Paper Overview
Field: Machine Learning Authors: Xin Chen, Sen Chen, Yujuan Ding, Jian Liu, Guoqing Wang, Wei Ye, Heng Tao Shen, Yi Bin Posted: 2026-09-17 arXiv: 2609.20776
Summary
Action chunking is widely used for action generation and execution in Vision-Language-Action (VLA) policies, yet existing approaches commonly use a fixed action horizon. During a rollout, different task stages may require different levels of action continuity, control precision, and closed-loop feedback, making a fixed horizon unable to accommodate changing control requirements.
The authors propose GeoAAC, a geometry-based adaptive action chunking method for flow-based VLA policies that adjusts the action horizon according to the reliability of the current action prediction. Key contributions:
- The geometry of Flow Matching denoising trajectories provides process-level information for characterizing prediction reliability; geometric variation across action prefixes remains positively correlated with prediction uncertainty.
- GeoAAC leverages this prefix-level geometry to construct horizon-level geometric profiles and adaptively determine the action horizon from a single generation pass, with no additional training required.
- Experiments on GR00T N1.5 and π0.5 across LIBERO, LIBERO-Pro, RoboCasa365, and real-world manipulation tasks show consistent improvements over fixed action horizon baselines and existing adaptive methods: up to 8.7 percentage points in simulation, and real-world average success rate improving from 53.3% to 74.4%.