VLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action Reuse

Sicheng He; Yi Pan; Qiuting Yu; Jiaming Huang; Qi Lu; Shuyan Wu; Hongyan Feng; Xuanyu Liu

cs.AIOct 18, 2026

VLA-Rolling: Fewer Denoising Steps, Better Control via Adaptive Action Reuse

Authors: Sicheng He, Yi Pan, Qiuting Yu, Jiaming Huang, Qi Lu, Shuyan Wu, Hongyan Feng, Xuanyu Liu,

Organizations: Zhejiang University

Abstract

Flow-matching vision-language-action (VLA) models generate an action chunk by iteratively denoising Gaussian noise, but typically execute only a short prefix before replanning with updated observations. Under this receding-horizon paradigm, the unexecuted remainder---the action tail---is discarded, and the next chunk is regenerated from fresh noise. We show that this discarded tail retains transferable denoising progress: action-tail initialization can match or outperform a native 10-step Gaussian baseline using fewer denoising steps. However, the optimal denoising budget varies substantially across samples, indicating that tail reuse should be adapted to the updated environment. Based on this observation, we introduce VLA-Rolling, a lightweight inference framework that reuses action tails through three stages: action-tail inheritance, per-action confidence estimation, and confidence-guided refinement through renoising and denoising budget allocation, while the underlying VLA remains frozen. Across $π_{0.5}$ , SmolVLA, X-VLA, and MolmoAct2 on LIBERO, MetaWorld, LIBERO-Plus, and RoboTwin 2.0, VLA-Rolling reduces action-chunk inference cost in all 16 model--benchmark pairs, achieving up to a $1.48\times$ speedup, while improving success rate in 14 pairs by up to $15.08$ percentage points. These results show that action tails are not merely discarded predictions, but reusable intermediate states that enable more efficient and accurate VLA inference. These results demonstrate that the action tail can support the generation of better actions with fewer denoising steps. Our project is available at https://chenbao123ww.github.io/VLA-Rolling/.

Figures & tables

Figure 1: VLA-Rolling reuses rather than discards action tails. Confidence controls both how strongly inherited actions are renoised and how many denoising steps refine the next chunk.

Figure 2: Action tails recover a 10-step Gaussian prediction in fewer steps. Tail initialization crosses the Gaussian baseline in four steps for $\pi_{0.5}$ and six for SmolVLA on both benchmarks. Subtle shaded bands mark these earliest shared operating points.

Figure 3: Action tails require sample-dependent refinement. Each bar shows the fraction of samples with a given optimal denoising step $N^{*}$ . Green denotes $N^{*}$ at or below the model-specific median (8 for $\pi_{0.5}$ , 9 for SmolVLA), while red denotes $N^{*}$ above it. Dashed lines mark the median boundaries; brackets report the fraction of samples in each region. Means and standard deviations appear inside the panels.

Figure 4: Overview of VLA-Rolling. VLA-Rolling reuses the unexecuted action tail from the previous inference round by estimating its confidence, which jointly determines tail preservation and the adaptive denoising budget while keeping the original VLA frozen. (a) VLA-Rolling inference. The Action Confidence Estimator predicts a confidence which controls its interpolation with Gaussian noise. The resulting initialization is refined using $N_{\mathrm{adapt}}$ denoising steps. The executable prefix is sent to the robot, while the remaining tail rolls into the next round. (b) VLA-Rolling training. Confidence labels are constructed from the quality of inherited actions and normalized with EMA-based quantile statistics.

Model	LIBERO		MetaWorld		LIBERO-Plus		RoboTwin
Model	SR (%) $\uparrow$	Cost (ms) $\downarrow$	SR (%) $\uparrow$	Cost (ms) $\downarrow$	SR (%) $\uparrow$	Cost (ms) $\downarrow$	SR (%) $\uparrow$	Cost (ms) $\downarrow$
$\pi_{0.5}$	96.38	168.09	67.22	222.69	84.73	167.65	66.54	172.92
+VLA-Rolling	97.83	129.60	67.88	184.21	85.42	147.40	67.26	164.27
$\Delta$ (pp/%)	+1.45	-22.90%	+0.66	-17.28%	+0.69	-12.08%	+0.72	-5.00%
X-VLA	96.03	135.24	57.08	123.31	66.04	136.80	60.68	131.09
+VLA-Rolling	96.65	112.79	61.46	90.05	65.91	121.93	60.40	96.99

Table 1: Performance and inference efficiency comparison with VLA-Rolling. SR denotes the average success rate across tasks, and Cost denotes the inference time per action chunk. Bold data indicates an improvement compared to the baseline.

Method	LIBERO		MetaWorld		LIBERO-Plus
Method	SR (%) $\uparrow$	Cost (ms) $\downarrow$	SR (%) $\uparrow$	Cost (ms) $\downarrow$	SR (%) $\uparrow$	Cost (ms) $\downarrow$
$\pi_{0.5}$	96.38	166.04	67.22	217.28	84.73	161.92
Variants	$\Delta$ SR (pp) $\uparrow$	$\Delta$ Cost (%) $\downarrow$	$\Delta$ SR (pp) $\uparrow$	$\Delta$ Cost (%) $\downarrow$	$\Delta$ SR (pp) $\uparrow$	$\Delta$ Cost (%) $\downarrow$
$+A$	-2.13	+0.02%	+0.63	+0.01%	-4.89	+0.02%
$+A+B_{\mathrm{fixed}}$	+0.20	+0.02%	+0.61	+0.02%	-0.16	+0.02%
$+A+B$	+0.37	+0.55%	+0.40	+0.42%	-0.22	+0.84%

Table 2: Ablation study of VLA-Rolling inference components with

\pi_{0.5}

on LIBERO, MetaWorld, and LIBERO-Plus. A, B, C denote different components of our model. A : Action Tail Inheritance; B : Confidence-Guided Renoising; C : Adaptive-Step Denoising.

B_{\mathrm{fixed}}

: Fixed 50% of renoising. The full VLA-Rolling model corresponds to A+B+C.

Execution Horizon $K$	$\Delta o_{t}$	$\Delta q_{t}$	$c_{min}$	$\rho(\Delta o_{t},c_{min})$	$\rho(\Delta q_{t},c_{min})$
$K=$ 10(default)	10.544	0.895	0.359	$-0.257$	$-0.237$
$K=$ 20	16.071	1.610	0.175	$-0.318$	$-0.207$
$K=$ 30	19.249	2.151	0.146	$-0.272$	$-0.200$
$K=$ 40	20.658	2.551	0.140	$-0.100$	$-0.155$

Table 3: Observation shift and confidence statistics under different execution horizon

K

. Correlation values

\rho

are Pearson correlations computed within each

K

group.

c_{min}

is the average minimum of predicted confidence.

Figure 5: Left: Average denoising steps across task difficulty levels on MetaWorld. Successful trajectories consistently require more denoising steps than failed trajectories. Right: Distribution of episodes across different difficulty levels

Figure 6: Real-world experiment of context mutations revealed a significant decrease in confidence at the mutation site, leading to an increase in the number of denoising steps.

Model	SR (%) $\uparrow$	Cost (ms) $\downarrow$
$\pi_{0.5}$	70%(14/20)	227.28
+VLA-Rolling	75%(15/20)	176.50

Table 4: Model performance on real-world experiments.

Appendix figures & tables5 assets

Supplementary material from the paper’s appendix.

Appendix

$\alpha$	Success Rate (%)	Average Denoising Steps	Model Inference (ms/chunk)
0.0	97.25	6.861	206.7
1.0	97.00	6.155	187.6
1.5 (default)	98.50	6.332	195.7
2.0	97.25	6.349	190.9

Appendix

Table 5: Sensitivity of VLA-Rolling performance to the calibration parameter

\alpha

on LIBERO. We test 40 tasks with 10 episodes per task.

Benchmark	variation	Mean Trajectory Error	Std.	Avg. Denoising Steps
LIBERO	min	53.739	17.311	6.33
	mean	54.420	17.964	4.36
	max	55.989	15.979	2.77
MetaWorld	min	6.046	2.772	7.23
	mean	6.165	2.909	5.63
	max	6.585	2.953	4.12

Appendix

Table 6: Offline trajectory-level action error under different denoising step convert strategies.

VLA-Rolling Model	Success Rate (%)	VLA-Rolling Cost (ms)
10M	$97.25\pm 5.47$	$1.267\pm 0.105$
30M	$98.00\pm 5.10$	$1.579\pm 0.335$
100M	$97.25\pm 7.41$	$1.731\pm 0.047$

Appendix

Table 7: Performance and inference cost of VLA-Rolling models on LIBERO. Success rate statistics are computed across 40 tasks, 10 episodes per task

Benchmark	$N-1$ $\uparrow$	$N$ $\uparrow$	$N+1$ $\uparrow$
LIBERO	97.00 ( $-0.83$ )	97.83	96.75 ( $-1.08$ )
MetaWorld	66.67 ( $-1.21$ )	67.88	67.27 ( $-0.61$ )

Appendix

Table 8: N* Locally Optimal Test.

Benchmark	Image Cosine	Image $(1-\mathrm{MAE})$	State Cosine	Standardized State Distance
LIBERO	$0.953\pm 0.028$	$0.951\pm 0.019$	$0.998\pm 0.003$	$0.309\pm 0.222$
MetaWorld	$0.992\pm 0.007$	$0.986\pm 0.009$	$0.991\pm 0.012$	$0.506\pm 0.307$

Appendix

Table 9: Observation and proprioceptive-state similarity across a 10-step execution horizon. We compare the source context at time

t

with the updated context at time

t+10

. Image Cosine denotes the cosine similarity between the two visual representations; Image

(1-\mathrm{MAE})

denotes one minus the mean absolute error between the normalized visual representations; State Cosine measures the cosine similarity between the corresponding proprioceptive states

Efficient VLA Model Inference

Efficient vision-language-action (VLA) model inference reduces robot-control latency, memory use, and energy while preserving task performance.

See all 190 papers

StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models
Oct 6, 2026Shangyuan Yuan, Xinda Qi, Yujiang Pu +2Efficient VLA Model Inference Vision-Language-Action Models
Michigan State University · Ant Technology U.S., Inc.
Toward Real-Time VLAs: Stage-Aware Two-Step Flow Denoising and System-Level Evaluation
Sep 30, 2026Di Wu, Rongtian Shen, Ping Liu +8Robotic Control Real-Time Robot Control
Magic-Lab Team, Magiclab Robotics Inc. · Southeast University · Harbin Institute of Technology +3
CF-VLA: Efficient Coarse-to-Fine Action Generation for Vision-Language-Action Policies
Apr 27, 2026Fan Du, Feng Yan, Jianxiong Wu +8Robotic Control Efficient VLA Model Inference
1Southern University of Science and Technology · 2Xi’an Jiaotong University · 3United Nova Technology +1

Oct 6, 2026J/K move · Enter open · S save

StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models

Shangyuan Yuan, Xinda Qi, Yujiang Pu, Wenliang Guo+1

Michigan State University · Ant Technology U.S., Inc.

Efficient VLA Model Inference Vision-Language-Action Models VLMs for Robotics

Open paper AskPDF arXiv

Vision-language-action (VLA) models increasingly rely on diffusion- or flow-matching-based action heads to generate continuous robot actions. These action heads typically process the denoising trajectory in a largely uniform manner. However, we observe that the conditioning focus naturally shifts across denoising stages: early stages combine language instructions and visual observations to establish a coarse action trajectory, whereas later stages place greater emphasis on current visual observations for action alignment. Based on this insight, we introduce StairVLA, a stage-aware hierarchical action generation framework that uses partially denoised actions as a natural interface between coarse long-horizon action generation and local refinement. A high-level VLA performs early denoising to produce a reusable long-horizon partially denoised action trajectory, while a lightweight refiner operates at a higher frequency to refine local action chunks using the latest observations. This design amortizes expensive high-level VLA computation while preserving frequent closed-loop correction. On LIBERO, our GR00T-style instantiation improves average success from 96.5% to 97.8% while reducing amortized inference latency from 115.0 ms to 44.2 ms per action chunk. More broadly, across two VLA backbones, simulation benchmarks, and real-robot tasks, StairVLA consistently reduces inference cost while maintaining strong task performance.

Abstract

Efficient VLA Model Inference

Explore similar work

StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models

Toward Real-Time VLAs: Stage-Aware Two-Step Flow Denoising and System-Level Evaluation

CF-VLA: Efficient Coarse-to-Fine Action Generation for Vision-Language-Action Policies