Pith. sign in

REVIEW 3 major objections 4 minor 57 references

OC-VLA++: Monocular Geometry-Guided Cross-View Consistency for Viewpoint-Robust Robotic Manipulation

T0 review · 3 major / 4 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 OC-VLA++ claims that explicitly teaching a robot policy how its predicted actions must transform when the camera moves—through synthetic paired views and a cross-view action-equivariance loss—makes it more robust to unseen camera poses…

desk verdict The cross-view equivariance idea is genuinely new and worth airing, but the reported gains sit within sampling noise, so the paper needs stronger evidence before its claims can be taken at face value. read the letter →

arxiv 2608.01066 v1 pith:JDVBOUCL submitted 2026-08-02 cs.RO

Tianyi Zhang , Ziyang Gong , Zhenjie Yang , Zhe Qian , Haonan Duan This is my paper · ORCID
classification cs.RO
keywords vision-language-actionmodelviewpointrobustnesscross-viewactionequivariancemonoculargeometrypaired-viewsupervisioncamera-spacegroundingroboticmanipulationdiffusionpolicy
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 the remaining weakness of camera-space action grounding is not the representation itself but the lack of cross-view supervision: a policy trained from one or a few fixed cameras has never been taught how its action predictions should change when the camera moves. OC-VLA++ therefore pairs every training observation with a nearby synthetic view, generated offline by reprojecting a monocular metric point map through a sampled camera perturbation, and gives each view its own camera-space action target. The two branches share parameters, and the recovered clean-action estimates are transformed into a shared robot frame and aligned by a cross-view action-equivariance loss. On a real robot with 20 rollouts per task, the largest camera shift sees average success of 48.3% with the Dita-based policy and 43.3% with the Qwen3-VL-2B-based policy, versus 40.8% and 37.5% for OC-VLA, while fixed-view success stays at 68.3% and 59.2%.

What carries the argument

The load-bearing object is the paired-view training signal built from monocular geometry. For each real observation, a frozen monocular geometry estimator produces a metric point map and validity mask; local camera perturbations within 15 cm translation and 10 degrees rotation are sampled, the point map is rigidly transformed and reprojected to synthesize a nearby view, and the known camera transform gives that view's camera-space action target. The original and synthesized views share model parameters and the same generative timestep, each is supervised by its own generative loss, and a differentiable clean-action estimate is recovered from each branch and transformed into the robot frame. The cross-view action-equivariance loss then compares translation, quaternion geodesic rotation discrepancy, and gripper command between the synthesized-view prediction and a stop-gradient reference from the original view, with capped Huber terms to limit the influence of synthesis artifacts. This mechanism explicitly teaches the model how actions must transform when the camera moves.

What would settle it

Take a real manipulation scene, capture RGB-D from two calibrated cameras, and train OC-VLA++ with paired views synthesized from the monocular point map; then compare each synthesized view's camera-space action target, computed from the known transform, against the action obtained by transforming the true robot action through the calibrated extrinsics. If the discrepancy exceeds the loss capping thresholds on a large fraction of pixels, the equivariance loss is aligning predictions to erroneous targets, and a variant trained with exact rendered pairs, as in simulation, should outperform it on the same real-robot displacement protocol.

Watch

Extended reading notes

Core claim

The central claim is that viewpoint generalization in robot manipulation can be improved by supervising how actions transform across viewpoints directly, rather than by relying on image-level augmentation or canonicalizing observations to a reference view. Starting from OC-VLA's camera-space action representation, the paper adds two complementary training signals: generative supervision on a synthesized nearby view with that view's correct camera-space target, and an equivariance loss that aligns recovered robot-frame actions from paired views. The design is architecture-agnostic, applying to both diffusion denoisers and flow-matching action experts, adds no new network modules, and is used only at training time. The paper supports the claim with real-robot experiments under increasing camera displacement, simulation experiments with exact paired views, and ablations showing that each training component contributes separately.

Load-bearing premise

The synthesized paired views, produced by reprojecting a monocular point map under a sampled camera transform, are geometrically accurate enough that the camera-space action target derived from the known transform is correct supervision; if the monocular metric geometry is off, the equivariance loss can train the model toward actions that are right for a view that does not exist.

Editorial extensions

If this is right

  • Policies trained from a single fixed camera can be deployed at camera positions tens of centimeters and tens of degrees away, with a smaller success-rate drop than the camera-space baseline.
  • The training recipe carries across generative model families: both a diffusion-transformer policy and a flow-matching action-expert policy benefit, so the method is not tied to one action head.
  • Fixed-view performance is essentially unchanged, so the robustness gain does not trade away in-distribution accuracy.
  • Local paired-view supervision within 15 cm and 10 degrees transfers to larger evaluation shifts up to about 30.5 cm and 23.4 degrees, suggesting the mechanism generalizes beyond the sampled neighborhood.
  • At inference time no geometry estimator or synthesis branch is needed, so deployment cost is identical to the base policy.

Reading between the lines

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

  • If the equivariance premise holds, the same paired-view objective could be applied to any camera-centric action representation, including ones that predict relative to the gripper or an object frame, where synthesizing views would be even cheaper.
  • The valid-pixel filter and capped loss terms imply the method should degrade gracefully as monocular geometry quality worsens; a direct stress test would be to perturb the point map accuracy and measure how much of the gain survives.
  • The finding that local perturbations improve large-displacement robustness suggests that sampling perturbations from a broader distribution, or an online curriculum over perturbation size, may yield further gains without new architecture.
  • The equivariance loss could be reused at test time as a self-check: disagreement between real and synthesized views about the same robot-frame action could flag viewpoint failures before execution, which the paper does not explore.
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

3 major / 4 minor

Summary. This manuscript proposes OC-VLA++, an extension of the OC-VLA policy that augments camera-space action grounding with geometry-guided paired-view supervision and a cross-view action-equivariance objective. For each training observation, the method uses a frozen monocular geometry estimator (MoGe-2) to synthesize nearby camera views offline, assigns each view its own camera-space action target, recovers clean action estimates from diffusion or flow-matching generative outputs, transforms both estimates into a shared robot frame, and penalizes their discrepancy. The authors evaluate on six real-robot tasks with a single training camera and on ManiSkill2, using Dita and Qwen3-VL-2B-based backbones. The central claim is that OC-VLA++ degrades more gracefully than OC-VLA as the evaluation camera moves away from the training pose, while preserving fixed-view performance.

Significance. The idea of explicitly teaching how camera-space action predictions should transform under viewpoint change is a sensible and potentially valuable complement to observation-centric grounding. The method is architecture-agnostic, requires no geometry or view synthesis at inference time, and the simulation setup with exact paired views is a clean way to isolate the effect. The authors should be credited for including two model families, a component ablation, and a loss-weight sensitivity analysis. If the claimed gains were statistically established, this would be a useful contribution. At present, however, the evidence does not clearly separate the reported improvements from sampling noise.

major comments (3)
  1. [Camera-Displacement Evaluation, Figure 4 and Table 6] The headline claim that OC-VLA++ exhibits 'consistently slower degradation' is not statistically supported. At the largest displacement (Position 3), the Dita-based variant succeeds in 58 of 120 rollouts versus 49 of 120 for OC-VLA, a 7.5-point difference; a two-proportion z-test gives p≈0.24. The Qwen-based comparison (52 vs 45 of 120) gives p≈0.36. Per-task counts are 20 rollouts, so the reported 5–10 point per-task gains correspond to 1–2 successful rollouts, and the appendix's own statement that 'a 5-point difference corresponds to one additional successful rollout' makes the fragility explicit. No confidence intervals, error bars, or multiple-seed variance are reported, and the fixed-view Qwen comparison is actually one trial worse than OC-VLA (59.2% vs 60.0%). To support the central claim, the authors need either substantially more rollouts, seed variance, or a pre-specified significance test on the degradation rate.
  2. [Simulation Experiments, Tables 2 and 3] The simulation results are similarly underpowered. Table 2 reports 56.8% versus 52.4% for OC-VLA++ versus OC-VLA; under the 100-episode-per-task protocol stated for Table 4, this is 284 vs 262 successes out of 500, giving p≈0.16. The ablations in Table 3 show margins of 1.0, 1.8, and 2.6 percentage points over the OC-VLA baseline, each consistent with chance at these counts. If the simulation is intended to provide a controlled setting with exact paired views, it is exactly the place where the effect should be visible at high power; currently it is not. Please report episode counts for Table 2, standard errors, and at least one independent seed per configuration.
  3. [Eq. (2) and Paired-View Generation Details] The correctness of the cross-view equivariance target depends on the metric accuracy of MoGe-2's point maps. If the reconstructed geometry is biased or miscalibrated, the synthesized-view camera-space action target A^{c'}_t computed via Eq. (3) will be wrong, and the equivariance loss (Eq. 11) will align predictions to erroneous robot-frame actions. The paper currently supports this premise with a binary valid-pixel filter and visual examples (Figure 8), which do not quantify geometric error. I suggest adding a quantitative validation of the synthesized-view action targets (e.g., reprojection or action-target error against known camera poses, or a simulation experiment with deliberately perturbed geometry) to bound the sensitivity of the method to monocular reconstruction error.
minor comments (4)
  1. [Abstract] The abstract contains a typo: 'We proposeOC-VLA++' is missing a space between 'propose' and 'OC-VLA++'.
  2. [Figure 4] The displacement curves would be much more informative with confidence intervals or per-point standard errors, given the small per-position sample sizes.
  3. [Table 3 caption] The abbreviations 'Synth.' and 'Equiv.' are not defined in the caption; please spell out 'synthesized-view generative objective' and 'cross-view action-equivariance loss' in the caption itself.
  4. [Appendix, Detailed Results under Camera Displacement] The caveat that 'a 5-point difference corresponds to one additional successful rollout' is placed only in the appendix; this limitation is important enough to be stated in the main text alongside the headline numbers.

Circularity Check

0 steps flagged · score 1.0 of 10

No material circularity: OC-VLA++'s central claim is an empirical training result with a new regularizer, and self-citations to OC-VLA are baselines rather than load-bearing derivations.

full rationale

The paper's derivation chain is not circular. The equations define a training objective rather than a derivation that reduces to its inputs: synthesized views are formed by reprojecting MoGe-2 point maps under sampled camera transforms (Eqs. 1-2); each view receives its own camera-space action target (Eq. 3); clean-action estimates are recovered from generative outputs (Eqs. 4-7), transformed into a shared robot frame (Eq. 8), and compared with a Huber/quaternion-based loss (Eqs. 9-11); the total objective combines per-view generative losses with this equivariance term (Eq. 12). The claimed viewpoint-robustness improvement is then obtained by training and measuring success rates under camera displacement, with no fitted parameter renamed as a prediction and no result defined in terms of the quantity it is said to predict. The paper does rely on the authors' own OC-VLA as the comparison baseline and copies some OC-VLA simulation numbers from the prior paper, but this is standard empirical baseline reuse, not a load-bearing self-citation: OC-VLA++ is independently trained and evaluated on the same tasks, and the comparison is externally falsifiable. Citations to MoGe-2, Dita, Qwen3-VL, and pi0 are external. The appendix note that 'a 5-point difference corresponds to one additional successful rollout' and the absence of significance tests are statistical-power limitations, not evidence that the result reduces to its inputs; likewise, the MoGe-2 metric-geometry assumption concerns input quality, not circularity. Score 1 reflects only the minor reuse of the authors' own OC-VLA baseline and evaluation setup.

Assumptions & free parameters 7 free parameters · 3 assumptions · 0 invented entities

The central claim depends on the accuracy of monocular geometry for synthesized views and on several hand-tuned loss weights that are shown to be benign in a sensitivity analysis. No entirely new physical entities are introduced.

free parameters (7)
  • lambda_syn = 0.5
    Weight of the synthesized-view generative loss L_c'_gen in Eq. (12); set by hand, sensitivity analysis shows moderate robustness.
  • lambda_eq = 0.5
    Weight of the cross-view action-equivariance loss L_eq in Eq. (12); set by hand.
  • lambda_p, lambda_R, lambda_g = 1.0, 0.2, 0.1
    Component weights for translation, rotation, and gripper discrepancies in Eq. (11).
  • Huber parameters beta_p, beta_R, beta_g = 0.01, 0.05, 1.0
    Used in the component-wise discrepancy functions in Eq. (10).
  • Discrepancy caps kappa_p, kappa_R = 0.5, 0.5
    Caps on translation and rotation losses to limit influence of synthesis artifacts in Eq. (10).
  • Camera perturbation range = translation <= 15 cm, rotation <= 10 deg
    Range for sampling paired-view camera poses in real-robot view synthesis.
  • Number of synthesized views per observation = 10
    Offline view pool size for real-robot training data.
assumptions (3)
  • domain assumption MoGe-2 provides metric-scale point maps accurate enough for local novel-view synthesis.
    The synthesized views and their camera-space action targets are computed from these point maps (Eq. 1-2); the paper does not quantitatively validate geometry accuracy.
  • domain assumption Camera calibrations (intrinsics and extrinsics) are accurate.
    All action transforms and view reprojections rely on calibration, stated in the paired-view construction and appendix.
  • domain assumption Recovered clean-action estimates from the generative models are suitable for enforcing action-level consistency.
    The equivariance loss (Eq. 11) uses estimates from Eqs. (5) and (7); no analysis of their fidelity is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of OC-VLA++: Monocular Geometry-Guided Cross-View Consistency for Viewpoint-Robust Robotic Manipulation." pith.science (2026). https://pith.science/paper/JDVBOUCL

@misc{pith2026260801066,
  author       = {Pith},
  title        = {Pith review of: OC-VLA++: Monocular Geometry-Guided Cross-View Consistency for Viewpoint-Robust Robotic Manipulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JDVBOUCL}},
  note         = {Machine review of arXiv:2608.01066}
}
read the original abstract

We propose OC-VLA++, an extension of OC-VLA for viewpoint generalization under limited camera coverage. While OC-VLA grounds robot actions in the camera coordinate system to align action supervision with visual observations, camera-space grounding alone can still overfit to the few viewpoints observed during training. OC-VLA++ addresses this limitation by introducing geometry-guided paired-view supervision and an explicit cross-view action-equivariance objective. Given paired observations of the same manipulation scene from geometrically related viewpoints, the model is trained such that their camera-space predictions correspond to the same robot-frame action. This objective explicitly supervises how action predictions should transform across viewpoints, rather than relying solely on image-level augmentation. Experiments demonstrate substantial improvements in unseen-view generalization under limited camera coverage, with performance degrading more gracefully under increasing camera displacement. These results establish cross-view action equivariance as an effective complement to observation-centric action grounding for robust real-world deployment.

Figures

Figures reproduced from arXiv: 2608.01066 by the authors.

Figure 1
Figure 1. Overview of OC-VLA++. Compared with OC-VLA (Zhang et al. 2026) and other methods, OC-VLA++ constructs [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Framework of OC-VLA++. Starting from a single RGB observation, a monocular geometry estimator reconstructs a [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Real-robot experimental setup. We use a Franka [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Real-robot success rates under increasing camera displacement. The left and right panels show the Dita-based and [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Qualitative real-robot results of OC-VLA++ under increasing camera displacement from top to bottom. [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Model structures used in our experiments. The Dita-based model (Hou et al. 2025) (left) uses a unified denoiser to [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Qualitative results on ManiSkill2 (Gu et al. 2023). [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Examples of geometry-guided synthesized views for real-robot training data with MoGe-2 (Wang et al. 2026b). [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 11 canonical work pages

  1. [1]

    arXiv preprint arXiv:2212.06817 , year=

    Rt-1: Robotics transformer for real-world control at scale , author=. arXiv preprint arXiv:2212.06817 , year=

    arXiv

  2. [2]

    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

  3. [3]

    arXiv preprint arXiv:2406.09246 , year=

    Openvla: An open-source vision-language-action model , author=. arXiv preprint arXiv:2406.09246 , year=

    arXiv

  4. [4]

    arXiv preprint arXiv:2502.19645 , year=

    Fine-tuning vision-language-action models: Optimizing speed and success , author=. arXiv preprint arXiv:2502.19645 , year=

    arXiv

  5. [5]

    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=

  6. [6]

    Intelligence, Physical and Black, Kevin and Brown, Noah and Darpinian, James and Dhabalia, Karan and Driess, Danny and Esmail, Adnan and Equi, Michael and Finn, Chelsea and Fusai, Niccolo and others , journal=

  7. [7]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Dita: Scaling diffusion transformer for generalist vision-language-action policy , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  8. [8]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Grounding actions in camera space: Observation-centric vision-language-action policy , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    work page

Show all 57 references
  1. [9]

    arXiv preprint arXiv:2405.12213 , year=

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

    arXiv

  2. [10]

    International Conference on Learning Representations , volume=

    Rdt-1b: a diffusion foundation model for bimanual manipulation , author=. International Conference on Learning Representations , volume=

  3. [11]

    arXiv preprint arXiv:2501.09747 , year=

    Fast: Efficient action tokenization for vision-language-action models , author=. arXiv preprint arXiv:2501.09747 , year=

    arXiv

  4. [12]

    arXiv preprint arXiv:2503.14734 , year=

    Gr00t n1: An open foundation model for generalist humanoid robots , author=. arXiv preprint arXiv:2503.14734 , year=

    arXiv

  5. [13]

    arXiv preprint arXiv:2505.06111 , year=

    Univla: Learning to act anywhere with task-centric latent actions , author=. arXiv preprint arXiv:2505.06111 , year=

    arXiv

  6. [14]

    arXiv preprint arXiv:2506.01844 , year=

    Smolvla: A vision-language-action model for affordable and efficient robotics , author=. arXiv preprint arXiv:2506.01844 , year=

    arXiv

  7. [15]

    arXiv preprint arXiv:2510.02268 , year=

    Do You Know Where Your Camera Is? View-Invariant Policy Learning with Camera Conditioning , author=. arXiv preprint arXiv:2510.02268 , year=

  8. [16]

    Peng, Yue and Zhao, Yongzhe and Habuda, Artur and Pham, Khuyen and Zhu, Yanheng and Le, Tran Nguyen and Abu-Dakka, Fares and Guo, Li , journal=. G

  9. [17]

    arXiv preprint arXiv:2607.11498 , year=

    See like a Robot: Robot-Centric Pointmaps for Vision-Language-Action Models , author=. arXiv preprint arXiv:2607.11498 , year=

    arXiv

  10. [18]

    arXiv preprint arXiv:2607.05396 , year=

    From Fixed to Free Cameras: Calibration-Free View-Robust Vision-Language-Action Model , author=. arXiv preprint arXiv:2607.05396 , year=

    arXiv

  11. [19]

    arXiv preprint arXiv:2409.03685 , year=

    View-invariant policy learning via zero-shot novel view synthesis , author=. arXiv preprint arXiv:2409.03685 , year=

    arXiv

  12. [20]

    arXiv preprint arXiv:2603.29192 , year=

    Efficient Camera Pose Augmentation for View Generalization in Robotic Policy Learning , author=. arXiv preprint arXiv:2603.29192 , year=

  13. [21]

    arXiv preprint arXiv:2603.26757 , year=

    Beyond Viewpoint Generalization: What Multi-View Demonstrations Offer and How to Synthesize Them for Robot Manipulation? , author=. arXiv preprint arXiv:2603.26757 , year=

  14. [22]

    arXiv preprint arXiv:2603.05868 , year=

    AnyCamVLA: Zero-Shot Camera Adaptation for Viewpoint Robust Vision-Language-Action Models , author=. arXiv preprint arXiv:2603.05868 , year=

    arXiv

  15. [23]

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

    Learning to see and act: Task-aware virtual view exploration for robotic manipulation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  16. [24]

    arXiv preprint arXiv:2604.21914 , year=

    VistaBot: View-Robust Robot Manipulation via Spatiotemporal-Aware View Synthesis , author=. arXiv preprint arXiv:2604.21914 , year=

    arXiv

  17. [25]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Towards robust monocular depth estimation: Mixing datasets for zero-shot cross-dataset transfer , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2020 , publisher=

    2020

  18. [26]

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

    Vision transformers for dense prediction , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  19. [27]

    Advances in Neural Information Processing Systems , volume=

    Depth anything v2 , author=. Advances in Neural Information Processing Systems , volume=

  20. [28]

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

    Unidepth: Universal monocular metric depth estimation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  21. [29]

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

    Moge: Unlocking accurate monocular geometry estimation for open-domain images with optimal training supervision , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  22. [30]

    Advances in Neural Information Processing Systems , volume=

    Moge-2: Accurate monocular geometry with metric scale and sharp details , author=. Advances in Neural Information Processing Systems , volume=

  23. [31]

    arXiv preprint arXiv:2410.15959 , year=

    Diffusion transformer policy , author=. arXiv preprint arXiv:2410.15959 , year=

    arXiv

  24. [32]

    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

  25. [33]

    arXiv preprint arXiv:2607.15330 , year=

    Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectories , author=. arXiv preprint arXiv:2607.15330 , year=

    arXiv

  26. [34]

    arXiv preprint arXiv:2607.06403 , year=

    From Foundation to Application: Improving VLA Models in Practice , author=. arXiv preprint arXiv:2607.06403 , year=

    arXiv

  27. [35]

    arXiv preprint arXiv:2607.17977 , year=

    RynnBrain 1.1: Towards More Capable and Generalizable Embodied Foundation Model , author=. arXiv preprint arXiv:2607.17977 , year=

    arXiv

  28. [36]

    arXiv preprint arXiv:2606.17846 , year=

    Qwen-robotmanip technical report: Alignment unlocks scale for robotic manipulation foundation models , author=. arXiv preprint arXiv:2606.17846 , year=

    arXiv

  29. [37]

    arXiv preprint arXiv:2605.30280 , year=

    Qwen-vla: Unifying vision-language-action modeling across tasks, environments, and robot embodiments , author=. arXiv preprint arXiv:2605.30280 , year=

    arXiv

  30. [38]

    arXiv preprint arXiv:2602.11236 , year=

    Abot-m0: Vla foundation model for robotic manipulation with action manifold learning , author=. arXiv preprint arXiv:2602.11236 , year=

    arXiv

  31. [39]

    5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization , author=

    InternVLA-A1. 5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization , author=. arXiv preprint arXiv:2607.04988 , year=

    arXiv

  32. [40]

    arXiv preprint arXiv:2607.04426 , year=

    ACE-Brain-0.5: A Unified Embodied Foundational Model for Physical Agentic AI , author=. arXiv preprint arXiv:2607.04426 , year=

    arXiv

  33. [41]

    arXiv preprint arXiv:2510.11027 , year=

    Vlaser: Vision-Language-Action Model with Synergistic Embodied Reasoning , author=. arXiv preprint arXiv:2510.11027 , year=

  34. [42]

    8th Annual Conference on Robot Learning , year=

    Rovi-aug: Robot and viewpoint augmentation for cross-embodiment robot learning , author=. 8th Annual Conference on Robot Learning , year=

  35. [43]

    arXiv preprint arXiv:2509.22578 , year=

    Egodemogen: Novel egocentric demonstration generation enables viewpoint-robust manipulation , author=. arXiv preprint arXiv:2509.22578 , year=

  36. [44]

    arXiv preprint arXiv:2601.09605 , year=

    Sim2real Image Translation Enables Viewpoint-Robust Policies from Fixed-Camera Datasets , author=. arXiv preprint arXiv:2601.09605 , year=

  37. [45]

    arXiv preprint arXiv:2501.15830 , year=

    Spatialvla: Exploring spatial representations for visual-language-action model , author=. arXiv preprint arXiv:2501.15830 , year=

    arXiv

  38. [46]

    arXiv preprint arXiv:2407.01479 , year=

    Equibot: Sim (3)-equivariant diffusion policy for generalizable and data efficient learning , author=. arXiv preprint arXiv:2407.01479 , year=

    arXiv

  39. [47]

    International Conference on Learning Representations , volume=

    Et-seed: Efficient trajectory-level se (3) equivariant diffusion policy , author=. International Conference on Learning Representations , volume=

  40. [48]

    arXiv preprint arXiv:2411.18623 , year=

    Lift3d foundation policy: Lifting 2d large-scale pretrained models for robust 3d robotic manipulation , author=. arXiv preprint arXiv:2411.18623 , year=

    arXiv

  41. [49]

    arXiv preprint arXiv:2601.08325 , year=

    ActiveVLA: Injecting Active Perception into Vision-Language-Action Models for Precise 3D Robotic Manipulation , author=. arXiv preprint arXiv:2601.08325 , year=

  42. [50]

    arXiv preprint arXiv:2302.04659 , year=

    Maniskill2: A unified benchmark for generalizable manipulation skills , author=. arXiv preprint arXiv:2302.04659 , year=

    arXiv

  43. [51]

    arXiv preprint arXiv:2511.21631 , year=

    Qwen3-vl technical report , author=. arXiv preprint arXiv:2511.21631 , year=

    arXiv

  44. [52]

    arXiv preprint arXiv:2509.15733 , year=

    GP3: A 3D Geometry-Aware Policy with Multi-View Images for Robotic Manipulation , author=. arXiv preprint arXiv:2509.15733 , year=

  45. [53]

    arXiv preprint arXiv:2402.10885 , year=

    3d diffuser actor: Policy diffusion with 3d scene representations , author=. arXiv preprint arXiv:2402.10885 , year=

    arXiv

  46. [54]

    Physical Intelligence and Amin, Ali and Aniceto, Raichelle and Balakrishna, Ashwin and Black, Kevin and Conley, Ken and Connors, Grace and Darpinian, James and Dhabalia, Karan and DiCarlo, Jared and others , journal=

  47. [55]

    Physical Intelligence and Ai, Bo and Amin, Ali and Aniceto, Raichelle and Balakrishna, Ashwin and Balke, Greg and Black, Kevin and Bokinsky, George and Cao, Shihao and Charbonnier, Thomas and others , journal=

  48. [56]

    arXiv preprint arXiv:2403.12945 , year=

    Droid: A large-scale in-the-wild robot manipulation dataset , author=. arXiv preprint arXiv:2403.12945 , year=

    arXiv

  49. [57]

    arXiv preprint arXiv:1711.05101 , year=

    Decoupled weight decay regularization , author=. arXiv preprint arXiv:1711.05101 , year=

    arXiv

Pith tools

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