pith. sign in

arxiv: 2605.01581 · v4 · pith:DIAXRENUnew · submitted 2026-05-02 · 💻 cs.RO

Hyper-DP3: Frequency-Aware Right-Sizing of 3D Diffusion Policies for Visuomotor Control

Jinhao Zhang , Zhexuan Zhou , Huizhe Li , Yichen Lai , Wenlong Xia , Haoming Song , Youmin Gong , Jie Mei This is my paper

Pith reviewed 2026-07-01 00:15 UTC · model grok-4.3

classification 💻 cs.RO
keywords diffusion policiesvisuomotor controlfrequency domain analysisaction denoisingrobot manipulationlightweight modelsDDIM sampling
0
0 comments X

The pith

Robot action smoothness bounds denoising error, allowing two-step inference in compact diffusion policies.

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

The paper shows that because robot action trajectories concentrate energy in a few low-frequency modes, the error in the optimal denoiser is limited by the size of that subspace and leftover high-frequency content. This bound means that denoising performance stops improving after only a small number of reverse diffusion steps. Consequently, the full complexity of image-generation diffusion models is unnecessary for visuomotor policies. The authors therefore introduce a much smaller architecture that still reaches high performance on manipulation tasks.

Core claim

Under the frequency structure of smooth robot actions, the error of the optimal denoiser is bounded by the low-frequency subspace dimension and residual high-frequency energy. This implies that denoising error saturates after very few reverse steps, so action denoising requires a much simpler denoising model than image generation. This motivates a pocket-scale 3D diffusion policy with a lightweight decoder that supports two-step inference while matching or exceeding prior performance.

What carries the argument

The frequency-domain bound on the optimal denoiser's error using low-frequency discrete cosine transform subspace dimension and residual high-frequency energy.

If this is right

  • Two-step DDIM sampling is sufficient for action denoising.
  • A lightweight Diffusion Mixer decoder replaces heavier image-style decoders.
  • State-of-the-art performance is achieved with under 1% of the parameters of prior 3D diffusion policies.
  • Substantially lower inference latency results from the reduced model size and step count.

Where Pith is reading between the lines

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

  • If action trajectories in other domains exhibit similar frequency concentration, the same right-sizing approach could apply.
  • Tasks with less smooth actions may require more steps or larger models, providing a way to predict model size from trajectory statistics.
  • The approach separates the design of visuomotor policies from general image diffusion techniques.

Load-bearing premise

Robot action trajectories are highly smooth, with most energy concentrated in a few low-frequency discrete cosine transform modes.

What would settle it

Collecting action trajectories from a robot task and computing their discrete cosine transform energy distribution; if the energy is not concentrated in the lowest few modes, the error bound and two-step sufficiency would not hold.

Figures

Figures reproduced from arXiv: 2605.01581 by Haoming Song, Huizhe Li, Jie Mei, Jinhao Zhang, Wenlong Xia, Yichen Lai, Youmin Gong, Zhexuan Zhou.

Figure 1
Figure 1. Figure 1: Frequency structure of action trajectories. view at source ↗
Figure 2
Figure 2. Figure 2: Overall architecture of the Proposed Method. In the figure, T denotes transpose. We adopt the efficient point-cloud encoder from DP3[34] and stacks K DiM blocks as the decoder. Each DiM block is built upon an MLP-Mixer style[29] architecture, enabling efficient information fusion with a small parameter budget, thereby improving decision-making performance. denoising is substantially simpler than that of hi… view at source ↗
Figure 3
Figure 3. Figure 3: MSE Under Different NFEs 0 10 20 30 40 50 60 time 2 1 0 1 2 value lowfreq steps 1 2 10 100 0 10 20 30 40 50 60 time broadband steps 1 2 10 100 0 10 20 30 40 50 60 time highfreq steps 1 2 10 100 view at source ↗
Figure 3
Figure 3. Figure 3: MSE Under Different NFEs 0 10 20 30 40 50 60 time 2 1 0 1 2 value lowfreq steps 1 2 10 100 0 10 20 30 40 50 60 time broadband steps 1 2 10 100 0 10 20 30 40 50 60 time highfreq steps 1 2 10 100 [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Examples of normalized synthetic trajectories from the low-frequency-dominant, broadband, and high-frequency view at source ↗
Figure 6
Figure 6. Figure 6: Decoding Error at Different Sampling Steps view at source ↗
Figure 7
Figure 7. Figure 7: Real-World Experiment Setup. target platform (a) (b) (c) (d) (e) view at source ↗
Figure 8
Figure 8. Figure 8: Real-world Experiments. The image sequence (top to bottom) illustrates the robot successfully performing three tasks: placing an object, uprighting a fallen cup, and stacking two blocks. D Frequency-domain Decomposition Details For each trajectory segment, we analyze the 14-dimensional action sequence X ∈ R T ×14, where Xt,d denotes the d-th action component at time step t. Each episode is partitioned into… view at source ↗
Figure 9
Figure 9. Figure 9: Fraction of Energy Contained in the First 5% of DCT Modes for Each Task view at source ↗
read the original abstract

Diffusion-based visuomotor policies perform well in robotic manipulation, yet current methods still inherit image-generation-style decoders and multi-step sampling. We revisit this design from a frequency-domain perspective. Robot action trajectories are highly smooth, with most energy concentrated in a few low-frequency discrete cosine transform modes. Under this structure, we show that the error of the optimal denoiser is bounded by the low-frequency subspace dimension and residual high-frequency energy, implying that denoising error saturates after very few reverse steps. This also suggests that action denoising requires a much simpler denoising model than image generation. Motivated by this insight, we propose Hyper-DP3 (HDP3), a pocket-scale 3D diffusion policy with a lightweight Diffusion Mixer decoder that supports two-step DDIM inference. Our synthetic experiments validate the theory and support the sufficiency of two-step denoising. Futhermore, across RoboTwin2.0, Adroit, MetaWorld, and real-world tasks, HDP3 achieves state-of-the-art performance with fewer than 1% of the parameters of prior 3D diffusion-based policies and substantially lower inference latency.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript claims that robot action trajectories are highly smooth with energy concentrated in a few low-frequency DCT modes. From this, it derives that the error of the optimal denoiser is bounded by the low-frequency subspace dimension and residual high-frequency energy, implying that denoising error saturates after very few reverse steps and that action denoising requires a simpler model than image generation. Motivated by this, the authors introduce Hyper-DP3, a pocket-scale 3D diffusion policy with a lightweight Diffusion Mixer decoder supporting two-step DDIM inference. Synthetic experiments are said to validate the theory, and the method is reported to achieve SOTA performance on RoboTwin2.0, Adroit, MetaWorld, and real-world tasks with under 1% of the parameters of prior 3D diffusion policies and lower inference latency.

Significance. If the frequency bound is rigorously derived and the smoothness assumption holds with quantitative support across the evaluated action spaces, the work would provide a principled way to right-size diffusion policies for robotics, substantially improving efficiency and deployability while maintaining performance.

major comments (2)
  1. [Abstract] Abstract: The central claim that 'robot action trajectories are highly smooth, with most energy concentrated in a few low-frequency discrete cosine transform modes' is asserted without quantitative energy distribution plots, DCT coefficient analysis, or checks on the specific action spaces (e.g., 7-DoF in Adroit/MetaWorld). This assumption is load-bearing for the error bound and the justification for two-step DDIM; its absence prevents assessment of whether residual high-frequency energy is negligible enough to imply early saturation.
  2. [Synthetic experiments] Synthetic experiments section: The validation of the theoretical bound is described without reference to error bars, data exclusion rules, or explicit metrics demonstrating saturation after two steps, leaving open whether the experiments confirm the bound or merely illustrate an empirical observation.
minor comments (1)
  1. [Abstract] Abstract: Typo in 'Futhermore' (should be 'Furthermore').

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our frequency-domain analysis and experimental validation. The comments highlight opportunities to strengthen the presentation of our central assumptions and results. We address each major comment below and commit to revisions that improve clarity and rigor without altering the core contributions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that 'robot action trajectories are highly smooth, with most energy concentrated in a few low-frequency discrete cosine transform modes' is asserted without quantitative energy distribution plots, DCT coefficient analysis, or checks on the specific action spaces (e.g., 7-DoF in Adroit/MetaWorld). This assumption is load-bearing for the error bound and the justification for two-step DDIM; its absence prevents assessment of whether residual high-frequency energy is negligible enough to imply early saturation.

    Authors: We agree that the smoothness assumption is foundational and benefits from explicit quantitative support in the main text. The manuscript derives the error bound in Section 3 and validates the frequency concentration via synthetic experiments in Section 4.1 across the relevant action spaces (including 7-DoF from Adroit and MetaWorld). To directly address the concern, we will add energy distribution plots, DCT coefficient histograms, and a concise quantitative summary to the main body (with references from the abstract and introduction). This will make the load-bearing nature of the assumption transparent for readers. revision: yes

  2. Referee: [Synthetic experiments] Synthetic experiments section: The validation of the theoretical bound is described without reference to error bars, data exclusion rules, or explicit metrics demonstrating saturation after two steps, leaving open whether the experiments confirm the bound or merely illustrate an empirical observation.

    Authors: We concur that additional statistical details will strengthen the validation. The current synthetic experiments demonstrate saturation consistent with the bound, but the revised version will incorporate error bars on all plots, explicitly state any data exclusion criteria, and report quantitative metrics (such as per-step denoising MSE curves with confidence intervals) showing that error plateaus after two DDIM steps. These changes will confirm that the results rigorously support the theoretical prediction rather than serving only as illustration. revision: yes

Circularity Check

0 steps flagged

No circularity; bound derived from standard DCT subspace properties under stated assumption

full rationale

The paper premises robot action smoothness and low-frequency DCT energy concentration as an empirical fact, then states that the optimal denoiser error is bounded by subspace dimension plus residual high-frequency energy. This is a direct mathematical consequence of the DCT basis truncation and does not reduce to a fitted parameter renamed as prediction, a self-citation chain, or a self-definitional loop. No equations are shown that equate the claimed saturation result to the input assumption by construction. Synthetic experiments are cited as validation, keeping the derivation self-contained against external frequency-analysis benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests primarily on the domain assumption of trajectory smoothness in the frequency domain and standard mathematical properties of DCT and diffusion processes; no new entities are postulated and the two-step choice is the main free parameter.

free parameters (1)
  • number of reverse steps = 2
    Selected as the point where denoising error saturates according to the frequency bound.
axioms (1)
  • domain assumption Robot action trajectories are highly smooth with most energy concentrated in a few low-frequency discrete cosine transform modes
    This premise is invoked to derive the bound on denoiser error and the sufficiency of few reverse steps.

pith-pipeline@v0.9.1-grok · 5755 in / 1332 out tokens · 48621 ms · 2026-07-01T00:15:25.745613+00:00 · methodology

Review history (4 revisions) →

discussion (0)

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

Forward citations

Cited by 2 Pith papers

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

  1. Ambient Diffusion Policy: Imitation Learning from Suboptimal Data in Robotics

    cs.RO 2026-06 unverdicted novelty 7.0

    Ambient Diffusion Policy enables better imitation learning from suboptimal robot data by leveraging spectral properties to restrict data usage to specific diffusion times.

  2. Latent Diffusion Policy: Shaping Latent Spaces for Diffusion-Based Robotic Manipulation

    cs.RO 2026-06 unverdicted novelty 6.0

    LDP shapes an observation-conditioned latent space with CVAE to simplify flow matching for diffusion policies, claiming substantial gains over DP3 on bimanual coordination tasks in simulation and real-world transfer.

Reference graph

Works this paper leans on

40 extracted references · 18 canonical work pages · cited by 2 Pith papers · 12 internal anchors

  1. [1]

    A survey of robot learning from demonstration.Robotics and autonomous systems, 57(5):469–483, 2009

    Brenna D Argall, Sonia Chernova, Manuela Veloso, and Brett Browning. A survey of robot learning from demonstration.Robotics and autonomous systems, 57(5):469–483, 2009

    2009

  2. [2]

    $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control

    Kevin Black, Noah Brown, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, Lachy Groom, Karol Hausman, Brian Ichter, et al. π0: A vision-language-action flow model for general robot control.arXiv preprint arXiv:2410.24164, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  3. [3]

    Score approximation, estimation and distribution recovery of diffusion models on low-dimensional data

    Minshuo Chen, Kaixuan Huang, Tuo Zhao, and Mengdi Wang. Score approximation, estimation and distribution recovery of diffusion models on low-dimensional data. InInternational Conference on Machine Learning, pages 4672–4712. PMLR, 2023

    2023

  4. [4]

    RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation

    Tianxing Chen, Zanxin Chen, Baijun Chen, Zijian Cai, Yibin Liu, Zixuan Li, Qiwei Liang, Xianliang Lin, Yiheng Ge, Zhenyu Gu, et al. Robotwin 2.0: A scalable data generator and benchmark with strong domain randomization for robust bimanual robotic manipulation.arXiv preprint arXiv:2506.18088, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    Diffusion policy: Visuomotor policy learning via action diffusion

    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. The International Journal of Robotics Research, 44(10-11):1684–1704, 2025

    2025

  6. [6]

    Diffusion Posterior Sampling for General Noisy Inverse Problems

    Hyungjin Chung, Jeongsol Kim, Michael T Mccann, Marc L Klasky, and Jong Chul Ye. Diffu- sion posterior sampling for general noisy inverse problems.arXiv preprint arXiv:2209.14687, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  7. [7]

    A Fourier Space Perspective on Diffu- sion Models,

    Fabian Falck, Teodora Pandeva, Kiarash Zahirnia, Rachel Lawrence, Richard Turner, Edward Meeds, Javier Zazo, and Sushrut Karmalkar. A fourier space perspective on diffusion models. arXiv preprint arXiv:2505.11278, 2025

    work page arXiv 2025

  8. [8]

    Implicit behavioral cloning

    Pete Florence, Corey Lynch, Andy Zeng, Oscar A Ramirez, Ayzaan Wahid, Laura Downs, Adrian Wong, Johnny Lee, Igor Mordatch, and Jonathan Tompson. Implicit behavioral cloning. InConference on Robot Learning, pages 158–168. PMLR, 2022

    2022

  9. [9]

    Mean Flows for One-step Generative Modeling

    Zhengyang Geng, Mingyang Deng, Xingjian Bai, J Zico Kolter, and Kaiming He. Mean flows for one-step generative modeling.arXiv preprint arXiv:2505.13447, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  10. [10]

    Act3d: 3d feature field transformers for multi-task robotic manipulation

    Theophile Gervet, Zhou Xian, Nikolaos Gkanatsios, and Katerina Fragkiadaki. Act3d: 3d feature field transformers for multi-task robotic manipulation. InConference on Robot Learning, pages 3949–3965. PMLR, 2023

    2023

  11. [11]

    Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    2020

  12. [12]

    Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow

    Xingchao Liu, Chengyue Gong, and Qiang Liu. Flow straight and fast: Learning to generate and transfer data with rectified flow.arXiv preprint arXiv:2209.03003, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  13. [13]

    Decoupled Weight Decay Regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  14. [14]

    What Matters in Learning from Offline Human Demonstrations for Robot Manipulation

    Ajay Mandlekar, Danfei Xu, Josiah Wong, Soroush Nasiriany, Chen Wang, Rohun Kulkarni, Li Fei-Fei, Silvio Savarese, Yuke Zhu, and Roberto Martín-Martín. What matters in learning from offline human demonstrations for robot manipulation.arXiv preprint arXiv:2108.03298, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  15. [15]

    An algorithmic perspective on imitation learning.Foundations and Trends® in Robotics, 7(1-2):1–179, 2018

    Takayuki Osa, Joni Pajarinen, Gerhard Neumann, J Andrew Bagnell, Pieter Abbeel, and Jan Peters. An algorithmic perspective on imitation learning.Foundations and Trends® in Robotics, 7(1-2):1–179, 2018

    2018

  16. [16]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF international conference on computer vision, pages 4195–4205, 2023. 11

    2023

  17. [17]

    $\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization

    Physical Intelligence, Kevin Black, Noah Brown, James Darpinian, Karan Dhabalia, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, et al. π0.5: A vision- language-action model with open-world generalization.arXiv preprint arXiv:2504.16054, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  18. [18]

    Consistency policy: Accelerated visuomotor policies via consistency distillation,

    Aaditya Prasad, Kevin Lin, Jimmy Wu, Linqi Zhou, and Jeannette Bohg. Consistency policy: Accelerated visuomotor policies via consistency distillation.arXiv preprint arXiv:2405.07503, 2024

    work page arXiv 2024

  19. [19]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space.Advances in neural information processing systems, 30, 2017

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space.Advances in neural information processing systems, 30, 2017

    2017

  20. [20]

    On the spectral bias of neural networks

    Nasim Rahaman, Aristide Baratin, Devansh Arpit, Felix Draxler, Min Lin, Fred Hamprecht, Yoshua Bengio, and Aaron Courville. On the spectral bias of neural networks. InInternational conference on machine learning, pages 5301–5310. PMLR, 2019

    2019

  21. [21]

    Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations

    Aravind Rajeswaran, Vikash Kumar, Abhishek Gupta, Giulia Vezzani, John Schulman, Emanuel Todorov, and Sergey Levine. Learning complex dexterous manipulation with deep reinforcement learning and demonstrations.arXiv preprint arXiv:1709.10087, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  22. [22]

    Hierarchical Text-Conditional Image Generation with CLIP Latents

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image generation with clip latents.arXiv preprint arXiv:2204.06125, 1(2):3, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  23. [23]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. InInternational Conference on Medical image computing and computer-assisted intervention, pages 234–241. Springer, 2015

    2015

  24. [24]

    A reduction of imitation learning and structured prediction to no-regret online learning

    Stéphane Ross, Geoffrey Gordon, and Drew Bagnell. A reduction of imitation learning and structured prediction to no-regret online learning. InProceedings of the fourteenth interna- tional conference on artificial intelligence and statistics, pages 627–635. JMLR Workshop and Conference Proceedings, 2011

    2011

  25. [25]

    Masked world models for visual control

    Younggyo Seo, Danijar Hafner, Hao Liu, Fangchen Liu, Stephen James, Kimin Lee, and Pieter Abbeel. Masked world models for visual control. InConference on Robot Learning, pages 1332–1344. PMLR, 2023

    2023

  26. [26]

    Mp1: Meanflow tames policy learning in 1-step for robotic manipulation

    Juyi Sheng, Ziyi Wang, Peiming Li, and Mengyuan Liu. Mp1: Meanflow tames policy learning in 1-step for robotic manipulation. InProceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 18532–18539, 2026

    2026

  27. [27]

    Perceiver-actor: A multi-task transformer for robotic manipulation

    Mohit Shridhar, Lucas Manuelli, and Dieter Fox. Perceiver-actor: A multi-task transformer for robotic manipulation. InConference on Robot Learning, pages 785–799. PMLR, 2023

    2023

  28. [28]

    Learning structured output representation using deep conditional generative models

    Kihyuk Sohn, Honglak Lee, and Xinchen Yan. Learning structured output representation using deep conditional generative models. InAdvances in Neural Information Processing Systems, volume 28, 2015

    2015

  29. [29]

    arXiv preprint arXiv:2505.21432 (2025)

    Haoming Song, Delin Qu, Yuanqi Yao, Qizhi Chen, Qi Lv, Yiwen Tang, Modi Shi, Guanghui Ren, Maoqing Yao, Bin Zhao, et al. Hume: Introducing system-2 thinking in visual-language- action model.arXiv preprint arXiv:2505.21432, 2025

    work page arXiv 2025

  30. [30]

    Denoising Diffusion Implicit Models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models.arXiv preprint arXiv:2010.02502, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  31. [31]

    Consistency models

    Yang Song, Prafulla Dhariwal, Mark Chen, and Ilya Sutskever. Consistency models. 2023

    2023

  32. [32]

    Fourier controller networks for real-time decision-making in embodied learning.arXiv preprint arXiv:2405.19885, 2024

    Hengkai Tan, Songming Liu, Kai Ma, Chengyang Ying, Xingxing Zhang, Hang Su, and Jun Zhu. Fourier controller networks for real-time decision-making in embodied learning.arXiv preprint arXiv:2405.19885, 2024. 12

    work page arXiv 2024

  33. [33]

    Mlp- mixer: An all-mlp architecture for vision.Advances in neural information processing systems, 34:24261–24272, 2021

    Ilya O Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Unterthiner, Jessica Yung, Andreas Steiner, Daniel Keysers, Jakob Uszkoreit, et al. Mlp- mixer: An all-mlp architecture for vision.Advances in neural information processing systems, 34:24261–24272, 2021

    2021

  34. [34]

    One-step diffusion policy: Fast visuomotor policies via diffusion distillation

    Zhendong Wang, Max Li, Ajay Mandlekar, Zhenjia Xu, Jiaojiao Fan, Yashraj Narang, Linxi Fan, Yuke Zhu, Yogesh Balaji, Mingyuan Zhou, Ming-Yu Liu, and Yu Zeng. One-step diffusion policy: Fast visuomotor policies via diffusion distillation. InForty-second International Conference on Machine Learning, 2025

    2025

  35. [35]

    One-step diffusion policy: Fast visuomotor policies via diffusion distillation,

    Zhendong Wang, Zhaoshuo Li, Ajay Mandlekar, Zhenjia Xu, Jiaojiao Fan, Yashraj Narang, Linxi Fan, Yuke Zhu, Yogesh Balaji, Mingyuan Zhou, et al. One-step diffusion policy: Fast visuomotor policies via diffusion distillation.arXiv preprint arXiv:2410.21257, 2024

    work page arXiv 2024

  36. [36]

    Stable Velocity: A Variance Perspective on Flow Matching

    Donglin Yang, Yongxing Zhang, Xin Yu, Liang Hou, Xin Tao, Pengfei Wan, Xiaojuan Qi, and Renjie Liao. Stable velocity: A variance perspective on flow matching.arXiv preprint arXiv:2602.05435, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  37. [37]

    Consistency flow matching: Defining straight flows with velocity consistency.arXiv preprint arXiv:2407.02398, 2024

    Ling Yang, Zixiang Zhang, Zhilong Zhang, Xingchao Liu, Minkai Xu, Wentao Zhang, Chenlin Meng, Stefano Ermon, and Bin Cui. Consistency flow matching: Defining straight flows with velocity consistency.arXiv preprint arXiv:2407.02398, 2024

    work page arXiv 2024

  38. [38]

    Meta-world: A benchmark and evaluation for multi-task and meta reinforcement learning

    Tianhe Yu, Deirdre Quillen, Zhanpeng He, Ryan Julian, Karol Hausman, Chelsea Finn, and Sergey Levine. Meta-world: A benchmark and evaluation for multi-task and meta reinforcement learning. InConference on robot learning, pages 1094–1100. PMLR, 2020

    2020

  39. [39]

    3d diffusion policy: Generalizable visuomotor policy learning via simple 3d representations

    Yanjie Ze, Gu Zhang, Kangning Zhang, Chenyuan Hu, Muhan Wang, and Huazhe Xu. 3d diffusion policy: Generalizable visuomotor policy learning via simple 3d representations. In 2nd Workshop on Dexterous Manipulation: Design, Perception and Control (RSS), 2024

    2024

  40. [40]

    Flowpolicy: Enabling fast and robust 3d flow-based policy via consistency flow matching for robot manipulation

    Qinglun Zhang, Zhen Liu, Haoqiang Fan, Guanghui Liu, Bing Zeng, and Shuaicheng Liu. Flowpolicy: Enabling fast and robust 3d flow-based policy via consistency flow matching for robot manipulation. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 14754–14762, 2025. 13 A Proof of Theorem 4.1 We first derive the closed-form ex...

    2025