Pith. sign in

REVIEW 4 major objections 5 minor 40 references

Spatial RoboGrasp: Generalized Robotic Grasping Control Policy

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a diffusion-based grasping policy generalizes across lighting, object, and grasp-type variation when its observations combine monocular depth, domain-randomized augmentation, and a depth-derived grasp prompt.

desk verdict A credible integration of known modules with real-robot ablations, but the metric-depth assumption behind the '6-DoF' prompt is unverified and undercuts the central spatial-grounding claim. read the letter →

arxiv 2505.20814 v1 pith:P4PRB4OS submitted 2025-05-27 cs.RO cs.CV

Yiqi Huang , Travis Davies , Jiahuan Yan , Jiankai Sun , Xiang Chen , Luhui Hu This is my paper
classification cs.ROcs.CV
keywords roboticgraspingdiffusionpolicymonoculardepthestimationgraspaffordancesdomainrandomizationvisualrobustnessimitationlearning6-DoFprompt
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

Robotic grasping policies trained on raw RGB tend to degrade when lighting, occlusion, or object appearance changes. This paper proposes to append two cheap spatial signals to the RGB stream: a monocular depth estimate and a structured grasp prompt that tells the policy where and how to contact the object, while also training with domain-randomized augmentations that push the model toward those signals. The claim is that this spatial perception stack, feeding a diffusion-based action model, is what lets a grasping policy generalize to new objects, few-shot examples, and extreme exposures. On three real-robot tasks, the full model averages 82%, 82%, and 47% task success across exposure levels, compared with 42%, 27%, and 7% for the diffusion-policy baseline. The paper reads these results as evidence that spatially grounded perception, not just more data, is the key to robust manipulation.

What carries the argument

The load-bearing object is the 6-DoF Grasp Prompt, defined in the paper as a predicted end-effector pose and gripper width derived from RGB-D input that explicitly conditions the diffusion policy. It is built from a lightweight detection network's 2D grasp box (x, y, w, h, θ), the predicted monocular depth at the box center, and known camera intrinsics: the 2D center is projected to a 3D position and the in-plane angle becomes a rotation matrix whose approach axis is fixed to the camera optical axis, z = [0, 0, 1]. The same pipeline generates grasp annotations for the training set, so the prompt is available at both training and inference. Around this prompt, the architecture includes AugFusion, a probabilistic mixer of realistic image corruptions; a monocular depth module that pre-extracts depth for training and runs a lighter network at inference; an observation encoder that pools multi-view RGB features with robot state and the prompt through a short temporal transformer; and a diffusion-based action head that denoises actions over sixteen steps with a cosine beta schedule.

What would settle it

Move the camera in the PickCup setup so that the sidewall and diameter grasp approach directions are oblique to the optical axis while keeping the same objects; if task success collapses below the reported 82% average, the fixed-axis assumption is the binding constraint, and if success holds, the policy is getting the missing orientation from elsewhere.

Watch

Extended reading notes

Core claim

The paper's central claim is that combining domain-randomized visual augmentation, monocular depth estimation, and a depth-derived grasp prompt into one observation representation makes a diffusion-based imitation policy substantially more robust to environmental variation and better at executing specified grasp strategies. The grasp prompt is computed by detecting a 2D oriented grasp box, lifting its center to 3D through the predicted depth map and camera intrinsics, and building a rotation matrix from the in-plane angle with a fixed approach direction along the camera axis; this prompt is fed alongside RGB, depth, robot state, and a task prompt into a transformer-conditioned denoising head. The authors report that each module contributes: depth alone adds 34% task success in PickBig, AugFusion adds 24% in PickCup at mid-range exposures, and the grasp prompt improves spatial precision in PickBig and PickGoods. Combined, the full model outperforms the diffusion-policy baseline by 15 to 30 percentage points across metrics and maintains high success at exposure extremes where the baseline collapses. The paper interprets the combined results as showing that spatial grounding is the active ingredient, not any single input modality.

Load-bearing premise

The approach direction of every grasp is fixed to the camera optical axis, so a grasp whose required approach is not parallel to that axis cannot be represented by the grasp prompt; if the camera is not aligned with those approaches, the reported success on sidewall and diameter grasps cannot be carried by the prompt.

Editorial extensions

If this is right

  • A grasping policy that consumes this spatial representation should keep high success across exposure extremes, because the depth map and grasp prompt carry geometric information that survives RGB corruption.
  • Few-shot transfer to new objects should improve when the policy receives an explicit grasp-type prompt, since the prompt disambiguates handle, wall, and diameter strategies.
  • Prompt-following in cluttered scenes becomes tractable: the grasp prompt tells the policy which object to contact, letting it disambiguate goal directionality.
  • The three modules address separate failure modes: depth adds geometric priors, augmentation widens the training distribution, and the grasp prompt adds contact-aware spatial guidance, so the combined stack is what the paper credits.
  • Robust grasping can be achieved without dedicated depth hardware or point-cloud processing, using only monocular depth and a lightweight grasp detector.

Reading between the lines

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

  • The paper's '6-DoF' label is stronger than the prompt's actual representational power: because the approach axis is fixed to the camera optical axis, the prompt encodes a 3D position plus yaw rather than a general 6-DoF pose, so true 6-DoF coverage would require varying the approach direction.
  • A direct test of this limitation is to rotate the camera relative to the grasp approach in the PickCup setup and remeasure task success; if success collapses, the fixed-axis assumption is the active constraint, and if it holds, the policy is getting the missing orientation from somewhere else.
  • The interaction between AugFusion and depth is likely the robustness engine: augmentations that degrade RGB may force the policy to rely on depth, so an ablation that removes each component separately would reveal which one is bearing the weight.
  • The paper's proposed future work of learning grasp prompts end-to-end from the same encoder would remove the separate detector and make the prompt joint with the policy, and the fixed-axis assumption could also be lifted by predicting a full rotation instead of only an in-plane angle.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes Spatial RoboGrasp, a diffusion-policy grasping controller whose observation encoder fuses domain-randomized RGB augmentation (AugFusion), monocular depth maps from Depth Anything V2, and a '6-DoF Grasp Prompt' obtained by lifting YOLO-predicted 2D grasp boxes with the predicted depth map and camera intrinsics. The policy is evaluated on three real-robot tasks (PickBig, PickCup, PickGoods) across ten exposure levels, with Task Success Rate (TSR) and Grasp Success Rate (GSR) as metrics. The central claim is that combining these perception modules yields spatially grounded, generalizable, and precise manipulation, with the full model achieving the highest TSR/GSR in all tasks.

Significance. If the reported results are correct, the paper would provide a practically useful integration of cheap monocular depth and grasp affordance prompts into diffusion policies, showing robustness to lighting variation without additional 3D sensors. The experimental design (multiple exposures, multiple grasp strategies, few-shot cups) is well suited to the claim. However, the significance is conditional on resolving the metric-depth and degree-of-freedom concerns below. The paper does not provide code, data, or detailed calibration information, so the central quantitative claims are currently not independently verifiable.

major comments (4)
  1. [§3.2, §3.3, §4.2] The 6-DoF Grasp Prompt is not shown to be metrically grounded. Section 3.2 adopts Depth Anything V2, whose output is relative depth with an unknown per-image scale and shift, yet Section 3.3 projects the 2D grasp center using x3D = (x − cx)·z/fx, y3D = (y − cy)·z/fy, z3D = z as if z were metric distance. No scale/shift calibration is described, either per camera, per scene, or globally. Consequently x3D, y3D, z3D, and any gripper width derived from box dimensions are in arbitrary image-dependent units, so the same physical grasp can produce very different prompt vectors across training and deployment. Since the paper's core claim is that a spatially grounded prompt enables precise manipulation, this missing calibration is load-bearing. Please either document the calibration procedure, use a metric depth sensor or metric-depth model, or explicitly show that the downstream policy is invariant to the unknown depth scale.
  2. [§3.3] The constructed pose is not 6-DoF. The rotation matrix is built with z = [0,0,1], so the orientation has only in-plane yaw; together with the 3D position this is at most a 4-DoF pose. Moreover, the approach direction is fixed to the camera optical axis, which makes sidewall and handle grasps in PickCup (Figure 7) geometrically unrepresentable unless the camera axis exactly aligns with the required approach direction for every grasp in all three tasks. The manuscript should state which camera view provides the depth used for each grasp prompt, justify that the fixed axis assumption holds for all reported grasp types, or rename the prompt to a lower-DoF spatial cue and avoid the claim of full 6-DoF grounding.
  3. [§4.1, §4.3, Table 1] The trial counts are internally inconsistent with the evaluation protocol. Section 4.3 says that for each exposure level, every model is evaluated over 100–150 trials per task, and Table 1 reports 10 exposure levels for 5 model variants. That implies at least 5 × 10 × 100 = 5000 trials per task, yet Section 4.1 reports only 600 (PickBig), 315 (PickCup), and 400 (PickGoods) total trials per task. Please clarify how the exposure-level trials were allocated and whether the reported totals include all model conditions or only part of them.
  4. [§5.1, Abstract, Table 1] Several numerical claims in the text are not supported by Table 1. Section 5.1 states that in PickBig the full model outperforms all four baselines 'by over 14%' in both TSR and GSR, but the best baseline (+Depth) has 76% TSR and 74% GSR versus Ours at 82% and 81%, differences of 6 and 7 points. The abstract claims 'up to 40% improvement in grasp success,' but the largest GSR improvement in Table 1 is 59 points (PickCup: 80% vs 21%), and the PickGoods improvement is 39 points. These discrepancies should be corrected or the statements should be rephrased with the actual ranges.
minor comments (5)
  1. [Algorithm 1, line 16] In the else branch (ξ > β is not taken), the expression ai(xt) uses ai before it is defined in that branch; the sequential-augmentation logic is unclear and should be rewritten with explicit notation.
  2. [§3.3] The paper mentions 'YOLOv11-m' but gives no reference or implementation detail; please cite the specific YOLO version and include the training configuration.
  3. [§2, §3.2] The two core modules, RoboGrasp [Huang et al. 2025] and SVP [Davies et al. 2024], are unpublished preprints by overlapping authors. The paper should include enough self-contained detail about these modules so that reviewers and readers can assess the novelty of their combination.
  4. [§4.2] No camera intrinsics, camera poses, or calibration results are reported, despite the central role of 'known camera intrinsics' in the 3D projection. Please provide these values or a reference to the full technical appendix.
  5. [Figure 1 and Figure 2 captions] The captions contain grammatical and clarity issues: Figure 1's caption says 'These observation conditions and robot state data to enhance...', and Figure 2's caption should specify which camera view and depth map were used to generate the 6-DoF prompt.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported success rates are empirical robot-trial measurements, and the cited prior work by the authors is contextual rather than load-bearing.

full rationale

The central results are physical trial outcomes measured under a fixed exposure protocol, with the same protocol applied to all baselines and ablations. The depth component is a third-party pretrained model (Depth Anything V2), and the baseline is an externally published Diffusion Policy. The self-citations to Huang et al. [2025] and Davies et al. [2024] are used as contextual motivation, not as a substitute for the experiments reported here: the architecture is specified in this paper (Algorithm 1, Section 3.3 projection equations, Section 3.4 encoder), and the claimed gains are not derived from those preprints by construction. The remaining concerns — that monocular depth from Depth Anything V2 is relative and not explicitly calibrated to metric units, and that the rotation construction assumes a top-down approach direction that may not represent sidewall grasps — are correctness or validity risks, not cases where a prediction reduces by definition to a fitted input or to a self-citation chain. Since the success-rate tables are empirical measurements with independent baseline comparisons, no circular step is exhibited.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central claim depends on five domain assumptions about depth quality, grasp-pose geometry, exposure distribution, grasp detection reliability, and human labeling, none of which are validated with independent data. No new physical entities are introduced; the 'Grasp Prompt' is a feature vector, not an independent ontological commitment.

free parameters (3)
  • AugFusion mixing parameters (α, β, λ, k) = not specified
    Algorithm 1 defines augmentation mixing via Dirichlet weights and a gate threshold; these are hand-chosen hyperparameters that control training distribution. No sensitivity analysis is given, so their effect on the reported robustness gains is unknown.
  • Number of diffusion denoising steps = 16
    The action head refines noise over 16 timesteps with a DDIM scheduler; this is a fixed architectural choice without ablation, and it affects action quality and inference speed.
  • Temporal self-attention context length = 2 timesteps
    The observation encoder applies self-attention over tokens from the previous two timesteps. This context length is chosen without reported ablations, and the policy's temporal behavior depends on it.
assumptions (5)
  • domain assumption Monocular depth estimates from Depth Anything V2 are accurate enough for 2D grasp boxes to be projected into correct 3D poses.
    Sections 3.2 and 3.3 compute x3D, y3D, z3D directly from the predicted depth map without any validation of depth accuracy; errors in depth propagate directly into the grasp prompt and the policy's actions.
  • domain assumption Grasp approach directions are always aligned with the camera optical axis, with only in-plane rotation.
    Section 3.3 sets z = [0,0,1] in the rotation matrix. This restricts the end-effector to approach along the camera viewing direction, which is unlikely to cover the sidewall and diameter grasps described in PickCup.
  • domain assumption The extreme exposure levels (10-40 ms and 160-170 ms) are out-of-distribution for the trained policy.
    The paper never states the exposure distribution of the training demonstrations or whether AugFusion includes those extremes. If the training augmentations already cover the full exposure range, the robustness result is partly interpolation rather than generalization.
  • domain assumption A single high-confidence grasp box per frame is sufficient to condition the policy in all three tasks.
    Section 3.3 outputs one grasp box per frame; in cluttered PickGoods scenes the correct target may not be the highest-confidence box, and the detector's precision/recall is never reported.
  • domain assumption Human consensus among 2-3 raters is a reliable success label.
    Section 4.3 defines success through subjective consensus; no grading rubric or inter-rater agreement statistic is given, so label noise could affect all reported numbers.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Spatial RoboGrasp: Generalized Robotic Grasping Control Policy." pith.science (2026). https://pith.science/paper/P4PRB4OS

@misc{pith2026250520814,
  author       = {Pith},
  title        = {Pith review of: Spatial RoboGrasp: Generalized Robotic Grasping Control Policy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P4PRB4OS}},
  note         = {Machine review of arXiv:2505.20814}
}
read the original abstract

Achieving generalizable and precise robotic manipulation across diverse environments remains a critical challenge, largely due to limitations in spatial perception. While prior imitation-learning approaches have made progress, their reliance on raw RGB inputs and handcrafted features often leads to overfitting and poor 3D reasoning under varied lighting, occlusion, and object conditions. In this paper, we propose a unified framework that couples robust multimodal perception with reliable grasp prediction. Our architecture fuses domain-randomized augmentation, monocular depth estimation, and a depth-aware 6-DoF Grasp Prompt into a single spatial representation for downstream action planning. Conditioned on this encoding and a high-level task prompt, our diffusion-based policy yields precise action sequences, achieving up to 40% improvement in grasp success and 45% higher task success rates under environmental variation. These results demonstrate that spatially grounded perception, paired with diffusion-based imitation learning, offers a scalable and robust solution for general-purpose robotic grasping.

Figures

Figures reproduced from arXiv: 2505.20814 by the authors.

Figure 1
Figure 1. An overview Spatial RoboGrasp architecture, demonstrating the integration of image [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Grasp prompt visualization in the Pick￾Big task. (a) Predicted oriented 2D grasp box from the RGB image. (b) Corresponding 6-DoF grasp prompt derived using camera intrinsics and a rota￾tion matrix. At both training and inference stages, the grasp detection module outputs a single high￾confidence grasp box per frame. During infer￾ence, this 2D box—comprising its center coordi￾nates, size, and in-plane angle—is transf… view at source ↗
Figure 3
Figure 3. Placement generalization setup for the PickBig task. (a) and (b) illustrate two of the eight object configurations. The goal is to identify and grasp the larger of two similarly shaped blocks along its diameter. The observation encoder fuses multiview RGB inputs, low-dimensional robot states, and grasp￾specific features into a unified spatial-temporal representation for diffusion-based policy learn￾ing. Each fixed c… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Few-shot PickCup setup. (a) Green mug with 5-shot handle grasping. (b) Blue plastic cup with 10-shot diameter grasping. Our action head adopts a diffusion-based policy Chi et al. [2024a], which models action genera￾tion as a denoising process from Gaussian noise to exp…
Figure 5
Figure 5. Figure 5: Promptable PickGoods setup. (a) Grasp prompt for a chocolate bar; (b) for a biscuit. The goal is to follow prompts to pick the target item. Each demonstration includes synchronized RGB and monocular depth frames. A representa￾tive subset is annotated with 2D grasp boxe…
Figure 6
Figure 6. Figure 6: Real-time orientated-2D grasp box predictions across tasks using Spatial RoboGrasp. It [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Illustration of grasping strategies for the PickCup task (top-down view). (a) shows a [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

40 extracted references · 13 canonical work pages

  1. [1]

    Apollo: An open autonomous driving platform

    ApolloAuto. Apollo: An open autonomous driving platform. https://github.com/ApolloAuto/apollo, 2023. Accessed: October 18, 2023

    work page 2023

  2. [2]

    Midas v3.1 -- a model zoo for robust monocular relative depth estimation, 2023

    Reiner Birkl, Diana Wofk, and Matthias Müller. Midas v3.1 -- a model zoo for robust monocular relative depth estimation, 2023. URL https://arxiv.org/abs/2307.14460

    arXiv 2023

  3. [3]

    _0 : A vision-language-action flow model for general robot control, 2024

    Kevin Black, Noah Brown, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, Lachy Groom, Karol Hausman, Brian Ichter, Szymon Jakubczak, Tim Jones, Liyiming Ke, Sergey Levine, Adrian Li-Bell, Mohith Mothukuri, Suraj Nair, Karl Pertsch, Lucy Xiaoyang Shi, James Tanner, Quan Vuong, Anna Walling, Haohuan Wang, and Ury Zhilinsky. _0 : A vis...

    arXiv 2024

  4. [4]

    Diffusion policy: Visuomotor policy learning via action diffusion

    Cheng Chi, Siyuan Feng, Yilun Du, Zhenjia Xu, Eric Cousineau, Benjamin Burchfiel, and Shuran Song. Diffusion policy: Visuomotor policy learning via action diffusion. In Proceedings of Robotics: Science and Systems (RSS), 2023

    2023

  5. [5]

    Diffusion policy: Visuomotor policy learning via action diffusion, 2024 a

    Cheng Chi, Zhenjia Xu, Siyuan Feng, Eric Cousineau, Yilun Du, Benjamin Burchfiel, Russ Tedrake, and Shuran Song. Diffusion policy: Visuomotor policy learning via action diffusion, 2024 a . URL https://arxiv.org/abs/2303.04137

    arXiv 2024

  6. [6]

    Universal manipulation interface: In-the-wild robot teaching without in-the-wild robots, 2024 b

    Cheng Chi, Zhenjia Xu, Chuer Pan, Eric Cousineau, Benjamin Burchfiel, Siyuan Feng, Russ Tedrake, and Shuran Song. Universal manipulation interface: In-the-wild robot teaching without in-the-wild robots, 2024 b . URL https://arxiv.org/abs/2402.10329

    arXiv 2024

  7. [7]

    Spatially Visual Perception for End-to-End Robotic Learning

    Travis Davies, Jiahuan Yan, Xiang Chen, Yu Tian, Yueting Zhuang, Yiqi Huang, and Luhui Hu. Spatially visual perception for end-to-end robotic learning, 2024. URL https://arxiv.org/abs/2411.17458

    work page Pith review arXiv 2024

  8. [8]

    Benchmarking Robustness of 3D Object Detection to Common Corruptions in Autonomous Driving

    Yinpeng Dong, Caixin Kang, Jinlai Zhang, Zijian Zhu, Yikai Wang, Xiao Yang, Hang Su, Xingxing Wei, and Jun Zhu. Benchmarking robustness of 3d object detection to common corruptions in autonomous driving, 2023. URL https://arxiv.org/abs/2303.11040

    work page Pith review arXiv 2023

Show all 40 references
  1. [9]

    Open x-embodiment: Robotic learning datasets and rt-x models, 2024

    Embodiment Collaboration et al. Open x-embodiment: Robotic learning datasets and rt-x models, 2024. URL https://arxiv.org/abs/2310.08864

    2024 arXiv

  2. [10]

    Zhao, and Chelsea Finn

    Zipeng Fu, Tony Z. Zhao, and Chelsea Finn. Mobile aloha: Learning bimanual mobile manipulation with low-cost whole-body teleoperation, 2024. URL https://arxiv.org/abs/2401.02117

    2024 arXiv

  3. [11]

    Deep residual learning for image recognition, 2015

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition, 2015. URL https://arxiv.org/abs/1512.03385

    2015 arXiv

  4. [12]

    Cubuk, Barret Zoph, Justin Gilmer, and Balaji Lakshminarayanan

    Dan Hendrycks, Norman Mu, Ekin D. Cubuk, Barret Zoph, Justin Gilmer, and Balaji Lakshminarayanan. Augmix: A simple data processing method to improve robustness and uncertainty, 2020. URL https://arxiv.org/abs/1912.02781

    2020 arXiv

  5. [13]

    Rekep: Spatio-temporal reasoning of relational keypoint constraints for robotic manipulation, 2024

    Wenlong Huang, Chen Wang, Yunzhu Li, Ruohan Zhang, and Li Fei-Fei. Rekep: Spatio-temporal reasoning of relational keypoint constraints for robotic manipulation, 2024. URL https://arxiv.org/abs/2409.01652

    2024 arXiv

  6. [14]

    Robograsp: A universal grasping policy for robust robotic control, 2025

    Yiqi Huang, Travis Davies, Jiahuan Yan, Xiang Chen, Yu Tian, and Luhui Hu. Robograsp: A universal grasping policy for robust robotic control, 2025. URL https://arxiv.org/abs/2502.03072

    2025 arXiv

  7. [15]

    Yeying Jin, Beibei Lin, Wending Yan, Yuan Yuan, Wei Ye, and Robby T. Tan. Enhancing visibility in nighttime haze images using guided apsf and gradient adaptive convolution, 2024. URL https://arxiv.org/abs/2308.01738

    2024 arXiv

  8. [16]

    A survey on Learning-Based robotic grasping

    Kilian Kleeberger, Richard Bormann, Werner Kraus, and Marco F Huber. A survey on Learning-Based robotic grasping. Current Robotics Reports, 1 0 (4): 0 239--249, December 2020

    2020

  9. [17]

    End-to-end planning of autonomous driving in industry and academia: 2022-2023, 2023

    Gongjin Lan and Qi Hao. End-to-end planning of autonomous driving in industry and academia: 2022-2023, 2023. URL https://arxiv.org/abs/2401.08658

    2022 arXiv

  10. [18]

    Okami: Teaching humanoid robots manipulation skills through single video imitation

    Jinhan Li, Yifeng Zhu, Yuqi Xie, Zhenyu Jiang, Mingyo Seo, Georgios Pavlakos, and Yuke Zhu. Okami: Teaching humanoid robots manipulation skills through single video imitation. In 8th Annual Conference on Robot Learning (CoRL), 2024

    2024

  11. [19]

    Robust visual imitation learning with inverse dynamics representations, 2023

    Siyuan Li, Xun Wang, Rongchang Zuo, Kewu Sun, Lingfei Cui, Jishiyu Ding, Peng Liu, and Zhe Ma. Robust visual imitation learning with inverse dynamics representations, 2023. URL https://arxiv.org/abs/2310.14274

    2023 arXiv

  12. [20]

    Data scaling laws in imitation learning for robotic manipulation

    Fanqi Lin, Yingdong Hu, Pingyue Sheng, Chuan Wen, Jiacheng You, and Yang Gao. Data scaling laws in imitation learning for robotic manipulation. arXiv preprint arXiv:2410.18647, 2024

    2024 arXiv

  13. [21]

    Feature pyramid networks for object detection, 2017

    Tsung-Yi Lin, Piotr Dollár, Ross Girshick, Kaiming He, Bharath Hariharan, and Serge Belongie. Feature pyramid networks for object detection, 2017. URL https://arxiv.org/abs/1612.03144

    2017 arXiv

  14. [22]

    Moka: Open-world robotic manipulation through mark-based visual prompting, 2024 a

    Fangchen Liu, Kuan Fang, Pieter Abbeel, and Sergey Levine. Moka: Open-world robotic manipulation through mark-based visual prompting, 2024 a . URL https://arxiv.org/abs/2403.03174

    2024 arXiv

  15. [23]

    The practice of mass production autonomous driving

    Langechuan Patrick Liu. The practice of mass production autonomous driving. Presented at the CVPR 2023 E2EAD Workshop, 2023. Available at https://opendrivelab.com/e2ead/cvpr23

    2023

  16. [24]

    Rdt-1b: a diffusion foundation model for bimanual manipulation, 2024 b

    Songming Liu, Lingxuan Wu, Bangguo Li, Hengkai Tan, Huayu Chen, Zhengyi Wang, Ke Xu, Hang Su, and Jun Zhu. Rdt-1b: a diffusion foundation model for bimanual manipulation, 2024 b . URL https://arxiv.org/abs/2410.07864

    2024 arXiv

  17. [25]

    Dex-net 2.0: Deep learning to plan robust grasps with synthetic point clouds and analytic grasp metrics, 2017

    Jeffrey Mahler, Jacky Liang, Sherdil Niyaz, Michael Laskey, Richard Doan, Xinyu Liu, Juan Aparicio Ojea, and Ken Goldberg. Dex-net 2.0: Deep learning to plan robust grasps with synthetic point clouds and analytic grasp metrics, 2017. URL https://arxiv.org/abs/1703.09312

    2017 arXiv

  18. [26]

    Improved denoising diffusion probabilistic models, 2021

    Alex Nichol and Prafulla Dhariwal. Improved denoising diffusion probabilistic models, 2021. URL https://arxiv.org/abs/2102.09672

    2021 arXiv

  19. [27]

    Dinov2: Learning robust visual features without supervision, 2024

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy Vo, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, Mahmoud Assran, Nicolas Ballas, Wojciech Galuba, Russell Howes, Po-Yao Huang, Shang-Wen Li, Ishan Misra, Michael Rabba...

    2024 arXiv

  20. [28]

    Alvinn: An autonomous land vehicle in a neural network

    Dean Pomerleau. Alvinn: An autonomous land vehicle in a neural network. In D.S. Touretzky, editor, Proceedings of (NeurIPS) Neural Information Processing Systems, pages 305 -- 313. Morgan Kaufmann, December 1989

    1989

  21. [29]

    Robust multimodal vehicle detection in foggy weather using complementary lidar and radar signals

    Kun Qian, Shilin Zhu, Xinyu Zhang, and Li Erran Li. Robust multimodal vehicle detection in foggy weather using complementary lidar and radar signals. In 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 444--453, 2021. doi:10.1109/CVPR46437.2021.00051

    2021

  22. [30]

    3d-outdet: A fast and memory efficient outlier detector for 3d lidar point clouds in adverse weather

    Abu Mohammed Raisuddin, Tiago Cortinhal, Jesper Holmblad, and Eren Erdal Aksoy. 3d-outdet: A fast and memory efficient outlier detector for 3d lidar point clouds in adverse weather. October 2023. doi:10.36227/techrxiv.24297166.v1. URL http://dx.doi.org/10.36227/techrxiv.24297166.v1

    2023 doi

  23. [31]

    Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra

    Ramprasaath R. Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad-cam: Visual explanations from deep networks via gradient-based localization. International Journal of Computer Vision, 128 0 (2): 0 336–359, October 2019. ISSN 15...

    2019 doi

  24. [32]

    Cliport: What and where pathways for robotic manipulation, 2021

    Mohit Shridhar, Lucas Manuelli, and Dieter Fox. Cliport: What and where pathways for robotic manipulation, 2021. URL https://arxiv.org/abs/2109.12098

    2021 arXiv

  25. [33]

    Kalie: Fine-tuning vision-language models for open-world manipulation without robot data, 2024

    Grace Tang, Swetha Rajkumar, Yifei Zhou, Homer Rich Walke, Sergey Levine, and Kuan Fang. Kalie: Fine-tuning vision-language models for open-world manipulation without robot data, 2024. URL https://arxiv.org/abs/2409.14066

    2024 arXiv

  26. [34]

    ALOHA 2 Team, Jorge Aldaco, Travis Armstrong, Robert Baruch, Jeff Bingham, Sanky Chan, Kenneth Draper, Debidatta Dwibedi, Chelsea Finn, Pete Florence, Spencer Goodrich, Wayne Gramlich, Torr Hage, Alexander Herzog, Jonathan Hoech, Thinh Nguyen, Ian Storz, Baruch Tabanpour, Leil...

    2024 arXiv

  27. [35]

    Grasp-anything: Large-scale grasp dataset from foundation models, 2023

    An Dinh Vuong, Minh Nhat Vu, Hieu Le, Baoru Huang, Binh Huynh, Thieu Vo, Andreas Kugi, and Anh Nguyen. Grasp-anything: Large-scale grasp dataset from foundation models, 2023. URL https://arxiv.org/abs/2309.09818

    2023 arXiv

  28. [36]

    Depth anything v2, 2024

    Lihe Yang, Bingyi Kang, Zilong Huang, Zhen Zhao, Xiaogang Xu, Jiashi Feng, and Hengshuang Zhao. Depth anything v2, 2024. URL https://arxiv.org/abs/2406.09414

    2024 arXiv

  29. [37]

    Primedepth: Efficient monocular depth estimation with a stable diffusion preimage, 2024

    Denis Zavadski, Damjan Kalšan, and Carsten Rother. Primedepth: Efficient monocular depth estimation with a stable diffusion preimage, 2024. URL https://arxiv.org/abs/2409.09144

    2024 arXiv

  30. [38]

    Coinrobot: Generalized end-to-end robotic learning for physical intelligence, 2025

    Yu Zhao, Huxian Liu, Xiang Chen, Jiankai Sun, Jiahuan Yan, and Luhui Hu. Coinrobot: Generalized end-to-end robotic learning for physical intelligence, 2025. URL https://arxiv.org/abs/2503.05316

    2025 arXiv

  31. [39]

    Autofed: Heterogeneity-aware federated multimodal learning for robust autonomous driving, 2023

    Tianyue Zheng, Ang Li, Zhe Chen, Hongbo Wang, and Jun Luo. Autofed: Heterogeneity-aware federated multimodal learning for robust autonomous driving, 2023. URL https://arxiv.org/abs/2302.08646

    2023 arXiv

  32. [40]

    Domain generalization: A survey

    Kaiyang Zhou, Ziwei Liu, Yu Qiao, Tao Xiang, and Chen Change Loy. Domain generalization: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, page 1–20, 2022. ISSN 1939-3539. doi:10.1109/tpami.2022.3195549. URL http://dx.doi.org/10.1109/TPAMI.2022.3195549

    2022

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.