bugliarello et al., “Paligemma: A versatile 3b vlm for transfer,”arXiv preprint arXiv:2407.07726, 2024

    Pith/arXiv arXiv 2024

  28. [28]

    What uncertainties do we need in bayesian deep learning for computer vision?

    A. Kendall and Y . Gal, “What uncertainties do we need in bayesian deep learning for computer vision?”Advances in neural information processing systems, vol. 30, 2017

    2017

  29. [29]

    Neural discrete representation learning,

    A. Van Den Oord, O. Vinyalset al., “Neural discrete representation learning,”Advances in neural information processing systems, vol. 30, 2017

    2017

  30. [30]

    Vector-quantized image modeling with improved vqgan,

    J. Yu, X. Li, J. Y . Koh, H. Zhang, R. Pang, J. Qin, A. Ku, Y . Xu, J. Baldridge, and Y . Wu, “Vector-quantized image modeling with improved vqgan,”arXiv preprint arXiv:2110.04627, 2021

    Pith/arXiv arXiv 2021

  31. [31]

    Generating diverse high-fidelity images with vq-vae-2,

    A. Razavi, A. Van den Oord, and O. Vinyals, “Generating diverse high-fidelity images with vq-vae-2,”Advances in neural information processing systems, vol. 32, 2019

    2019

  32. [32]

    Soundstream: An end-to-end neural audio codec,

    N. Zeghidour, A. Luebs, A. Omran, J. Skoglund, and M. Tagliasac- chi, “Soundstream: An end-to-end neural audio codec,”IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 30, pp. 495–507, 2021

    2021

  33. [33]

    Film: Visual reasoning with a general conditioning layer,

    E. Perez, F. Strub, H. De Vries, V . Dumoulin, and A. Courville, “Film: Visual reasoning with a general conditioning layer,” inProceedings of the AAAI conference on artificial intelligence, vol. 32, no. 1, 2018

    2018

  34. [34]

    Diffusion policy: Visuomotor policy learning via action diffusion,

    C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song, “Diffusion policy: Visuomotor policy learning via action diffusion,”The International Journal of Robotics Research, vol. 44, no. 10-11, pp. 1684–1704, 2025

    2025

  35. [35]

    Visualizing data using t-sne

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.” Journal of machine learning research, vol. 9, no. 11, 2008

    2008