Adaptive verification improves quality of robotic actions in manipulation
VeriVLA: Adaptive Test-Time Verification for Flow-Based VLAs with Latent World Models
Robotics
Summary
Robotic systems that decide actions by sampling can sometimes pick a poor choice, which can cause tasks like inserting a pen or placing an apple to fail. The authors designed a way to check action quality during execution by comparing multiple candidate actions and predicting their effects in a simplified internal model. This step helps the robot choose better actions without retraining, leading to more successful task completion in both real and simulated tests.
What this means in practice
- •For robotic automation teams: Choose higher-quality robot actions during manipulation tasks by verifying action outcomes without retraining the policy.
- •For industrial robotics engineers: Reduce task failures in fine manipulation by dynamically ranking candidate action sequences based on predicted task success.
Authors
Xueli Sun, Yuxiang Lu, Yantong Liu, Xianzhe Fan, Xi Chen, Zhe Liu, Hengshuang Zhao
Abstract
Flow-based Vision-Language-Action (VLA) policies are inherently stochastic: different noise seeds can produce action chunks of varying quality from the same observation. At test time, however, the policy typically commits to a single sample without an explicit estimate of its quality. In fine-grained manipulation, executing a suboptimal action can irreversibly disrupt task progress and lead to failure. To exploit policy stochasticity, we introduce a training-free two-axis trigger that uses two initial samples to measure within-chunk temporal variation and cross-sample variance, detecting the critical states. Once triggered, a JEPA-based latent world model ranks candidate action chunks by predicting action-conditioned future latent states and the corresponding changes in task progress. By operating in latent space, the verifier avoids pixel reconstruction and enables efficient and more accurate ranking. Across three real-world tasks, $\textit{i.e.,}$ pen insertion, pipetting, and apple placement), our method raises the average completion score from 50.8% to 65.5%, while adding 326 ms for sampling and verification per triggered decision on a single RTX 4090. Consistent gains are also observed in simulation, where the average success rate rises from 66.4% to 73.6% in the clean setting of RoboTwin 2.0, from 62.8% to 69.6% in its randomized setting, and from 96.9% to 97.6% on LIBERO.