arXiv Robotics — research abstracts· Jasper Gerigk, Kenzo Aspuru-Takata, Chin-Hsuan Wu, Mohammad Mohammadi, Shuhong Zheng, Igor Gilitschenski·· 2 days agoEditorial score42
Enhancing VLA Generalization via Task Scrubbing
When Listening Becomes Easier: Scrubbing Visual Cues for Shortcut-Free VLAs
Summary
VLA models face shortcut learning from visual cues. Task scrubbing reduces reliance on visual shortcuts, improving out-of-distribution robustness without policy rollouts.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org