cs.RONov 19, 2026

Dune: A Zeroth-Order Optimizer with Time-Conditioned Historical Error Regression for VLA Test-Time Adaptation

Authors: Chengyang Dong, Ze Chen, Kai Ye, Chutian Wang, Heng Xu, Wei Mo, Maomao Li, Jiayan Xu, 

Organizations: Statistics and Mathematics, University of Glasgow · Shenzhen University of Advanced Technology · The University of Hong Kong · Beijing Normal University · 

Abstract

Test-Time Adaptation (TTA) enables Vision-Language-Action (VLA) models to handle distribution shifts during deployment without source-domain retraining. However, backpropagation-based adaptation incurs heavy memory overhead on resource-constrained robots, whereas zeroth-order (ZO) methods suffer from high estimation variance under limited perturbation budgets. We observe that while ZO estimation enables low-memory adaptation with rapid initial gains, it often stalls or degrades over long-horizon updates due to noisy gradient estimates. To address these limitations, we propose Dune, a zeroth-order optimizer that stabilizes long-horizon adaptation by coupling online historical error prediction with current perturbation observations. Specifically, Dune learns a lightweight, time-conditioned error predictor from past optimization dynamics and fuses its predictions with instantaneous ZO estimates, achieving effective variance reduction without backpropagation or extra forward passes. Additionally, we theoretically characterize how estimation variance restricts convergence in high-dimensional, small-population regimes, motivating the use of historical statistics. Extensive experiments on LIBERO and real-world robots demonstrate that Dune achieves superior success rates under a low memory footprint, with ablations confirming improved closed-loop stability and post-disturbance recovery.

Figures & tables

Explore similar work

CardsList
  1. Curvature-Aware Zeroth-Order Optimization for Memory-Efficient Test-Time Adaptation

    Aug 12, 2026Junming Zhang, Shuyu Yin, Peilin Liu +2On-Device TrainingTest-Time Adaptation