Pith. sign in

REVIEW 4 major objections 6 minor 38 references

Integrating Diffusion-based Multi-task Learning with Online Reinforcement Learning for Robust Quadruped Robot Control

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read DMLoco claims diffusion policies can master quadruped gaits and language commands when offline imitation is followed by a simple two-term PPO finetune.

desk verdict A practical diffusion-PPO recipe with real-robot gains, but the multi-gait claim is not yet proven because gait identity is never measured. read the letter →

arxiv 2507.05674 v2 pith:ZXBYAV44 submitted 2025-07-08 cs.RO

Xinyao Qin , Xiaoteng Ma , Yang Qi , Qihan Liu , Chuanyi Xue , Ning Gui , Qinyu Dong , Jun Yang
show 1 more author
Bin Liang
This is my paper
classification cs.RO
keywords diffusionpolicyquadrupedlocomotionPPOfinetuninglanguage-conditionedcontrolgaittransitionsDDIMsamplingsim-to-realtransfermulti-taskimitationlearning
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 tries to show that diffusion-based imitation learning, which has lagged behind in legged locomotion because of compounding errors and missing transition data, becomes practical when followed by a short online reinforcement-learning phase. The proposed DMLoco first pretrains a diffusion policy on offline expert data for four quadruped gaits, conditioned on either structured velocity commands or natural language. It then finetunes that policy with PPO in simulation using only two reward terms—speed-tracking error and fall detection—which lets the robot switch between gaits without any explicit transition data. The authors report that this two-stage recipe reaches perfect success across gaits, beats four baselines on tracking error, and runs onboard at 50 Hz using deterministic DDIM sampling and hardware acceleration. If true, it offers a low-reward-engineering route to language-guided, multi-skill locomotion on resource-constrained robots.

What carries the argument

The load-bearing mechanism is the formulation of a DDPM policy as a two-layer MDP, whose inner denoising loop can be optimised by standard policy-gradient algorithms; this is what lets PPO finetune the pretrained diffusion policy directly. Around that core sit three supporting design choices: conditioning the denoiser on a window of past states and goals (state horizon 30) while predicting only the current action rather than action chunks; a two-term finetuning reward (speed-tracking error and fall detection) with a higher denoising noise schedule to encourage exploration; and DDIM deterministic sampling with five denoising steps, which removes the stochastic delays of DDPM and, with hardware acceleration, reaches the 50 Hz control frequency needed for real-time quadruped control.

What would settle it

Run the exact two-term PPO finetuning recipe (same noise schedule, learning rate, five DDIM steps, and reward weights) on a second quadruped model in simulation, then test the same four gait transitions at 1.0 m/s in at least 20 trials per transition. If success rates fall well below the reported near-100% or the policy catastrophically forgets one of the pretrained gaits, the finetuning phase—not the pretraining—is not the source of the claimed robustness.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that the gap between offline diffusion policies and deployable quadruped control is closed by a particular division of labour. Offline pretraining supplies multi-modal coverage of four gaits (trotting, bounding, pacing, pronking) and aligns language embeddings with structured commands; online PPO finetuning supplies robustness and gait-transition ability that pure imitation lacks. Crucially, finetuning needs no transition demonstrations and no complex reward stack—two terms, speed-tracking error and fall detection, suffice. The paper further finds that predicting a single action per timestep rather than action chunks is important for dynamic locomotion, and that DDIM's deterministic few-step sampling (5 steps) gives both the speed and stability that DDPM cannot match at low step counts. Together these choices yield 100% success across gaits in simulation, the lowest tracking error among the compared methods, and stable transitions on the physical robot, with language-conditioned performance nearly matching structured-command performance.

Load-bearing premise

The load-bearing premise is that a short PPO finetuning phase using only speed-tracking error and fall detection can turn an imitation-only diffusion policy into one that transfers to a real robot with stable gait transitions, despite never seeing transition data and without forgetting the pretrained skills.

Editorial extensions

If this is right

  • A diffusion policy pretrained by imitation can be made robust enough for real-robot deployment with a finetuning phase that needs no gait-transition demonstrations and only two reward terms.
  • The same policy can follow both structured velocity commands and natural-language instructions, with language-conditioned control staying close to structured-command performance on unseen test instructions.
  • Gait transitions that are absent from the training distribution—trotting to bounding, bounding to pacing, pacing to pronking, pronking to trotting—become reliable after online finetuning, even at 1.0 m/s.
  • DDIM with five sampling steps replaces 100-step DDPM sampling without losing stability, cutting inference cost enough to run the policy onboard at 50 Hz on embedded hardware.
  • Predicting single actions instead of action chunks is preferable in locomotion because it avoids delayed responses to rapidly changing states.

Reading between the lines

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

  • The two-term-reward finetuning recipe is stated generally enough that it should transfer to other quadruped platforms and other skill sets; a direct test would be applying the same hyperparameters to a different robot model without reward redesign.
  • If the division of labour between imitation pretraining and RL finetuning is the real cause of robustness, then the same recipe should also improve non-diffusion policies—for instance, a transformer policy pretrained on the same dataset and finetuned with the same two-term PPO reward.
  • The language interface's apparent generalization suggests the diffusion policy stores skills in a shared representation aligned to both structured and free-form commands; an untested corollary is that novel combinations of commands and gaits might be composable, such as requesting a new gait at speeds outside the training ranges.
  • The dependence on DDIM's determinism for stability implies a sensitivity to sampling stochasticity; a useful stress test would be perturbing the denoising trajectory at deployment and measuring how much the gait-transition success degrades.
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 DMLoco, a framework for quadruped robot locomotion that combines diffusion-based multi-task imitation pretraining with online PPO finetuning. The policy is first trained on an offline dataset of four gaits (trotting, bounding, pacing, pronking) with structured velocity/gait commands or natural-language embeddings; it is then finetuned in simulation with a simple two-term reward (speed-tracking error and fall detection) to improve robustness and enable gait transitions. The authors report high velocity-tracking success and low tracking error in simulation, transition-success results on a real Unitree Aliengo robot, and real-time 50 Hz onboard inference via DDIM and TensorRT. The central claims are that DMLoco enables language-conditioned multi-gait control and robust gait transitions without explicit transition data.

Significance. If the central claims hold, DMLoco offers a practical recipe for converting offline diffusion policies for legged locomotion into robust, deployable controllers with a minimal reward design, and it demonstrates language-conditioned control plus real-time onboard execution. The paper provides a code/video link and includes both simulation and real-robot experiments, which is valuable for reproducibility. The main technical risk is that the finetuning procedure may destroy the multi-gait structure of the pretrained policy, and the current evaluation metrics do not resolve this uncertainty. The contribution is incremental but potentially useful; the evidence as presented does not yet fully support the multi-gait and transition claims.

major comments (4)
  1. [Section V-A and Figure 4] The finetuning reward contains only speed-tracking error and fall detection, with no term that rewards matching the commanded gait, no KL constraint toward the pretrained policy, and no auxiliary behavior-cloning loss. The paper states that a relatively high diffusion noise schedule is used to encourage exploration, which further increases the risk of mode collapse. The evaluation metrics in Table I (success rate defined as 'correctly follows the instruction and remains upright' and MSE velocity error) do not measure foot-contact pattern, gait phase, or any other gait-identity quantity. Consequently, the reported 100% success rates and low tracking errors are compatible with a policy that has collapsed to a single robust gait (e.g., trotting) while the velocity command is tracked. The paper's central claims of multi-gait execution and task transitions therefore require either an explicit gait-preservation mechanism in the finetuning objective or an evaluation that verifies the executed gait matches the command; without one of these, the evidence in Table I and Figure 4 does not support the claims.
  2. [Section III-C] The gait-transition experiment is underpowered and incompletely reported. The text says 'Each transition type is repeated 5 times, resulting in 20 trials in total,' but the experiment is run at three speeds (0.1, 0.5, 1.0 m/s) and four transition types, which would require 60 trials if five repetitions per condition are used; the discrepancy is unexplained. Figure 4 shows no error bars or statistical tests, and two baselines (BeT and DMLoco-lang) are excluded with only a brief textual justification. The reader cannot assess whether the reported transition-success differences are significant or whether the selected baselines would change the comparison. The paper should clarify the trial count, report per-speed and per-transition breakdowns with confidence intervals, and either include all baselines or justify their exclusion with quantitative results.
  3. [Section V-C and Figure 5(a)] The paper claims to use PPO for finetuning but only shows a REINFORCE-style gradient: ∇θ J = E[∑ ∇θ log π̄θ r̄]. The clipped surrogate objective, advantage estimation (GAE), and the treatment of the two-layer MDP for diffusion policies are not specified. Since the integration of online PPO finetuning is one of the paper's main contributions, the missing derivation makes the method difficult to reproduce and obscures whether the implementation deviates from standard PPO. The authors should provide the full PPO objective as applied to their diffusion policy, or explicitly refer to the derivation in [8] and state any modifications.
  4. [Section V-B, Table I] The learning curve in Figure 5(a) shows the success rate increasing from about 0.6 to nearly 1.0 during finetuning, but the paper does not specify the evaluation protocol used to compute this curve. If the success rate is measured on the same velocity-tracking task as Table I, it contradicts the near-perfect performance of the pretrained policy (DiffuseLoco, which is described as similar to the pretraining stage) and the high success rates reported in Table II for the pretrained policy with DDIM. If the evaluation during finetuning includes gait transitions or other more difficult conditions, this should be stated explicitly. The inconsistency between the learning curve and the pretrained-policy performance needs to be resolved for the finetuning benefit to be convincing.
minor comments (6)
  1. [General] The manuscript contains several typos and stylistic inconsistencies, e.g., 'First We' in the Figure 1 caption and 'preform' instead of 'perform'; the abstract uses 'finetuning' while the main text uses both 'finetuning' and 'fine-tuning'.
  2. [References] References [17] and [18] are identical (both list 'Learning quadrupedal locomotion over challenging terrain'); one should be removed or replaced with a distinct relevant work.
  3. [Table I] Table I reports point estimates without standard deviations, confidence intervals, or the number of seeds; the 100.00 success rates across all four gaits are suspiciously clean and would be more credible with per-seed variation.
  4. [Section V-A] The definition of 'success' in the velocity-tracking task is vague; the paper should specify the threshold on tracking error, the required time horizon, and whether the robot must stay within a bounded region.
  5. [Table III] The finetuning hyperparameter 'Sampling denoising std 0.04' is not tied to the statement in Section IV.3 that 'we employ a relatively high diffusion noise scheduling to encourage exploration'; the paper should state the pretraining sampling std and clarify how the finetuning value is relatively high.
  6. [Section IV.4] The description of the GPT-4-generated language dataset lacks details such as the prompt template, filtering criteria, and examples of the 1,000 instructions; additional detail would improve reproducibility of the language-conditioning results.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: DMLoco's derivation is empirical and independently validated.

full rationale

The paper's core claims are empirical and do not reduce to their inputs by construction. The pretraining stage trains a diffusion policy by behavior cloning on an expert dataset using an MSE loss; the finetuning stage optimizes a two-term reward (speed tracking error and fall detection) via PPO. Neither stage defines a target metric in terms of the method's own outputs in a way that forces the reported success. Language-conditioned generalization is evaluated on a held-out test split of 1,000 GPT-4-generated instructions (8:1:1 train/validation/test), providing an independent check of the language encoder's alignment. The gait-transition results are empirical simulation and real-robot measurements compared against external baselines (BeT, DiffuseLoco, WTW, Cassi). The paper cites prior work (e.g., DiffuseLoco) but does not rely on any self-citation chain to justify its central premise; the diffusion-policy formulation and PPO for diffusion are standard results from the literature. The absence of gait-identity metrics or the small number of real-robot transition trials are experimental-validity concerns, not circularity. No equation in the paper is equivalent to an input fit or a renamed prediction, so the circularity score is 0.

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

The paper relies on standard diffusion and RL machinery, plus a small number of tuned hyperparameters (state horizon, denoising steps, learning rate, sampling std, and undisclosed reward weights). The main domain assumptions are the adequacy of Isaac Gym for sim-to-real transfer and the generalizability of the language encoder. No new physical entities are introduced.

free parameters (5)
  • State horizon h_s = 30
    Tuned in ablation (Section V-C): shorter horizons degrade performance, longer ones add redundancy; the best value 30 is chosen before finetuning.
  • Denoising steps during finetuning = 5
    Selected for latency with DDIM; the ablation in Table II justifies 5 steps vs 100 for DDPM.
  • Sampling denoising std = 0.04
    Set during finetuning to encourage exploration; no sensitivity analysis is reported.
  • Finetuning learning rate = 1e-5
    Hyperparameter from Table III; no sensitivity analysis reported.
  • Reward weights for speed tracking and fall penalty = not disclosed
    The reward is two terms but weights are not given; this is a free parameter that materially affects the finetuned behavior.
assumptions (5)
  • standard math DDPM denoising as defined in Ho et al. [36] produces correct samples for the policy distribution.
    Section III-B adopts the standard DDPM reverse process without modification.
  • domain assumption A UNet conditioned on state and goal sequences can represent the multi-modal action distribution of the four gaits.
    Assumed in Section IV.2; no theoretical guarantee that the model covers all gaits without mode collapse.
  • standard math The two-layer MDP formulation (Ren et al. [8]) allows PPO to optimize the diffusion policy.
    Section III-C cites the formulation; the gradient expression is the score function gradient, and the paper claims PPO but does not derive the clipped objective.
  • domain assumption Isaac Gym simulation is a sufficiently accurate proxy for the Unitree Aliengo real robot for sim-to-real transfer of the finetuned policy.
    Section V-A uses Isaac Gym for finetuning; real robot tests confirm, but limited to 5 trials per condition.
  • domain assumption The learned language encoder maps free-form language to the same structured goal space used for reward computation.
    Section IV.4 trains El to approximate g_struct; evaluation on held-out GPT-generated instructions assumes this mapping generalizes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Integrating Diffusion-based Multi-task Learning with Online Reinforcement Learning for Robust Quadruped Robot Control." pith.science (2026). https://pith.science/paper/ZXBYAV44

@misc{pith2026250705674,
  author       = {Pith},
  title        = {Pith review of: Integrating Diffusion-based Multi-task Learning with Online Reinforcement Learning for Robust Quadruped Robot Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZXBYAV44}},
  note         = {Machine review of arXiv:2507.05674}
}
read the original abstract

Recent research has highlighted the powerful capabilities of imitation learning in robotics. Leveraging generative models, particularly diffusion models, these approaches offer notable advantages such as strong multi-task generalization, effective language conditioning, and high sample efficiency. While their application has been successful in manipulation tasks, their use in legged locomotion remains relatively underexplored, mainly due to compounding errors that affect stability and difficulties in task transition under limited data. Online reinforcement learning (RL) has demonstrated promising results in legged robot control in the past years, providing valuable insights to address these challenges. In this work, we propose DMLoco, a diffusion-based framework for quadruped robots that integrates multi-task pretraining with online PPO finetuning to enable language-conditioned control and robust task transitions. Our approach first pretrains the policy on a diverse multi-task dataset using diffusion models, enabling language-guided execution of various skills. Then, it finetunes the policy in simulation to ensure robustness and stable task transition during real-world deployment. By utilizing Denoising Diffusion Implicit Models (DDIM) for efficient sampling and TensorRT for optimized deployment, our policy runs onboard at 50Hz, offering a scalable and efficient solution for adaptive, language-guided locomotion on resource-constrained robotic platforms.

Figures

Figures reproduced from arXiv: 2507.05674 by the authors.

Figure 1
Figure 1. The overall framework of DMLoco. First We perform imitation learning on an offline multi-skill dataset and encode [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Snapshots of DMLoco showing different gaits, with red points indicating ground contact points. This figure highlights [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Network architecture of DMLoco. (Left) Natural [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Gait Transition Success Rate over Baselines [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Compared to the pretrained policy, the finetuned [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 7
Figure 7. Figure 7: Inference frequency of DMLoco across different plat [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

38 extracted references · 23 canonical work pages

  1. [8]

    Diffusion Policy Policy Optimiza- tion

    Allen Z. Ren et al. “Diffusion Policy Policy Optimiza- tion”. In:arXiv preprint arXiv:2409.00588. 2024

    arXiv 2024

  2. [1]

    High-resolution image syn- thesis with latent diffusion models

    Robin Rombach et al. “High-resolution image syn- thesis with latent diffusion models”. In:Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022, pp. 10684–10695. TABLE III:Hyperparameters in Training Stage Hyperparameter Value Pretrain Batch size 512 Learning rate 3e-4 Denoising steps 100 State horizon 30 Diffusion step e...

    work page 2022

  3. [2]

    Scaling rectified flow transformers for high-resolution image synthesis

    Patrick Esser et al. “Scaling rectified flow transformers for high-resolution image synthesis”. In:Forty-first international conference on machine learning. 2024

    work page 2024

  4. [3]

    Video generation models as world simulators

    Tim Brooks et al. “Video generation models as world simulators”. In:OpenAI Blog1.8 (2024), p. 1

    work page 2024

  5. [4]

    Condi- tional variational autoencoder with adversarial learn- ing for end-to-end text-to-speech

    Jaehyeon Kim, Jungil Kong, and Juhee Son. “Condi- tional variational autoencoder with adversarial learn- ing for end-to-end text-to-speech”. In:International Conference on Machine Learning. PMLR. 2021, pp. 5530–5540

    work page 2021

  6. [5]

    Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robots

    Cheng Chi et al. “Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robots”. In:Proceedings of Robotics: Science and Systems (RSS). 2024

    2024

  7. [6]

    Diffusion Policy: Visuomotor Policy Learning via Action Diffusion

    Cheng Chi et al. “Diffusion Policy: Visuomotor Policy Learning via Action Diffusion”. In:Proceedings of Robotics: Science and Systems (RSS). 2023

    work page 2023

  8. [7]

    3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Repre- sentations

    Yanjie Ze et al. “3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Repre- sentations”. In:Proceedings of Robotics: Science and Systems (RSS). 2024

    work page 2024

Show all 38 references
  1. [9]

    π0: A vision-language-action flow model for general robot control. CoRR, abs/2410.24164, 2024. doi: 10.48550

    Kevin Black et al. “π0: A vision-language-action flow model for general robot control. CoRR, abs/2410.24164, 2024. doi: 10.48550”. In:arXiv preprint ARXIV .2410.24164()

    2024 arXiv

  2. [10]

    RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation

    Songming Liu et al. “RDT-1B: a Diffusion Foundation Model for Bimanual Manipulation”. In:arXiv preprint arXiv:2410.07864(2024)

    2024 arXiv

  3. [11]

    Agibot world colosseo: A large-scale manipulation platform for scalable and intelligent embodied systems

    Qingwen Bu et al. “Agibot world colosseo: A large-scale manipulation platform for scalable and intelligent embodied systems”. In:arXiv preprint arXiv:2503.06669(2025)

    2025 arXiv

  4. [12]

    Genesis Authors.Genesis: A Universal and Genera- tive Physics Engine for Robotics and Beyond. Dec. 2024.URL:https://github.com/Genesis- Embodied-AI/Genesis

    2024

  5. [13]

    Orbit: A Unified Simulation Framework for Interactive Robot Learning Environ- ments

    Mayank Mittal et al. “Orbit: A Unified Simulation Framework for Interactive Robot Learning Environ- ments”. In:IEEE Robotics and Automation Letters8.6 (2023), pp. 3740–3747

    2023

  6. [14]

    Isaac gym: High perfor- mance gpu-based physics simulation for robot learn- ing

    Viktor Makoviychuk et al. “Isaac gym: High perfor- mance gpu-based physics simulation for robot learn- ing”. In:arXiv preprint arXiv:2108.10470(2021)

    2021 arXiv

  7. [15]

    SAPIEN: A SimulAted Part-based Interactive ENvironment

    Fanbo Xiang et al. “SAPIEN: A SimulAted Part-based Interactive ENvironment”. In:The IEEE Conference on Computer Vision and Pattern Recognition (CVPR). June 2020

    2020

  8. [16]

    Mu- JoCo: A physics engine for model-based control

    Emanuel Todorov, Tom Erez, and Yuval Tassa. “Mu- JoCo: A physics engine for model-based control”. In: 2012 IEEE/RSJ International Conference on Intelli- gent Robots and Systems. IEEE. 2012, pp. 5026–5033

    2012

  9. [18]

    Learning quadrupedal locomotion over challenging terrain

    Joonho Lee et al. “Learning quadrupedal locomotion over challenging terrain”. In:Science robotics5.47 (2020), eabc5986

    2020

  10. [19]

    Walk these ways: Tuning robot control for generalization with multiplicity of behavior

    Gabriel B Margolis and Pulkit Agrawal. “Walk these ways: Tuning robot control for generalization with multiplicity of behavior”. In:Conference on Robot Learning. PMLR. 2023, pp. 22–31

    2023

  11. [20]

    Adversarial motion priors make good substitutes for complex reward functions

    Alejandro Escontrela et al. “Adversarial motion priors make good substitutes for complex reward functions”. In:2022 IEEE/RSJ International Conference on In- telligent Robots and Systems (IROS). IEEE. 2022, pp. 25–32

    2022

  12. [21]

    From Experts to a Generalist: Toward General Whole-Body Control for Humanoid Robots

    Yuxuan Wang et al. “From Experts to a Generalist: Toward General Whole-Body Control for Humanoid Robots”. In:arXiv preprint arXiv:2506.12779(2025)

    2025 arXiv

  13. [22]

    Rldg: Robotic generalist policy dis- tillation via reinforcement learning

    Charles Xu et al. “Rldg: Robotic generalist policy dis- tillation via reinforcement learning”. In:arXiv preprint arXiv:2412.09858(2024)

    2024 arXiv

  14. [23]

    Sim-to-real learning of all common bipedal gaits via periodic reward com- position

    Jonah Siekmann et al. “Sim-to-real learning of all common bipedal gaits via periodic reward com- position”. In:2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE. 2021, pp. 7309–7315

    2021

  15. [24]

    Learning a single policy for di- verse behaviors on a quadrupedal robot using scalable motion imitation

    Arnaud Klipfel et al. “Learning a single policy for di- verse behaviors on a quadrupedal robot using scalable motion imitation”. In:2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE. 2023, pp. 2768–2775

    2023

  16. [25]

    Versatile skill control via self- supervised adversarial imitation of unlabeled mixed motions

    Chenhao Li et al. “Versatile skill control via self- supervised adversarial imitation of unlabeled mixed motions”. In:arXiv preprint arXiv:2209.07899(2022)

    2022 arXiv

  17. [26]

    Learning mul- tiple gaits within latent space for quadruped robots

    Jinze Wu, Yufei Xue, and Chenkun Qi. “Learning mul- tiple gaits within latent space for quadruped robots”. In:arXiv preprint arXiv:2308.03014(2023)

    2023 arXiv

  18. [27]

    Skill latent space based multigait learning for a legged robot

    Xin Liu et al. “Skill latent space based multigait learning for a legged robot”. In:IEEE Transactions on Industrial Electronics(2024)

    2024

  19. [28]

    DiffuseLoco: Real-Time Legged Locomotion Control with Diffusion from Of- fline Datasets

    Xiaoyu Huang et al. “DiffuseLoco: Real-Time Legged Locomotion Control with Diffusion from Of- fline Datasets”. In:arXiv preprint arXiv:2404.19264 (2024)

    2024 arXiv

  20. [29]

    Language-Guided Manipulation with Diffusion Policies and Constrained Inpainting

    Ce Hao et al. “Language-Guided Manipulation with Diffusion Policies and Constrained Inpainting”. In: arXiv preprint arXiv:2406.09767(2024)

    2024 arXiv

  21. [30]

    Render and Diffuse: Align- ing Image and Action Spaces for Diffusion- based Behaviour Cloning

    Vitalis V osylius et al. “Render and Diffuse: Align- ing Image and Action Spaces for Diffusion- based Behaviour Cloning”. In:arXiv preprint arXiv:2405.18196(2024)

    2024 arXiv

  22. [31]

    Consistency Policy: Accelerated Visuomotor Policies via Consistency Distillation

    Aaditya Prasad et al. “Consistency Policy: Accelerated Visuomotor Policies via Consistency Distillation”. In: Robotics: Science and Systems. 2024

    2024

  23. [32]

    3d diffuser actor: Policy diffusion with 3d scene representations

    Tsung-Wei Ke, Nikolaos Gkanatsios, and Katerina Fragkiadaki. “3d diffuser actor: Policy diffusion with 3d scene representations”. In:arXiv preprint arXiv:2402.10885(2024)

    2024 arXiv

  24. [33]

    Diffusion reward: Learning rewards via conditional video diffusion

    Tao Huang et al. “Diffusion reward: Learning rewards via conditional video diffusion”. In:European Con- ference on Computer Vision. Springer. 2024, pp. 478– 495

    2024

  25. [34]

    Extracting reward functions from diffusion models

    Felipe Nuti, Tim Franzmeyer, and Jo ˜ao F Henriques. “Extracting reward functions from diffusion models”. In:Advances in Neural Information Processing Sys- tems36 (2023), pp. 50196–50220

    2023

  26. [35]

    Robot Motion Diffusion Model: Motion Generation for Robotic Characters

    Agon Serifi et al. “Robot Motion Diffusion Model: Motion Generation for Robotic Characters”. In:SIG- GRAPH Asia 2024 Conference Papers. SA ’24. As- sociation for Computing Machinery, 2024.ISBN: 9798400711312

    2024

  27. [36]

    De- noising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. “De- noising diffusion probabilistic models”. In:Advances in neural information processing systems33 (2020), pp. 6840–6851

    2020

  28. [37]

    Learning fine-grained biman- ual manipulation with low-cost hardware

    Tony Z Zhao et al. “Learning fine-grained biman- ual manipulation with low-cost hardware”. In:arXiv preprint arXiv:2304.13705(2023)

    2023 arXiv

  29. [38]

    Minilmv2: Multi-head self- attention relation distillation for compressing pretrained transformers

    Wenhui Wang et al. “Minilmv2: Multi-head self- attention relation distillation for compressing pretrained transformers”. In:arXiv preprint arXiv:2012.15828(2020)

    2020 arXiv

  30. [39]

    Behavior transform- ers: Cloningkmodes with one stone

    Nur Muhammad Shafiullah et al. “Behavior transform- ers: Cloningkmodes with one stone”. In:Advances in neural information processing systems35 (2022), pp. 22955–22968

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.