RAVEL: Asynchronous Rolling Inference for Flow-Based Vision-Language-Action Models

Yuhan Chen; Ke Yu; Pengfei Liu; Shuxun Wang; Yi Yang; Linchao Zhu

cs.ROSep 28, 2026

RAVEL: Asynchronous Rolling Inference for Flow-Based Vision-Language-Action Models

Authors: Yuhan Chen, Ke Yu, Pengfei Liu, Shuxun Wang, Yi Yang, Linchao Zhu

Organizations: Zhejiang University · Nanyang Technological University

Abstract

Flow-based vision-language-action (VLA) models are highly effective for generalist robot manipulation, yet their reliance on computationally expensive VLM encoding and multi-step iterative action generation imposes a significant latency bottleneck. The resulting inference latency makes it difficult for robots to respond quickly, especially in dynamic environments. We address this limitation with RAVEL (Rolling Asynchronous VLA Enabling Low-Latency Control), an asynchronous inference framework that addresses the computational bottlenecks of both the VLM backbone and the action expert. To reduce the delay from multi-step action denoising, RAVEL allows near-term actions to be executed after a single denoising step by carrying partially denoised future actions forward in a rolling buffer. To avoid blocking on slow VLM encoding, RAVEL decouples VLM encoding from rolling action generation, allowing the action expert to operate continuously using the latest available VLM context, while a lightweight Fast Observation Pathway (FOP) directly conditions the action expert on current observations. Across simulated and real-world manipulation tasks, RAVEL consistently achieves substantially lower response latency while maintaining the task capability of the underlying VLA, enabling high-frequency and responsive closed-loop control.

Figures & tables

Figure 1: Responsive control with RAVEL . Synchronous inference waits for VLM encoding and multi-step action denoising before executing new actions. RAVEL runs VLM context updates in parallel with rolling action denoising and execution, enabling faster responses to new observations.

Figure 2: Overview of RAVEL . The VLM begins encoding observation $o_{t}$ at time $t$ and produces VLM context $\mathbf{c}_{t}$ between $t+1$ and $t+2$ , while the action expert continues using the previously available context. At $t+2$ , the expert uses $\mathbf{c}_{t}$ with a temporal offset of two control steps. Meanwhile, FOP directly conditions the action expert on the current observation $o_{t+2}$ . Conditioned on these inputs, the expert advances the rolling buffer by one denoising step and releases the first action. The remaining actions are retained, and a new noisy action is appended.

Figure 3: Synchronous evaluation of the baseline and RAVEL under matched generation budgets. Results aggregate 2,000 LIBERO trials and 5,000 RoboTwin 2.0 trials.

Model	Full denoising (ms) $\downarrow$	Rolling inference (ms) $\downarrow$	Speedup
X-VLA	142.58	54.97	2.59 $\times$
$\pi_{0.5}$	272.86	93.45	2.92 $\times$

Table 1: Median first-action latency for X-VLA and

\pi_{0.5}

using PyTorch inference on one RTX 4090, measured from observation receipt to the first predicted action becoming available for execution.

	Interaction			Perception			Generalization			Average
Method	CR	DA	LS	VU	SR	MP	VG	MG	DR	SR $\uparrow$	Time $\downarrow$
DynamicVLA	25.5	22.0	13.5	38.5	20.0	23.0	22.0	53.5	16.5	26.06	7.85
+ CI & LAAS	46.0	23.5	28.5	38.5	39.0	23.0	39.5	58.5	23.0	35.50	6.58
+ VLASH	18.5	16.0	9.5	13.5	13.5	17.0	23.5	40.0	17.5	18.78	7.13
+ FASTER	37.0	19.0	14.5	20.5	22.5	24.0	19.0	45.0	12.5	23.78	7.11
+ RAVEL	49.0	29.0	38.0	48.0	46.0	38.5	50.0	70.5	27.5	44.06	6.04

Table 2: Success rates (%) and mean completion time over common-success episodes on DOM. All results are obtained from local evaluations using two NVIDIA RTX 4090 GPUs, with simulation and policy inference on separate GPUs. Bold and underlined values indicate the best and second-best results, respectively.

Figure 4: Real-world table-tennis return sequences for RAVEL (top) and Naive Async (bottom). The faster response of RAVEL enables a timely paddle swing that contacts the incoming ball and returns it farther. Slow inference in Naive Async delays the swing, so the ball hits the paddle handle before the robot starts to move. Frames progress from left to right within each rollout; columns do not indicate synchronized timestamps across methods.

Method	Table-tennis return		Tennis-ball grasping
	Success (%) $\uparrow$	Obs. age (ms) $\downarrow$ P50 / P95	Success (%) $\uparrow$	$E_{\mathrm{HF}}\downarrow$ (mm 2 )	TCP jerk $\downarrow$ (mm/s 3 )
Sync	0	327.05 / 482.54	40	14.45	1983
Naive Async	0	221.88 / 287.35	45	19.43	3208
RTC	0	418.88 / 563.39	40	101.70	8398
VLASH	25	283.53 / 441.56	–	–	–
FASTER	25	149.63 / 212.44	–	–	–

Table 3: Real-world evaluation. Success rates are computed over 20 trials per method for each task. Observation age measures visual-input freshness at action dispatch. P50 and P95 denote the median and 95th percentile of observation age, respectively.

E_{\mathrm{HF}}

and TCP jerk RMS characterize trajectory oscillations and motion smoothness, respectively. Bold indicates the best result for each metric.

Figure 5: Ablation of temporal alignment components. (a) Training-time alignment for

\pi_{0.5}

on LIBERO and RoboTwin 2.0: the base checkpoint under rolling inference, noise alignment only, shifted-horizon training without temporal-offset conditioning, and full RAVEL . All variants use 2,000 episodes on LIBERO and 5,000 episodes on RoboTwin. (b) Robustness to delayed observations on LIBERO with X-VLA under synchronous rollouts; each point uses 400 trials. (c) Temporal-offset bias for execution-delay compensation in real-world table-tennis return with a fixed policy checkpoint. Success peaks at 90% with

\delta_{0}=0.3

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

Benchmark	Backbone	Action representation	$H$	$h$	Steps	Batch
LIBERO	$\pi_{0.5}$	Relative EE pose	10	1	5K	128
LIBERO	X-VLA	Absolute EE pose	30	3	5K	128
RoboTwin 2.0	$\pi_{0.5}$	Relative joint position	30	3	10K	128
RoboTwin 2.0	X-VLA	Absolute EE pose	30	3	10K	128

Appendix

Table 4: Training and evaluation configurations for synchronous evaluation. Batch size is global; steps denote fine-tuning steps for RAVEL . EE denotes end-effector pose.

Component / inference path	P50 (ms)	P95 (ms)
VLM encoding	63.47	84.34
Action expert	31.75	48.75
Action expert with FOP	41.4	58.6
Full model	370.7	433.5

Appendix

Table 5: DynamicVLA inference latency measured locally on an NVIDIA RTX 4090 GPU under our configuration. All values are in milliseconds; P50 and P95 denote the median and 95th percentile.

Figure 6: Real-world table-tennis setup.

Setting	Task policy	VLASH	FASTER	RAVEL
Initialization	$\pi_{0.5}$	Task policy (10K)
Training scope	LoRA	Full model	Full model	Action expert
Trainable parameters	29.89M	3.35B	3.35B	431.15M
Evaluation checkpoint step	10K	10K	10K	10K
Batch size	128	128	128	64
Learning rate	$5\times 10^{-5}$	$5\times 10^{-5}$	$2.5\times 10^{-5}$	$5\times 10^{-5}$

Appendix

Table 6: Training configurations for table-tennis return.

Setting	Task policy	RAVEL
Initialization	$\pi_{0.5}$	Task policy (10K)
Training scope	LoRA	Action expert
Trainable parameters	29.89M	431.15M
Evaluation checkpoint step	10K	5K
Batch size	128	128
Learning rate	$5\times 10^{-5}$	$10^{-5}$

Appendix

Table 7: Training configurations for tennis-ball grasping.

Figure 7: Action–state delay on the xArm7. Joint-space execution traces from four table-tennis trials with temporal-offset bias

\delta_{0}=0.3

. Each panel shows Joints 2, 4, and 6 from 2 s after the first command, omitting the initial stationary period. These joints have the largest commanded excursions; each joint uses the same vertical scale across trials. Blue: dispatched action; dashed green: controller-reported target; orange: measured state.

Explore similar work

Understanding Asynchronous Inference Methods for Vision-Language-Action Models
May 4, 2026Ayoub AgouzoulInference Latency Fast Inference
École polytechnique · Systems Group, ETH Zürich
ReactVLA: Fast and Lightweight Reactive Robot Manipulation via Improved Mean Flow Action Generation
Jun 12, 2026Yanzhao Guo, Wenkai Chen, Jianwei ZhangDiffusion-Based Vision-Language-Actions Action Generation
Shanghai Jiao Tong University · Technical Aspects of Multimodal Systems (TAMS), Department of Informatics, Universität Hamburg
Acting While Understanding: Asynchronous Semantic-Action Decoupling for Real-Time Vision-Language-Action Models
Jun 13, 2026Shenhao Yan, Ge Wang, Qi Liu +7Diffusion-Based Vision-Language-Actions Robotic Manipulation
Northeastern University · Ising AI · CUHK-Shenzhen

May 4, 2026, cs.ROJ/K move · Enter open · S save

Understanding Asynchronous Inference Methods for Vision-Language-Action Models

Ayoub Agouzoul

École polytechnique · Systems Group, ETH Zürich

Inference Latency Fast Inference Libero Manipulation Benchmark

Open paper AskPDF arXiv

Vision-Language-Action (VLA) models offer a promising path to generalist robot control, but their inference latency causes observation staleness when generated actions are executed asynchronously. Several methods have been proposed concurrently to mitigate this problem: inference-time inpainting (IT-RTC), training-time delay simulation (TT-RTC), future-state-aware conditioning (VLASH), and lightweight residual correction (A2C2). Each takes a fundamentally different approach, but they have so far been evaluated independently with different codebases, base policies, and protocols. We present a systematic comparison of these four methods under controlled conditions. We develop two unified codebases that integrate all methods with harmonized library and dataset versions, and we benchmark them on the Kinetix suite with MLPMixer policies and on the LIBERO manipulation benchmark with SmolVLA, sweeping inference delays up to

d=20

control steps. A2C2's per-step residual correction is the most effective method on Kinetix, holding above 90% solve rate up to

d=8

, and also leads on LIBERO from

d=4

onwards. IT-RTC is competitive at low delays but degrades sharply under long chunks (

H=30

) and high delays. TT-RTC is the most robust training-based method: stable across

d_\max

choices, generalizes beyond its training delay distribution, and adds zero inference overhead. VLASH exhibits a clear low-delay vs. high-delay trade-off governed by the fine-tuning delay range

[0,d_\max]

. Code is available at https://github.com/TheAyos/async-vla-inference

Abstract

Explore similar work

Understanding Asynchronous Inference Methods for Vision-Language-Action Models

ReactVLA: Fast and Lightweight Reactive Robot Manipulation via Improved Mean Flow Action Generation

Acting While Understanding: Asynchronous Semantic-Action Decoupling for Real-Time Vision-Language-Action Models