paper/arxiv/2608.25585

Robotics Dataset

This paper introduces RA-VLA, a retrieval-augmented Vision-Language-Action model for training-free test-time adaptation in robotic manipulation, evaluated on LIBERO benchmark and real-world UR5e environment.

Catalog metadata

Repository owner

Sanghwan Jang, Minjin Jeon, Minsoo Kim et al.

Source

Research paper

Task

RA-VLA: Retrieval-Augmented VLA for Test-Time Adaptation

Cameras

0

Browse the catalog to compare robots, tasks, sensors, and licenses. Sign in to open protected download and model links.