Pith. sign in

REVIEW 3 major objections 5 minor 15 cited by

DextrAH-RGB: Visuomotor Policies to Grasp Anything with Dexterous Hands

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read RGB-only dexterous grasping policies trained purely in simulation transfer to the real world and rival depth-based methods.

desk verdict The first convincing RGB-only dexterous grasping transfer is real, but the 'competitive' claim is softer than the abstract suggests. read the letter →

arxiv 2412.01791 v2 pith:6F5E4K6K submitted 2024-11-27 cs.RO

Ritvik Singh , Arthur Allshire , Ankur Handa , Nathan Ratliff , Karl Van Wyk This is my paper · ORCID
classification cs.RO
keywords dexterousgraspingsim-to-realtransferRGB-basedpolicyteacher-studentdistillationgeometricfabricsdomainrandomizationstereovisionreinforcementlearning
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 sets out to show that a dexterous hand-arm system can grasp a wide range of objects using only RGB camera images, with the policy trained entirely in simulation. The authors argue this matters because most prior visuomotor grasping policies rely on depth maps or object poses, which fail on translucent and reflective objects and in direct sunlight. They build a two-stage pipeline: a state-based teacher policy trained by reinforcement learning inside a geometric fabric controller, then an RGB-based student policy distilled from the teacher via DAgger using photorealistic rendered images. In real-world tests, the RGB policy matches most depth-based performance, succeeds on novel objects and lighting conditions, and does so with faster cycle times. The paper claims this is the first robust sim-to-real demonstration of an end-to-end RGB policy for contact-rich dexterous grasping.

What carries the argument

The load-bearing components are the geometric fabric controller and the stereo RGB student architecture. The geometric fabric is a second-order dynamical system that turns the policy's outputs -- a palm pose and a five-dimensional PCA hand action -- into safe, reactive, speed-independent motions, giving the policy a strong inductive bias for grasping. The student encoder processes left and right images through a shared ResNet backbone, splits each into tokens, and lets tokens attend across images (plus a learnable embedding token) so the policy can estimate depth implicitly. Training uses online DAgger with a KL-divergence loss to the teacher plus an auxiliary head that predicts the object's 3D position. The photorealistic tiled rendering and automatic domain randomization of physics and visual parameters make the simulated images close enough to real camera frames that the policy works without real-world fine-tuning.

What would settle it

Place the robot in a scene with an object texture, a camera viewpoint, or a lighting condition that lies well outside the randomized ranges used in training (for instance, a mirror-finish object or a camera angle with strong glare), and measure whether grasp success collapses while a depth-based policy remains reliable; such an asymmetry would show the sim-to-real RGB claim does not generalize.

Watch

Extended reading notes

Core claim

The central claim is that end-to-end visuomotor policies for dexterous grasping can be learned from RGB images alone and still transfer from simulation to reality. The authors demonstrate this with DextrAH-RGB, which first trains a privileged teacher policy that observes object state and acts on a geometric fabric controller, then distills it into a student policy whose only visual input is one or two RGB images. Domain randomization of lighting, materials, textures, and camera conditions, together with a stereo cross-attention encoder that lets the student infer depth implicitly, produces a policy that grasps and lifts unseen objects in the real world. In controlled bin-packing trials, the RGB policy reached success rates of 73--77 percent and cycle times 1--2 seconds faster than the depth-based baseline, while maintaining similar performance under harsh sunlight that degrades depth sensors. The authors acknowledge the RGB policy is somewhat less reliable than the depth baseline, but they treat this as a consequence of sim-to-real variance rather than a fundamental limit of RGB perception.

Load-bearing premise

The student policy sees only rendered images during training, so the transfer succeeds only if the domain randomization covers the real camera's appearance, lighting, and object variety closely enough that deployment-time images are not out of distribution.

Editorial extensions

If this is right

  • If the central claim is right, dexterous grasping systems no longer need depth cameras or object pose estimates, sidestepping the failure modes of IR depth sensors with transparent, reflective, or sunlit objects.
  • The same two-stage teacher-student pipeline with geometric fabrics could be applied to other contact-rich manipulation skills, since the fabric exposes a safe action space and the student only needs RGB.
  • Because the student is trained entirely in simulation and runs at 60 Hz, the approach scales to many cameras and object sets without additional real-world data collection.
  • The observed speed improvement over depth-based bin packing suggests RGB policies can make dexterous grasping fast enough for practical pick-and-place cycles, close to human cycle times.
  • The authors' note that repeated distillation runs with different seeds transfer unevenly implies that scaling up training runs may yield even stronger policies than the ones reported.

Reading between the lines

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

  • Beyond the paper's claims, the stereo cross-attention module is a generic way to fuse views that could be reused for other perception tasks such as object localization or in-hand manipulation, not just grasping.
  • The auxiliary object-position prediction head could serve as a confidence signal for detecting out-of-distribution scenes during deployment, since its error should grow when the input is far from the training distribution.
  • If robustness to lighting is as strong as reported, follow-up work could move these policies to mobile platforms or outdoor settings, where depth sensors are even more unreliable.
  • A testable extension is to evaluate the same pipeline with only monocular input but with a heavier temporal model, since the paper finds stereo's advantage largely disappears in the real world, suggesting temporal context may compensate for missing depth.
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

3 major / 5 minor

Summary. The paper presents DextrAH-RGB, a two-stage system for dexterous arm-hand grasping from RGB images, trained entirely in simulation. A privileged state-based teacher policy is trained with PPO on a geometric fabric controller, and an RGB student policy is distilled via online DAgger using a stereo ResNet-18 encoder with cross-attention, plus an auxiliary object-position prediction loss. The student is deployed on a Kuka IIWA/Allegro platform and evaluated in single-object grasping (11 objects, 5 poses each) and a 36-object bin-packing task under indoor and HDR lighting, with comparisons to the depth-based DextrAH-G baseline. The authors claim the first robust sim2real transfer of an end-to-end RGB policy for dexterous grasping and competitiveness with depth-based methods.

Significance. If the empirical claims are supported, the work is a meaningful advance: it demonstrates that photorealistic rendering combined with domain randomization can support RGB-only visuomotor policies in a contact-rich dexterous manipulation task, and it shows a concrete path to removing depth-camera dependence. The system-level engineering is substantial, including 60 Hz deployment of a ResNet+transformer policy with CUDA graphs, real-time ray-traced tiled rendering, and an honest discussion of sim2real variance. However, the comparative and robustness claims currently rest on thinly sampled real-world statistics, so the significance is qualified until the statistical evidence is strengthened.

major comments (3)
  1. [Section IV-B, Table II] The single-object success rates are based on five poses per object, so each 20 percentage-point step corresponds to a single trial, and the table reports no confidence intervals. The stopping rule stated in the same section (running the policy continually until either the grasp succeeds or a failure from which the robot cannot recover) is ambiguous: if multiple attempts are allowed within a pose trial, the reported percentages are not single-attempt success rates and cannot be directly compared with methods that report first-attempt success. Please report per-pose first-attempt versus final success separately, give exact trial counts, and include confidence intervals (for example, Jeffreys or Wilson intervals) for each object.
  2. [Section IV-B, Table III] The table reports consecutive successes, cycle time, and success rate for DextrAH-G and DextrAH-RGB but does not state how many independent sessions produced these numbers. The Discussion states that repeated distillation runs with different seeds transfer differently in the real world; under that acknowledged variance, the 10-14 percentage-point success-rate gap between DextrAH-RGB and DextrAH-G may be within session-to-session noise. Please report the number of sessions per configuration, per-session results (or per-object results), and distributions for consecutive successes and success rate so that the 'competitive' claim can be assessed.
  3. [Section IV-B, Table II (comparison to prior work)] The comparison with DexDiffuser, ISAGrasp, and Matak uses success rates from prior publications without a matched evaluation protocol; object sets, pose sampling, number of trials, and the definition of a trial are not controlled across methods. Given the sensitivity of success rates to trial counts (see the first major comment), the wording 'state-of-the-art performance for most objects' is stronger than the evidence supports. Please either run the baselines under the same protocol or explicitly frame the comparison as indicative and soften the state-of-the-art claim.
minor comments (5)
  1. [Section IV-A, Table I] The ablation results are averaged across three seeds, but no variance or confidence intervals are reported; adding them would strengthen the claims that stereo, attention, and fine-tuning are beneficial.
  2. [Section III-C] The claim that L40S GPUs are used because they 'contain the RTX cores necessary for high-fidelity rendering' is imprecise; RTX cores accelerate ray tracing but are not strictly necessary for high-fidelity rendering. Please rephrase.
  3. [Table III caption] The caption uses 'bin picking' while the text uses 'bin packing'; please make the terminology consistent.
  4. [Section III-C] The sentence 'This is expected, as stereo vision provides additional information for depth perception' is stated as an explanation after the results are introduced; it would be better placed as a hypothesis before the ablation or as a discussion point.
  5. [Section V] The Discussion usefully acknowledges high variance, but the phrase 'an improved top performer' suggests selection among many seeds; please clarify how many teacher and student seeds were trained and how many were deployed in the real world, since this affects the interpretation of the reported results.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central sim2real claim rests on external real-world benchmarks, not on self-referential equations; self-citations are prior tooling and methodology, not load-bearing reductions.

full rationale

The paper's derivation chain is empirical rather than mathematical: a state-based teacher policy is trained by RL with hand-defined rewards, then distilled into an RGB student via online DAgger, and the resulting policy is evaluated on real-world grasp success against external baselines. The reported success rates are measured after deployment (e.g., objects physically transported to a bin), not computed from the training objective or from fitted parameters. The auxiliary object-position prediction is used both as a training loss and as a trigger in the bin-packing state machine, but the final metric (SR) is whether the object actually reaches the bin, so success is not defined by the policy's own prediction. The paper uses several self-citations (geometric fabrics [14,20,30], Synthetica rendering [23], ADR [8,18], Isaac Lab [17]) as engineering tools whose properties were established in prior work; the central claim—that RGB-based dexterous grasping transfers from simulation to the real world—does not reduce to any of these citations. The Discussion honestly reports high run-to-run variance in sim2real transfer, which is a robustness limitation, not a circularity. There is no equation in which a predicted quantity is defined in terms of itself, and no benchmark result is obtained by construction from the training data.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The policy is an empirical RL system; the load-bearing free parameters are the reward weights and the domain randomization ranges, which are hand-set and not swept in the paper. The central sim2real claim additionally depends on several domain assumptions about the fidelity of simulation and rendering, none of which are independently verified beyond the final real-world results.

free parameters (5)
  • Reward weights whand_obj, wobj_goal, wlift, wcurl = not reported
    Section III-B defines the final reward as a weighted sum; the weights are hand-chosen and not stated, and they shape the teacher policy that the student distills.
  • βobj_goal, βcurl (sharpness coefficients) = βobj_goal terminal -20, βcurl terminal -0.05; βlift unspecified
    Table IV lists these as ADR-controlled from initial to terminal values; they determine how peaked the goal and finger-curling rewards are.
  • ADR physics parameter terminal ranges = e.g., object mass scale U(0.5,3), friction U(0.3,1.2), disturbance acceleration U(0,10); full list in Table IV
    These randomized ranges are hand-set to make the policy robust to uncertain real-world physics; they are a key input to the policy's transfer.
  • Visual domain randomization ranges and augmentation probabilities = HDRI intensity U(1000,4000), texture and material ranges in Table V; augmentation probabilities 0.5, 1.0, 0.1 in Table…
    Hand-selected to make simulated RGB images look like real D415 frames; the student's visual generalization depends on them.
  • PD velocity target and fabric damping gain = velocity target annealed 1 to 0; damping gain 10 to 20
    Section III-B and Table IV: these ADR settings control exploration dynamics and are chosen to produce behavior that is fast in training and stable in deployment.
assumptions (4)
  • domain assumption The geometric fabric controller from [14] behaves the same in simulation and on the real Kuka/Allegro hardware, without retuning.
    Section III-A states the same fabric and action space as [14] are used; the entire policy action layer depends on this transfer.
  • domain assumption Photorealistic tiled rendering in Isaac Lab plus the domain randomization in Tables V and VI produces RGB images whose distribution covers the real D415 camera images at deployment.
    Section III-C trains the student exclusively on rendered images; there is no real-image validation or fine-tuning.
  • domain assumption The stereo cross-attention encoder can recover object 3D position and teacher actions from the rendered image pair.
    Figure 7 and Section III-C: the stereo embedding is the only visual input to the policy; the auxiliary object-position loss assumes this information is learnable.
  • domain assumption Isaac Lab's physics simulator, with ADR on friction, mass, restitution, and disturbance, adequately models real contact interactions for the tested objects.
    Section III-B: all teacher and student training is in simulation; the real-world success depends on this fidelity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DextrAH-RGB: Visuomotor Policies to Grasp Anything with Dexterous Hands." pith.science (2026). https://pith.science/paper/6F5E4K6K

@misc{pith2026241201791,
  author       = {Pith},
  title        = {Pith review of: DextrAH-RGB: Visuomotor Policies to Grasp Anything with Dexterous Hands},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6F5E4K6K}},
  note         = {Machine review of arXiv:2412.01791}
}
read the original abstract

One of the most important, yet challenging, skills for a dexterous robot is grasping a diverse range of objects. Much of the prior work has been limited by speed, generality, or reliance on depth maps and object poses. In this paper, we introduce DextrAH-RGB, a system that can perform dexterous arm-hand grasping end-to-end from RGB image input. We train a privileged fabric-guided policy (FGP) in simulation through reinforcement learning that acts on a geometric fabric controller to dexterously grasp a wide variety of objects. We then distill this privileged FGP into a RGB-based FGP strictly in simulation using photorealistic tiled rendering. To our knowledge, this is the first work that is able to demonstrate robust sim2real transfer of an end2end RGB-based policy for complex, dynamic, contact-rich tasks such as dexterous grasping. DextrAH-RGB is competitive with depth-based dexterous grasping policies, and generalizes to novel objects with unseen geometry, texture, and lighting conditions in the real world. Videos of our system grasping a diverse range of unseen objects are available at \url{https://dextrah-rgb.github.io/}.

Figures

Figures reproduced from arXiv: 2412.01791 by the authors.

Figure 1
Figure 1. DextrAH-RGB (Dexterous Arm-Hand RGB) is an end-to-end RGB-based policy that can dexterously grasp a wide variety of objects. Abstract—One of the most important, yet challenging, skills for a dexterous robot is grasping a diverse range of objects. Much of the prior work has been limited by speed, generality, or reliance on depth maps and object poses. In this paper, we introduce DextrAH-RGB, a system that can perform… view at source ↗
Figure 2
Figure 2. The first stage of our pipeline involves training a state-based teacher policy in simulation using PPO. We adopt an asymmetric actor critic framework whereby the teacher policy receives noisy state observations whereas the critic receives privileged (and perfect) state observations. This is done to ensure that the policy is not overly reliant on behaviors that require accurate state estimates as this can make it har… view at source ↗
Figure 3
Figure 3. The second stage of our pipeline involves distilling the pre￾viously trained state-based teacher policy into a vision-based student policy. We use an online implementation of DAgger [21] where at each step, the observations for the teacher and student are queried and fed into the respective networks. The student is supervised to minimize the KL-divergence between its action distribution and that of the teacher. Furt… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: For all physics parameters p i , the initial values for p i_lo and p i_hi are initialized to p i init. As the policy starts performing better, p i_lo is decremented by ∆n and p i_hi is incremented by ∆n . The parameter value ranges are constantly increased until they r…
Figure 5
Figure 5. Figure 5: (a) shows an example subset of object meshes with no texture. (b) Shows those meshes with random textures binded to them. The architecture of the student is shown in [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: The top row shows the left camera renderings for different environments in simulation. The bottom row shows various data augmentations applied to these sim renderings that are passed to the student policy. can learn important recovery behaviors if it is unable to grasp…
Figure 7
Figure 7. Figure 7: (a) Our stereo encoder first starts with a pre-trained ResNet-18 encoder where the last two layers are removed. Each image is passed into the encoder and outputs a 40960-dimensional vector, which is projected down to a 16384-dimensional layer. These are then split into…
Figure 8
Figure 8. Figure 8: (a) Our real-world robot setup: an Allegro Hand mounted onto a Kuka iiwa robot arm and two Intel RealSense D415 cameras in a stereo configuration (55 mm baseline). (b) The stereo image pair from the real world, along with robot proprioceptive states, are fed into the t…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 15 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. SILO: Simulation-in-the-Loop Sim-to-Real Transfer for Multi-Stage Cable Routing

    cs.RO 2026-07 conditional novelty 7.0 of 10

    SILO enables the first reported zero-shot sim-to-real RL transfer for multi-stage cable routing by approximating cables as articulated rigid links and executing policy actions inside a synchronized digital twin.

  2. HUGS: Guiding Unified Dexterous Grasp Synthesis Across Modes and Scales via Learned Human Priors

    cs.RO 2026-07 conditional novelty 7.0 of 10

    An object-conditioned human prior over contact modes and wrists guides force-closure optimization to synthesize diverse multi-mode dexterous grasps across object scales more efficiently than heuristics.

  3. MuJoCo Playground

    cs.RO 2025-02 conditional novelty 7.0 of 10

    An open-source, MJX-based robot learning framework with integrated batch rendering that provides fast training and demonstrates sim-to-real transfer on six robot platforms.

  4. World Translation: Minimizing Sim-to-Real Gap with Backward Dynamics Extraction and Unpaired Domain Translation

    cs.RO 2026-07 conditional novelty 6.0 of 10

    World Translation predicts a robot's next state by encoding hidden dynamics from the observed transition and cycle-translating that latent code from simulation to reality.

  5. Cross-Embodiment Robot Manipulation via a Unified Hand Action Space

    cs.RO 2026-07 conditional novelty 6.0 of 10

    UHAS maps hand actions to deformations of a shared unit sphere and recovers joint commands via cascade IK, enabling multi-hand RL, zero-shot transfer, and modest real-world cube reorientation on LEAP and Allegro.

  6. Play2Perfect: What Matters in Dexterous Play Pretraining for Precise Assembly?

    cs.RO 2026-06 unverdicted novelty 6.0 of 10

    Task-agnostic RL play pretraining on diverse objects yields a reusable dexterous prior that makes sparse-reward assembly learning ~33× more sample-efficient and enables zero-shot sim-to-real transfer on tight insertio...

  7. PTLD: Sim-to-real Privileged Tactile Latent Distillation for Dexterous Manipulation

    cs.RO 2026-03 unverdicted novelty 6.0 of 10

    PTLD distills real privileged tactile data into a state estimator to boost sim-to-real performance of proprioceptive dexterous manipulation policies, yielding 182% improvement on in-hand rotation and 57% on reorientat...

  8. Grasp to Act: Dexterous Grasping for Tool Use in Dynamic Settings

    cs.RO 2026-02 conditional novelty 6.0 of 10

    Combining wrench-tested grasp optimization with real-time RL finger adjustments lets a 16-DoF robot hand keep tools stable during hammering, sawing, cutting, stirring, and scooping.

  9. StereoVLA: Enhancing Vision-Language-Action Models with Stereo Vision

    cs.RO 2025-12 conditional novelty 6.0 of 10

    A vision-language-action model that fuses stereo-derived geometric features with semantic features improves real-world grasping success and camera-pose robustness over single-view baselines.

  10. Dexplore: Scalable Neural Control for Dexterous Manipulation from Reference-Scoped Exploration

    cs.RO 2025-09 conditional novelty 6.0 of 10

    Dexplore learns dexterous robotic hand control from human MoCap demonstrations by treating them as soft, adaptively shrinking spatial references, then distills the policy into a vision-based controller.

  11. Grasp-MPC: Closed-Loop Visual Grasping via Value-Guided Model Predictive Control

    cs.RO 2025-09 conditional novelty 6.0 of 10

    A value-guided MPC policy trained on 2 million synthetic trajectories improves closed-loop 6-DoF grasping in clutter and adapts to object perturbations.

  12. ClutterDexGrasp: A Sim-to-Real System for General Dexterous Grasping in Cluttered Scenes

    cs.RO 2025-06 conditional novelty 6.0 of 10

    A simulation-trained teacher-student policy achieves zero-shot sim-to-real closed-loop target-oriented dexterous grasping in cluttered scenes, with 83.9 percent real-world success.

  13. HERMES: Human-to-Robot Embodied Learning from Multi-Source Motion Data for Mobile Dexterous Manipulation

    cs.RO 2025-08 conditional novelty 5.0 of 10

    HERMES converts a single human motion demonstration into a deployable mobile bimanual dexterous manipulation policy, using RL, depth-image distillation, and closed-loop PnP pose refinement.

  14. SimLauncher: Launching Sample-Efficient Real-world Robotic Reinforcement Learning via Simulation Pre-training

    cs.RO 2025-07 conditional novelty 5.0 of 10

    Simulation-pretrained policies, with digital-twin demos for critic bootstrapping and action proposals, cut real-world RL training time while reaching near-perfect success on three manipulation tasks.

  15. A Survey: Learning Embodied Intelligence from Physical Simulators and World Models

    cs.RO 2025-07 conditional novelty 4.0 of 10

    Embodied intelligence learning is reviewed through the complementary lenses of physical simulators and world models, with a proposed IR-L0 to IR-L4 robot capability taxonomy.

Reference graph

Works this paper leans on

32 extracted references · 9 canonical work pages · cited by 15 Pith papers

  1. [1]

    Dexterous functional grasping, 2023

    Ananye Agarwal, Shagun Uppal, Kenneth Shaw, and Deepak Pathak. Dexterous functional grasping, 2023. URL https://arxiv.org/abs/2312.02975

    arXiv 2023

  2. [2]

    Berk Calli, Aaron Walsman, Arjun Singh, Siddhartha Srinivasa, Pieter Abbeel, and Aaron M. Dollar. Bench- marking in manipulation research: Using the yale-cmu- berkeley object and model set. IEEE Robotics & Automation Magazine , 22(3):36–52, September 2015. ISSN 1070-9932. doi: 10.1109/mra.2015.2448951. URL http://dx.doi.org/10.1109/MRA.2015.2448951

    arXiv 2015

  3. [3]

    Visual dexter- ity: In-hand reorientation of novel and complex object shapes

    Tao Chen, Megha Tippur, Siyang Wu, Vikash Kumar, Edward Adelson, and Pulkit Agrawal. Visual dexter- ity: In-hand reorientation of novel and complex object shapes. Science Robotics , 8(84), November 2023. ISSN 2470-9476. doi: 10.1126/scirobotics.adc9244. URL http://dx.doi.org/10.1126/scirobotics.adc9244

    doi:10.1126/scirobotics.adc9244 2023

  4. [4]

    Learning robust real-world dexterous grasping policies via implicit shape augmentation, 2022

    Zoey Qiuyu Chen, Karl Van Wyk, Yu-Wei Chao, Wei Yang, Arsalan Mousavian, Abhishek Gupta, and Dieter Fox. Learning robust real-world dexterous grasping policies via implicit shape augmentation, 2022. URL https://arxiv.org/abs/2210.13638

    arXiv 2022

  5. [5]

    Ciocarlie, Corey Goldfeder, and Peter K

    Matei T. Ciocarlie, Corey Goldfeder, and Peter K. Allen. Dexterous grasping via eigengrasps : A low-dimensional approach to a high-complexity problem. 2007. URL https://api.semanticscholar.org/CorpusID:6853822

    work page 2007

  6. [6]

    An image is worth 16x16 words: Transformers for image recognition at scale, 2021

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition at scale, 2021. URL https://arxiv.org/abs/ 2010.11929

    arXiv 2021

  7. [7]

    Ferrari and J

    C. Ferrari and J. Canny. Planning optimal grasps. In Proceedings 1992 IEEE International Conference on Robotics and Automation , pages 2290–2295 vol.3, 1992. doi: 10.1109/ROBOT.1992.219918

    work page arXiv 1992

  8. [8]

    Dex- treme: Transfer of agile in-hand manipulation from sim- ulation to reality, 2024

    Ankur Handa, Arthur Allshire, Viktor Makoviychuk, Aleksei Petrenko, Ritvik Singh, Jingzhou Liu, Denys Makoviichuk, Karl Van Wyk, Alexander Zhurke- vich, Balakumar Sundaralingam, Yashraj Narang, Jean- Francois Lafleche, Dieter Fox, and Gavriel State. Dex- treme: Transfer of agile in-hand manipulation from sim- ulation to reality, 2024. URL https://arxiv.or...

    work page 2024

Show all 32 references
  1. [9]

    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

  2. [10]

    Weinberger

    Gao Huang, Zhuang Liu, Laurens van der Maaten, and Kilian Q. Weinberger. Densely connected convolutional networks, 2018. URL https://arxiv.org/abs/1608.06993

    2018 arXiv

  3. [11]

    Gendexgrasp: Generalizable dexterous grasping, 2023

    Puhao Li, Tengyu Liu, Yuyang Li, Yiran Geng, Yixin Zhu, Yaodong Yang, and Siyuan Huang. Gendexgrasp: Generalizable dexterous grasping, 2023. URL https:// arxiv.org/abs/2210.00722

    2023 arXiv

  4. [12]

    Dexrep- net: Learning dexterous robotic grasping network with geometric and spatial hand-object representations, 2023

    Qingtao Liu, Yu Cui, Qi Ye, Zhengnan Sun, Haoming Li, Gaofeng Li, Lin Shao, and Jiming Chen. Dexrep- net: Learning dexterous robotic grasping network with geometric and spatial hand-object representations, 2023. URL https://arxiv.org/abs/2303.09806

    2023 arXiv

  5. [13]

    Li, Preston Culbertson, Krishnan Srinivasan, Aaron D

    Tyler Ga Wei Lum, Albert H. Li, Preston Culbertson, Krishnan Srinivasan, Aaron D. Ames, Mac Schwager, and Jeannette Bohg. Get a grip: Multi-finger grasp evaluation at scale enables robust sim-to-real transfer,

  6. [14]

    Ratliff, and Karl Van Wyk

    Tyler Ga Wei Lum, Martin Matak, Viktor Makoviy- chuk, Ankur Handa, Arthur Allshire, Tucker Hermans, Nathan D. Ratliff, and Karl Van Wyk. DextrAH-G: Pixels-to-Action Dexterous Arm-Hand Grasping with Geometric Fabrics, 2024. URL https://arxiv.org/abs/ 2407.02274

    2024 arXiv

  7. [15]

    Planning visual- tactile precision grasps via complementary use of vision and touch, 2022

    Martin Matak and Tucker Hermans. Planning visual- tactile precision grasps via complementary use of vision and touch, 2022. URL https://arxiv.org/abs/2212.08604

    2022 arXiv

  8. [16]

    Miller and P.K

    A.T. Miller and P.K. Allen. Graspit! a versatile simulator for robotic grasping. IEEE Robotics & Automation Magazine, 11(4):110–122, 2004. doi: 10.1109/MRA. 2004.1371616

    2004 arXiv

  9. [17]

    Orbit: A unified simulation framework for interactive robot learning environments

    Mayank Mittal, Calvin Yu, Qinxi Yu, Jingzhou Liu, Nikita Rudin, David Hoeller, Jia Lin Yuan, Ritvik Singh, Yunrong Guo, Hammad Mazhar, Ajay Mandlekar, Buck Babich, Gavriel State, Marco Hutter, and Animesh Garg. Orbit: A unified simulation framework for interactive robot learni...

    2023

  10. [18]

    Solving rubik’s cube with a robot hand, 2019

    OpenAI, Ilge Akkaya, Marcin Andrychowicz, Maciek Chociej, Mateusz Litwin, Bob McGrew, Arthur Petron, Alex Paino, Matthias Plappert, Glenn Powell, Raphael Ribas, Jonas Schneider, Nikolas Tezak, Jerry Tworek, Peter Welinder, Lilian Weng, Qiming Yuan, Wojciech Zaremba, and Lei Zh...

    2019 arXiv

  11. [19]

    Dexpoint: Generalizable point cloud reinforcement learning for sim-to-real dexterous manip- ulation, 2022

    Yuzhe Qin, Binghao Huang, Zhao-Heng Yin, Hao Su, and Xiaolong Wang. Dexpoint: Generalizable point cloud reinforcement learning for sim-to-real dexterous manip- ulation, 2022. URL https://arxiv.org/abs/2211.09423

    2022 arXiv

  12. [20]

    Fabrics: A foundation- ally stable medium for encoding prior experience, 2023

    Nathan Ratliff and Karl Van Wyk. Fabrics: A foundation- ally stable medium for encoding prior experience, 2023. URL https://arxiv.org/abs/2309.07368

    2023 arXiv

  13. [21]

    Gordon, and J

    Stephane Ross, Geoffrey J. Gordon, and J. Andrew Bagnell. A reduction of imitation learning and structured prediction to no-regret online learning, 2011. URL https://arxiv.org/abs/1011.0686

    2011 arXiv

  14. [22]

    Hand-object interaction pretraining from videos, 2024

    Himanshu Gaurav Singh, Antonio Loquercio, Carmelo Sferrazza, Jane Wu, Haozhi Qi, Pieter Abbeel, and Ji- tendra Malik. Hand-object interaction pretraining from videos, 2024. URL https://arxiv.org/abs/2409.08273

    2024 arXiv

  15. [23]

    Synthetica: Large scale synthetic data for robot perception, 2024

    Ritvik Singh, Jingzhou Liu, Karl Van Wyk, Yu-Wei Chao, Jean-Francois Lafleche, Florian Shkurti, Nathan Ratliff, and Ankur Handa. Synthetica: Large scale synthetic data for robot perception, 2024. URL https://arxiv.org/abs/ 2410.21153

    2024 arXiv

  16. [24]

    D2rl: Deep dense architectures in reinforcement learning, 2020

    Samarth Sinha, Homanga Bharadhwaj, Aravind Srinivas, and Animesh Garg. D2rl: Deep dense architectures in reinforcement learning, 2020. URL https://arxiv.org/abs/ 2010.09163

    2020 arXiv

  17. [25]

    Fast-grasp’d: Dexterous multi-finger grasp generation through differentiable simulation, 2023

    Dylan Turpin, Tao Zhong, Shutong Zhang, Guanglei Zhu, Jingzhou Liu, Ritvik Singh, Eric Heiden, Miles Macklin, Stavros Tsogkas, Sven Dickinson, and Ani- mesh Garg. Fast-grasp’d: Dexterous multi-finger grasp generation through differentiable simulation, 2023. URL https://arxiv.o...

    2023 arXiv

  18. [26]

    Unidexgrasp++: Im- proving dexterous grasping policy learning via geometry- aware curriculum and iterative generalist-specialist learn- ing, 2023

    Weikang Wan, Haoran Geng, Yun Liu, Zikang Shan, Yaodong Yang, Li Yi, and He Wang. Unidexgrasp++: Im- proving dexterous grasping policy learning via geometry- aware curriculum and iterative generalist-specialist learn- ing, 2023. URL https://arxiv.org/abs/2304.00464

    2023 arXiv

  19. [27]

    Dexgraspnet: A large-scale robotic dexterous grasp dataset for general objects based on simulation, 2023

    Ruicheng Wang, Jialiang Zhang, Jiayi Chen, Yinzhen Xu, Puhao Li, Tengyu Liu, and He Wang. Dexgraspnet: A large-scale robotic dexterous grasp dataset for general objects based on simulation, 2023. URL https://arxiv. org/abs/2210.02697

    2023 arXiv

  20. [28]

    Dust3r: Geometric 3d vision made easy, 2024

    Shuzhe Wang, Vincent Leroy, Yohann Cabon, Boris Chidlovskii, and Jerome Revaud. Dust3r: Geometric 3d vision made easy, 2024. URL https://arxiv.org/abs/2312. 14132

    2024

  21. [29]

    Dexdiffuser: Generating dexterous grasps with diffusion models, 2024

    Zehang Weng, Haofei Lu, Danica Kragic, and Jens Lundell. Dexdiffuser: Generating dexterous grasps with diffusion models, 2024. URL https://arxiv.org/abs/2402. 02989

    2024

  22. [30]

    Karl Van Wyk, Ankur Handa, Viktor Makoviychuk, Yijie Guo, Arthur Allshire, and Nathan D. Ratliff. Geometric fabrics: a safe guiding medium for policy learning, 2024. URL https://arxiv.org/abs/2405.02250

    2024 arXiv

  23. [31]

    Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned policy, 2023

    Yinzhen Xu, Weikang Wan, Jialiang Zhang, Haoran Liu, Zikang Shan, Hao Shen, Ruicheng Wang, Haoran Geng, Yijia Weng, Jiayi Chen, Tengyu Liu, Li Yi, and He Wang. Unidexgrasp: Universal robotic dexterous grasping via learning diverse proposal generation and goal-conditioned polic...

    2023 arXiv

  24. [2024]

    URL https://arxiv.org/abs/2410.23701

    arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.