AnchorRefine: Synergy-Manipulation Based on Trajectory Anchor and Residual Refinement for Vision-Language-Action Models

TL;DR

AnchorRefine introduces a hierarchical approach for vision-language-action tasks, separating trajectory planning from local refinement to enhance manipulation precision and success rates.

Contribution

The paper presents a novel hierarchical framework that factorizes action modeling into trajectory anchors and residual refinement, improving manipulation accuracy.

Findings

Achieves up to 7.8% success rate improvement in simulation.

Yields up to 18% success rate increase in real-world tasks.

Enhances both regression-based and diffusion-based VLA models.

Abstract

Precision-critical manipulation requires both global trajectory organization and local execution correction, yet most vision-language-action (VLA) policies generate actions within a single unified space. This monolithic formulation forces macro-level transport and micro-level refinement to be optimized under the same objective, causing large motions to dominate learning while suppressing small but failure-critical corrective signals. In contrast, human manipulation is structured by global movement planning together with continuous local adjustment during execution. Motivated by this principle, we propose AnchorRefine, a hierarchical framework that factorizes VLA action modeling into trajectory anchor and residual refinement. The anchor planner predicts a coarse motion scaffold, while the refinement module corrects execution-level deviations to improve geometric and contact precision. We further introduce a decision-aware gripper refinement mechanism to better capture the discrete and boundary-sensitive nature of gripper control. Experiments on LIBERO, CALVIN, and real-robot tasks demonstrate that AnchorRefine consistently improves both regression-based and diffusion-based VLA backbones, yielding gains of up to 7.8% in simulation success rate and 18% in real-world success rate.