General
RePlan-VLA: Progress-Aware Replanning for Long-Horizon Vision-Language-Action Models
arXiv cs.RO0 views
arXiv:2611.23843v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models have emerged as a promising paradigm for general-purpose robotic manipulation, yet their performance often degrades on long-horizon tasks that require multiple interdependent manipulations. Instead of burdening the VLA policies with long-horizon reasoning, memory, action generation, progress tracking, and error recovery, we argue for leveraging their short-horizon manipulation capabilities within an execution-aware framework. To this end, we introduce \textnormal{\textbf{RePlan-VLA}}, a VLA-agnostic framework that translates a VLA's short-horizon capabilities into reliable long-horizon execution. RePlan-VLA explicitly monitors the robot's motion for progress evaluation, and adaptively revises the future plan based on observed outcomes. We preserve the underlying VLA policy entirely, requiring no additional fine-tuning. Experiments on RoboCasa365 and CALVIN benchmarks show an improvement of 21.9% and 10.4% in success rate over the next strongest baseline. We also validate our approach for real-world deployment on a Franka Research 3 robot.
Source:https://arxiv.org/abs/2611.23843PreviousEstimation-Aware ReachabilityNextLAGR: Language-Aware Grasp Re-Ranking for Task Oriented Grasping
Related posts
Uber tests rides for kids as young as 10, with extra pay for drivers
0TechCrunch
MAHA PAC spent more on steak, luxury club than candidates, report shows
0Ars Technica
Man hired by Russia spied on Silicon Valley tech and defense company, feds say
0Ars Technica

FCC greenlights rule allowing prisons to jam cellphone signals
0The Verge
Tuyển Anh và hệ quả kỷ luật từ UEFA sau trận thắng Tây Ban Nha tại Nations League
0dev.to Data
SakaQL: A new general-purpose relational programming language
0Hacker News New