Pith. sign in

REVIEW 4 major objections 6 minor 39 references

Prior Reinforce: Goal-Conditioned Dynamic Manipulation with Limited Trials

T0 review · 4 major / 6 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 Prior Reinforce reaches new dynamic-manipulation goals in fewer than 10 real-world trials, starting from a handful of demonstrations.

desk verdict A practical two-stage framework for low-trial dynamic manipulation, with a solid empirical core but an uncharacterized smoothness assumption and thin real-world evidence. read the letter →

arxiv 2505.21916 v3 pith:7DELWPNZ submitted 2025-05-28 cs.RO

Yihang Hu , Pingyue Sheng , Yuyang Liu , Shengjie Wang , Yang Gao This is my paper
classification cs.RO
keywords goal-conditionedmanipulationdynamicconditionaldiffusionmodelBayesianoptimizationfew-shotadaptationreal-worldroboticssemanticfeedbackdataforgetting
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

The paper proposes Prior Reinforce (P.R.), a two-stage framework for goal-conditioned dynamic manipulation tasks such as throwing a basketball into a hoop, pushing a curling stone, or swinging a fishing rod to a target. P.R. first learns a motion pattern from just 6 to 8 prior demonstrations using a conditional diffusion model, then adapts to a previously unseen goal through iterative real-world rollouts guided by low-dimensional semantic feedback and Bayesian optimization. The central claim is that this separation of motion generation from outcome-driven adaptation lets a robot reach a new goal in about three additional trials, for a total of fewer than ten real trials, without large-scale interaction, reward engineering, or accurate dynamic models. The paper reports this performance across three real-world tasks with strict success criteria, under both human-eye and vision-language-model perception, and with robustness to perception noise and hardware changes.

What carries the argument

The central mechanism is a conditional diffusion motion generator (U-Net backbone with 1D convolutions over the time axis) trained as $\pi: \tilde{r} \to A$, coupled with a Gaussian-process-regression Bayesian optimization condition adapter $M$ that proposes the next generation condition $c = M(\tilde{g})$ given the goal. The diffusion model supplies the motion pattern and a smooth, low-dimensional condition space, while the Bayesian adapter performs feedback-driven search in that space; the data-forgetting rule keeps only the most recent $m$ Stage-2 trials so that outlier rollouts do not permanently poison the adaptation. Together they turn sparse semantic feedback into a rapid closed-loop search over plausible motions without requiring a dynamic model.

What would settle it

Run P.R. on a task where the condition-to-outcome map is deliberately rough or multi-modal, such as throwing over a low obstacle where a slightly too-soft throw hits the obstacle and fails completely, and count whether the agent still reaches an unseen goal within ten total trials; a significant trial-count increase in such settings would show that the smoothness assumption is load-bearing.

Watch

Extended reading notes

Core claim

The core discovery is that a goal-conditioned dynamic manipulation task can be solved with human-like trial efficiency by treating the generation condition of a diffusion-based motion generator as a low-dimensional latent space, then searching that space with a Gaussian-process Bayesian optimization adapter. The framework assumes that after each rollout the agent receives a low-dimensional, semantically meaningful feedback vector (e.g., the 2D landing offset from the hoop). The conditional motion generator $\pi: \tilde{r} \to A$ maps a perceived result $\tilde{r}$ back to an action plan, and the condition adapter $M: \tilde{r} \to c$ refines the generation condition $c$ using the accumulated dataset $D_M$, so that the perceived error $\tilde{e}(c) = P(\mathrm{rollout}(\pi(c)) - g)$ converges to zero within a few iterations. A data-forgetting technique retains only the latest $m$ Stage-2 trials (with $m=2$) to prevent occasional outliers from corrupting the adaptation process. Real-world experiments show that P.R. reaches an unseen goal on average in 2.0 to 3.7 Stage-2 trials across tasks and perception modes, and simulator ablations show it outperforms interpolation-based and Dynamic-Movement-Primitive baselines, with the Bayesian adapter and data forgetting both contributing to faster convergence and higher final success rates.

Load-bearing premise

The method hinges on the assumption that each rollout produces a low-dimensional, semantically meaningful feedback vector (like a 2D landing offset) and that the mapping from the diffusion generation condition to the rollout outcome is smooth enough for Bayesian optimization to drive the error to zero within a few queries.

Editorial extensions

If this is right

  • If the fewer-than-10-trials claim holds, dynamic manipulation can be adapted to new goals with the same order of interaction cost as a human novice, making real-world deployment far more practical for tasks where each trial is expensive.
  • The same prior demonstrations can be reused for modified hardware or slightly changed task versions, since the self-correction stage absorbs the transfer gap; the paper demonstrates this with two hardware variants of the basketball-shot task.
  • Robustness to coarse human-eye feedback suggests that expensive or brittle perception pipelines are not required, so the method could work in settings where only approximate visual estimates are available.
  • The separation of motion generation from outcome-driven adaptation implies that improvements in either the diffusion prior or the optimizer could transfer directly into lower trial counts across a family of similar agile tasks.
  • In simulator experiments, the full P.R. reaches more than 90% of goals within four additional trials even under amplified perception noise and timeline-misaligned demonstrations, indicating the approach tolerates realistic imperfections.

Reading between the lines

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

  • A plausible testable extension is to apply P.R. to tasks whose feedback is not a simple Euclidean offset (e.g., contact forces or soft-body shapes); the method's assumption of a low-dimensional semantic vector would be violated, and observing whether the trial count degrades would clarify the boundary of the claim.
  • The paper's success criterion counts trials to the first success, not success rate; a natural follow-up is to measure how often the adapted motion continues to succeed on later repetitions and whether the data-forgetting window affects repeatability.
  • The condition-space-search view suggests that other generative architectures (e.g., flow matching or action VAEs) could replace the diffusion backbone while preserving the same trial-efficiency, and that the key ingredient is a smooth, low-dimensional condition manifold rather than the specific diffusion formulation.
  • The authors note that VLM perception gives fewer trials but higher variance; this hints that robust automatic feedback extraction, rather than ever-more-precise perception, may be the bottleneck for scaling the method to tasks without a human in the loop.
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 / 6 minor

Summary. The paper proposes Prior Reinforce (P.R.), a two-stage framework for goal-conditioned dynamic manipulation under very limited trials. In Stage 1, a conditional diffusion model is trained on 6-8 prior demonstrations to map a low-dimensional result/condition vector to a full action plan. In Stage 2, a GPR-based Bayesian optimizer adapts the generation condition using semantic feedback from real rollouts, with a data-forgetting mechanism that retains only the latest m=2 trials. The authors evaluate on three real-world tasks (basketball shot, curling, fishing rod swinging) and report reaching new goals in roughly 2-4 stage-2 trials after the prior rollouts, totaling 8.1-10.0 trials, plus simulator ablations comparing against I.N.N., DMP, a naive compensator, and a no-forgetting variant.

Significance. If the <10-trials claim holds, P.R. is a useful contribution: it offers a low-trial adaptation recipe for a class of open-loop dynamic manipulation tasks without reward engineering, large-scale interaction, or accurate dynamic models. The real-world validation across three physically distinct tasks, the systematic simulator ablations with 100 goals, and the explicit treatment of perception noise and outlier trials are concrete strengths. The method's inductive bias of searching in a low-dimensional condition space rather than raw action space is sensible and well-motivated by the human-learning analogy. However, the quantitative headline claim is not yet fully substantiated by the reported statistics, and the core smoothness assumption that makes Bayesian optimization work is asserted rather than characterized. These issues are addressable but currently prevent the paper from being fully convincing.

major comments (4)
  1. [§V-A, Table II] The headline claim that the agent needs 'fewer than 10 trials' is not directly supported by Table II. The Fishing Rod Swinging row reports 8 prior trials plus 2.0 stage-2 trials on average, giving a total of exactly 10.0, not fewer than 10. For the other rows, the reported mean±std values imply that a nontrivial fraction of individual goals exceed 10 total trials (e.g., Basketball Shot-V3: 6 + 3.7±1.5, so one standard deviation above the mean is already 11.2). The table aggregates only 2 groups of 5 goals per task and provides no per-goal counts, maxima, or confidence intervals. Please report the full distribution of total trials, the fraction of goals reached within ≤10 trials, and a bootstrap or equivalent confidence interval for the mean; the abstract and introduction make a precise quantitative claim that the current table does not fully establish.
  2. [§IV-B.2, Eq. (6)] The adaptation mechanism rests on the claim that the condition space of the diffusion generator is 'low-dimensional and continuous', so that the composition f(c) = P(rollout(pi(c))) can be minimized by GPR-based Bayesian optimization with only 6-8 initial points and m=2 retained trials. The manuscript asserts this property but provides no empirical or theoretical characterization of f. The real-world tasks contain success boundaries (the ball must enter the hoop, the iron ball must contact the magnet) where f can be steep, flat, or discontinuous, so the smoothness assumption is genuinely load-bearing for the fewer-than-10-trials guarantee. Please add evidence that the GP model adequately captures f within the explored region: for example, report GP predictive error on held-out conditions, plot error magnitude versus condition dimension, show cases where the optimizer stalls or requires more than 5 iterations, and study sensitivity to the kernel choice, acquisition function, and forgetting window. Without this, the result remains a demonstration on specific configurations rather than a supported property of the framework.
  3. [§V-A, real-world experimental results] The real-world evidence is statistically thin: only 2 groups of prior demonstrations per task and 5 goals per group, with no randomization protocol, no independent evaluator, and no confidence intervals. The paper states that goals were 'deliberately set far from the rollout results of the prior demonstrations', which is good for avoiding trivial interpolation, but the selection process is not described in enough detail to assess bias. At minimum, the authors should acknowledge the small sample size as a limitation, report the worst-case goal in each task, and provide per-goal trial counts so readers can verify the distribution rather than relying on aggregate means.
  4. [§V-B, Figure 6] The simulator ablations are reported as success-rate curves without error bars, confidence intervals, or information about the number of independent random seeds. The text makes comparative statements such as 'our full version of P.R. decreases this fraction to less than 5%' and 'more than 90% of goals in only 4 additional trials', but without uncertainty quantification the superiority over the baselines is not statistically demonstrated. Since the simulator allows many parallel goals, please provide means over multiple seeds or bootstrap intervals for each curve.
minor comments (6)
  1. [Algorithm 1 and §IV-B.2] The notation for the Condition Adapter is inconsistent: the text defines M: r~ → c, but Algorithm 1 calls c_new = M(g~) and Eq. (7) initializes D_M with pairs (c_i, r~_i). Please clarify whether M is a forward Gaussian-process model of r(c) combined with an acquisition function that returns c, or an inverse model trained on (r, c) pairs.
  2. [Abstract and Introduction] The wording is inconsistent between 'within as few as ten total trials' (abstract) and 'fewer than 10 trials' (introduction); given Table II, the 'fewer than 10' phrasing is not strictly correct for the Fishing Rod Swinging task.
  3. [§IV-B.2] The data-forgetting window m=2 and the diffusion learning rate 2e-4 are free parameters, but no sensitivity analysis is reported; a brief study of how these choices affect stage-2 trial counts would strengthen the claim of robustness.
  4. [§V-A, Perception Process] The noise model for human-eye perception is described only qualitatively (values of 0, 2, 5, and 10 cm); please specify the actual distribution used in the experiments and report the VLM detection accuracy or failure rate for the key-frame bounding boxes.
  5. [Figure 6] The caption should define 'round 2' explicitly (e.g., success within ≤2 trials) and state the exact comparison conditions for each panel, since panel (a) and panel (b) use different baseline configurations.
  6. [General] The paper would benefit from a dedicated limitations paragraph that names situations where the assumptions of low-dimensional semantic feedback and smooth condition-to-outcome mapping are likely to break down, and how the method would behave in those cases.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the trial-count result is measured, not encoded in the inputs.

full rationale

We find no circularity in the paper's derivation chain. Prior Reinforce is an empirical feedback loop: Stage 1 rolls out the prior demonstration set D_prior and trains a conditional diffusion motion generator; Stage 2 initializes a Gaussian-process condition adapter with the identity pairs (c_i, r_i) = (r_i, r_i) from those same rollouts, then iteratively rolls out new actions, observes the perceived error e_new, and augments the adapter dataset with (c_new, g + e_new). The central few-trials claim is reported as a measured quantity in Table II (6 or 8 stage-1 rollouts plus 2.0-3.7 stage-2 rollouts), not as a quantity derived from an equation that already contains it. The identity initialization is a warm start, not an assumed conclusion, because the adapter is subsequently corrected by real outcomes through Eq. (8). The low-dimensionality and continuity statement about the condition space in Section IV-B.2 is an explicit assumption and empirical premise, not a result imported from prior work by the same authors; even if unverified, that is a correctness or generalization risk rather than circularity. No uniqueness theorem is invoked, no fitted parameter is relabeled as a prediction, and the self-citations ([8], [12]-[14], [35]) are not load-bearing for the main result. The framework is self-contained against external benchmarks, and the outcome stands or falls on real-world measurement and on the smoothness of f(c) = rollout(pi(c)), which is an empirical matter.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a handful of hand-chosen hyperparameters and four explicit assumptions about feedback, prior demonstrations, open-loop control, and smoothness of the condition space. No new physical entities are introduced; the Condition Adapter and Conditional Motion Generator are algorithmic modules, not new forces or objects.

free parameters (2)
  • data forgetting window m = 2
    Retains only the latest m stage-2 trials in D_M to forget outliers; m=2 is chosen for all experiments with no sensitivity study, so performance may depend on this ad hoc value.
  • diffusion model learning rate = 2e-4
    Increased from the Diffusion Policy default to 2e-4 'for faster convergence'; no ablation reported, so the effect on the few-demo regime is unquantified.
assumptions (4)
  • domain assumption A small set (6-8) of prior demonstrations is available and spans the reachable goal region roughly evenly.
    Section IV-A.1 (assumption 'Few Prior Motion Demonstrations') and Section V.A 'Prior Demonstrations'; without this, the diffusion model has no manifold to learn.
  • domain assumption After each trial, the agent receives a low-dimensional, semantically meaningful feedback vector (e.g., 2D landing offset).
    Section IV-A.1 'Semantic Feedback Perception' and Eq. (2); the whole adaptation loop depends on this error vector being informative.
  • domain assumption Executed motions are transient open-loop trunks with no need for mid-motion perception or feedback.
    Section IV-A.1 'Transient Open-loop Control'; the method cannot handle tasks where closed-loop reaction is required.
  • ad hoc to paper The condition space of the trained diffusion generator is low-dimensional and smooth enough that GPR-based Bayesian optimization can find a good condition within a few queries.
    Section IV-B.2 states the core idea that the generation condition forms a low-dim, continuous latent space; this is asserted, not proven, and is not verified for general tasks.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Prior Reinforce: Goal-Conditioned Dynamic Manipulation with Limited Trials." pith.science (2026). https://pith.science/paper/7DELWPNZ

@misc{pith2026250521916,
  author       = {Pith},
  title        = {Pith review of: Prior Reinforce: Goal-Conditioned Dynamic Manipulation with Limited Trials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7DELWPNZ}},
  note         = {Machine review of arXiv:2505.21916}
}
read the original abstract

Embodied robots have achieved strong performance in many real-world manipulation tasks, yet agile dynamic manipulation remains challenging due to high sensitivity to motion parameters and sparse outcome-level feedback. Tasks such as shooting a basketball into a hoop require precise control of fast open-loop motions, where small trajectory variations can lead to large outcome deviations, making data-efficient adaptation difficult for existing methods that rely on large-scale interaction, reward engineering, or accurate dynamic modeling. We propose Prior Reinforce (P.R.), a simple and practical framework for goal-conditioned dynamic manipulation. The method first learns a structured motion manifold from a small set of demonstrations using a conditional diffusion model, and then adapts motions toward new goals through feedback-driven optimization in a low-dimensional condition space. By separating motion generation from outcome-driven adaptation, the framework enables efficient refinement using only a small number of real-world trials under noisy perception. Experiments on multiple real-world dynamic manipulation tasks demonstrate that P.R. reliably achieves new goals within as few as ten total trials while remaining robust to perception noise and hardware uncertainty, suggesting a practical approach for low-trial real-world robot adaptation. Project website: https://adap-robotics.github.io/.

Figures

Figures reproduced from arXiv: 2505.21916 by the authors.

Figure 1
Figure 1. Overview of task layouts & processes. The table is a 1 meter sized square and the yellow area indicates the rough range of [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Overview of Prior Reinforce with its two stages: Motion Pattern Learning (left) and Iterative Rollout & Adaption (right). Algorithm 1 Prior Reinforce Inputs: Environment Env, estimated goal g˜, rough error perceptron P, prior demonstration dataset Dprior = {Ai}, Conditional Action Planner π : ˜r → A, Condition Adapter M and corresponding dataset DM. Stage 1: Motion Pattern Learning 1: Rollout all action plans Ai in … view at source ↗
Figure 3
Figure 3. Iterative Rollout & Adaption process of Prior Reinforce on the three real-world tasks. The small black errors indicates the vectorized error to be estimated. The yellow star for the Fishing Rod Swinging task indicates the goal location (a tiny magnet), which is too small to be obviously seen [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Hardware involved in real-world tasks: Left: Hardware for Basketball Shot (V1) and its 2 modified versions V2 and V3. The original version V1 uses the hoop (diameter = 9.0 cm), white spoon to throw the bigger ball (diameter = 6.0 cm) into the hoop; V2 changes to the sh…
Figure 5
Figure 5. Figure 5: 2D position of important objects detected by GLM-4.5V, for [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Comparison results of P.R. and ablative baselines. Figure 6b shows comparisons of P.R. with the ablative baselines with changed Condition Adapter, under Amplified Noise condition. We can see that the other baselines miss about 10% of goals even with 8 round of trials, …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 23 canonical work pages

  1. [1]

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

    K. Black, N. Brown, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fusai, L. Groom, K. Hausman, B. Ichter, S. Jakubczak, T. Jones, L. Ke, S. Levine, A. Li-Bell, M. Mothukuri, S. Nair, K. Pertsch, L. X. Shi, J. Tanner, Q. Vuong, A. Walling, H. Wang, and U. Zhilinsky, “π 0: A vision-language-action flow model for general robot control,”arXiv preprint arXiv: 241...

    arXiv

  2. [2]

    Catch it! learning to catch in flight with mobile dexterous hands,

    Y . Zhang, T. Liang, Z. Chen, Y . Ze, and H. Xu, “Catch it! learning to catch in flight with mobile dexterous hands,”IEEE International Conference on Robotics and Automation, 2024

    work page 2024

  3. [3]

    Reinforcement learning with foundation priors: Let the embodied agent efficiently learn on its own,

    W. Ye, Y . Zhang, H. Weng, X. Gu, S. Wang, T. Zhang, M. Wang, P. Abbeel, and Y . Gao, “Reinforcement learning with foundation priors: Let the embodied agent efficiently learn on its own,”arXiv preprint arXiv:2310.02635, 2023

    arXiv 2023

  4. [4]

    Roboagent: Generalization and efficiency in robot manipulation via semantic augmentations and action chunking,

    H. Bharadhwaj, J. Vakil, M. Sharma, A. Gupta, S. Tulsiani, and V . Ku- mar, “Roboagent: Generalization and efficiency in robot manipulation via semantic augmentations and action chunking,”IEEE International Conference on Robotics and Automation, 2023

    work page 2023

  5. [5]

    Diffusion policy: Visuomotor policy learning via ac- tion diffusion,

    C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song, “Diffusion policy: Visuomotor policy learning via ac- tion diffusion,”The International Journal of Robotics Research, p. 02783649241273668, 2023

    2023

  6. [6]

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

    C. Chi, Z. Xu, C. Pan, E. Cousineau, B. Burchfiel, S. Feng, R. Tedrake, and S. Song, “Universal manipulation interface: In-the-wild robot teaching without in-the-wild robots,”arXiv preprint arXiv:2402.10329, 2024

    arXiv 2024

  7. [7]

    UMI on legs: Making manipulation policies mobile with manipulation-centric whole-body controllers,

    H. Ha, Y . Gao, Z. Fu, J. Tan, and S. Song, “UMI on legs: Making manipulation policies mobile with manipulation-centric whole-body controllers,” inProceedings of the 2024 Conference on Robot Learn- ing, 2024

    work page 2024

  8. [8]

    Skil: Semantic keypoint imitation learning for generalizable data-efficient manipulation,

    S. Wang, J. You, Y . Hu, J. Li, and Y . Gao, “Skil: Semantic keypoint imitation learning for generalizable data-efficient manipulation,”arXiv preprint arXiv:2501.14400, 2025

    arXiv 2025

Show all 39 references
  1. [9]

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

    A. Brohan, N. Brown, J. Carbajal, Y . Chebotar, J. Dabis, C. Finn, K. Gopalakrishnan, K. Hausman, A. Herzog, J. Hsu, J. Ibarz, B. Ichter, A. Irpan, T. Jackson, S. Jesmonth, N. J. Joshi, R. Julian, D. Kalash- nikov, Y . Kuang, I. Leal, K.-H. Lee, S. Levine, Y . Lu, U. Malla, D....

    2022

  2. [10]

    Openvla: An open-source vision-language-action model,

    M. J. Kim, K. Pertsch, S. Karamcheti, T. Xiao, A. Balakrishna, S. Nair, R. Rafailov, E. Foster, G. Lam, P. R. Sanketi, Q. Vuong, T. Kollar, B. Burchfiel, R. Tedrake, D. Sadigh, S. Levine, P. Liang, and C. Finn, “Openvla: An open-source vision-language-action model,”Conference ...

    2024

  3. [11]

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

    S. Liu, L. Wu, B. Li, H. Tan, H. Chen, Z. Wang, K. Xu, H. Su, and J. Zhu, “Rdt-1b: a diffusion foundation model for bimanual manipulation,”arXiv preprint arXiv: 2410.07864, 2024

    2024 arXiv

  4. [12]

    Copa: General robotic manipulation through spatial constraints of parts with foundation models,

    H. Huang, F. Lin, Y . Hu, S. Wang, and Y . Gao, “Copa: General robotic manipulation through spatial constraints of parts with foundation models,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024, pp. 9488–9495

    2024

  5. [13]

    Onetwovla: A unified vision-language-action model with adaptive reasoning,

    F. Lin, R. Nai, Y . Hu, J. You, J. Zhao, and Y . Gao, “Onetwovla: A unified vision-language-action model with adaptive reasoning,”arXiv preprint arXiv:2505.11917, 2025

    2025

  6. [14]

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

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

    2025 arXiv

  7. [15]

    Itera- tive residual policy: for goal-conditioned dynamic manipulation of deformable objects,

    C. Chi, B. Burchfiel, E. Cousineau, S. Feng, and S. Song, “Itera- tive residual policy: for goal-conditioned dynamic manipulation of deformable objects,”The International Journal of Robotics Research, vol. 43, no. 4, pp. 389–404, 2024

    2024

  8. [16]

    Dynamic handover: Throw and catch with bimanual hands,

    B. Huang, Y . Chen, T. Wang, Y . Qin, Y . Yang, N. Atanasov, and X. Wang, “Dynamic handover: Throw and catch with bimanual hands,” Conference on Robot Learning, 2023

    2023

  9. [17]

    Tossing- bot: Learning to throw arbitrary objects with residual physics,

    A. Zeng, S. Song, J. Lee, A. Rodriguez, and T. Funkhouser, “Tossing- bot: Learning to throw arbitrary objects with residual physics,”IEEE Transactions on Robotics, vol. 36, no. 4, pp. 1307–1319, 2020

    2020

  10. [18]

    A stochastic dynamic motion planning al- gorithm for object-throwing,

    A. Sintov and A. Shapiro, “A stochastic dynamic motion planning al- gorithm for object-throwing,” in2015 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2015, pp. 2475–2480

    2015

  11. [19]

    Optimal shape and motion planning for dynamic planar manipulation,

    O. Taylor and A. Rodriguez, “Optimal shape and motion planning for dynamic planar manipulation,”Autonomous Robots, vol. 43, pp. 327–344, 2019

    2019

  12. [20]

    Learning agile robotic locomotion skills by imitating animals,

    X. B. Peng, E. Coumans, T. Zhang, T.-W. Lee, J. Tan, and S. Levine, “Learning agile robotic locomotion skills by imitating animals,” Robotics: Science and Systems, 2020

    2020

  13. [21]

    Tool-as-interface: Learning robot policies from human tool usage through imitation learning,

    H. Chen, C. Zhu, Y . Li, and K. Driggs-Campbell, “Tool-as-interface: Learning robot policies from human tool usage through imitation learning,”arXiv preprint arXiv:2504.04612, 2025

    2025

  14. [22]

    Conditional variational auto encoder based dynamic motion for multi-task imitation learning,

    B. Xu, M. U. Din, and I. Hussain, “Conditional variational auto encoder based dynamic motion for multi-task imitation learning,” arXiv preprint arXiv: 2405.15266, 2024

    2024 arXiv

  15. [23]

    Whole- body dynamic throwing with legged manipulators,

    H. Munn, B. Tidd, D. Howard, and M. Gallagher, “Whole- body dynamic throwing with legged manipulators,”arXiv preprint arXiv:2410.05681, 2024

    2024 arXiv

  16. [24]

    Mockus, V

    J. Mockus, V . Tiesis, and A. Zilinskas,The application of Bayesian methods for seeking the extremum, 09 2014, vol. 2, pp. 117–129

    2014

  17. [25]

    Gaussian processes in machine learning,

    C. Rasmussen, O. Bousquet, U. Luxburg, and G. Rätsch, “Gaussian processes in machine learning,”Advanced Lectures on Machine Learn- ing: ML Summer Schools 2003, Canberra, Australia, February 2 - 14, 2003, Tübingen, Germany, August 4 - 16, 2003, Revised Lectures, 63- 71 (2004), ...

    2004

  18. [26]

    A ball-throwing robot with visual feedback,

    J.-S. Hu, M. Chien, Y . Chang, S. Su, and C. Kai, “A ball-throwing robot with visual feedback,” inIEEE/RSJ 2010 International Con- ference on Intelligent Robots and Systems, IROS 2010 - Conference Proceedings, ser. IEEE/RSJ 2010 International Conference on Intel- ligent Robots...

    2010

  19. [27]

    Dynamic task execution using active parameter identification with the baxter research robot,

    A. D. Wilson, J. A. Schultz, A. R. Ansari, and T. D. Murphey, “Dynamic task execution using active parameter identification with the baxter research robot,”IEEE Transactions on Automation Science and Engineering, vol. 14, no. 1, pp. 391–397, 2017

    2017

  20. [28]

    Dynamic movement primitives in robotics: A tutorial survey,

    M. Saveriano, F. J. Abu-Dakka, A. Kramberger, and L. Peternel, “Dynamic movement primitives in robotics: A tutorial survey,” The International Journal of Robotics Research, vol. 42, no. 13, p. 1133–1184, Sept. 2023. [Online]. Available: http://dx.doi.org/10.1177/02783649231201196

    2023 doi

  21. [29]

    Probabilistic movement primitives,

    A. Paraschos, C. Daniel, J. R. Peters, and G. Neumann, “Probabilistic movement primitives,”Advances in neural information processing systems, vol. 26, 2013

    2013

  22. [30]

    Prodmp: A unified perspective on dynamic and probabilistic move- ment primitives,

    G. Li, Z. Jin, M. V olpp, F. Otto, R. Lioutikov, and G. Neumann, “Prodmp: A unified perspective on dynamic and probabilistic move- ment primitives,”IEEE Robotics and Automation Letters, vol. 8, no. 4, pp. 2325–2332, 2023

    2023

  23. [31]

    Learning table tennis with a mixture of motor primitives,

    K. Muelling, J. Kober, and J. Peters, “Learning table tennis with a mixture of motor primitives,” in2010 10th IEEE-RAS International Conference on Humanoid Robots, 2010, pp. 411–416

    2010

  24. [32]

    Residual learning from demonstration: Adapting dmps for contact-rich manipulation,

    T. Davchev, K. S. Luck, M. Burke, F. Meier, S. Schaal, and S. Ra- mamoorthy, “Residual learning from demonstration: Adapting dmps for contact-rich manipulation,”IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 4488–4495, 2022

    2022

  25. [33]

    Residual robot learning for object-centric probabilistic movement primitives,

    J. Carvalho, D. Koert, M. Daniv, and J. Peters, “Residual robot learning for object-centric probabilistic movement primitives,”arXiv preprint arXiv:2203.03918, 2022

    2022 arXiv

  26. [34]

    Optimizing robot striking movement primitives with iterative learning control,

    O. Koç, G. Maeda, G. Neumann, and J. Peters, “Optimizing robot striking movement primitives with iterative learning control,” in2015 IEEE-RAS 15th International Conference on Humanoid Robots (Hu- manoids), 2015, pp. 80–87

    2015

  27. [35]

    Hub: Learning extreme humanoid balance,

    T. Zhang, B. Zheng, R. Nai, Y . Hu, Y .-J. Wang, G. Chen, F. Lin, J. Li, C. Hong, K. Sreenath,et al., “Hub: Learning extreme humanoid balance,”arXiv preprint arXiv:2505.07294, 2025

    2025 arXiv

  28. [36]

    Serl: A software suite for sample- efficient robotic reinforcement learning,

    J. Luo, Z. Hu, C. Xu, Y . L. Tan, J. Berg, A. Sharma, S. Schaal, C. Finn, A. Gupta, and S. Levine, “Serl: A software suite for sample- efficient robotic reinforcement learning,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 16 961–16 969

    2024

  29. [37]

    Precise and dexterous robotic manipulation via human-in-the-loop reinforcement learning,

    J. Luo, C. Xu, J. Wu, and S. Levine, “Precise and dexterous robotic manipulation via human-in-the-loop reinforcement learning,”arXiv preprint arXiv:2410.21845, 2024

    2024 arXiv

  30. [38]

    Asap: Aligning simulation and real-world physics for learning agile humanoid whole-body skills,

    T. He, J. Gao, W. Xiao, Y . Zhang, Z. Wang, J. Wang, Z. Luo, G. He, N. Sobanbab, C. Pan, Z. Yi, G. Qu, K. Kitani, J. Hodgins, L. J. Fan, Y . Zhu, C. Liu, and G. Shi, “Asap: Aligning simulation and real-world physics for learning agile humanoid whole-body skills,” Robotics: Sci...

    2025

  31. [39]

    Glm-4.5v and glm-4.1v-thinking: Towards versatile multimodal reasoning with scalable reinforcement learning,

    V . Team, W. Hong, W. Yu, X. Gu, G. Wang, G. Gan, H. Tang, J. Cheng, J. Qi, J. Ji, L. Pan, S. Duan, W. Wang, Y . Wang, Y . Cheng, Z. He, Z. Su, Z. Yang, Z. Pan, A. Zeng, B. Wang, B. Chen, B. Shi, C. Pang, C. Zhang, D. Yin, F. Yang, G. Chen, J. Xu, J. Zhu, J. Chen, J. Chen, J. ...

    2025 arXiv

Pith tools

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