Real-Time Execution of Action Chunking Flow Policies
Open accessThe DOI matches its Crossref record; 61 of the 61 references it cites were found in a registry or an index; 34 of the 47 sources its claims cite were checked for support.
Abstract
Modern AI systems, especially those interacting with the physical world, increasingly require real-time performance. However, the high latency of state-of-the-art generalist models, including recent vision-language action models (VLAs), poses a significant challenge. While action chunking has enabled temporal consistency in high-frequency control tasks, it does not fully address the latency problem, leading to pauses or out-of-distribution jerky movements at chunk boundaries. This paper presents a novel inference-time algorithm that enables smooth asynchronous execution of action chunking policies. Our method, real-time chunking (RTC), is applicable to any diffusion- or flow-based VLA out of the box with no re-training. It generates the next action chunk while executing the current one, "freezing" actions guaranteed to execute and "inpainting" the rest. To test RTC, we introduce a new benchmark of 12 highly dynamic tasks in the Kinetix simulator, as well as evaluate 6 challenging real-world bimanual manipulation tasks. Results demonstrate that RTC is fast, performant, and uniquely robust to inference delay, significantly improving task throughput and enabling high success rates in precise tasks $\unicode{x2013}$ such as lighting a match $\unicode{x2013}$ even in the presence of significant latency. See https://pi.website/research/real_time_chunking for videos.
Overview
CoChat Summary
TL;DR
The paper introduces real-time chunking (RTC), an inference-time method that executes action-chunking diffusion/flow policies asynchronously by treating the next chunk as an inpainting problem. Across a 12-task Kinetix benchmark and 6 real-world bimanual tasks, RTC improves robustness to inference delay and task throughput without retraining.
Key findings
- In simulation, RTC is the most robust to inference delay among compared methods, outperforming bidirectional decoding (BID) and temporal ensembling (TE); soft masking improves performance especially at lower delays and execution horizons.
- In real-world experiments with π0.5, RTC achieves the best average task throughput at all inference delays, with statistically significant gains at +100 ms and +200 ms injected latency; synchronous inference degrades linearly, while both TE variants fail at those delays due to high oscillations triggering protective stops.
- RTC is completely robust to injected delay in the real-world setting, showing no degradation, and completes tasks faster than synchronous inference even when inference pauses are removed; in the precision-sensitive light candle task, RTC shows a large final-score advantage.
- Figure 1 reports RTC performs the same robot motion 20% faster than synchronous inference and smoother than all competing methods, including temporal ensembling.
Methods
- RTC is an inference-time algorithm for asynchronous execution of action-chunking policies: it generates the next action chunk while executing the current one, freezes actions guaranteed to execute due to inference delay, and inpaints the rest using a guidance-based flow-matching inpainting procedure with soft masking and guidance weight clipping (β = 5).
- Simulated benchmark: 12 highly dynamic tasks in the Kinetix simulator, with expert policies trained via RPO, action-chunking flow policies with prediction horizon H = 8 and a 4-layer MLP-Mixer, and 2048 rollouts per data point; delays simulated between 0 and 4.
- Real-world evaluation: π0.5 VLA with H = 50, Δt = 20 ms, n = 5 denoising steps, model latency 76 ms for baselines and 97 ms for RTC; 6 challenging bimanual manipulation tasks, 10 trials per task and method for 480 episodes and 28 hours of robot execution time, with injected latency of +100 ms and +200 ms.
Limitations
- The paper states RTC adds significant computational overhead compared to sampling directly from the base policy, is applicable only to diffusion- and flow-based policies, and its real-world experiments do not cover legged locomotion even though it is represented in simulation.
- The paper states that the data used to train π0.5 and the real robot runtime code are not released because they are proprietary; only the simulated benchmark code is provided.
- Evident from the text: real-world results use a single base policy (π0.5) and 10 trials per task/method, so generalization across policies and statistical precision are limited beyond what is reported.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures
![Figure 1 Top: Real-time chunking (RTC) enables the robot to perform highly dexterous and dynamic tasks, such as lighting a match-even in the presence of inference delays in excess of 300 milliseconds, corresponding to more than 30% of the model's prediction horizon. Bottom: RTC performs the same robot motion 20% faster than synchronous inference [5, 30, 8, 24, 31, 59], and smoother than all competing methods, including temporal ensembling [68]. The shown positions, velocities, and accelerations correspond to the shoulder joint of one arm, and are taken from the first 10 seconds of a real autonomous match-lighting rollout.](https://figures.cochat.ai/paper-figures/3cd4ba2056b24cc4b146393d27abcbea/real-time-execution-of-action-chunking-flow-policies-figure-1.png)
![Figure 2 An illustration of a typical bifurcation between consecutive chunks. Inference is started between timesteps 3 and 4. The original chunk that was executing, { a t } (black), had planned to go above the obstacle while the newly generated chunk { a ′ t } (red) goes below the obstacle. However, { a ′ t } is not available until d = 7 steps later. A naive asynchronous algorithm might jump from a 10 to a ′ 11 , inducing a very high, outof-distribution acceleration. Temporal ensembling [68], i.e., interpolating between chunks, reduces the acceleration but produces poor actions.](https://figures.cochat.ai/paper-figures/d3bac9cd3b5f4edc91d2704f2802ba8e/real-time-execution-of-action-chunking-flow-policies-figure-2.png)





![Figure 8 Left: Simulated ablation over different schedules for soft masking weights (Eq. 5). Exponential decay performs the best overall, although linear decay is very close behind. Right: Comparison with the inpainting algorithm from Diffuser [26], which overwrites a portion of the action chunk with the desired actions at each denoising step. While this simpler (and cheaper) inpainting method still provides some benefit, it is outperformed by our guidance-based approach.](https://figures.cochat.ai/paper-figures/536a75e2a3ad4b978b379cb9b7f0de58/real-time-execution-of-action-chunking-flow-policies-figure-8.png)
Verification
Claims the paper makes
14 Valid · 7 Not valid · 4 Not checked · 76 Own results and interpretations
Some cited sources support this claim and some don’t — read each source’s verdict below.
In the paper
“Though recent works combining learning methods with MPC have demonstrated real-time control capabilities in narrow domains [53, 21], they rely on explicit, hand-crafted dynamics models and cost functions.”
What the sources say
- Supportedcites [53] [53] Tim Salzmann, Elia Kaufmann, Jon Arrizabalaga, Marco Pavone, Davide Scaramuzza, and Markus Ryll. Real-time neural mpc: Deep learning model predictive control for quadrotors and agile robotic platforms. IEEE Robotics and Automation Letters , 8(4):2397-2404, 2023. - source
The Problem Setup (p. 3) says the MPC dynamics combine first-principles equations with a learned model. The Experimental Setup (pp. 5–6) specifies a nominal rigid-body dynamics model and a quadratic tracking cost. Thus the paper supports reliance on explicitly specified dynamics and cost functions, although its dynamics also include a learned residual. (LLM, conf=0.88)
- Disputedcites [21] [21] Nicklas Hansen, Xiaolong Wang, and Hao Su. Temporal difference learning for model predictive control, 2022. URL https://arxiv.org/abs/2203.04955 . - sourceExperimental — accuracy not yet measured
Section 3 (p. 3) describes TD-MPC as combining MPC with a learned latent dynamics model and learned reward model. The Introduction (p. 1) also says its terminal value function is learned. As an example of the cited recent work, it contradicts the claim that such methods rely on explicit, hand-crafted dynamics models and cost functions. (LLM, conf=0.90)
Checked by a reasoning model — may be wrong.
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“While action chunking has enabled temporal consistency in high-frequency control tasks, it does not fully address the latency problem, leading to pauses or out-of-distribution jerky movements at chunk boundaries.”
What the sources say
- Partly supportedcites [68] [68] Tony Z Zhao, Vikash Kumar, Sergey Levine, and Chelsea Finn. Learning fine-grained bimanual manipulation with low-cost hardware. arXiv preprint arXiv:2304.13705 , 2023. - sourceExperimental — accuracy not yet measured
In “Action Chunking and Temporal Ensemble” (page not given), the paper says chunking helps model temporally correlated behavior in high-frequency trajectories. It also says chunking alone can produce jerky motion at chunk boundaries and credits temporal ensembling with improving smoothness, so it does not establish that chunking by itself enables temporal consistency. (LLM, conf=0.84)
- Could not verifycites [33] [33] Lucy Lai, Ann Zixiang Huang, and Samuel J Gershman. Action chunking as policy compression. 2022. - sourceExperimental — accuracy not yet measured
The Introduction and §4.1 discuss action chunking in human sequence-learning tasks, including faster and more accurate responses to patterned stimuli. Our copy does not address temporal consistency in high-frequency control tasks; searches for that phrase and for high-frequency control, latency, and chunk-boundary movements found no supporting passage. (LLM, conf=0.76)
- Supportedcites [11] [11] 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 , page 02783649241273668, 2023. - source
Section 5.3 (p. 8) reports that an action horizon greater than one helps the policy predict consistent actions; Section 4.3 (p. 5) explains how sequence prediction avoids jitter between consecutive actions. The paper applies this approach to real-time robot control, with learned policies running at 10 Hz (D.1, p. 18). (LLM, conf=0.87)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Kim et al. [31], who optimize the 7B OpenVLA model [30] specifically for inference speed, achieve no better than 321ms of latency on a server-grade A100 GPU.”
What the sources say
- Partly supportedcites [31] [31] Moo Jin Kim, Chelsea Finn, and Percy Liang. Fine-tuning vision-language-action models: Optimizing speed and success. arXiv preprint arXiv:2502.19645 , 2025. - sourceExperimental — accuracy not yet measured
Table III (ALOHA Inference Efficiency Comparisons, p. 10) reports 0.321 seconds on an NVIDIA A100 for OpenVLA-OFT+, an optimized OpenVLA variant, in a bimanual task with three input images. The claim’s “no better than” wording extends that task-specific result to the model generally: Related Work (p. 2) reports lower latency for a single-arm task with one input image. (LLM, conf=0.88)
- Partly supportedcites [30] [30] Moo Jin Kim, Karl Pertsch, Siddharth Karamcheti, Ted Xiao, Ashwin Balakrishna, Suraj Nair, Rafael Rafailov, Ethan Foster, Grace Lam, Pannag Sanketi, et al. Openvla: An open-source vision-language-action model. arXiv preprint arXiv:2406.09246 , 2024. - sourceExperimental — accuracy not yet measured
Section 5.4 (p. 10) identifies OpenVLA as a 7B-parameter model and discusses inference speed on server-grade GPUs. But the text in our copy does not give a 321 ms A100 latency or establish it as a best achievable latency; the Figure 6 data are not present in the text. Section 3.5 (p. 6) describes its reported inference speed as measured without speed-up techniques, rather than for a model specifically optimized for inference speed. (LLM, conf=0.84)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“we devise a new benchmark based on the Kinetix simulator [43] consisting of 12 highly dynamic manipulation and locomotion tasks.”
What the sources say
- Supportedcites [43] [43] Michael Matthews, Michael Beukman, Chris Lu, and Jakob Foerster. Kinetix: Investigating the training of general agents through open-ended physics-based control tasks. arXiv preprint arXiv:2410.23208 , 2024. - source
The citation identifies Kinetix as the simulator on which the citing authors base their new benchmark; it does not attribute the 12 tasks to the cited paper. The Introduction describes Kinetix as a framework for 2D physics-based environments, and Section 3.1 identifies its underlying physics engine. (LLM, conf=0.91)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Inspecting the per-task results (Figure 5, top), we can conclude that RTC helps with more than just execution speed: it completes tasks faster than synchronous inference even when inference pauses are removed .”
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
42 more claims
Create a free account to see all 101 claims and to verify the 9 references still open.
- Probably not a substantial claim. 33
- It may state two separate claims as one. 25
- The quote only partly states this claim. 2
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Baselines. We compare to the following baselines: · p. 8
“Inspecting the per-task results (Figure 5, top)”
§ Baselines. We compare to the following baselines: · p. 8
“Figure 6: Top: Controller steps (equivalent to elapsed time with inference pauses removed multiplied by 50Hz) vs. cumulative progress for each task, aggregated across all delays.”
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Research works printed but not linked: 6
- [2][2] Atilim Gunes Baydin, Barak A Pearlmutter, Alexey Andreyevich Radul, and Jeffrey Mark Siskind. Automatic differentiation in machine learning: a survey. Journal of machine learning research , 18(153):1-43, 2018. -
- [8][8] Anthony Brohan, Noah Brown, Justice Carbajal, Yevgen Chebotar, Xi Chen, Krzysztof Choromanski, Tianli Ding, Danny Driess, Avinava Dubey, Chelsea Finn, et al. Rt-2: Vision-languageaction models transfer web knowledge to robotic control. arXiv preprint arXiv:2307.15818 , 2023. - · arXiv
- [14][14] Donald R. Coughanowr and Steven E. LeBlanc. Process Systems Analysis and Control , chapter 18. McGraw-Hill, New York, 3rd edition, 2009. ISBN 978-0073397894. -
- [22][22] Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems , 33:6840-6851, 2020. -
- [44][44] Atharva Mete, Haotian Xue, Albert Wilcox, Yongxin Chen, and Animesh Garg. Quest: Selfsupervised skill abstractions for learning continuous control, 2024. URL https://arxiv. org/abs/2407.15840 . -
- [55][55] Jiaming Song, Arash Vahdat, Morteza Mardani, and Jan Kautz. Pseudoinverse-guided diffusion models for inverse problems. In International Conference on Learning Representations , 2023. -
Cited by
Items 101–200 of 207
Static In, Dynamic Out: Counterfactual Action Augmentation for Moving Object ManipulationPreprint
Shin, W. C., Chen, Z., Cueva, A., Arachchige, N. R., Lin, Y. C., Joffe, B., Kousik, S., & Xu, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.27890 · Cited by 3
Long-WAM: Scaling the Context of World-Action ModelsPreprint
Huang, W., Zhang, B., Liu, C., Liu, I., Yang, S., Mao, W., Wang, L., Xiao, Y., Lin, W., Hu, Q., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.10528 · Cited by 3
Towards Agentic Robots: Model, Data, Environment, and HarnessPreprint
Qin, C., Wu, D., Feng, K., Wang, J., Yu, X., Sun, P., Fan, Y., He, X., Chen, B., Lian, W., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.20442 · Cited by 4
Zhang, Z., Huang, R., Liu, Y., Zhu, S., Mou, L., & Zhao, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.17834 · Cited by 2
Position: Embodied AI Requires a Privacy-Utility Trade-offPreprint
Fan, X., Chen, J., Liu, Z., Yang, Z., Xu, P., Shen, R., Liu, J., Qi, J., & Wang, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.05017 · Cited by 3
Frequency-Guided Action Diffusion via Sub-Frequency Manifold TraversalPreprint
Wang, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.27919 · Cited by 2
DynamicManip: Enabling Dynamic Manipulation from a Single Static DemonstrationPreprint
Liao, H., Wang, P., Chen, S., Cheng, K., Chen, X., Lin, Y., Lin, M., Liang, Z., Fan, X., Xing, C., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.01452 · Cited by 2
Deng, Y., & Xu, F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.12308 · Cited by 2
FluxVLA Engine: A One-Stop VLA Engineering Platform for Embodied IntelligencePreprint
Li, Y., Mao, W., Lan, Z., Rong, J., Hu, Q., Zhang, Y., Deng, W., Shen, B., Zhu, M., Mao, Y., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.17210 · Cited by 4
Watch, Recall, Act: Always-On Robots in Concurrent Embodied StreamsPreprint
Yi, D., Sun, P., Rong, C., Wang, J., Dai, X., Yue, X., & Lin, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.28429 · Cited by 2
Realtime-Bench: Can Frontier Models Solve Real-time Interactive Problems?Preprint
Tang, J., Yang, X., Zhang, Z., Wang, Z., Cheng, Q., Zhao, H., Li, M., Li, W., Xia, X., & Pei, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.22155 · Cited by 2
Large Language Models for UAV Autonomy from a Perception–Cognition–Action Perspective
Xiong, T., Zhan, J., Deng, Q., Wang, X., Fan, C., Liu, X., & Zhang, T. (2026) · Drones · doi:10.3390/drones10090669 · Cited by 1
Data Scaling for Navigation in Unknown Environments
Suomela, L., TAKAHATA, N., Arachchige, S. K., Edelman, H., & Kämäräinen, J.-K. (2026) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2026.3677718 · Cited by 4
You, H., Lu, H., Dong, W., Shaohui, L., Li, Z., Yang, L., Jie, Z., & Ruan, S. (2026) · Journal of Image and Graphics · doi:10.11834/jig.260042 · Cited by 1
Beyond LLM Serving: Characterizing Vision-Language-Action Workloads for Embodied AI System Design
Jung, S., Moon, S., Jeong, J., Lee, J., & Huh, J. (2026) · International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) · doi:10.1145/3845814.3850089 · Cited by 2
iFlyBot-VLA Technical ReportPreprint
Zhang, Y., Xue, C., Xu, W., Ji, C., wu, J., & Pan, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2511.01914 · Cited by 1
Hechtl, J., Schmitt, P. S., Wichert, G. V., & Burgard, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.16218 · Cited by 1
Steering Autoregressive Vision-Language-Action Policies via Action Token InterventionPreprint
Chan, J., & Kao, J. C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.15021 · Cited by 1
Learning to Fold: prizewinning solution at LeHome Challenge 2026 (1st place online, 2nd offline)Preprint
Larchenko, I. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.27163 · Cited by 3
SUNTA: Hierarchical Video Prediction with Surprise-based ChunkingPreprint
Iiyama, T., Suzuki, M., & Matsuo, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.02087 · Cited by 1
Spatial Attention: Adapting Execution Horizons for Diffusion Policies via Observation SensitivityPreprint
Park, C. G., Ha, J., Fu, J., & Park, F. C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.04739 · Cited by 1
Shang, M., Huang, Y., Chen, J., Hu, H., Yu, H., Song, L., 冯绿匀, Bao, S., Dong, W., Wang, X., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.13924 · Cited by 1
SmoothRL: Online Reinforcement Learning During Asynchronous ExecutionPreprint
Gao, G., Nong, Y., Huang, B., & Wang, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.29768 · Cited by 1
FIERCE: From Generalist Robot Policies to Fast Specialists via Progress-Failure FeedbackPreprint
Tan, R., Tu, Y., Yan, Y., Yu, L., Tian, X., & Lv, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.18651 · Cited by 1
ForceDelta-VLA: Distilling Force-Conditioned Action Corrections for Contact-Rich ManipulationPreprint
Dong, J., Fu, Y., Chen, J., Liu, Y., Zhao, H., Zhang, L., Bai, K., Zhang, L., Zheng, D., Knoll, A., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.18242 · Cited by 1
Dynamic Manipulation with World-Action Models via Counterfactual PlanningPreprint
Park, S., Lee, W., Jin, S., 김영민, Park, J., & Ye, J. C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.33172 · Cited by 1
ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuningPreprint
Zhu, D., Yan, Z., & Wan, F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34982 · Cited by 1
Workhorse: Learning Robust Whole-Body Humanoid Loco-Manipulation from Human DataPreprint
Hu, S., Liao, Q., Chi, Y., Zakka, K., Shao, Y. S., Abbeel, P., & Sreenath, K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.09117 · Cited by 1
TrajWAM: An Efficient Trajectory-Guided World Action Model for Dynamic Robotic ManipulationPreprint
Geng, H., Zhang, S., Sun, W., Ye, G., Yu, W., Xie, Y., Zeng, Q., Ye, P., Deng, J., & Ji, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.25763 · Cited by 1
LIBERO-Dynamic: Benchmarking Robotic Manipulation in Dynamic EnvironmentsPreprint
Ren, S., Sun, J., Liu, Z., Chen, H., Jin, X., Zou, W., & Chen, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.24579 · Cited by 2
Gym-Anything-RT: Scaling Real-Time Environments for Vision-Language AgentsPreprint
Aggarwal, P., & Welleck, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.37567 · Cited by 3
Cowling, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.37170 · Cited by 1
Yang, X., & Wu, N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.40210 · Cited by 1
SegDiff: Segmented Trajectory Diffusion for Consistent and Adaptive Robot Manipulation
Cao, H., Cao, W., Li, Q., Xie, S., Du, Z., Leng, J., Wu, Z., & Jiang, Y.-G. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-37577-3_36
Khan, M. T. A., & Bombardelli, J. (2026) · Journal of Robotic Surgery · doi:10.1007/s11701-026-03866-9 · Cited by 2
VTP: A Task-Aware Transmission Protocol for Edge Vision-Language-Action Robotic Systems
Huo, R., Wang, Y., Wu, Y., Dong, Y., & Yang, P. (2026) · IEEE/CIC International Conference on Communications in China (ICCC) · doi:10.1109/iccc70295.2026.11680263
Qian, C., Xie, H., & Wang, W. (2026) · International Symposium on Control Engineering and Robotics (ISCER) · doi:10.1109/iscer70245.2026.11681422
Adaptive Mean Flow for Responsive Closed-Loop Robot Control
Vaaler, A., Job, M., Holden, C., & Egeland, O. (2026) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2026.3734902
SMJ-ACT: Chunk-Aware Stable Minimum-Jerk Execution Refinement for ACT Policies
Wu, S., Li, W., Feng, Y., Xie, J., Dai, H., & Xing, F. (2026) · International Conference on Machine Manufacturing and Intelligent Control (ICMMIC) · doi:10.1109/icmmic70515.2026.11690522
Toward a Multi-Embodied Grasping Agent
Freiberg, R., Qualmann, A., Vien, N. A., & Neumann, G. M. (2026) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2026.3686666 · Cited by 1
Recent advances in diffusion-based embodied imitation learning: a survey
Mu, L. I., Liu, X., Li, Q., Yu, X., Jingzhi, C. U. I., Zhao, L., Feng, H. A. N., Yang, X., Zhou, B., Niu, J., et al. (2025) · Scientia Sinica Informationis · doi:10.1360/ssi-2025-0307 · Cited by 2
Compositional Visual Planning via Inference-Time Diffusion ScalingPreprint
Zhang, Y., Luo, Y., Mishra, U. A., Shin, W. C., Chen, Y., & Xu, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.02646
GoldenStart: Q-Guided Priors and Entropy Control for Distilling Flow PoliciesPreprint
Zhang, H., Sun, Y., & Xiong, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.14245 · Cited by 1
RIO: Flexible Real-time Robot I/O for Cross-Embodiment Robot Learning
Ortega-Kral, P., Xing, E., Bucker, A., Luk, V., Kim, J., Kwon, O., Xie, A., Sobanbabu, N., Yuan, Y., Lee, M., et al. (2026) · Robotics: Science and Systems · doi:10.15607/rss.2026.xxii.164 · Cited by 3
Implicit Action Chunking for Smooth Continuous ControlPreprint
Liang, B., Pei, S., Chen, Z.-R., Fan, C., Sun, C., Wu, Y.-K., Tan, H.-C., & Wang, Y. (2026) · arXiv.org · doi:10.48550/arxiv.2605.19592
Abstraction for Offline Goal-Conditioned Reinforcement Learning
Wibault, C., Goldie, A. D., Villares, A. L., Osborne, M., & Foerster, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.22711 · Cited by 1
Real-World Human-Robot Interaction Behavior Generation using Latent Diffusion Models
Stanovcic, S. (2026) · reposiTUm (TU Wien) · doi:10.34726/hss.2026.123206
Truongcao, K., Nhu, C., An, Z., Nguyen, P. V., Cai, S., & Zhou, L. (2026) · IEEE International Conference on Automation Science and Engineering (CASE) · doi:10.1109/case69030.2026.11704444
Zhang, Z., Zhang, N., Li, Y., Cai, J., & Wang, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.11019 · Cited by 1
A Glimpse into Long-term Physical Coexistence with Intelligent RobotsPreprint
Jin, W., Tang, P., Luo, K., Huang, B., Sun, B., Yang, H., Xie, S., & Wang, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.11377
Wang, T., Wang, W., Wang, J., Ye, Z., Wang, Q., Huang, W., & Xu, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.15924
Mandischer, N., Böckmann, N., Holl, L., & Mikelsons, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.23320
Otake, A., Hartmann, F., Igari, K., & Murata, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.25284
Enwerem, C. O., Baras, J. S., & Belta, C. A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.01453
WAVE-Go: World-Model Navigation with Adaptive Execution for Wheel-Legged RobotsPreprint
Li, M., Li, J., Ouyang, Z., He, Y., & Karlsson, B. F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.18193
rMuscle: Robotic Muscle Memory for Efficient Vision-Language-Action Model InferencePreprint
Zhou, K., Li, Z., Chen, L., & Gu, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.19104
KerColle: Unlocking Fine-Grained GPU Concurrency in Vision-Language-Action ModelsPreprint
Li, A., Giannoula, C., & Vijaykumar, N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.22335 · Cited by 1
DualWAM: Dual-System World Action Models for Asynchronous Global Planning and Local RefinementPreprint
Zheng, Y., Lyu, J., Deng, Y., Liu, K., Zhou, Y., Wang, Y., Zhao, X., Wang, H., & Zhang, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.24868
Mao, Y., Liu, F., Yuan, S., Hu, T., Liu, R., & Su, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.25969
Zhu, H., Ma, Y., He, S., Li, L., Gong, S., Li, M., Xiong, X., Wang, C., Tian, H., Liu, Q., et al. (2026) · IEEE Transactions on Circuits and Systems I Regular Papers · doi:10.1109/tcsi.2026.3732698
Aarya, S., Xi-Jia, Z., Huang, C., Kim, J., Xue, H., Leen, H., Yakunin, R., Garg, A., & Kira, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.33007
D$^2$-VLA: Dual-Memory Dual-Frequency Vision-Language-Action Model For Long Dynamic ManipulationPreprint
Ye, Z., Wei, C., Huang, W., Zheng, A., Zou, C., Wu, L., Zhao, Z., Peng, Z., Yang, Y., Zhao, S., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.34792
Reactive Real-Time Flow Policies via Asynchronous Distribution AlignmentPreprint
Zoellner, M., O'Mahoney, R., Havoutis, I., & Paleja, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.36540
Rho: A Foundation for Efficiently Adaptable VLA ModelsPreprint
Rho Team, Bagaria, S., Chen, D., Fortier, D., Fu, J., Harrison, M., Hellebrekers, T., Joshi, N., Kolobov, A., Moore, D., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.38164
Action Chunking Proximal Policy Optimization with Feedback CorrectionPreprint
Hahn, S., & Choi, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.36250
Staircase Policy: Streaming Inference for World-Action Models with Large Action ChunksPreprint
Sun, G., Chen, C., Wang, J., Li, A., & Lv, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.36471
Toward Real-Time VLAs: Stage-Aware Two-Step Flow Denoising and System-Level EvaluationPreprint
Wu, D., Shen, R., Liu, P., Shen, Y., 尹貞漢, Zuo, S., Chen, X., Zheng, H., Zhang, L., Zhang, J., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.39822
Looking Back to Move Forward: Temporal Verification for Generative Robot PoliciesPreprint
Wang, H., Wu, W., Yan, Y., & Zhou, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.39038
Li, W., Su, X., Han, Y., Cao, Y., You, S., & Xu, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.39198 · Cited by 1
Generative World Models Enable Predictive Control of Laser Melt Pool DynamicsPreprint
Yan, Y., Bambach, M., & Afrasiabi, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.06250
RealtimeWAM: One-Step Asynchronous World Action ModelsPreprint
Lv, C., Du, J., Feng, S., Yong, Y., Gu, S., Yang, S., Gong, R., Ren, S., Zhang, T., & Wang, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.06617
The Unexpired Plan: A Free Monitor for Accelerated Diffusion PoliciesPreprint
Zhao, Y., & Scherer, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.05747
Wu, D., Liu, P., Chen, X., Zheng, H., Zhang, L., & Zhang, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.06090
RealtimeWAM: How Fast Can I Run My World Action Model?Preprint
Liu, H., Li, Y., Ji, K., Chen, X., Cui, H., Shen, Y., Meng, Y., Wu, C., Jiang, J., Li, B., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.10079
PathTime-VLA: Path-Time Decoupling for Factorized Post-Training of Vision-Language-Action PoliciesPreprint
Huang, Q., Yang, Y., Zou, Z., Chen, A., Zhu, Z., Wei, Y., Xiong, R., & Wang, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.11771
JitterVLA: A Replay Benchmark for Motion Quality in VLA-Driven Human-Robot ManipulationPreprint
Ogunwoye, F., Seyidov, D., Gubernatorov, K. N., & Tsetserukou, D. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.13628
SmolWAM: A Latent World-Action Model for Robotic Manipulation on Edge DevicesPreprint
Owolabi, T., Olawore, H., Adelakun, F., & Oyewumi, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.13740
PACE: Temporal-Role Tactile Conditioning for Chunk-Based Insertion PoliciesPreprint
Chen, Z., Xu, S., Chen, S., & You, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.16308
Foundation Models for Autonomous Space Robotics: Architectures, Physical Adaptation, and ValidationPreprint
Wang, Z., Deng, J., Lu, X., Jia, Z., Huang, Y., Zhu, Y., & Li, J. (2026) · arXiv (Cornell University) · doi:10.13140/rg.2.2.27769.22885
Spark-0: Scaling Spatially Aligned Tactile Pretraining for Robot Dexterous ManipulationPreprint
Chen, T., Lin, W., Lyu, C., Xiang, P., Tang, J., Nian, T., Cai, Z., Ye, K., Yu, C., Zhu, W., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.22342
Physical AI: A Survey of Models, Learning, Systems, and Open ChallengesPreprint
Gao, B., Gautam, A., Bagchi, A., Liang, Y., Li, W., Li, R., Zhou, Z., Mao, Y., Yan, X., He, X., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.21171
VeriVLA: Adaptive Test-Time Verification for Flow-Based VLAs with Latent World ModelsPreprint
Sun, X., Lu, Y., Liu, Y., Fan, X., Chen, X., Liu, Z., & Zhao, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.21985
Han, D., & Niranjani, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.22576
Learning Robot Metareasoning: Adaptive Computation Budgets for Real-Time ControlPreprint
Magnino, L., Marino, A., Amir, M., Shishika, D., & Prorok, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.23662
REACT: Responsive Execution-aware Action Coordination And Tuning For Moving Object Manipulation
Qin, X., Zhang, J., & Yi, A. (2026) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn69854.2026.11729198
FlashRT-PAI: Fast Kernels Are Not Fast Pipelines for Physical AIPreprint
Su, L., Chen, D., Li, J., Guo, K., & Lv, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.24349 · Cited by 1
Duan, S., Wang, Z., Pace, F. D., & Liarokapis, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.25006
Plausible Future, Wrong Choice: Steering World-Action Models toward Intended FuturePreprint
Li, Z., He, J., Yan, Y., Zhou, L., Li, Y., Li, Y., Zhang, T., Song, Y.-Y., Wei, X. W. A.-L., Hong, F., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.28431
Training in Deployment: Human-in-the-loop Online Post-Training for Robot ManipulationPreprint
Gu, C., Wei, X., Chen, Y., Zhang, B., Zhang, J., Zhu, X., & Zhang, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.29264
Lee, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2611.04392
Preserve or Repair? Selective Boundary Policy for a World Action Model
Xue, H., Wen, G., Liang, L., & Lang, C. (2026) · International Workshop on Human-centric Multimedia Analysis (HCMA) · doi:10.1145/3841192.3841755
Zero-HLM: Generalizable Humanoid Loco-Manipulation without Real Robot DataPreprint
Yang, D., Liu, M., Wang, Z., Liu, Y., Zhang, X., Li, Y., Ma, L., Yang, Y., Huang, W., Liang, W., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.35075
Rollback: Experience Backtracking in World Models for Policy ImprovementPreprint
Luo, J., Zhang, Y., Chen, L., Wang, N., Gao, C., Qiu, B., Li, Z., Peng, B., Li, C., Yan, Y., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.36068
Continuous-Action World Models: Learning Cross-Embodiment Spatiotemporal Dynamics PriorsPreprint
Li, L., & Liu, Q. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.40401
Srinivasan, V. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.39282
What Actually Matters? A Paired Evaluation of Inference Settings for $π_{0.5}$ on BEHAVIOR-1KPreprint
Hammele, R., Huber, S., Nazeri, M., Payandeh, A., Card, A. T., Nguyen, A., Pokhrel, A., Song, D., Xiao, X., & Pirk, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.37917
TaskFrame: Connecting Task-Level Planning with Frame-Level Control for Game AgentsPreprint
Wang, G.-Z., Pu, J., Wang, Y.-R., Yin, X., Chen, H.-H., Cheng, C., Li, L.-G., Zhou, Y.-N., Wu, J.-R., Jiang, Y.-N., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.40276 · Cited by 2
STAR: Steering Action Representations with Proprioception for VLA InferencePreprint
Zhao, G., Zhang, Y., Ru, H., He, K., Wen, H., Zhan, K., Zhang, Z., Li, Y., & Zhong, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.41009
Leap-VLA: High-Rate Robot Control from a Low-Rate VLM BackbonePreprint
Yang, L., Shen, Y., Li, Z., Tian, Z., Wang, Y., Huang, D., Wang, W., Zhu, Y., Jiang, Z., Xie, G., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2611.01011
Chen, F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2611.01390
Metadata
- Authors
- Black, K., Galliker, M., & Levine, S.
- Affiliations
- Physical Intelligence
- University of California, Berkeley (US)
- ORCID
- Levine, Sergey
- Venue
- Advances in Neural Information Processing Systems 38
- Published
- 2025
- Pages
- 37596-37620
- ISSN
- 1049-5258 (ISSN-L)
- OpenAlex
- W4417097926
- Cited by
- 12
- References
- 0
- Citations by year
- 2026: 12
- FWCI
- 26.12
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- neural information processing systems (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University) (cc-by)
- Topics
- Physical Sciences › Engineering › Control and Systems Engineering › Robot Manipulation and Learning
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › AI-based Problem Solving and Planning
- Keywords
- action chunking, real-time execution, vision-language-action models, flow-based generative models, diffusion models, inference latency, bimanual manipulation, temporal coherence
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.